Neuroimaging: Structural Imaging: Magnetic Resonance Imaging, Computed Tomography

Published on 12/04/2015 by admin

Filed under Neurology

Last modified 22/04/2025

Print this page

rate 1 star rate 2 star rate 3 star rate 4 star rate 5 star
Your rating: none, Average: 0 (0 votes)

This article have been viewed 5879 times

Chapter 33A Neuroimaging

Structural Imaging: Magnetic Resonance Imaging, Computed Tomography

Computed Tomography

Computed tomography (CT; other terms include computer assisted tomography [CAT]) has been commercially available since 1973. The term tomography (i.e., to slice or section) refers to a process for generating two-dimensional (2D) image slices of an examined organ of three dimensions (3D). CT imaging is based on the differential absorption of x-rays by various tissues. X-rays are electromagnetic waves with wavelengths falling in the range of 10 to 0.01 nanometers on the electromagnetic spectrum. X-rays can also be described as high-energy photons, with corresponding energies varying between 124 and 124,000 electron volts, respectively. X-rays in the higher range of energies, known as hard x-rays, are used in diagnostic imaging because of their ability to penetrate tissue yet (to an extent) also be absorbed or scattered differentially by various tissues, allowing for the generation of image contrast.

Owing to their high energy, x-rays are also a form of ionizing radiation, and the health risks associated with their use, although minimal, should always be accounted for in diagnostic imaging. The x-rays generated by the x-ray source of the CT scanner are shaped into an x-ray beam by a collimator, a rectangular opening in a lead shield. The beam penetrates the slab of tissues to be imaged, which will absorb/deflect it to a varying degree depending on their atomic composition, structure, and density (photoelectric effect and Compton-scattering). The remaining x-rays emerge from the imaged slab and are measured by detectors located opposite the collimator. In fourth-generation CT scanners, the detectors are in a fixed position and the x-ray source rotates about the patient. As the beam of x-rays is transmitted through the imaged body part, sweeping a 360-degree arc for each slice imaged, the emerging x-rays are collected, then a computer analyzes the output of the detectors and calculates the x-ray attenuation of each individual tissue volume (voxel).

The degree of x-ray absorption by the various tissues is expressed and displayed as shades of gray in the CT image. Darker shades correspond to less attenuation. The attenuation by each voxel of tissue is projected on the flat image of the scanned slice as a tiny quadrilateral, generally square, called a pixel or picture element. Depending on the reconstruction matrix, a slice will be represented by more or fewer pixels, corresponding to more or less resolution. The shade of gray in each pixel corresponds to a number on an arbitrary linear scale, expressed as Hounsfield units (HU). This number varies between approximately −1000 and 3000+, with values of greater magnitude corresponding to tissues or substances of greater radiodensity, which are depicted in lighter tones. The −1000 value is for air, 0 is for water. Bone is greater than several hundred units, but cranial bone can be 2000 or even more. Fresh blood (with a normal hematocrit) is about 80 units, fat is −50 to −80. Tissues or materials with higher degrees of x-ray absorption, shown in white or lighter shades of gray, are referred to as hyperdense, whereas those with lower x-ray absorption properties are hypodense; these are relative terms compared to other areas of any given image.

By changing the settings of the process of transforming the x-ray attenuation values to shades on the grayscale, it is possible to select which tissues to preferentially display in the image. This is referred to as windowing. Utilizing a bone window, for instance, is very useful for evaluating fractures in cases of craniofacial trauma (Fig. 33A.1).

In CT, imaging contrast agents are frequently used for the purpose of detecting abnormalities that cause disruption of the blood-brain barrier (BBB) (e.g., certain tumors, inflammation, etc.). The damaged BBB allows for the net diffusion of contrast material into the site of pathology, where it is detected; this is referred to as contrast enhancement. Contrast materials used in CT scanning contain iodine in an injectable water-soluble form. Iodine is a heavy atom; its inner electron shell absorbs x-rays through the process of photoelectric capture. Even a small amount of iodine effectively blocks the transmitted x-rays so they will not reach the detector. The high x-ray attenuation/absorption will result in hyperdense appearance in the image. Other CT techniques requiring contrast administration are CT angiography, CT myelography, and CT perfusion studies.

More than 20 years ago, a fast-imaging technique called spiral (or helical) CT scanning was introduced to clinical practice. With this technique, the x-ray tube in the gantry rotates continuously, but data acquisition is combined with continuous movement of the patient through the gantry. The circular rotating path of the x-rays, combined with the linear movement of the imaged body, results in a spiral or helix-shaped x-ray path, hence the name. These scanners can acquire data rapidly, and a large volume can be scanned in 20 to 60 seconds. This technique offers several advantages, including more rapid image acquisition. During the short scan time, patients can usually hold their breath, which reduces/minimizes motion artifacts. Timing of contrast bolus administration can be optimized, and less contrast material is sufficient. The short scan time, optimal contrast bolus timing, and better image quality are very useful in CT angiography, where cervical and intracranial blood vessels are visualized. These images can also be reformatted as 3D views of the vasculature, which are often displayed in color and can be depicted along with reformatted bone or other tissues in the region of interest (Fig. 33A.2).

Superfast CT scanners have become available in the past 5 years. By multiplying by 4 the number of detectors, they can obtain 64 slices of an organ in a fraction of a second. They are particularly useful in cardiology and also allow for the acquisition of perfusion images of the entire brain. One shortcoming is a greater exposure to ionizing radiation per scan.

Magnetic Resonance Imaging

Basic Principles

Magnetic resonance imaging (MRI) is based on the magnetic characteristics of the imaged tissue. It involves creation of tissue magnetization (which can then be manipulated in several ways) and detection of tissue magnetization as revealed by signal intensity. The various degrees of detected signal intensity provide the image of a given tissue.

In clinical practice, MRI uses the magnetic characteristics inherent to the protons of hydrogen nuclei in the tissue, mostly in the form of water but to a significant extent in fat as well. The protons spin about their own axes, which creates a magnetic dipole moment for each proton (Fig. 33A.3). In the absence of an external magnetic field, the axes of these dipoles are arranged randomly, and therefore, the vectors depicting the dipole moments cancel each other out, resulting in a zero net magnetization vector and a zero net magnetic field for the tissue.

This situation changes when the body is placed in the strong magnetic field of a scanner (see Fig. 33A.3, A). The magnetic field is generated by an electric current circulating in wire coils that surround the open bore of the scanner. Most MRI scanners used in clinical practice are superconducting magnets. Here the electrical coils are housed at near–absolute zero temperature, minimizing their resistance and allowing for the strong currents needed to generate the magnetic field without undue heating. The low temperature is achieved by cryogens (liquid nitrogen or helium). Most clinical scanners in commercial production today produce magnetic fields at strengths of 1.5 or 3.0 tesla (T).

When the patient is placed in the MRI scanner, the magnetic dipoles in the tissues line up relative to the external magnetic field. Some dipoles will point in the direction of the external field (“north”), some will point in the opposite direction (“south”), but the net magnetization vector of the dipoles (the sum of individual spins) will point in the direction of the external field (“north”), and this will be the tissue’s acquired net magnetization. At this point, a small proportion of the protons (and therefore the net magnetization vector of the tissue) is aligned along the external field (longitudinal magnetization), and the protons precess with a certain frequency. The term precession describes a proton spinning about its own axis and its simultaneous wobbling about the axis of the external field (see Fig. 33A.3, B). The frequency of precession is directly proportional to the strength of the applied external magnetic field.

As a next step in obtaining an image, a radiofrequency pulse is applied to the part of the body being imaged. This is an electromagnetic wave, and if its frequency matches the precession frequency of the protons, resonance occurs. Resonance is a very efficient way to give or receive energy. In this process, the protons receive the energy of the applied radiofrequency pulse. As a result, the protons flip, and the net magnetization vector of the tissue ceases transiently to be aligned with that of the external field but flips into another plane, thereby transverse magnetization is produced. One example of this is the 90-degree radiofrequency pulse that flips the entire net magnetization vector by 90 degrees to the transverse (horizontal) plane (Fig. 33A.4). What we detect in MRI is this transverse magnetization, and its degree will determine the signal intensity. Through the process of electromagnetic induction, rotating transverse magnetization in the tissue induces electrical currents in receiver coils, thus accomplishing signal detection. Several cycles of excitation pulses by the scanner with detection of the resulting electromagnetic signal from the imaged subject are repeated per imaged slice. This occurs while varying two additional magnetic field gradients along the x and y axes for each cycle. Varying the magnetic field gradient along these two additional axes, known as phase and frequency encoding, is necessary to obtain sufficient information to decode the spatial coordinates of the signal emitted by each tissue voxel. This is accomplished using a mathematical algorithm known as a Fourier transform. The final image is produced by applying a gray scale to the intensity values calculated by the Fourier transform for each voxel within the imaging plane, corresponding to the signal intensity of individual tissue elements.

T1 and T2 Relaxation Times

During the process of resonance, the applied 90-degree radiofrequency pulse flips the net magnetization vectors of the imaged tissues to the transverse (horizontal) plane by transmitting electromagnetic energy to the protons. The radiofrequency pulse is brief, and after it is turned off the magnitude of the net magnetization vector starts to decrease along the transverse or horizontal plane and return (“recover or relax”) toward its original position, in which it is aligned parallel to the external magnetic field. The relaxation process, therefore, changes the magnitude and orientation of the tissue’s net magnetization vector. There is a decrease along the horizontal or transverse plane and an increase (recovery) along the longitudinal or vertical plane (Fig. 33A.5).

To understand the meaning of T1 and T2 relaxation times, the decrease in the magnitude of the horizontal component of the net magnetization vector and its simultaneous increase in magnitude along the vertical plane should be analyzed independently. These processes are in fact independent and occur at two different rates, T2 relaxation always occurring more rapidly than T1 relaxation (Fig. 33A.6). The T1 relaxation time refers to the time required by protons within a given tissue to recover 63% of their original net magnetization vector along the vertical or longitudinal plane immediately after completion of the 90-degree radiofrequency pulse. As an example, a T1 time of 2 seconds means that 2 seconds after the 90-degree pulse is turned off, the given tissue’s net magnetization vector has recovered 63% of its original magnitude along the vertical (longitudinal) plane. Different tissues may have quite different T1 time values (T1 recovery or relaxation times). T1 relaxation is also known as spin-lattice relaxation.

image

Fig. 33A.6 This diagram illustrates the simultaneous recovery of longitudinal magnetization (T1 relaxation) and decay of horizontal magnetization (T2 relaxation) after the RF pulse is turned off.

(Reprinted with permission from Hashemi, R.H., Bradley, W.G., Lasanti, C.J., 2004. MRI—The Basics. 2nd ed. Lippincott Williams & Wilkins.)

While T1 relaxation relates to the longitudinal plane, T2 relaxation refers to the decrease of the transverse or horizontal magnetization vector. When the 90-degree pulse is applied, the entire net magnetization vector is flipped in the horizontal or transverse plane. When the pulse is turned off, the transverse magnetization vector starts to decrease. The T2 relaxation time is the time it takes for the tissue to lose 63% of its original transverse or horizontal magnetization. As an example, a T2 time of 200 ms means that 200 ms after the 90-degree pulse has been turned off, the tissue will have lost 63% of its transverse or horizontal magnetization. The decrease of the net magnetization vector in the horizontal plane is due to dephasing of the individual proton spins as they precess at slightly different rates owing to local inhomogeneities of the magnetic field. This dephasing of the individual proton magnetic dipole vectors causes a decrease of the transverse component of the net magnetization vector and loss of signal. T2 relaxation is also known as spin-spin relaxation. Just like the T1 values, the T2 time values of different tissues may also be quite different. Tissue abnormalities may alter a given tissue’s T1 and T2 time values, ultimately resulting in the signal changes seen on the patient’s MR images.

Repetition Time and Time to Echo

As mentioned before, the amount of the signal detected by the receiver coils depends on the magnitude of the net magnetization vector along the transverse or horizontal plane. Using certain operator-dependent parameters, it is possible to influence how much net magnetization strength (in other words, vector length) will be present in the transverse plane for the imaged tissues at the time of signal acquisition. During the imaging process, the initial 90-degree pulse flips the entire vertical or longitudinal magnetization vector into the horizontal plane. When this initial pulse is turned off, recovery along the longitudinal plane begins (T1 relaxation). Subsequent application of a second radiofrequency pulse at a given time after the first pulse will flip the net magnetization vector that recovered so far along the longitudinal plane back to the transverse plane. As a result, we can measure the magnitude of the net longitudinal magnetization that had recovered within each voxel at the time of application of the second pulse, provided that signal acquisition is begun immediately afterwards. The time between these radiofrequency pulses is referred to as repetition time, or TR (Fig. 33A.7). It is important to realize that contrary to the T1 and T2 times, which are properties of the given tissue, the repetition time is a controllable parameter. By selecting a longer TR, for instance, we allow more time for the net magnetization vector to recover before we flip it back to the transverse plane for measurement. A longer TR, because it increases the amount of signal that can potentially be detected, will also result in a higher signal-to-noise ratio, with higher image quality.

image

Fig. 33A.7 Repetition time. This pulse sequence diagram demonstrates the concept of repetition time (TR), which is the time interval between two sequential radio frequency pulses.

(Reprinted with permission from Hashemi, R.H., Bradley, W.G., Lasanti, C.J., 2004. MRI—The Basics. 2nd ed. Lippincott Williams & Wilkins.)

As described earlier, the other process that begins after the initial radiofrequency pulse is turned off is the decrease of net horizontal or transverse magnetization, owing to dephasing of the proton spins (T2 relaxation). Time to echo (TE) refers to the time we wait until we measure the magnitude of the remaining transverse magnetization. TE, just like TR, is a parameter controlled by the operator. If we use a longer TE, tissues with significantly different T2 values (i.e., different rates of loss of transverse magnetization component) will show more difference in the measured signal intensity (transverse magnetization vector size) when the signals are collected. However, there is a tradeoff. If the TE is set too high, the signal-to-noise ratio of the resulting image will drop to a level that is too low, resulting in poor image quality.

Tissue Contrast (T1, T2, and Proton Density Weighting)

By using various TR and TE values, it is possible to increase (or decrease) the contrast between different tissues in an MR image. Achieving this contrast may be based on either the T1 or the T2 properties of the tissues in conjunction with their proton density. Selecting a long TR value reduces the T1 contrast between tissues (Fig. 33A.8). Thus, if we wait long enough before applying the second 90-degree pulse, we allow enough time for all tissues to recover most of their longitudinal or vertical magnetization. Because T1 is relatively short, even for tissues with the longest T1, this is possible without resulting in excessively long scan times. Since after a long TR, the longitudinally oriented net magnetization vectors of separate tissue types are all of similar magnitudes prior to being flipped into the transverse plane by the second pulse, a long TR will result in little T1 tissue contrast. Conversely, by selecting a short TR value, there will be significant variation in the extent to which tissues with different T1 relaxation times will have recovered their longitudinal magnetization prior to being flipped by the second 90-degree pulse (see Fig. 33A.8). Therefore, with a short TR, the second pulse will flip magnetization vectors of different magnitudes into the transverse plane for measurement, resulting in more T1 contrast between the tissues.

