Image manipulation

Published on 12/06/2015 by admin

Filed under Radiology

Last modified 22/04/2025

Print this page

This article have been viewed 1500 times

7 Image manipulation

Aim of Image Manipulation

To produce an excellent image to maximise diagnostic accuracy

Terminology

Analogue	Represents a quantity changing in steps which are continuous, i.e. a sine wave
Brightness	The intensity values of the individual pixels in an image, the lower the brightness the darker the image
Compression	The reduction in size (in bytes) of an image to save storage space
Contrast	The density difference between two adjacent areas on the image
Digital	An image comprised of discrete areas or pixels
Edge Enhancement	The highlighting of a straight line or edge of an object to visually increase the sharpness of the image
Fourier Transform	A method of mathematically changing data, e.g. changing spatial data to frequency data
Frequency Data	The number of times a specific value occurs in an image
Heuristic	When an image is automatically improved because the program has changed due to a previous imaging experience
Hough Transform	A method of highlighting areas of a specific shape within an image
Noise	Anything that may detract from the image
Resolution (Sharpness)	The size of the smallest object or distance between two objects that must exist before the imaging system will record that object or objects as separate entities.
Segmentation	Selection of an area of interest and eliminating unwanted data. Can be done manually or automatically with an appropriate software package
Signal	The information required from the imaging system, e.g. the radiograph, the minimum size of the object that must be visible
Spatial Data	Gives the position of the varying intensities (brightness) across an image
Spatial Frequency	Object size, measured in line pairs per millimetre
Spatial Resolution	The smallest part of an image that can be seen
Window	The range of colour (or grey) scale values displayed on a digital image

Digitising an Analogue Image

An Analogue Image	• Two dimensional image • Different shades of grey • The shading is continuous throughout the image
A Digital Image	• Two dimensional image • Different shades of grey (or colours – red, green blue) • The grey is made up of discrete areas or pixels
Changing an Analogue Image to a Digital Image	• Take the analogue image • Imagine a grid superimposed over the image – usually 512 × 512 or 1024 × 1024 • The brightness (amount of grey) in each square (pixel) in the image is measured • Each pixel can then be given a number representing the brightness/shade of grey • Usually each pixel is given a value between 0 and 256 • This gives the image numerical values, i.e. digitises it
Nyquist Theorem	States that an analogue signal waveform may be reconstructed without error from a sample which is equal to, or greater than, twice the highest frequency in the analogue signal, e.g. • To digitally convert a 2 MHz signal, a sample must be taken at 4 MHz • To give a resolution of 5 line pairs per millimeter, each line pair equals 0.2 mm therefore a pixel sample must be measured at 0.1 mm intervals
Fourier Transform	• A method of mathematically changing data, e.g. changing spatial data to frequency data • Spatial data give the position of the varying intensities (brightness) across an image

Image Enhancement

	Methods of manipulating the pixel values to improve or enhance the area of interest in the image
Windowing	• Process of using the pixels to make an image • 256 shades of grey are usually assigned • But the human eye can only determine about 100 shades of grey • The shades of grey can be distributed over a wide or a narrow range of pixels
Narrow Window	• The grey is distributed over a narrow range of units • The central unit is the average pixel number for the structure of interest • If the average pixel was 50 and a narrow window of 170 was selected, then pixels of 85 (half 170) above and below 50 would be used • Therefore the grey scale would extend from – 35 to 135 • Any readings below – 35 would be pure black • Any readings above 135 would be pure white
Wide Window	• The grey is distributed over a wide range of pixels • The central unit is the average pixel number for the structure of interest • If the average pixel was 400 and a wide window of 2000 was selected, then pixels of 1000 (half 2000) above and below 400 would be used • Therefore the grey scale would extend from – 600 to 1400 • Any readings below – 600 would be pure black • Any readings above 1400 would be pure white

Adjusting Noise and Contrast

Signal to Noise Ratio	Image quality may be defined as the signal to noise ratio: The signal is the information required from the imaging system • The signal can be defined as the minimum size of the object that must be visible The noise is anything that may detract from that signal • The noise, on the monitor, could be defined as the graininess of the image
Image Quality	• If the sharpness of the system is increased, the visually disturbing noise will also increase, as now the system is resolving the noise better as well as giving