During T2 relaxation in the transverse plane, selecting a short TE will give higher measured signal intensities (as a short TE will not allow enough time for significant dephasing, i.e., transverse magnetization loss), but tissues with different T2 relaxation times will not show much contrast (Fig. 33A.9). This is because by selecting a short time until measurement (short TE) we do not allow significant T2-related magnitude differences to develop. If we select longer TE values, tissues with different T2 relaxation times will have time to lose different amounts of transverse magnetization, and therefore by the time of signal measurement, different signal intensities will be measured from these different tissues (see Fig. 33A.9). This is referred to as T2 contrast.

Based on the described considerations, selecting TR and TE values that are both short will increase the T1 contrast between tissues, referred to as T1 weighting. Selecting long TR and long TE values will cause increased T2 contrast between tissues, referred to as T2 weighting.

On T1-weighted images, substances with a longer T1 relaxation time (such as water) will be darker. This is because the short TR does not allow as much longitudinal magnetization to recover, so the vector flipped to the transverse plane by the second 90-degree pulse will be smaller with a lower resulting signal strength. Conversely, tissues with shorter T1 relaxation times (such as fat or some mucinous materials) will be brighter on T1-weighted images, as they recover more longitudinal magnetization prior to their proton spins being flipped into the transverse plane by the second 90-degree pulse (Fig. 33A.10). Among many other applications of T1-weighted images, they allow for evaluation of BBB breakdown: areas with abnormally permeable BBB show increased signal after the intravenous administration of gadolinium. Gadolinium administration is contraindicated in pregnancy. Breast-feeding immediately after receiving gadolinium is generally regarded to be safe (Chen et al., 2008). Renally impaired patients are susceptible to an uncommon but serious adverse reaction to gadolinium, nephrogenic systemic fibrosis (Marckmann et al., 2006).

On T2-weighted images, substances with longer T2 relaxation times (e.g., water) will be brighter because they will not have lost as much transverse magnetization magnitude by the time the signal is measured (Fig. 33A.11). The T1 and T2 signal characteristics of various tissues or substances found in neuroimaging are listed in Table 33A.1.

Table 33A.1 MRI Signal Intensity of Some Substances Found in Neuroimaging

  T1-Weighted Image T2-Weighted Image
Air ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓
Free water/CSF ↓ ↓ ↓ ↑ ↑ ↑
Fat ↑ ↑ ↑
Cortical bone ↓ ↓ ↓ ↓ ↓ ↓
Bone marrow (fat) ↑ ↑
Edema ↑ ↑
Calcification ↓ (Heavy amounts of Ca++)
↑ (Little Ca++, some Fe+++)
Mucinous material
Gray matter Lower than in T2-WI  
White matter Higher than in T2-WI  
Muscle Similar to gray matter Similar to gray matter
Blood products:    
• Oxyhemoglobin Similar to background
• Deoxyhemoglobin
• Intracellular methemoglobin ↑ ↑
• Extracellular methemoglobin ↑ ↑ ↑ ↑
• Hemosiderin ↓ ↓ ↓

CSF, Cerebrospinal fluid; MRI, magnetic resonance imaging; T2-WI, T2-weighted image.

What happens if we select long TR and short TE values? With the longer TR, the T1 differences between the tissues diminish, whereas the short TE does not allow much T2 contrast to develop. The signal intensity obtained from the various tissues, therefore, will mostly depend on their relative proton densities. Tissues having more proton density, and thereby larger net magnetization vectors, will have greater signal intensity. This set of imaging parameters is referred to as proton density (PD) weighting.

Magnetic Resonance Image Reconstruction

To construct an MR image, a slice of the imaged body part is selected, then the signal coming from each of the voxels making up the given slice is measured. Slice selection is achieved by setting the external magnetic field to vary linearly along one of the three principal axes perpendicular to the axial, sagittal, and coronal planes of the subject being imaged. As a result, protons within the slice to be imaged will precess at a Larmor frequency different from the Larmor frequency within all other imaging planes perpendicular to the axis along which the magnetic field gradient is applied. The Larmor frequency is the natural precession frequency of protons within a magnetic field of a given strength and is calculated simply as the product of the magnetic field, B0, and the gyromagnetic ratio, gamma. The precession frequency of a hydrogen proton is therefore directly proportional to the strength of the applied magnetic field. The gyromagnetic ratio for any given nucleus is a constant, with a value for hydrogen protons of 42.58 MHz/T. In slices at lower magnetic strengths of the gradient, the protons precess more slowly, whereas in slices at higher magnetic field strengths, the protons precess more quickly. Based on the property of nuclear magnetic resonance, the applied radiofrequency pulse (which flips the magnetization vector to the transverse plane) will stimulate only those protons with a precession frequency that matches the frequency of the applied radiofrequency pulse. By selecting the frequency of the stimulating radiofrequency pulse during the application of the slice selection gradient, we can choose which protons (those with a specific Larmor frequency) to stimulate (“make resonate”), and thereby we can select which slice of the body to image (Fig. 33A.12).

After excitation of the slice to be imaged, using the slice selection gradient, the spatial coordinates of each voxel within the slice must be encoded to determine how much signal is coming from each voxel of that slice. This is achieved by means of two additional gradients that are orthogonal to each other within the imaging plane, known as the frequency encoding gradient and the phase encoding gradient. The phase encoding gradient briefly alters the precession frequency of the protons along the axis to which it is applied, thereby changing the relative phases of the precessing protons along this in-plane axis. The frequency encoding gradient, applied orthogonally to the phase encoding gradient within the imaging plane, alters the precession frequency of the protons along the axis to which it is applied, during the acquisition of the MRI signal. As a result of these encoding steps, each voxel will have its own unique frequency and its own unique phase shift, which upon repeating the acquisition with several incremental changes in the phase encoding gradient, will allow for deduction of the spatial localization of different intensity values for each voxel using a mathematical algorithm known as a Fourier transform. Phase encoding takes time; it has to be performed for each row of voxels in the image along the phase encoding axis. Therefore, the higher the resolution of the image along the phase encoding axis, the longer the time required to acquire the image for that slice of tissue.

In the online version of this chapter (available at www.expertconsult.com), there is a discussion of the nature and application of the following MRI sequences or techniques: spin echo and fast (turbo) spin echo; gradient-recalled echo (GRE) sequences, partial flip angle; inversion recovery sequences (FLAIR, STIR); fat saturation; echoplanar imaging; diffusion-weighted magnetic resonance imaging (DWI); perfusion-weighted magnetic resonance imaging (PWI); susceptibility-weighted imaging (SWI); diffusion tensor imaging (DTI); and magnetization transfer contrast imaging.

Magnetic Resonance Imaging

Spin Echo and Fast (Turbo) Spin Echo Techniques

Conventional spin echo imaging is time consuming because the individual echoes are obtained one by one, using a unique strength for the phase encoding gradient at each step in the acquisition of a given slice. The signal from each echo is acquired after a time period equal to one repetition time (TR) after the prior echo. During acquisition and digitization of the signal, with each such step, one row of data space (k-space) is filled. To fill the entire data space for one image, this process has to be repeated as many times as the number of phase encoding steps that are needed. To express this time in seconds, the number of phase encoding steps are multiplied by the TR. Distinct from the conventional spin echo technique, in fast (turbo) spin echo imaging (FSE), within each TR period, multiple echoes at various TE values are obtained, and a new phase encoding step is applied before each of these echoes. The number of echoes obtained for the encoding of each line of k-space in the FSE technique is referred to as the echo train length. Each echo will fill a new line within the k-space data set. Therefore, instead of filling just one line with each TR, multiple lines are filled, and the data space acquisition is completed much more quickly. It is important to realize that even though only a single TE is typically displayed on the MRI technician’s imaging console (this is sometimes referred to as effective TE) during acquisition of FSE images, multiple TE times are actually used. The obvious advantage of fast spin echo imaging is that by filling up k-space much more quickly, the scan time is significantly reduced. This improves image quality by increasing the signal-to-noise ratio. The increased signal, however, may at times also be a disadvantage (e.g., identifying a PV hyperintense lesion adjacent to brighter CSF).

Gradient-Recalled Echo Sequences, Partial Flip Angle

As described earlier, in spin echo imaging, the 90-degree pulse flips the longitudinal magnetization vector into the horizontal plane. After this pulse, the transverse magnetization starts to decay as a result of dephasing. This would in theory result in a decrease of signal by the time (TE) the signal is read by the receiver coils. To prevent this, at a time point equal to one-half of the echo time (TE/2) a 180-degree refocusing pulse is applied that will result in reversal of the directions in which the individual precessing protons are dephasing, so that at a time point equal to TE they will once again be in phase, maximizing the acquired signal by the receiver coils. Thus a signal can be collected that is close in strength to the original. This method only compensates for the dephasing caused by magnetic field inhomogeneities, not for the loss of signal caused by spin-spin interactions, so the recorded signal will not be as large as the original.

In GRE, or gradient echo imaging, instead of “letting” the transverse magnetization dephase and then using the 180-degree refocusing pulse to rephase, a dephasing-refocusing gradient is applied. This gradient will initially dephase the spins of the transverse magnetization. This is followed by the refocusing component of the gradient, which will rephase them at time TE as a readable echo at the receiver coils. Because of greater spin dephasing, GRE is more susceptible to local magnetic field inhomogeneities. This may cause increased artifacts within and near interfaces between tissues with significantly different degrees of magnetic susceptibility, such as at bone/soft tissue or air/bone/brain interfaces near the ethmoid sinuses and medial temporal lobes. However, it is very useful when looking specifically for pathology involving tissue components or deposits exhibiting significant paramagnetism. For example, in the case of chronic hemorrhage, the iron in hemosiderin causes magnetic susceptibility artifact by distorting the magnetic field, resulting in very dark signal voids with an apparent size greater than the spatial extent of the iron deposition, thereby increasing sensitivity for such lesions on the specific pulse sequences designed to maximize this effect. Such pulse sequences include 3D spoiled gradient echo, T2* (pronounced T2-star) and SWI techniques. T2* imaging, in which signal is obtained from transversely magnetized precessing protons without a preceding echo, allows for the detection of hemorrhage as well as deoxyhemoglobin, as in the blood oxygen level dependent (BOLD) effect used to assess relative brain perfusion levels in functional MRI.

Another term that should be explained in conjunction with gradient echo imaging is the partial flip angle. Instead of applying a 90-degree pulse to flip the entire magnetization vector into the horizontal plane, a pulse is used that only partially flips the vector, at a smaller angle. As a result, only a component of the magnetization vector will be in the horizontal plane after application of the excitation pulse. Utilizing a smaller flip angle allows use of a shorter TR, since there will already be a significant longitudinal component of the net magnetization vector after excitation, requiring less time for sufficient recovery of longitudinal magnetization prior to the next excitation pulse.

The T1-weighted signal generated by a tissue in a GRE sequence can be optimized for any given TR by varying the flip angle according to a mathematical relationship known as the Ernst equation. The optimal flip angle for a given tissue at a particular TR is thus known as the Ernst angle.

Use of shorter longitudinal relaxation times in gradient echo imaging has the obvious advantage of decreasing scan time. By changing the flip angle (which, just like TR and TE is an operator-controlled parameter) the tissue contrast may be manipulated. Selecting a small flip angle in conjunction with a sufficiently long TR will decrease the T1 weighting of the image, as the longitudinal magnetization will be nearly maximized for all tissues. This effect is similar to that for a conventional spin echo sequence, when selecting a long TR allows the longitudinal magnetization to recover more, thereby reducing or eliminating T1 weighting from the resulting image.

The generation of image contrast in GRE imaging is similar to that in spin-echo imaging. One important difference is that T2-weighted images cannot be generated, owing to lack of a refocusing pulse in the GRE technique. Instead, the shorter T2* decay is used to generate T2-like image contrast while minimizing T1 effects. Therefore, T2*-weighted images are obtained using a small flip angle, a long TR, and long TE. A small flip angle in conjunction with a long TR and a short TE will result in proton density weighting, because the T1 and T2* effects upon image contrast are minimized. Selecting a large flip angle together with a short TR and a short TE will result in T1 weighting. Advantages of GRE imaging include speed, less contamination of signal in the slice to be imaged by signal from adjacent slices, and higher spatial resolution. Disadvantages include greater susceptibility to inhomogeneities in the magnetic field such as magnetic susceptibility artifact (although this may also be an advantage, as outlined earlier) and the requirement for higher gradient field strengths. One very useful application of GRE imaging is in volumetric analysis of imaged tissues; the shorter TR and resultant speed allow for rapid data acquisition in three dimensions, which can be used to format and display images in any plane.

Inversion Recovery Sequences (FLAIR, STIR)

For better detection and visualization of abnormalities on MR images, it is often useful to suppress the signal from certain tissues, thereby increasing the contrast between the region of pathology and the surrounding and/or background tissue. Examples of this include visualization of hyperintense lesions adjacent to bright CSF spaces on T2-weighted images, or whenever there is a need to eliminate the hyperintense signal coming from fatty background.

Inversion recovery techniques use a unique pulse sequence to avoid signal detection from the selected tissues (fat or CSF). Initially, a 180-degree radiofrequency pulse is applied. This will flip the longitudinal magnetization vectors of all tissues by 180-degrees, so that the vectors will point downward (south). Next, the flipped vectors are allowed to start recovering according to their respective T1 times. As the downward-pointing vectors recover, they become progressively smaller, eventually reaching zero magnitude, and from that point they start growing and pointing upward (north). Without interference, they recover the original longitudinal magnetization. However, during the process of recovery, after a time period referred to as inversion time (TI), a 90-degree pulse is applied. This will flip the longitudinal vectors to the transverse plane, where signal detection occurs. The amount of magnetization flipped by this pulse depends on how far the longitudinal recovery has been allowed to proceed. If the 90-degree pulse is applied when a given tissue’s vector happens to be zero (this is the so-called null point), no magnetization will be flipped from that tissue to the transverse plane, and therefore no signal will be detected from that tissue. Different tissues recover their longitudinal magnetization at different rates according to their specific T1 times. Knowing a given tissue’s T1 time, we can calculate when it will reach the null point (when its longitudinal magnetization is zero), and if we apply the 90-degree pulse at that point, we will not detect any signal from that particular tissue. The inversion time is linearly dependent upon a given tissue’s T1 value, being calculated as 0.69 multiplied by the T1 value. In the FLAIR (fluid-attenuated inversion recovery) sequence, the inversion time (when the 90-degree pulse is applied) occurs when the magnetization vector for the CSF is at the null point, so no signal will be detected from the CSF (Fig. 33A.13). In FLAIR images, the dark CSF is in sharp contrast with the hyperintensity of PV lesions, allowing their better identification. In STIR (short TI, or tau inversion, recovery) imaging, which is a fat-suppression technique, the methodology is essentially the same as for FLAIR. However, instead of CSF, the signal from fat is nulled. The TI for the STIR technique is set to 0.69 times the T1 of fat, which results in application of the final 90-degree pulse when the fat tissue’s magnetization is at the null point, so no signal from fat will be detected.

Diffusion-Weighted Magnetic Resonance Imaging

Diffusion of water molecules within tissues has a random molecular (Brownian) motion, which varies in a tissue- and pathology-dependent manner. It may have a directional preference in some tissues; for instance, there is greater diffusion in the longitudinal than in the transverse plane of an axon. Water diffusion may occur more rapidly in aqueous compartments such as CSF, relative to water that is largely intracellular, as in regions of cytotoxic edema secondary to brain ischemia or water present in fluid compartments with high viscosity, such as abscesses or epidermoid cysts. DWI is an imaging technique that is able to differentiate areas of low from high diffusion. The imaging sequence used for this purpose is a T2-weighted sequence, with the addition of transient gradients applied before TE. The spin-echo sequence can be conventional, or more commonly, a single-shot spin-echo echoplanar imaging sequence that allows for much faster imaging. The purpose of the gradients is to sensitize the pulse sequence to diffusion occurring during the time interval between their application. In tissues where more diffusion occurred during application of the gradient (such as in normal tissues), the diffusion causes dephasing of transverse magnetization, resulting in signal loss and, therefore, a darker appearance on the image. In areas with less diffusion (for example in acutely ischemic brain areas), no significant dephasing or signal change occurs. Therefore, the detected signal is higher, and these areas appear bright on the image.

The degree of the applied diffusion encoding gradient is referred to as the B value. In a regular conventional T2 or FLAIR image, the B value is zero (i.e., no gradient). As the B value is increased by the gradient being stronger, the diffusion of the water molecules will cause more and more dephasing and signal loss. As a result, if the B value is high enough, as in DWI, the areas of higher diffusion rates, such as CSF and normal brain tissue, will be dark due to the dephasing and signal loss related to water diffusion. In contrast, ischemic areas with little or no water molecule diffusion will appear bright because they lack dephasing and signal loss. In imaging protocols where more T2 weighting (longer TE values) and smaller B values are used, areas with long T2 values may appear relatively bright in the diffusion-weighted images, despite their considerable diffusion. This phenomenon is referred to as T2 shine-through, and it is due to the low applied B value, which means a weaker diffusion gradient and less diffusion weighting. This shine-through can be decreased by applying a stronger diffusion gradient, leading to higher B values and more diffusion weighting.

Based on the differences in the change of signal intensity in different areas at different applied B values, it is possible to calculate the apparent diffusion coefficient (ADC) in various areas/tissues in the image. The term apparent is used because in a tissue there are other factors besides this coefficient that contribute to signal loss, including patient motion and blood flow. The higher the diffusion rate, the higher the ADC value of the given tissue, and the brighter it will appear on the ADC image or map. As an example, CSF, where the diffusion is highest, will be bright on the ADC map, whereas areas of little (restricted) diffusion, such as ischemic areas, will be dark.

One of the most obvious practical uses of DWI is the delineation of acutely ischemic areas, which appear bright against a dark background in diffusion-weighted images and dark on the corresponding ADC maps. According to the most appealing theory, the reason for restricted diffusion in acutely ischemic brain tissue is the evolving cytotoxic edema (cellular swelling), which decreases the relative size of the extracellular space, thereby limiting water diffusion.

Although in neurological practice, the term restricted diffusion usually refers to cerebral ischemia, and this imaging modality remains most important for acute stroke imaging, it is to be noted that there are other abnormalities that also restrict diffusion and appear bright on diffusion-weighted images. Examples include abscesses, hypercellular tumors such as lymphoma, some meningiomas, epidermoid cysts, aggressive demyelinating disease, and proteinaceous material, such as in association with sinusitis.

Perfusion-Weighted Magnetic Resonance Imaging

Perfusion-weighted imaging utilizes MRI sequences that are able to generate different signal intensities between tissues with different degrees of perfusion. Although there are techniques (like spin-labeled perfusion imaging) that provide information about tissue perfusion without injecting contrast material, the most common technique uses a rapid bolus of paramagnetic contrast agent (gadolinium) which, while passing through the tissues, causes distortion of the magnetic field and signal loss in the applied gradient echo or echo planar image. This signal loss only occurs in tissues that are perfused, whereas nonperfused regions do not have such signal loss, or in cases of decreased but not absent perfusion, the signal loss is not as prominent as seen in the healthy tissue. When the selected slice is imaged repeatedly in rapid succession, parameters related to perfusion (e.g., relative cerebral blood volume [rCBV], time to peak signal loss [TTP], mean transit time of the contrast bolus [MTT]) can be calculated for each voxel within the slice being imaged. Estimates of cerebral blood flow (CBF) can be calculated for each voxel as well.

The main clinical application of PWI is in the setting of acute stroke, primarily for visualization of tissue at risk, the ischemic penumbra. When used in conjunction with diffusion-weighted images, which delineate the acutely infarcting area, it is frequently seen that perfusion-weighted images reveal a more extensive area, beyond the extent of the zone of infarction, that exhibits decreased or absent perfusion. This is the ischemic penumbra, tissue at risk that is potentially salvageable, prompting use of thrombolytic therapy. If the perfusion deficit appears the same as the zone of restricted diffusion (area in the process of infarction), the chance for saving tissue is likely to be lower than that for an ischemic infarction exhibiting a significant perfusion-diffusion mismatch.

Susceptibility-Weighted Imaging

As described earlier, factors that distort magnetic field homogeneity, such as paramagnetic or ferromagnetic substances, cause local signal loss. Signal loss occurs because in the altered local magnetic field, protons will precess with different frequencies, resulting in dephasing and thus decreasing the net magnetization vector that translates into a detectable signal. Gradient echo images are especially sensitive to magnetic field distortions, which appear as areas of decreased signal due to the magnetic susceptibility artifact.

SWI (Haacke et al., 2009; Mittal et al., 2009) uses a high spatial resolution 3D gradient echo imaging sequence. The contrast achieved by this sequence distinguishes the magnetic susceptibility difference between oxygenated and deoxygenated hemoglobin. Since the applied phase postprocessing sequence accentuates the paramagnetic properties of deoxyhemoglobin and blood degradation products such as intracellular methemoglobin and hemosiderin, this technique is very sensitive for intravascular venous deoxygenated blood as well as extravascular blood products. It has been used for evaluation of venous structures, hence the earlier name high-resolution blood oxygen level–dependent venography, but the clinical application is now much broader. Its exquisite sensitivity for blood degradation products makes this technique very useful when evaluating any lesion (e.g., stroke, AVM, cavernoma or neoplasm) for associated hemorrhage (Fig. 33A.15). It is also used for imaging microbleeds associated with traumatic brain injury, diffuse axonal injury, or cerebral amyloid angiopathy.

Diffusion Tensor Imaging

Diffusion tensor imaging is a more advanced type of diffusion imaging capable of quantifying anisotropy of diffusion in white matter. Diffusion is isotropic when it occurs with the same intensity in all directions. It is anisotropic when it occurs preferentially in one direction, as along the longitudinal axis of axons. For this reason, DTI finds its greatest current application in MRI examinations of the white matter. As opposed to characterizing diffusion within each voxel with just a single apparent diffusion coefficient, as in DWI, in DTI intravoxel diffusion is measured along three, six, or more gradient directions. The measured values and their directions are called eigenvectors. The vector that corresponds to the principal direction of diffusion (the direction in which diffusion is greatest in magnitude) is called the principal eigenvector. In normal white matter, diffusion anisotropy is high because diffusion is greatest parallel to the course of the nerve fiber tracts. Therefore, the principal eigenvector delineates the course of a given nerve fiber pathway. Diffusion tensor images can be displayed as maps of the principal eigenvectors which will show the direction/course of the given white matter tract (tractography). These images can also be color coded, allowing for more spectacular visualization of nerve fiber tracts (Fig. 33A.16). Any disruption of a given nerve fiber tract (e.g., MS, trauma, gliosis) will reduce anisotropy, and the disruption of the white matter tract can be visualized. Tensor imaging/tractography is useful in imaging of degenerating white matter tracts and also in surgical resection planning, when the anatomical relationship of the resectable lesion and the adjacent fiber tracts has to be evaluated to avoid or reduce surgical injury to critical pathways.

Magnetization Transfer Contrast Imaging

As the name indicates, magnetization transfer contrast imaging is a technique that produces increased contrast within an MR image, specifically on T1-weighted gadolinium-enhanced images and in magnetic resonance angiography (Henkelman et al., 2001). In water, hydrogen atoms are relatively loosely bound to oxygen atoms, and they move frequently between them, binding to one oxygen atom then switching to another. In other tissues (e.g., lipids, proteins), the hydrogen atoms are more tightly bound and tend to stay in one place for longer periods of time. Nevertheless, it does happen that a “bound” hydrogen in lipid or protein is exchanged with a “more free” hydrogen from water. In magnetization transfer imaging, at the beginning of the sequence a radiofrequency pulse is applied that saturates the bound protons in lipids and proteins but does not affect the free protons in water. In regions where magnetization transfer (i.e., exchange of saturated protons with free protons) occurs, the saturated protons will decrease the signal obtained from the imaged free protons. The more frequently this magnetization transfer occurs, the less signal is obtained from the region and the darker the region will be in the image. Magnetization transfer happens more frequently in the white matter, resulting in signal loss, and therefore on magnetization transfer images, the white matter appears darker. The CSF on the other hand, where magnetization transfer does not occur, does not lose signal. Magnetization transfer is minimal in blood because of the high amount of free water protons.

This technique is useful when gadolinium-enhanced T1-weighted images are obtained, because enhancing lesions stand out better against the darker background of the more hypointense white matter. In fact, applying a magnetization transfer sequence to single-dose gadolinium-enhanced T1-weighted images results in contrast enhancement intensity comparable to giving a double dose of gadolinium. This sequence is also used in time-of-flight magnetic resonance angiography. There is no signal change in the blood, but the background tissue becomes darker, so the imaged blood vessels stand out better, and smaller branches are better visualized. This benefit comes at the expense of a significantly prolonged scan time, because it takes additional time to apply the magnetization transfer pulse.

Another application of magnetization transfer imaging is in the assessment of “normal-appearing” tissues that in fact contain abnormalities, albeit not visible on conventional MR pulse sequences. By selecting a region of interest (ROI, essentially a quadrilateral that is selected to enclose the tissue of interest within an image) corresponding to the “normal-appearing” tissue and calculating the degree to which magnetization transfer occurs within each voxel of the ROI, a histogram plot can be generated. On such magnetization transfer ratio (MTR) histograms, tissues with no apparent lesional signal on conventional images, such as the “normal-appearing white matter” of MS, may exhibit a decreased peak height. Such histograms in MS patients may also exhibit a larger proportion of voxels with low MTR values than normal tissues, reflecting a microscopic and macroscopic lesion load that is otherwise undetectable by conventional imaging techniques.

Structural Neuroimaging in the Clinical Practice of Neurology

Brain Diseases*

Brain Tumors

Epidemiology, pathology, etiology, and management of cancer in the nervous system are discussed in Chapter 52A, Chapter 52B, Chapter 52C, Chapter 52D, Chapter 52E, Chapter 52F, Chapter 52G . From the standpoint of structural neuroimaging, a useful anatomical classification distinguishes two main groups: intraaxial and extraaxial tumors. Intraaxial tumors are within the brain parenchyma, extraaxial tumors are outside the brain parenchyma (involving the meninges or, less commonly, the ventricular system). Intraaxial tumors are usually infiltrative with poorly defined margins. Conversely, extraaxial tumors, even though they often compress or displace the adjacent brain, are usually demarcated by a cerebrospinal (CSF) cleft or another tissue interface between tumor and brain parenchyma. For differential diagnostic purposes, intraaxial primary brain neoplasms can be further divided into the anatomical subgroups of supratentorial and infratentorial tumors (Table 33A.2).

For evaluation of brain tumors, the structural imaging modality of choice is MRI. Due to their gradual expansion and often infiltrative nature, most brain tumors are already visible on MRI by the time patients become symptomatic. Exceptions to this rule are tumors that tend to involve the cortex or corticomedullary junction, such as small oligodendrogliomas or metastases, which may cause seizures early, even before being clearly visible on noncontrast MRI. Meningeal involvement is also often symptomatic, for instance by causing headaches and confusion, but may not be appreciated on noncontrast images. Higher magnetic field strength (e.g., a 3-T scanner) and contrast administration (in double or triple dose if necessary) can improve detection of small or clinically silent neoplastic lesions.

Neuroimaging is particularly useful in the assessment of brain tumors. Unlike destructive lesions such as ischemic strokes, brain tumors often cause clinical manifestations that are difficult to interpret. Sometimes the clinical presentation may provide clues to localization—for example, a seizure is suggestive of an intraaxial tumor, whereas cranial nerve involvement tends to signal an extraaxial pathology. But edema, mass effect, obstructive hydrocephalus, and elevated intracranial pressure (ICP) can give rise to nonspecific symptoms (e.g., headache, visual disturbance, altered mental status), and false localizing signs may also appear, such as oculomotor or abducens nerve compression due to an expanding intraaxial mass.

Neoplastic tissues most commonly prolong the T1 and T2 relaxation times, appearing hypointense on T1 and hyperintense on T2-weighted images, but different tumors differ in this property, facilitating tumor identification on MRI. MRI is also very sensitive for detection of other pathologic changes that can be associated with tumors, such as calcification, hemorrhage, necrosis, and edema. The structural detail provided by MRI is useful for assessing involved structures and determining the number and macroscopic extent of the neoplasms, thereby guiding surgical planning or other treatment modalities.

Intraaxial Primary Brain Tumors

Pilocytic Astrocytomas

Pilocytic astrocytomas have two major groups: juvenile and adult. These tumors are classified as WHO grade I. Juvenile pilocytic astrocytomas are the most common posterior fossa tumors in children. The most common locations are the cerebellum, at the fourth ventricle, third ventricle, temporal lobe, optic chiasm, and hypothalamus (Koeller and Rushing, 2004). The appearance is often lobulated, and the lesion appears well demarcated on MRI. Hemorrhage and necrosis are uncommon. Areas of calcification may be present. The tumor usually exhibits solid as well as cystic components, with or without a mural nodule. The adult form is usually well circumscribed, often calcified, and typically exhibits a large cyst with a mural nodule. On MRI, the solid portions of the tumor are iso- to hypointense on T1 and iso- to hyperintense on T2-weighted images (Arai et al., 2006). The cystic component usually exhibits CSF signal characteristics. The associated edema and mass effect is usually mild, sometimes moderate. With gadolinium, the solid components (including the mural nodule) enhance intensely, but not the cyst, which rarely may show rim enhancement.

Pleomorphic Xanthoastrocytoma

Pleomorphic xanthoastrocytoma is a rare variant of astrocytic tumors. It is thought to arise from the subpial astrocytes and typically affects the cerebral cortex and adjacent meninges and may cause erosion of the skull. The most common location is the temporal lobe. It is classified as WHO grade II. It usually occurs in the second and third decades of life, and patients often present with seizures. On MRI (Tien et al., 1992) usually a well-circumscribed cystic mass appears in a superficial cortical location. A solid portion or mural nodule is often seen, and the differential diagnosis includes pilocytic astrocytoma and ganglioglioma. The signal characteristics are hypointense or mixed on T1, and hyperintense or mixed on T2-weighted images. With contrast, the solid portions and sometimes the adjacent meninges enhance. Calcification may be present. There is mild or no mass effect associated with this tumor.

Low-Grade Astrocytomas

Fibrillary astrocytomas, also termed diffuse astrocytomas, represent approximately 10% of all gliomas. Low-grade (WHO grade I and II) astrocytomas belong to this group (Figs. 33A.17 and 33A.18). These are well-differentiated tumors, usually arising from the fibrillary astrocytes of the white matter. Even though imaging may show a fairly well-defined boundary, these tumors are infiltrative and usually spread beyond their macroscopic border. Two-thirds of cases are supratentorial. A subgroup of these astrocytomas involves specific regions such as the optic nerves/tracts or the brainstem.

Low-grade astrocytomas are iso- or hypointense on T1-weighted images and hyperintense on T2-weighted images. Tissue expansion may be seen, and mass effect (if present) is generally modest. There is little to no associated edema. Fibrillary grade I astrocytomas do not enhance; grade II tumors may exhibit enhancement. The appearance of enhancement in a previously nonenhancing tumor is a worrisome sign of malignant transformation, often due to anaplastic astrocytoma.

Oligodendroglioma

Oligodendroglioma accounts for 5% to 10% of all gliomas. It arises from the oligodendroglia that form the myelin sheath of the central nervous system (CNS) pathways. Oligodendroglioma occurs most commonly in young and middle-aged adults, with a median age of onset within the fourth to fifth decades and a male predominance of up to 2 : 1. Seizure is often the presenting symptom. The most common location is the supratentorial hemispheric white matter, and it also involves the cortical mantle. The tumor often has cystic components and at least microscopically, in 90% of cases also shows calcification. Hemorrhage and necrosis are rare, and the mass effect is not impressive. On MRI (Koeller and Rushing, 2005) the appearance is heterogenous, and the tumor is hypo- and isointense on T1 and hyperintense on T2. With gadolinium, the enhancement is variable, usually patchy, and the periphery of the lesion tends to enhance more intensely. Oligodendrogliomas are hypercellular and have been noted to appear hyperintense on diffusion-weighted images (Fig. 33A.19).

Gliomatosis Cerebri

Gliomatosis cerebri, a rare glial neoplasm, usually presents in the third decade of life. The glial tumor cells are disseminated throughout the parenchyma and infiltrate large portions of the neuraxis. Macroscopically it appears homogenous and is seen as enlargement/expansion of the parenchyma; the gray/white matter interface may become blurred, but the architecture is otherwise not altered. The hemispheric white matter is involved first, then the pathology spreads to the corpus callosum, followed by both hemispheres. Later, the deep gray matter (basal ganglia, thalamus, massa intermedia) may be affected as well. Diffuse tumor infiltration often extends into the brainstem, cerebellum, and even the spinal cord. Histologically, most cases of gliomatosis cerebri are WHO grade III.

The MRI appearance is iso- to hypointense on T1 and hyperintense on T2. Hemorrhage is uncommon, and enhancement is also rare, at least in the early stages (Fig. 33A.20). Later, multiple foci of enhancement may appear, signaling more malignant transformation. The imaging appearance is similar to that of encephalitis, lymphoma, or subacute sclerosing panencephalitis, but in these disorders, clinical findings are more pronounced.

Glioblastoma Multiforme

Glioblastoma multiforme is a highly malignant tumor classified as grade IV by the WHO. It is most common in older adults, usually appearing in the fifth and sixth decades. GBM is the most common primary brain neoplasm, representing 40% to 50% of all primary neoplasms and up to 20% of all intracranial tumors. It forms a heterogenous mass exhibiting cystic and necrotic areas and often a hemorrhagic component as well. The most common locations are the frontal and temporal lobes. The tumor is highly infiltrative and has a tendency to spread along larger pathways such as the corpus callosum and invade the other hemisphere, resulting in a characteristic “butterfly” appearance. GBM has also been described to spread along the ventricular surface in the subarachnoid space and may also invade the meninges. There are reported cases of extracranial glioblastoma metastases.

Structural neuroimaging distinguishes between multifocal and multicentric glioblastomas. The term multifocal glioblastoma refers to multiple tumor islands in the brain that arose from a common source via continuous parenchymal spread or meningeal/CSF seeding; therefore, they are all connected, at least microscopically. Multicentric glioblastoma refers to multiple tumors that are present independently, and physical connection between them cannot be proven, implying they are separate de novo occurrences. This is less common, having been noted in 6% of cases.

On MRI (Fig. 33A.21) glioblastomas usually exhibit mixed signal intensities on T1- and T2-weighted images. Cystic and necrotic areas are present, appearing as markedly decreased signal on T1-weighted and hyperintensity on T2-weighted images. Mixed hypo- and hyperintense signal changes due to hemorrhage are also seen. The hemorrhagic component can also be well demonstrated by gradient echo sequences or by SWI. The core of the lesion is surrounded by prominent edema, which appears hypointense on T1-weighted and hyperintense on T2-weighted images. Besides edema, the signal changes around the core of the tumor reflect the presence of infiltrating tumor cells and, in treated cases, postsurgical reactive gliosis and/or post-irradiation changes. Following administration of gadolinium, intense enhancement is noted, which is inhomogenous and often ringlike, also including multiple nodular areas of enhancement. The surrounding edema and ringlike enhancement at times makes it difficult to distinguish glioblastoma from cerebral abscess. DWI is helpful in these cases; glioblastomas are hypointense with this technique, whereas abscesses exhibit remarkable hyperintensity on diffusion-weighted images.

Owing to its aggressive growth (the tumor size may double every 10 days) and infiltrative nature, the prognosis for patients with glioblastoma is very poor. Despite surgery, irradiation, and chemotherapy the median survival is 1 year.

Ependymoma

Although ependymomas are primarily extraaxial tumors (within the fourth ventricle), intraparenchymal ependymomas arising from ependymal cell remnants of the hemispheric parenchyma are also well known, so this tumor type is discussed here. Ependymomas comprise 5% to 6% of all primary brain tumors; 70% of cases occur in childhood and the first and second decades, and ependymoma is the third most common posterior fossa tumor in children. Ependymomas arise from differentiated ependymal cells, and the most common location (70%) is the fourth ventricle. The tumor is usually well demarcated and is separated from the vermis by a CSF interface. The tumor may be cystic and may contain calcification and hemorrhage, but these features are more common in supratentorial ependymomas. It may extrude from the cavity of the fourth ventricle through the foramina of Luschka and Magendie. Spreading via CSF to the spinal canal (drop-metastases) may occur, but on spine imaging ependymoma is more commonly noted to arise from the ependymal lining of the central canal, presenting as an intramedullary spinal cord tumor. A subtype, myxopapillary ependymoma, is almost always restricted to the filum terminale.

Ependymomas are hypo- to isointense on T1-weighted, and iso- to hyperintense on T2-weighted images. With gadolinium, intense enhancement is seen, mostly involving the solid components of the tumor, whereas the cystic components tend to exhibit rim enhancement. The differential diagnosis for infratentorial ependymoma includes medulloblastoma, pilocytic astrocytoma, and choroid plexus papilloma.

Lymphoma

Primary CNS lymphoma (PCNSL) is a non-Hodgkin lymphoma, which in 98% of cases is a B-cell lymphoma. It once accounted for only 1% to 2% of all primary brain tumors, but this percentage has been increasing, mostly because of the growing acquired immunodeficiency syndrome (AIDS) population. The peak age of onset is 60 in the immunocompetent population and age 30 in immunocompromised patients. Lesions may occur anywhere within the neuraxis, including the cerebral hemispheres, brainstem, cerebellum, and spinal cord, although the most common location (90% of cases) is supratentorial. PCNSL lesions are highly infiltrative and exhibit a predilection for sites that contact subarachnoid and ependymal surfaces as well as the deep gray nuclei.

The imaging appearance of PCNSL depends on the patient’s immune status. The tumor is hypo- to isointense on T1-weighted and hypo- to slightly hyperintense on T2-weighted images. Contrast enhancement is usually intense. In immunocompetent patients (Zhang et al., 2010) the lesion is often single, tends to abut the ventricular border (Costa et al., 2006), and ring enhancement is uncommon (Fig. 33A.22). In immunocompromised patients, usually multiple, often ring-enhancing lesions are seen, which are most commonly located in the periventricular white matter and the gray/white junction of the lobes of the hemispheres, but the deep central gray matter structures and the posterior fossa may be involved as well. Overall, the imaging appearance appears more malignant in the immunocompromised cases and may be difficult to differentiate from toxoplasmosis. Other components of the differential diagnosis in patients with multiple PCNSL lesions include demyelination, abscesses, neurosarcoidosis, and metastatic disease.

Extraaxial Primary Brain Tumors

Descriptions of the rarer extraaxial primary brain tumor types—esthesioneuroblastoma, central neurocytoma, and subependymoma—are available in the online version of this chapter (www.expertconsult.com).

Meningiomas

Meningiomas are the most common primary brain tumors of non-glial origin and make up 15% of all intracranial tumors. The peak age of onset is the fifth decade, and there is a striking female predominance that may be related to the fact that some meningiomas contain estrogen and progesterone receptors. These tumors arise from meningothelial cells. In 1% to 9% of cases, multiple tumors are seen. The most common locations are the falx (25%), convexity (20%), sphenoid wing, petrous ridge (15% to 20%), olfactory groove (5% to 10%), parasellar region (5% to 10%), and the posterior fossa (10%). Rarely an intraventricular location has been reported. Meningiomas often appear as smooth hemispherical or lobular dural-based masses (Fig. 33A.23). Calcification is common, seen in at least 20% of these tumors. Meningiomas also often exhibit vascularity. The extraaxial location of the tumor is usually well appreciated owing to a visible CSF interface between tumor and adjacent brain parenchyma. Meningiomas may become malignant, invading the brain and eroding the skull. In such cases, prominent edema may be present in the brain parenchyma, to the extent that the extraaxial nature of the tumor is no longer obvious.

On T1-weighted images, meningiomas are usually iso- to slightly hypointense. The appearance on T2 can be iso-, hypo-, or hyperintense to the gray matter. Although MRI does not reveal the histological subtypes of meningiomas with absolute certainty, there have been observations according to which fibroblastic and transitional meningiomas tend to be iso- to hypointense on T2-weighted images, whereas the meningothelial or angioblastic type is iso- or more hyperintense. Not uncommonly, the skull adjacent to a meningioma will exhibit subtle thickening, a useful diagnostic clue in some cases.

After gadolinium administration, meningiomas typically exhibit intense homogenous enhancement. A quite typical imaging finding on postcontrast images is the dural tail sign, which refers to the linear extension of enhancement along the dura, beyond the segment on which the tumor is based. Earlier this had been attributed to en plaque extension of the meningioma along these dural segments and was thought to be specific for this type of tumor. However, recently it has been recognized that this imaging appearance is not specific to this situation and may be seen in other tumors, secondary to increased vascularity/hyperperfusion or congestion of the dural vessels after irradiation and as a postsurgical change.

Schwannoma

Schwannomas arise from the Schwann cells of the nerve sheath, and the most commonly affected nerve is the vestibular portion of the vestibulocochlear nerve. They are typically bilateral in neurofibromatosis (NF) type 2 (see Fig. 33A.69). The unilateral form sporadically occurs in non-NF patients, with slight female predominance. Schwannomas typically arise in the intracanalicular segment of the eighth cranial nerve where myelin transitions from central (oligodendroglia) to peripheral (Schwann cell) type. If untreated, the tumor grows toward the internal auditory meatus and eventually bulges into the cerebellopontine angle, where it may deform and displace the brainstem. The intra- and extracanalicular parts of the tumor together result in a mushroom or ice cream cone–like appearance. The tumor is iso- to hypointense on T1-weighted images and iso- to hyperintense on T2-weighted images. This pattern may be modified by the presence of cystic changes or calcification. Gadolinium administration causes homogenous enhancement that, together with the performance of axial and coronal thin-slice T2-weighted images, allows for the visualization of even very small intracanalicular schwannomas.

Medulloblastoma

Medulloblastomas arise from the undifferentiated neuroectodermal cells of the roof of the fourth ventricle (superior or inferior medullary velum, vermis). They represent 25% of all cerebral tumors in children, usually presenting in the first and second decade. The tumor fills the fourth ventricle, extending rostrally toward the aqueduct and caudally to the cisterna magna, frequently resulting in obstructive hydrocephalus. Leptomeningeal and CSF spread may also occur, resulting in spinal drop metastases. Cystic components and necrosis may be present. Calcification is possible. On CT, medulloblastoma typically appears as a heterogeneous, generally hyperdense midline tumor occupying the fourth ventricle, with mass effect and variable contrast enhancement. The MRI signal (Koeller and Rushing, 2003) is heterogenous; the tumor is iso- or hypointense on T1 and hypo-, iso-, or hyperintense on T2. Contrast administration induces heterogenous enhancement (Fig. 33A.24). Restricted diffusion may be seen on DWI/ADC (Gauvain et al., 2001). Consistent with its site of origin, indistinct borders between the tumor and the roof of the fourth ventricle may be observed, aiding in the differential diagnosis, which in children includes atypical rhabdoid-teratoid tumor, brainstem glioma, pilocytic astrocytoma, choroid plexus papilloma, and ependymoma. The adult differential diagnosis includes the latter two entities in addition to metastasis and hemangioblastoma. Medulloblastoma does not tend to extrude via the foramina outside of the fourth ventricle, facilitating differentiation from ependymoma. In children, choroid plexus papilloma is more likely to occur within the lateral ventricle.

Extraaxial Primary Brain Tumors

Other Pineal Region Tumors

Besides pineoblastomas, which histologically belong to the group of primitive neuroectodermal tumors, the pineal gland may also develop tumors of pinealocyte origin (pineocytoma) and germ cell tumors.

Central Neurocytoma

This neuron-derived tumor accounts for less than 1% of all primary brain tumors. It tends to appear in the fourth decade. The tumor is intraventricular, most commonly in the lateral ventricles anteriorly at the foramen of Monro, close to the septum and the columns of the fornix. Even though the tumor is relatively benign histologically, this location frequently leads to obstructive hydrocephalus. The MRI signal is heterogenous (Chang et al., 1993); the signal is isointense on T1 and iso- or hyperintense on T2 relative to the cortical gray matter. Calcification is possible, and cystic regions are also seen. Sometimes multiple cysts are noted, resulting in a “bubbly” appearance. The enhancement pattern is variable, but usually moderate and heterogenous.

Tumors in the Sellar and Parasellar Region

The sellar and parasellar group of extraaxial masses include pituitary micro- and macroadenomas and craniopharyngiomas. Meningiomas, arachnoid cysts, dermoid and epidermoid cysts, optic pathway gliomas, hamartomas, metastases, and aneurysms are also encountered in the para- and suprasellar region.

Pituitary Adenomas

The distinction between micro- and macroadenomas is based on their size: tumors less than 10 mm are microadenomas, the larger tumors are macroadenomas. These tumors may arise from hormone-producing cells, such as prolactinomas or growth hormone–producing adenomas, resulting in characteristic clinical syndromes. Pituitary adenomas are typically hypointense on T1-weighted and hyperintense on T2-weighted images, relative to the surrounding parenchyma. This signal change, however, is not always conspicuous, especially in the case of small microadenomas. Gadolinium administration helps in these cases, when the microadenoma is visualized as relative hypointensity against the background of the normally enhancing gland (Fig. 33A.25). Following a delay, this difference in enhancement is often no longer apparent, and if the postcontrast images are obtained in a later phase, a reversal of contrast may be noted. The adenoma takes up contrast in a delayed fashion and is seen as hyperintense against the more hypointense gland from where the contrast has washed out. Sometimes when the signal characteristics are not conspicuous, only alteration of the size and shape of the pituitary gland or shifting of the infundibulum may indicate the presence of a microadenoma. Because of this, it is important to be familiar with the normal range of pituitary gland sizes, which depend on age and gender. In adults, a gland height of more than 9 mm is worrisome. In the younger population, however, different normal values have been established. Before puberty, the normal height is 3 to 5 mm. At puberty in girls, the gland height may be 10 to 11 mm and may exhibit an upward convex morphology. In boys at puberty, the height is 6 to 8 mm, and the upward convex morphology can be normal. The size and shape of the gland may also change during pregnancy: convex morphology may appear, and a gland height of 10 mm is considered normal.

While microadenomas are localized to the sellar region, macroadenomas may become invasive and extend to the suprasellar region and displace/compress the optic chiasm or even the hypothalamus. Extension to the cavernous sinus is also possible.

Craniopharyngioma

Craniopharyngiomas are believed to originate from the epithelial remnants of the Rathke pouch. This WHO grade I tumor may be encountered in children, and a second peak incidence is in the fifth decade (Eldevik et al., 1996). The most common location is the suprasellar cistern (Fig. 33A.26), but intrasellar tumors are also possible. The tumor may cause expansion of the sella or erosion of the dorsum sellae. In the suprasellar region, displacement of the chiasm, the anterior cerebral arteries, or even the hypothalamus is possible. Craniopharyngiomas have both solid and cystic components. Histologically, the more common adamantinomatous and the less common papillary forms are distinguished. The adamantinomatous type frequently exhibits calcification. The MRI signal is heterogeneous. Solid portions are iso- or hypointense on T1, whereas cystic components exhibit variable signal characteristics depending on the amount of protein or the presence of blood products. On T2, the solid and cystic components are sometimes hard to distinguish, as they are both usually hyperintense. Areas of calcification may appear hypointense on T2. With contrast, the solid portions of the tumor exhibit intense enhancement.

Metastatic Tumors

Intracranial metastases are detected in approximately 25% of patients who die of cancer. Cerebral metastases comprise over half of brain tumors (Vogelbaum and Suh, 2006) and are the most common type of brain tumor in adults (Klos and O’Neill, 2004). Most (80%) metastases involve the cerebral hemispheres, and 20% are seen in the posterior fossa. Pelvic and colon cancer have a tendency to involve the posterior fossa. Intracranial metastases, depending on the type of tumor, may involve the skull and the dura, the brain, and also the meninges in the form of meningeal carcinomatosis. Among all tumors that metastasize to the bone, breast and prostate cancer and multiple myeloma are especially prone to spread to the skull and dura. Most often, carcinomas involve the brain and get there by hematogenous spread. Systemic tumors with the greatest tendency to metastasize to brain are lung (as many as 30% of lung cancers give rise to brain metastases), breast (Fig. 33A.27), and melanoma (Fig. 33A.28). Cancers of the gastrointestinal tract (especially colon and rectum) and the kidney are the next most common sources. Other possibilities include gallbladder, liver, thyroid gland, pancreas, ovary, and testicles. Tumors of the prostate, esophagus, and skin (other than melanoma) hardly ever form brain parenchymal metastases.

It is important to highlight the potential imaging differences between primary and metastatic brain tumors, since a significant percentage of patients found to have brain metastasis have no prior diagnosis of cancer. Cerebral parenchymal metastases can be single (usually with kidney, breast, thyroid, and lung adenocarcinoma) or (more commonly) multiple (in small cell carcinomas and melanoma) and tend to involve the gray/white matter junction. Seeing multiple tumors at the corticomedullary junction favors the diagnosis of metastatic lesions over a primary brain tumor. The size of metastatic lesions is variable, and the mass effect and peritumoral edema is usually prominent and, contrary to that seen with primary brain tumors, frequently out of proportion to the size of the tumor itself. The edema is vasogenic, persistent, and involves the white matter, highlighting the intact cortical sulci as characteristic fingerlike projections. It is hypointense on T1 and hyperintense on T2 and FLAIR. The tumor itself exhibits variable, often heterogenous signal intensity, especially if the metastasis is hemorrhagic (15% of brain metastases). Tumors that tend to cause hemorrhagic metastases include melanoma; choriocarcinoma; and lung, thyroid, and kidney cancer. The tumor signal characteristic can be unique in mucin-producing colon adenocarcinoma metastases, where the mucin and protein content cause a hyperintense signal on T1-weighted images.

Detection of intracerebral metastases is facilitated by administration of gadolinium, and every patient with neurological symptoms and a history of cancer needs to have a gadolinium-enhanced MRI study. The enhancement pattern of metastatic tumors can be solid or ringlike. To improve the diagnostic yield, triple-dose gadolinium or magnetization transfer techniques have been used, which improve detection of smaller metastases that are not so conspicuous with single-dose contrast administration. A triple dose of gadolinium improves metastasis detection by as much as 43% (van Dijk et al., 1997). Meningeal carcinomatosis can also be detected by contrast administration, which can reveal thickening of the meninges and/or meningeal deposits of the metastatic tumor.

Ischemic Stroke

Acute Ischemic Stroke

With the introduction of thrombolytic therapy in the treatment of acute ischemic stroke, timely diagnosis of an ischemic lesion, determining its location and extent, and demonstrating the amount of tissue at risk has become essential (see Chapter 51A). CT imaging remains of great value in the evaluation of acute stroke; it is readily available, and newer CT modalities including CT angiography and CT perfusion imaging are coming into greater use. The applicability of CT to acute stroke continues to be enhanced by the ever-increasing rapidity with which scans can be acquired, allowing for greater coverage of tissues with thinner slices. The technological advances allowing for rapid acquisition of data have led to 4D imaging, where complete 3D data sets of the brain are serially obtained over very short time intervals, allowing for higher temporal and spatial resolutions in brain perfusion studies of acute ischemic stroke patients.

CT is very useful in detecting hyperdense hemorrhagic lesions as the cause of stroke. Early ischemic stroke, however, may not cause any change on unenhanced CT, making it difficult to determine the extent of the ischemic lesion and the amount of tissue at risk. CT is especially limited in evaluating ischemia in the posterior fossa, owing to streak artifacts at the skull base. Despite these limitations, early signs of acute ischemia on unenhanced CT may be helpful in the first few hours after stroke. CT signs of acute ischemia include blurring of the gray/white junction and effacement of the sulci due to ischemic swelling of the tissues. Blurring of the contours of the deep gray matter structures is of similar significance. In cases of internal carotid artery occlusion, middle cerebral artery main segment (M1) occlusion, or more distal occlusions, intraluminal clot may be seen as a focal hyperdensity, sometimes referred to as a hyperdense MCA, or hyperdense dot sign (Fig. 33A.29).

Several MRI modalities as well as CT perfusion studies are capable of providing data regarding cerebral ischemia and perfusion to assist in the evaluation for possible thrombolytic therapy very early after symptom onset. DWI with ADC mapping is considered to be the most sensitive method for imaging acute ischemia (Figs. 33A.30 to 33A.33 [Figs. 33A.31, B; 33A.32, B; and 33A.33 available online only]). In humans, the hyperintense signal indicating restriction of diffusion is detected within minutes after onset (Hossmann and Hoehn-Berlage, 1995).

Temporal Evolution of Ischemic Stroke on Magnetic Resonance Imaging

Acute Stroke

Initially, the hyperintense signal on DWI is caused by decreased water diffusivity due to swelling of the ischemic nerve cells (for the first 5 to 7 days), then it increasingly results from the abnormal T2 properties of the infarcted tissue (T2 shine-through). For this reason, a reliable estimation of the age of the ischemic lesion is not possible by looking at DWI images alone. Imaging protocols for acute ischemic stroke usually include T1- and T2-weighted fast spin echo images, FLAIR sequences, and DWI with ADC maps. These sequences together confirm the diagnosis of ischemia, determine its extent, and allow for an approximate estimation of the time of onset (Srinivasan et al., 2006). On ADC maps, the values decrease initially after the onset of ischemia (i.e., the signal from the affected area becomes progressively more hypointense). This reaches a nadir at 3 to 5 days but remains significantly low until the seventh day after onset. After this time, the values increase (the signal gets more and more hyperintense) and return to the baseline values in 1 to 4 weeks (usually in 7 to 10 days). Therefore, ADC maps are quite useful for the estimation of the age of the lesion: if the signal of the area is hypointense on an ADC map, the lesion is likely less than 7 to 10 days old. If the area is isointense or hyperintense on the ADC map, the onset was likely more than 7 to 10 days ago. As already noted, although these signal changes take place on ADC maps, the DWI images remain hyperintense, without noticeable changes of intensity by visual inspection.

On T2-weighted (including FLAIR) images, the signal intensity of the ischemic area is normal in the initial hyperacute stage, increases markedly over the first 4 days, then becomes stable. In a research setting, computing the numerical values of hyperintensity in infarcted tissue on serial T2-weighted scans can demonstrate a consistent sharp signal increase after 36 hours, distinguishing lesions younger or older than 36 hours. This is certainly not possible by visual inspection used in clinical practice.

One purpose of MRI in the evaluation of acute stroke is to determine the extent of irreversible tissue damage and to identify tissue that is at risk but potentially salvageable. The combination of DWI and PWI is frequently used for this purpose (Fig. 33A.34). Evaluation is based on the premise that diffusion-weighted images delineate the tissue that suffered permanent damage (although in some cases, restricted diffusion is reversible, corresponding to ischemia without infarction), whereas areas without signal change on DWI but abnormal signal on perfusion-weighted images represent tissue at risk, the so-called ischemic penumbra. If there is a mismatch between the extent of DWI changes and perfusion deficits, the latter being larger, reperfusion treatment with intravenous or intraarterial thrombolysis or other intravascular techniques is justified to salvage the brain tissue at risk. If the extent of diffusion and perfusion abnormalities is similar or the same, the tissue is thought to be irreversibly injured, with no penumbra, and therefore the potential benefit from reperfusion treatment may not be high enough to justify the risk of hemorrhage associated with thrombolytic treatment.

Chronic Ischemic Stroke (3 Weeks and Older)

Areas of complete tissue destruction with death not only of neurons but of glia and necrosis of other supporting tissues as well, will eventually appear as cavitary lesions filled with fluid that have signal characteristics identical to CSF: hyperintensity on T2-weighted images and marked hypointensity on T1 images and FLAIR sequences. The region of encephalomalacia is bordered by a glial scar (reactive gliosis) that is hyperintense on T2 and FLAIR images (Fig. 33A.35; Fig. 33A.35, B available online only). Although the initial signal changes on DWI frequently predict the final extent of tissue destruction, changes on DWI can also disappear, and the final size of tissue cavitation can be best determined on T1-weighted images, which should be part of every stroke follow-up imaging protocol. Tissue in the margins of the cavitary lesion, and often in other areas of the brain as well, may have undergone extensive neuronal loss resulting only in atrophy but not in signal intensity changes, even on T2-weighted images (partial infarction).

Besides signal changes, chronic ischemic infarcts lead to secondary changes in the brain. Owing to the loss of tissue, ex vacuo enlargement of the adjacent CSF spaces (sulci and adjacent ventricular segments) occurs. Pathways that originate from or pass through the infarcted area undergo wallerian degeneration, which is seen as T2-hyperintense signal change along the course of these pathways (Fig. 33A.36). Later, the hyperintensity may resolve, but the loss of pathways may result in volume loss of the structures they pass through (e.g., cerebral peduncle, pons, medullary pyramid), noted as decreased cross-sectional area.

Stroke Etiology

Ischemic Stroke of Thromboembolic Origin

This stroke type results from occlusion of one or more major cerebral arteries or their branches. The occlusion may be due to in situ thrombus formation or embolization from a distant source. Emboli can be of cardiac origin, but they may also be the result of artery-to-artery embolization, commonly due to carotid or aortic arch atherosclerotic disease. Structural imaging studies are essential for the evaluation of potential embolic sources. Since embolic strokes usually involve major intracranial arteries and their branches, typical branch patterns can be identified with imaging. Interpretation of the branch pattern seen on imaging helps distinguish between strokes in the anterior (supplied by the internal carotid arteries) and posterior circulation (supplied by the vertebral and basilar arteries); imaging may also provide information regarding an embolic source. Unilateral anterior strokes often are due to embolization from the proximal internal carotid artery, a preferential site for atherosclerotic plaque formation. Likewise, unilateral embolic stroke in the posterior circulation necessitates evaluation of the vertebrobasilar system. It should be kept in mind that in case of the quite common anatomical variant of fetal origin of the posterior cerebral arteries (termed fetal PCA when they are predominantly fed by large posterior communicating arteries, which are variably present and arise from the internal carotids), posterior circulation stroke may result from embolization from the anterior circulation. Multiple, especially bilateral, cortical ischemic strokes almost always suggest an embolic origin. If the strokes are bilateral and/or involve both the anterior and posterior circulation, a more proximal embolic source such as the aortic arch or heart can be suspected. Reperfusion injury is a common phenomenon in embolism, and in this stroke type, hemorrhagic transformation of varying degree is often seen.

Lacunar Ischemic Stroke

Lacunar ischemic strokes constitute 20% to 25% of all strokes and are typically seen in patients with hypertension and diabetes. This stroke type is thought to be due to narrowing and in situ thrombosis of the small, deep-penetrating arteries such as the lenticulostriate arteries. The most common locations include basal ganglia, internal capsule, and thalamus. According to structural imaging criteria, their size is usually less than 15 mm in diameter. Acutely, lacunar infarctions may exhibit restricted diffusion if the resolution of the ADC map is high enough to differentiate such from background signal variation. Chronic lacunes have a smoothly rounded, well-defined appearance. The encephalomalacic core of chronic lacunar infarctions follows CSF signal on all pulse sequences, appearing markedly hyperintense on T2 and hypointense on both T1 and FLAIR. There is often a thin rim of hyperintense signal on FLAIR due to gliosis, which helps differentiate lacunes from large Virchow-Robin spaces (Fig. 33A.37).

Venous Stroke

Venous stroke may follow the thrombosis of cerebral veins (cortical draining veins and the cerebral deep venous system) or of one or more intracranial venous sinuses. The pathogenesis of venous ischemia/stroke is fundamentally different from arterial strokes. Thrombosis of the efferent channels (veins or sinuses) causes elevation of venous pressure, leading to congestion/dilatation of upstream capillaries and venules. This results in interstitial edema, which makes the area of venous infarction/ischemia hyperintense on T2-weighted and FLAIR pulse sequences. Rupture of the vessels may occur, leading to the frequently observed hemorrhagic component of these lesions, best visualized on GRE images. Further changes depend on the severity and duration of venous occlusion. Often the congestion is brief or transient, and the ischemic tissue recovers. In these cases, the sometimes very prominent signal changes can resolve, and no residual deficits will remain. In more severe cases that progress to infarction, restriction of diffusion (hyperintense signal on DWI and hypointense signal on ADC maps) is a common finding due to cytotoxic edema. Cytotoxic and vasogenic edema also results in hypointense signal on T1-weighted images. The venous etiology of the stroke is suggested by the morphological appearance of the lesion. Its distribution does not follow an arterial branch pattern. The appearance of the hyperintense signal changes on T2-weighted images and FLAIR sequences is also different; oftentimes heterogeneous signal changes are noted within the venous infarction, consisting of a “curly cue” or “fudge-swirl” pattern. Tumor-like appearances are also possible.

In cases of ischemia/stroke that are suspected to be of venous origin, it is important to carefully evaluate the draining veins in the area, and the sinuses as well, to look for thrombosis. The normal flow voids on MRI may be absent, replaced in some cases by hyperintense signal changes on FLAIR or hyperdensities on CT that exhibit a tubular or curvilinear string-like morphology. However, the pattern and distribution of cortical draining veins is very variable, which makes it difficult to pinpoint abnormalities of individual veins. Sometimes there is a striking absence of visualizable draining veins. Conversely, in cases of sinus thrombosis, massive engorgement of the veins may be seen. Venous thrombosis frequently starts at the level of a draining vein. In these cases, MR venography (MRV) may be initially unremarkable. MRV will become abnormal only later when the thrombosis progresses to the venous sinuses. Suspected cases of venous stroke are often best evaluated with two modalities: conventional MRI or CT in conjunction with MRV or a CT venogram.

Other Cerebrovascular Occlusive Disease

Microvascular Ischemic White Matter Lesions, “White Matter Disease,” Binswanger Disease

Diffuse or patchy T2-hyperintense signal changes in the deep hemispheric and subcortical white matter are probably the most common abnormal findings on MRI in the adult and elderly patient population. The terms microvascular ischemic changes or chronic small vessel disease are frequently used to describe these lesions on imaging studies. Their etiology and clinical significance have been debated extensively.

Certain hyperintense signal changes are considered normal incidental findings, with no clinical relevance. A uniformly thin, linear, T2 hyperintensity that has a smooth outer border along the border of the body of the lateral ventricles is often seen in the elderly population and likely represents fluid or gliotic changes in the subependymal zone. It tends to be more pronounced at the tips of the frontal horns (ependymitis granularis). This finding is thought potentially to be due to focal loss of the ependymal lining with gliosis and/or influx of interstitial fluid into these regions.

Patchy signal changes within the white matter of the cerebral hemispheres beyond a relatively low threshold (generally, one white matter hyperintensity per decade of life is felt to fall within the normal range) are pathological and are most commonly of ischemic origin. According to the most accepted hypothesis, these hyperintensities are the result of gradual narrowing or occlusion of the small vessels of the white matter, the diameters of which are less than 200 micrometers (hence the terms microvascular lesions or small vessel disease). Pathologically, these lesions are composed of focal demyelination and gliosis. The lumen of the involved vessels is narrow or occluded; their walls may exhibit arteriosclerotic changes and commonly amyloid deposits. On imaging studies, they have a chronic appearance, with diffuse borders and no surrounding edema or evidence of mass effect. They are generally associated with some degree of central atrophy, which tends to worsen with higher lesion loads. The distribution of these lesions changes only very gradually on serial scans, often showing minimal to no significant difference on studies spaced several years apart.

While age by itself can cause such changes, and the incidence of these lesions increases with age in people 40 years or older, there are several other risk factors that can make them more numerous. These include hypertension, diabetes, hypercholesterolemia, and smoking. Indeed, patients with these medical problems are more likely to have an elevated number of ischemic white matter lesions.

Chronic ischemic white matter lesions are hypodense on CT, but MRI is much more sensitive and reveals more extensive lesions (Fig. 33A.38; Fig. 33A.38, B-C available online only). On MRI, the lesions are hyperintense on T2 and FLAIR sequences. They may or not be visible as T1 hypointensities. It is possible that only lesions visible on T1-weighted images may be clinically significant. Common locations are the periventricular (PV) and more commonly, the deep white matter, but subcortical lesions are also common, with sparing of the U-fibers. The lesions can be isolated, scattered, or more confluent, especially in the PV zone. Morphologically, individual lesions generally exhibit indistinct borders with a diffuse “cotton-wool” appearance and range in size from punctate to small. Regions of confluent lesions may appear large and more commonly affect the deep white matter anterior and posterior to the bodies of the lateral ventricles, symmetrically within the parietal and frontal lobes. Deep white matter lesions also often occur in a distribution parallel to the bodies of the lateral ventricles on axial views, with an irregular band-like or “beads-on-string” appearance often separated from the PV lesions by an intervening band of relatively unaffected white matter. Involvement of the external capsules is also characteristic. These patterns of lesion distribution and morphology are often best seen on FLAIR. Contrary to the lesions of multiple sclerosis (MS), microvascular ischemia tends not to involve the temporal lobes or the corpus callosum. Besides the hemispheric white matter, microvascular ischemic lesions often also involve the basis pontis.

The clinical significance of ischemic white matter lesions depends on their extent and location. The presence of a few small, scattered, ischemic white matter lesions on T2-weighted images is clinically meaningless, and these are usually considered a normal imaging manifestation of aging. Patients may feel more comfortable with descriptions such as “age spots of the brain” to convey their benign nature when verbally discussing results. More extensive lesions also visible on T1-weighted sequences, however, are more likely to be associated with neurological abnormalities such as abnormal gait, dementia, and incontinence. In ischemic arteriolar encephalopathy or Binswanger disease, there is pronounced, widely distributed, and confluent PV and deep white matter signal change. In more severe cases, the confluent hyperintensity also involves the internal and external capsules or subcortical white matter. Besides confluent lesions, coexisting multiple scattered T2 hyperintensities are also very common. Ischemic white matter lesions are often intermixed with lacunar ischemic strokes and generalized cerebral volume loss is also frequently noted.

Scattered small, nonspecific-appearing, seemingly microvascular white matter hyperintensities have a broader differential diagnosis in the younger patient population. Multiple small T2 hyperintense lesions in the hemispheric white matter can be caused by migraine, trauma, inborn errors of metabolism, vasculitis (including Sjögren syndrome, lupus, Behçet disease, and primary CNS vasculitis), Lyme disease, and MS. Since the MRI appearance of these is nonspecific, clinical correlation is always warranted. In many instances, these white matter lesions are idiopathic, and future serial imaging studies are needed for follow-up.

Cerebral Venous Sinus Thrombosis

Acute cerebral venous sinus thrombosis results in diminished or absent flow in the involved sinuses. Cerebral venous sinus thrombosis usually causes typical signal changes on MRI (Fig. 33A.40) and severely attenuated or absent flow signal on magnetic resonance venography (MRV). MRV techniques include flow-sensitive modalities such as 2D time of flight and phase contrast imaging, as well as postcontrast high-resolution 3D-SPGR, which offers excellent visualization of the sinuses with a very high spatial resolution and contrast-to-noise ratio.

In the appropriate clinical context, a useful sign of venous sinus thrombosis is the absence of a normal hypointense flow void in the involved sinuses on T1- and T2-weighted images and absent flow in the involved sinus on MRV. Nonflowing blood generally results in increased signal intensity on T1 and T2. In the early acute stage, however, the sinuses may still be hypointense. This is followed by signal that is isointense to the gray matter. The typical hyperintense signal on T1- and T2-weighted images appears when methemoglobin is present in the clot. At all stages, therefore, simultaneous review of the MRV or CT angiogram for lack of flow signal and lack of contrast filling in conjunction with conventional MRI may be particularly useful to increase the sensitivity and specificity of detection of sinus thrombosis while also adding information regarding the age of the clot.

Following administration of gadolinium, there may be enhancement of the dural wall of the sinus and along the periphery of the clot, but not within the clot itself, resulting in an “empty delta” appearance. This is classically a CT finding, but the same concept also applies to MRI in the context of the T1-weighted clot signal that varies with clot age. MR demonstrates lack of flow, appearing as absence of contrast-related signal in the involved sinuses. CT angiogram reveals no contrast filling in the thrombosed sinuses. The cortical veins that drain into the involved sinuses may appear engorged on MRV. However, if the thrombosis also involves these draining veins, they too may exhibit lack of signal on MRV, lack of filling on CT angiogram, and lack of flow voids in conjunction with iso- or hyperintense signal on T1- and T2-weighted images.

Variations in the speed of blood flow and anatomical variants of the venous sinuses may change their usual signal characteristics, leading to a false diagnosis of venous sinus thrombosis. Slow flow in a venous sinus may cause increased signal on T1- and T2-weighted images, potentially leading to a false assumption of thrombosis. Gadolinium-enhanced images help in these cases, demonstrating contrast filling/enhancement in the sinuses and confirming the absence of thrombosis. A normal variant of venous sinus hypoplasia/aplasia may result in decreased/absent flow signal on MRV, falsely interpreted as thrombosis. T1- and T2-weighted images, however, are usually able to demonstrate the absence of thrombus in the sinus. These examples highlight the importance of reviewing all necessary image modalities (MRV, T2-weighted images, T1-weighted images with and without contrast) to make or reject a diagnosis of venous sinus thrombosis.

Hemorrhagic Cerebrovascular Disease

Structural neuroimaging is crucial in the evaluation of hemorrhagic cerebrovascular disease. Besides detection of the hematoma itself, its location can provide useful information regarding its etiology. Lobar hematomas, especially along with small, scattered, parenchymal microbleeds, raise the possibility of cerebral amyloid angiopathy, whereas putaminal, thalamic, or cerebellar hemorrhages are more likely to be of hypertensive origin. Other underlying lesions such as brain tumors causing hemorrhages can be detected by structural imaging. This section discusses hemorrhagic cerebrovascular disease and cerebral intraparenchymal hematoma, whereas other causes of hemorrhage such as trauma or malignancy are discussed in other sections. Please also refer to Chapter 51B for a clinical neurological review of intracerebral hemorrhages.

For decades, noncontrast CT scanning has been (and in most emergency settings still is) the essential tool for initial evaluation of intracerebral hemorrhage. In hyperacute (<12 hours after onset) and acute hemorrhage (12 to 48 hours), the patient’s hematocrit largely determines the lesion’s degree of density on CT. With a normal hematocrit, both retracted and unretracted clots exhibit hyperdensity that contrasts sufficiently with the isodense background of brain parenchyma to be easily detectable. In cases of anemia, however, small hemorrhagic lesions may potentially be overlooked owing to their lower CT density and may even be isodense to brain. The following sections describing the appearance of hemorrhage on CT and MRI studies all assume a normal hematocrit.

In the acute stage, the hematoma is seen as an area of hyperdensity on CT. The associated mass effect depends on the size and location. Effacement of the ventricles, cortical sulci, or basal cisterns is often seen. Various degrees of midline shift or subtypes of herniation (transtentorial, subfalcine, etc.) may occur. The surrounding edema is seen as hypodensity and tends to appear irregular with varying thickness depending on the degree of involvement of adjacent white matter tracts, which are preferentially affected. The initially distinct border of the hematoma changes within days to a few weeks after onset and becomes irregular and “moth-eaten” due to the phagocytic activity of macrophages. Small hematomas may disappear on CT within 1 week; in the case of larger hematomas, the process may take more than a month. Small hemorrhages may resolve without any residual change, while those that are larger are gradually replaced by an encephalomalacic cavity of decreased density and ex vacuo enlargement of the adjacent CSF spaces.

The appearance of hemorrhagic cerebrovascular disease on MRI is very complex regarding both signal heterogeneity on individual scans and subsequent changes in appearance over successive imaging studies. Signal characteristics of hemorrhage vary widely across different pulse sequences (T1, T2, T2* gradient echo) depending on the age of the hemorrhage; presence of oxyhemoglobin, deoxyhemoglobin, methemoglobin, and hemosiderin; changing water content within the clot; and integrity of erythrocyte membranes. Understanding the typical MRI appearance of each stage in the evolution of a hemorrhage allows one to estimate its age, because biochemical and structural changes characteristic of each stage (macroscopic and microscopic) occur along a predictable time line. In addition to conventional (T1- and T2-weighted) images, the gradient echo technique has been used to detect even small intracerebral hemorrhages, given its sensitivity to the paramagnetic properties (magnetic field distorting effects) of various blood products. More recently introduced into clinical practice, the technique of SWI offers the greatest sensitivity for chronic hemorrhage to date and is particularly useful in evaluating punctate hemorrhages in patients with diffuse shear injury secondary to prior head trauma.

A discussion of the MR imaging features of hemorrhage is best organized according to the stages of hemorrhage evolution as follows.

Hyperacute Hemorrhage (0 to 24 Hours)

In the early (hyperacute) phase of intraparenchymal hemorrhage (<24 hours) the red blood cells are intact, and a mixture of oxy- and deoxyhemoglobin is present (Bakshi et al., 1998). In this stage, the signal on T1-weighted images is isointense to the brain, so even larger hematomas may be missed on this pulse sequence. On T2-weighted images, the oxyhemoglobin portion is hyperintense and deoxyhemoglobin is hypointense, resulting in the gradual appearance of a hypointense rim and gradually increasing hypointense foci within the hematoma as the amount of deoxyhemoglobin increases from the periphery. Such hypointense foci are also seen on FLAIR. Between the clot and the deoxyhemoglobin-containing rim, thin intervening clefts of fluid-like T2 hyperintensity may be seen as an initial manifestation of clot retraction. On gradient echo images, hyperacute hemorrhage will exhibit heterogeneously isointense to markedly hypointense signal, the latter corresponding to deoxyhemoglobin content in more peripheral portions of the clot. The amount of edema is mild in this stage, usually seen as a thin rim that is hyperintense on T2 and FLAIR images and hypointense on T1-weighted images (Atlas and Thulborn, 1998).

Chronic Hemorrhage (>4 Weeks)

In the chronic stage (Bakshi et al., 1998), the core of larger hematomas turns into a slitlike or linear cavity with CSF signal characteristics, being hypointense on T1 and FLAIR and hyperintense on T2-weighted images. At the periphery of the lesion, macrophages continue to remove iron from the extracellular methemoglobin; hemosiderin and ferritin are deposited in their lysosomes, resulting in a rim of hypointense signal on T2-weighted and GRE images. This hypointense rim becomes progressively more prominent during the transition from the late subacute to chronic stage (Fig. 33A.43).

If the hemorrhage is small, eventually its entire area will be occupied by hemosiderin deposition. Smaller hemorrhages or microbleeds, such as those seen in amyloid angiopathy or after head trauma, are visualized as multiple uniformly hypointense foci on GRE images. Susceptibility-weighted images are even more sensitive to magnetic filed distortion due to blood products and can reveal microbleeds that are missed even by conventional gradient echo images. It is important to keep in mind that because of magnetic field distortion, the area of hypointensity on GRE or susceptibility-weighted images is larger than the actual size of the bleed. GRE or, ideally, SWI should be part of every MRI protocol for brain trauma.

Hemorrhage, like many other lesions to the brain, provokes reactive gliosis. In the chronic stage, surrounding gliosis is seen as mildly hyperintense signal on T2 and FLAIR images.

Superficial siderosis, a chronic sequela of bleeding into the subarachnoid space, and cerebral amyloid angiopathy, a hemorrhage-prone condition, are discussed in the online version of this book. Please visit www.expertconsult.com for more information.

Infection

Structural neuroimaging can provide useful information for evaluating infectious diseases of the CNS. The imaging modality of choice is MRI, which is able to demonstrate even subtle parenchymal abnormalities and inflammatory involvement of the meninges. For a review of the etiology, clinical presentation, and treatment of infections of the nervous system, see Chapter 53. And in addition to the neuroimaging features of infections discussed in the print version of this text, the online version includes features of CNS tuberculosis, cysticercosis, and cytomegalovirus. Please visit www.expertconsult.com for more information.

Cerebritis, Abscess

Cerebritis and abscess can arise as complications of bacterial meningitis, but they may also spread to the brain hematogenously from another source such as endocarditis or pulmonary abscess. Cerebritis and abscess refer to different stages of the parenchymal infection. In the cerebritis stage, the lesion has poorly defined hyperintensity on T2 and FLAIR sequences and is iso- to hypointense on T1-weighted images. Foci of necrosis may be present. There is surrounding edema, appearing as hypointensity on T1 and hyperintensity on T2 and FLAIR. With gadolinium, a heterogenous irregular enhancement pattern may or may not be present.

If the process continues to cerebral abscess formation, after an average of 2 weeks, the core appears more demarcated, and fibrotic capsule formation is noted. The center of the abscess contains liquefied, necrotic, and purulent material. This is usually hypointense on T1 (but may appear more hyperintense, depending on the protein content) and hyperintense on T2. The rim is iso- to hypointense on T1 and iso- to hypointense on T2. Often the T2 hypointense rim is well seen, separating the hyperintense core from the usually less hyperintense surrounding edema. With gadolinium, the capsule exhibits ring enhancement, which is typically a smooth, thin, complete ring. Sometimes the deeper segment of the enhancing ring is thinner than the superficial. A characteristic feature that supports the diagnosis of abscess is the so-called daughter abscess, which is seen as a smaller ring-enhancing lesion connected to the parent abscess.

Cerebral abscesses are part of the differential diagnosis when ring-enhancing cerebral lesions are encountered. Besides the described ring morphology and the potential presence of daughter abscesses, cerebral abscesses exhibit restriction of diffusion centrally, appearing as hyperintense signal on diffusion-weighted images and hypointensity on ADC maps, which distinguishes them from metastatic and most primary brain tumors (Fig. 33A.45).

Infection

Herpes Simplex Encephalitis

In adults, herpes encephalitis is caused by herpes simplex virus type 1. The imaging diagnosis is suggested by the location of the lesions. Herpes encephalitis typically involves the medial temporal and limbic frontal regions, including the basal frontal and cingulate gyri and insula. The cerebral cortex is more affected than the white matter. Herpes simplex encephalitis is usually hyperintense on T2 and FLAIR images and mildly hypointense on T1-weighted images (Fig. 33A.46). The encephalitis is frequently hemorrhagic, causing additional signal changes depending on the age of the hemorrhage. Areas of necrosis may be seen as well. Typically, a few days after onset, variable patterns of enhancement may be seen (gyriform, nodular, leptomeningeal, or intravascular). In the chronic stage, varying degrees of encephalomalacia, atrophy, calcification, and gliosis are seen in the affected lobes. Early successful treatment may minimize such sequelae.

Creutzfeldt-Jakob Disease

Creutzfeldt-Jakob disease is a rapidly progressing, fatal dementing illness caused by prions—self-replicating, infectious protein particles. In advanced cases, MRI usually reveals prominent atrophy and gray-matter hyperintense signal changes on T2 and FLAIR sequences. The hyperintensity typically involves the cerebral cortex, basal ganglia, and cerebellum. There is no mass effect, and enhancement is rare. In the early stage of the disease, conventional T2-weighted images may be entirely normal, but diffusion-weighted images may show ribbonlike hyperintensity along the cerebral cortex (Fig. 33A.48), commonly involving the insula and cingulate cortex. Hyperintensity may be also seen in the caudate nucleus, lentiform nucleus, and in the thalamus as well, in a nonvascular distribution. Involvement of the pulvinar is characteristic of the mad-cow variant. Cerebellar cortical involvement is also common. The abnormalities may be uni- or bilateral. FLAIR sequences have been known to display these abnormalities earlier than conventional T2-weighted images, but not as conspicuously as DWI (Shiga et al., 2004; Young et al., 2005).

Multiple Sclerosis and Other Inflammatory or White Matter Diseases

Inflammatory and noninflammatory lesions of the corpus callosum, leukodystrophy (Krabbe disease, metachromatic leukodystrophy, adrenoleukodystrophy), radiation leukoencephalopathy, posterior reversible encephalopathy syndrome, and central pontine myelinolysis are discussed only in the online version of this chapter. Please visit www.expertconsult.com for more information.

Multiple Sclerosis

Multiple sclerosis is a demyelinating disease with autoimmune inflammatory reaction against the myelin sheath of CNS pathways (see Chapter 54). MRI is essential for the diagnosis of MS by demonstrating the typical inflammatory demyelinating lesions disseminated in time and space (Fig. 33A.49). It is also used for disease monitoring and assessment of response to therapy.

MS white matter lesions may occur supratentorially or infratentorially, as well as within the spinal cord (imaging of spinal cord MS lesions is described later). Best evaluated on T2-weighted images, infratentorial lesions may be seen within the medulla, pons, midbrain, or cerebellum. Characteristic locations include the pontine tegmentum, periaqueductal region, cerebral peduncles, middle and superior cerebellar peduncles, and the white matter of the cerebellar hemispheres. Punctate or small lesions that are present directly adjacent to the fourth ventricle or cisterns are sometimes difficult to detect on T2-weighted images but are not uncommon. Infratentorial lesions are generally smaller than supratentorial lesions and are also less frequently hypointense on conventional T1-weighted images; they commonly appear hypointense on T1-weighted 3D spoiled gradient echo pulse sequences.

Supratentorial white matter lesions are usually best appreciated using the FLAIR pulse sequence, which nulls out CSF signal that may obscure periventricular abnormalities on conventional T2-weighted images. PV and subcortical white matter lesions typically are small in size and morphologically are generally ovoid or round on axial images. On sagittal views, PV lesions often exhibit a thin linear or fingerlike morphology (Dawson fingers), with the long axis of the lesion oriented perpendicularly to the wall of the lateral ventricle in a PV distribution. The PV distribution of many MS lesions is well demonstrated on SWI imaging, which reveals a single tiny, profoundly hypointense dot or thin linearity at the center of a significant proportion of demyelinating lesions. It represents a venule and is visible because of the magnetic susceptibility effects of deoxygenated venous blood, to which SWI is particularly sensitive.

Although the distribution of white matter lesions in MS has a somewhat random appearance, the hemispheres characteristically match each other in terms of lesion load, which typically is highest around the ventricles. On sagittal views, PV lesions are usually most numerous adjacent to the bodies and atria of the lateral ventricles, with less involvement of the white matter adjacent to the occipital and temporal horns. The deep white matter of the frontal and parietal lobes typically also tends to exhibit a greater number of lesions than either the occipital or temporal lobes. However, the presence of lesions adjacent to the temporal horns favors a diagnosis of MS. Juxtacortical demyelination is less commonly seen. Juxtacortical lesions, which involve the U-fibers, often exhibit a crescentic morphology and are usually seen only on FLAIR, proton density, or T2-weighted images. Occasionally they are hypointense on T1 as well. When present, this type of lesion favors the diagnosis of MS. Corpus callosum lesions are also relatively specific for MS. They are best visualized on sagittal FLAIR images, typically as punctate hyperintensities along the septocallosal margin (undersurface of the corpus callosum). Thin hyperintense linearities that are contiguous with and perpendicular to the undersurface may also be present. These findings are often superimposed upon a thin irregular band of T2 hyperintensity running along the undersurface of the corpus callosum rostrocaudally. Isolated lesions within the central fibers of the corpus callosum that are noncontiguous with the septocallosal margin are less typical and should raise the level of suspicion for alternate differential diagnoses, discussed in the following section.

Although it is well known from histopathological studies that MS affects not only the hemispheric white matter but also the cortex and deep gray nuclei, cortical gray matter lesions are uncommonly seen on conventional MRI studies. They may be seen more often with the use of high-field scanners (3.0 T or higher). Of the conventional pulse sequences, high spatial resolution FLAIR images are the most sensitive for cortical gray matter lesions. Detection is limited because the subtle hyperintensity of cortical lesions on FLAIR is only slightly greater than the already relatively hyperintense background of the cortical gray matter. A cortical gray matter lesion may be verified by correlation of the finding on separate FLAIR image stacks acquired in orthogonal planes or by detection of a hypointense lesion of identical morphology and location on T1-weighted 3D spoiled gradient echo pulse sequences. Cortical gray matter lesions generally have a curvilinear contour that conforms to the topology of the cortex but may also overlap with the adjacent white matter. Like deep gray-matter lesions, cortical gray-matter hyperintensities visible on MRI are usually small, in the millimeter range. Deep gray-matter MS lesions tend to be round or oval and are most frequently seen in the thalami.

T1-weighted images are useful for the detection of “black holes,” markedly hypointense lesions that have been shown histopathologically to exhibit more extensive demyelination and axonal loss than other lesions. They always exhibit a correlating hyperintensity on T2-weighted images but may appear centrally hypointense on FLAIR owing to their relatively high free water content. Conversely, not all T2-weighted lesions exhibit T1 hypointensity, and therefore the T1 lesion load is always less extensive than the T2 lesion load. Some MS cases, usually earlier in disease progression, exhibit no T1 hypointense lesions.

T1-weighted postcontrast images are useful for the detection of enhancing lesions in patients with clinical exacerbations. Enhancement may be solid or ringlike. Open-ring configurations are more typical for MS than other disease entities that also exhibit ring enhancement, such as tumors and infections (Masdeu et al., 2000). Five minutes between gadolinium injection and the acquisition of postcontrast images is the minimum acceptable delay for detecting acute demyelinating lesions, but longer delays of up to 30 minutes can significantly increase sensitivity, as can incorporation of magnetization transfer techniques and double or triple doses of gadolinium.

Chronic MS lesions do not exhibit restricted diffusion, appearing on DWI as either isointense to the surrounding brain parenchyma or, less commonly, hyperintense due to T2 shine-through artifact. On the corresponding ADC maps, chronic lesions exhibit normal or high apparent diffusion coefficients. However, as recently described, acute enhancing demyelinating lesions may on rare occasions exhibit high signal on DWI, with corresponding low pixel values (hypointense) on the ADC map, consistent with restricted diffusion (Balashov et al., 2009). Rapid resolution of restricted diffusion in these acute lesions is the rule, and they often evolve to exhibit increased ADC values on follow-up studies.

Nonconventional MRI pulse sequences are commonly used to assess MS in the research environment. SWI is a newer technique that is exquisitely sensitive to the small venules as well as iron deposition, the latter thought to play a pathophysiological role in MS. Magnetization transfer imaging may be used to generate histograms of pixel values within the normal-appearing white matter; such histograms are typically shifted to the left, with lower peak values than in normal individuals. Diffusion tensor imaging is a more advanced application of DWI that is used to measure the directional diffusion of water molecules within white matter tracts. Like magnetization transfer imaging, DTI is useful for the quantification of pathology within normal-appearing white matter.

Inflammatory and Noninflammatory Lesions of the Corpus Callosum

In general neurological practice, the most commonly seen corpus callosum lesions are secondary to MS. FLAIR pulse sequences in the sagittal and axial planes are generally sufficient to assess corpus callosum pathology, although coronal views are also quite helpful. Isolated T2-hyperintense lesions within the central fibers of the genu, body, or splenium of the corpus callosum that are noncontiguous with the septocallosal margin are an uncommon finding. If such lesions exhibit a punctate or “snowball-like” appearance on sagittal views, Susac syndrome should be considered (see Chapter 15). Other small-vessel vasculitides can also result in small or punctate lesions within the corpus callosum.

An isolated medium-sized, bilaterally symmetrical lesion within the central fibers of the splenium that often also exhibits restricted diffusion is a transient finding in epilepsy (Polster et al., 2001) or metabolic encephalopathy (Takanashi et al., 2006). If patchy white-matter T2-hyperintensities are present in conjunction with a reversible splenial lesion, CNS lupus should be considered. When accompanied by multiple acute-appearing white and gray matter lesions, ADEM is more likely, and the corpus callosum lesion in this case is more likely to evolve to a chronic appearance. A T2-hyperintense lesion involving most of the central fibers of the splenium while sparing the periphery is typical of Marchiafava-Bignami disease; its typical morphology is best demonstrated on sagittal views.

The differential diagnosis (Uchino et al., 2006) for an isolated corpus callosum lesion also includes low-grade neoplasm if there is little to no mass effect and absent enhancement. If enhancement or mass effect is present, high-grade malignancies with a predilection for the corpus callosum (e.g., glioblastoma, primary CNS lymphoma), as well as metastases, are differential diagnostic considerations. Focal atrophy of the corpus callosum with overlapping T2-hyperintensity, contiguous with or near a region of cerebral hemispheric encephalomalacia and gliosis, is suggestive of wallerian degeneration. Diffuse axonal injury may also result in lesions of the corpus callosum and if suspected should be further evaluated with T2* or SWI to detect possible microhemorrhages.

Neurosarcoidosis

The prevalence of nervous system involvement in sarcoidosis is approximately 5%. In the brain parenchyma, multiple periventricular T2 hyperintense lesions are frequently noted, which may be due to a vasculitic process and often cannot be distinguished from MS or ischemic microvascular changes. Sarcoidosis may also involve the pituitary infundibulum and hypothalamus (resulting in endocrine symptoms). The granulomatous inflammation may affect the cranial nerves and/or their meningeal coverings as well, resulting in enlargement, hyperintense signal change, and abnormal enhancement. Although sarcoidosis may involve the dura, following gadolinium administration, leptomeningeal enhancement is more commonly noted. It is typically seen along the penetrating blood vessels and the adjacent leptomeninges, and is due to perivascular spread of the granulomatous process (Fig. 33A.50). Sometimes larger intraparenchymal or meningeal enhancing lesions are noted which may be mistaken for primary or metastatic tumors. When involved, the pituitary infundibulum and hypothalamus may also exhibit enhancement. Neurosarcoidosis may also lead to hydrocephalus, either by interfering with CSF absorption through the arachnoid granulations or by obstructing the ventricular system.

Leukodystrophy

Leukodystrophies are genetic degenerative diseases of the white matter. The various genetic defects result in deficiencies of enzymes necessary for myelin production and maintenance. This leads to degeneration/destruction of the myelin sheath of the neural pathways, seen as extensive white matter lesions and progressive atrophy.

Radiation Leukoencephalopathy

Radiation therapy may result in damage to the brain parenchyma, and this may happen at various time intervals after treatment (Bakshi et al., 1998). Acute radiation injury that happens during or a few days after treatment may not be seen on MRI. Sometimes, increased mass effect or more intense gadolinium enhancement may be seen in the irradiated area. In early delayed radiation damage, the changes appear in a few weeks or months after irradiation. Signal changes that are hypointense on T1 and hyperintense on T2 and FLAIR appear in the radiated region or in the PV white matter and basal ganglia in cases of whole brain radiation. These may enhance with gadolinium and tend to resolve in a few months. Late delayed radiation injury (which may be focal or diffuse) happens months or several years after treatment and usually follows longer treatment periods or higher radiation doses. The white matter changes (postirradiation leukoencephalopathy) start to appear in PV locations, and with time they become confluent around the borders of the ventricles. The changes can spread to other deep white-matter areas, seen confluently in the centrum semiovale (Fig. 33A.52). The involved white matter may exhibit areas of calcification. Ventricular and cortical sulcal enlargement is common. These changes are typically permanent.

Posterior Reversible Encephalopathy Syndrome

Posterior reversible encephalopathy syndrome (PRES) is thought to be due to abnormal cerebral vascular autoregulation resulting in vasogenic edema. A common etiology is severe hypertension (often in the setting of eclampsia or renal failure), but various medications can also cause it as a toxic side effect (tacrolimus, cyclosporin, FK-506, cyclophosphamide, interferons, and various chemotherapeutic agents). There is damage to the capillary endothelial junctions, vasodilatation resulting in fluid extravasation, and vasogenic edema. This condition has a predilection for the white matter of the posterior regions of the hemispheres, hence the former name of reversible posterior leukoencephalopathy syndrome (RPLS). In typical cases, variably extensive T1 hypointense and T2 hyperintense vasogenic edema is seen, usually symmetrically in the white matter of the parietal and occipital lobes (Fig. 33A.53). The name of the disease was changed after structural imaging revealed that gray matter, such as the basal ganglia and cerebral cortex, may also be involved (McKinney et al., 2007). It is also to be noted that at times the signal changes may be seen in the more anterior regions in the frontal and temporal lobes as well. With treatment of the underlying condition, such as hypertensive crisis or after withdrawal of the inciting agent, the clinical and imaging findings typically resolve.

Central Pontine Myelinolysis

Central pontine myelinolysis has been described in cases of rapid correction of hyponatremia but also in malnutrition, alcoholism, and cases of hepatic insufficiency. MRI is the study of choice, and on T2-weighted images, striking hyperintense signal change is seen in the pontine tegmentum and basis pontis that characteristically spares the periphery and the corticospinal tract (Allen, 1984). At times, gadolinium enhancement may be seen at the periphery of the lesion, but usually no enhancement is seen. Although the typical location is the pons, regions of extrapontine demyelination can also be seen in a minority of patients, including the basal ganglia, thalamus, cerebellum, and within the hemispheres.

Trauma

Both CT and MRI have a pivotal role in the evaluation of craniocerebral trauma (Chapter 50B). In the emergency room setting, the first imaging modality is usually a noncontrast CT scan. CT bone windowing is the best tool for evaluating skull fractures (Fig. 33A.54), whereas brain windowing is of great value for visualizing subarachnoid, epidural, subdural, intraventricular, or intraparenchymal hemorrhages. MRI is very useful in detecting traumatic lesions of the brain parenchyma, especially when more subtle changes are present, such as small contusions, parenchymal microbleeds, and the small or punctate lesions of diffuse axonal injury. The various consequences of trauma that we will review next seldom occur in isolation but tend to occur in various combinations (e.g., cortical contusion is frequently associated with subarachnoid hemorrhage).

Traumatic Subarachnoid Hemorrhage

Traditionally, a noncontrast CT scan has been the first-line imaging study to demonstrate traumatic subarachnoid hemorrhage. Acute blood appears hyperdense in the subarachnoid space. The hemorrhage may be seen in the sylvian fissures, interhemispheric fissure, basal cisterns, or in the cortical sulci at the convexities, depending on the site of trauma. Subarachnoid hyperdensity tends to resolve within 5 to 7 days after the hemorrhage, but depending on the amount of subarachnoid blood, this may be a shorter or longer process.

Until FLAIR pulse sequences became available, MRI was inferior to CT in detecting subarachnoid blood, especially in the first few days. Conventional T1- and T2-weighted images may completely miss subarachnoid blood in these early stages. However, subarachnoid blood appears hyperintense on FLAIR images (see Fig. 33A.54), and this pulse sequence is considered equal or superior to CT in detecting subarachnoid blood (Woodcock et al., 2001), especially in the posterior fossa and at the skull base, where CT images are often compromised by beam-hardening artifacts. The hyperintense signal change is due to the presence of blood, which changes the zero point of the inversion time of CSF, and therefore signal attenuation in subarachnoid regions containing hemorrhage will not be complete. An important caveat is that subarachnoid hyperintensity on FLAIR may also result from other causes. A common cause is magnetic susceptibility artifact, seen in patients who have braces or other metal devices that distort the magnetic field, which results in a lack of CSF signal suppression and the artifactually hyperintense appearance of CSF adjacent to the anterior frontal lobes.

Besides detection of blood in the acute phase, structural neuroimaging is also very useful to evaluate the potential later complications of subarachnoid hemorrhage. Subarachnoid blood may occlude the arachnoid granulations, leading to impaired CSF absorption, communicating hydrocephalus, and ventriculomegaly. Another late phenomenon that tends to follow repeated episodes of subarachnoid hemorrhage is superficial siderosis. In this condition, hemosiderin is deposited along the leptomeninges and appears as linear areas of hypointensity on T2-weighted images (see Fig. 33A.44 online at www.expertconsult.com).

Subdural Hemorrhage

Subdural hematomas are common sequelae of head trauma and are thought to result from rupture of the bridging veins (veins that drain from the cerebral surface and pierce the dura to enter the adjacent venous sinus). Morphologically they follow the contour of the cerebral surface and can cross the cranial suture lines but not the midline at the falx cerebri and cerebelli. Depending on the size, there is a varying degree of mass effect on the adjacent brain; in the more severe cases, effacement of the adjacent ventricles, midline shift, and various herniation syndromes may occur.

As subdural hematomas age, their imaging appearance changes both on CT and MRI. On CT, acute subdural hematomas appear hyperdense. If the patient remains in a recumbent position, the cellular elements settle to the lower part of the hematoma, which will appear more hyperdense, whereas the “supernatant” is less so. With time, hemoglobin degradation occurs and the density of the hematoma will decrease, eventually becoming hypodense. During this process there is a transitional stage when the density of the hematoma will be very similar or the same as that of the brain, rendering its detection more difficult. Just as with intraparenchymal hematomas, the density depends on the hematocrit, and in severely anemic patients, even acute subdural hematomas may appear iso- or hypodense, leading to erroneous dating.

On MRI, subdural hematomas exhibit a signal evolution similar to that seen with intraparenchymal hemorrhages, but the pace of evolution is different due to a slower decrease of the oxygen content within the hematoma. Acute subdural hematomas (Fig. 33A.55) are initially isointense on T1 and hyperintense on T2, but as deoxyhemoglobin appears, the signal on T2-weighted images becomes hypointense. In the subacute phase, the signal is hyperintense on T1 and hypointense on T2, but in the late subacute stage, the signal will be hyperintense on both T1- and T2-weighted images because of extracellular methemoglobin (Fig. 33A.56). It is important to remember that these stages are not separated sharply, and mixed patterns are often seen; this is due to the presence of oxy- and deoxyhemoglobin in the acute stage and intra- and extracellular methemoglobin in the chronic stage. Rebleeding into an existing subdural hematoma may also occur, resulting in the presence of clots of various ages.

Chronic subdural hematomas (Fig. 33A.57) are hypointense relative to the brain but, having higher protein content, are mildly hyperintense relative to the CSF on T1-weighted images and hyperintense on T2-weighted images. Hemosiderin deposition is not as prominent as in parenchymal hemorrhages because macrophages tend to be cleared by the meningeal circulation. Chronic subdural hematomas may look similar to hyperacute ones on noncontrast images, but because of their vascular membrane, with gadolinium they exhibit enhancement along their periphery. On CT, chronic subdural hematomas appear as hypodense subdural collections. Mass effect is variable depending on the size of the hematoma and degree of cerebral atrophy. If repeated hemorrhage occurs into the subdural collection, the hyperdense fresh blood is seen within the chronic hypodense collection (Fig. 33A.58).

Cortical Contusion

Cerebral contusions result from the brain hitting against the inner table of the skull or sliding against the bony ridges of the base of the skull. The most common locations include the poles and inferior surfaces of the frontal, temporal, and occipital lobes. The injured brain parenchyma exhibits foci of hemorrhage and varying degrees of edema, which may progress later to more confluent hematoma and more swelling. On CT, acute contusions appear as foci of hyperdense hemorrhage with or without swelling. The hematomas get reabsorbed later, and the swelling decreases. In the chronic stage, no residual findings or varying degrees of encephalomalacia may be seen. On MRI, with the appearance of deoxyhemoglobin, a hypointense signal is seen on T2; the surrounding edema is hyperintense on T2 and FLAIR sequences (Fig. 33A.59). Later, in the subacute stage, extracellular methemoglobin is hyperintense on T1- and T2-weighted images. Gradient echo and susceptibility-weighted images are very useful to show the hemorrhagic component of the lesion. Chronic contusions are associated with encephalomalacic cavities of various sizes, exhibiting CSF signal characteristics and surrounded by a hyperintense rim of reactive gliosis on T2 and FLAIR sequences. Variable degrees of hemosiderin deposition appear hypointense on FLAIR, gradient echo, and SWI.

Metabolic and Toxic Disorders

This section is in the online-only version of this chapter. Please visit www.expertconsult.com for more information, and also refer to Chapters 56 through 58 and 62 for further discussions of these topics.

Metabolic and Toxic Disorders

Disorders Due to Ethanol

Alcohol abuse may lead to various CNS disorders that are detectable by structural neuroimaging. These can be either due to direct toxic effects on the brain or due to toxic/metabolic complications of long-term alcoholism.

Brain degeneration with atrophy is a nonspecific but common imaging finding in chronic alcoholics. On CT and MRI, it appears as enlargement of the central and superficial CSF spaces, often prominently involving the frontal lobes. It is important to note that apparent cerebral volume loss initially may be due to the dehydrating effects of alcohol and, therefore, a reversible phenomenon: with abstinence, the size of the CSF spaces may normalize. Infratentorially the characteristic finding is cerebellar atrophy, typically involving the anterior and superior vermis.

Wernicke encephalopathy is a rapid-onset metabolic syndrome that may occur in alcoholism but also in other cases of severe malnutrition; the cause is thiamine deficiency. In the absence of thiamine, pyruvate cannot enter the Krebs cycle, resulting in nerve cell loss, demyelination, glial proliferation, and often microhemorrhages. These lesions have a characteristic distribution. The typical locations include PV areas of the thalamus (e.g., dorsal medial nucleus) and hypothalamus, periaqueductal region of the dorsal midbrain (including the nuclei of the third and fourth cranial nerves), mamillary bodies, rhomboid fossa (dorsal motor nucleus of the vagus nerve, vestibular nuclei) and superior cerebellar vermis. On MRI, more severe lesions are iso- to hypointense on T1 and hyperintense on T2-weighted images. They usually enhance, and often the abnormal enhancement is subtle and the only abnormality seen. Thin-slice coronal and axial images are often used to visualize these lesions.

Hepatic failure with hepatic or portosystemic encephalopathy is another frequent complication of long-term alcohol abuse. Typically on T1-weighted MRI, hyperintense signal changes appear, most commonly in the globus pallidus but sometimes also in the internal capsule, putamen, caudate nucleus, subthalamus, and mesencephalon. T2-weighted images are usually unremarkable, and enhancement is uncommon. The T1 hyperintense signal changes are attributed to deposits of paramagnetic toxins (such as manganese) and are potentially reversible after treatment of the underlying condition.

Mitochondrial Disease

Mitochondrial diseases (Chapter 63) cause cellular metabolic disturbances. Most often responsible are mutations in mitochondrial DNA or in nuclear genes that regulate mitochondrial function. Tissues with high metabolic activity such as brain and skeletal muscle are preferentially affected. Various syndromes belong to this group of metabolic disorders, the most common being MELAS (mitochondrial encephalomyopathy with lactic acidosis and stroke-like episodes), Kearns-Sayre syndrome, MERRF (myoclonic epilepsy with ragged-red fibers), and Leigh disease.

In MELAS, stroke-like events of acute onset are often clinically transient but may result in persistent deficits. The pathophysiological correlate of an acute stroke-like episode can be seen on MRI as a T2-hyperintense lesion that exhibits gyriform enhancement following contrast administration. Such lesions often resolve but may persist, progressing to cortical laminar necrosis and encephalomalacia within the affected region. The brain lesions associated with MELAS appear similar to ischemia, but they preferentially affect the cortex, with relative sparing of the deep white matter. The lesions usually cross vascular territories, and although they may be hyperintense acutely on DWI, the corresponding ADC value is generally not as low as in typical acute ischemic strokes. Magnetic resonance spectroscopy (MRS) of the ventricular CSF and brain parenchyma reveals a lactate peak in the majority of cases. Multiple episodes over time may result in a random-appearing distribution of chronic-appearing ischemic lesions within the brain parenchyma. In later stages, cerebral atrophy is common. MRI also reveals nonspecific T1 hypointense and T2 hyperintense lesions in the white matter, with some predilection for the parietal and occipital lobes. The cerebellum and basal ganglia may be involved as well.

Leigh disease is a severe mitochondrial disease that results in necrotic T1 hypointense and T2 hyperintense lesions bilaterally in the putamen, caudate nucleus, and globus pallidus. It may also affect the brainstem, including the medulla. In contrast to MELAS, the lesions seen in Leigh disease are strikingly symmetrical and well circumscribed. MRS may reveal elevated lactate levels within the basal ganglia.

Genetic and Degenerative Disorders Primarily Causing Ataxia (Cerebellar Disorders)

Multiple System Atrophy, Cerebellar Subtype

Multiple system atrophy (MSA) is a neurodegenerative disorder that results in symptoms and signs referable to various CNS structures. Depending on the systems involved, three different MSA subtypes are distinguished: MSA-a (autonomic subtype, Shy-Drager syndrome), MSA-p (parkinsonian subtype), and MSA-c (cerebellar subtype).

In the cerebellar subtype of MSA, predominantly infratentorial structures are affected, hence the other name for this entity, olivo-ponto-cerebellar atrophy (OPCA). There may be atrophy of the olivary eminences of the medulla, which is well appreciated on thin-slice axial images. In the pons, a horizontal T2 hyperintense signal change may be noted, characteristically involving the transverse pontine fibers, and often a vertical linear hyperintensity is also seen, corresponding to the degenerating neurons of the raphe nuclei. The constellation of these linear T2 hyperintense areas that cross each other perpendicularly is referred to as the hot-cross-bun sign. Although this sign has been thought to be specific for MSA-c, it has also been described in the other two MSA subtypes, spinocerebellar ataxia (types 2 and 3), parkinsonism with vasculitis (Muqit et al., 2001), fragile X premutation tremor/ataxia syndrome, and in variant Creutzfeldt-Jakob disease (Soares-Fernandes et al., 2009). Another potential finding in MSA-c is T2 hyperintense signal change, extending into the middle cerebellar peduncles in a symmetrical fashion. This finding is not pathognomonic—for instance, occurring also in fragile X premutation tremor ataxia syndrome.

Imaging features of MSA-a and MSA-p, including findings that help distinguish the three subtypes of MSA, are described in the section covering degenerative diseases that primarily cause parkinsonism.

Spinocerebellar Ataxias

Spinocerebellar ataxias (SCA, types 1-29) are genetically heterogeneous disorders resulting in expansion of glutamine-coding CAG triplets of the involved genes. The resultant abnormal proteins (containing polyglutamine) are felt to have a pathogenetic role. In these diseases, variable degrees of cerebellar, brainstem, and spinal cord atrophy are seen with imaging (Fig. 33A.61). The structural imaging appearance is not specific but supports the diagnosis. SCA type 6 is a pure cerebellar syndrome that stands out from the structural imaging standpoint in that there is little extracerebellar atrophy, and the atrophy predominantly involves the vermis, especially the superior vermis. This imaging appearance is similar to that seen in chronic alcohol abuse or phenytoin-related cerebellar toxicity, which are the main differential diagnostic considerations.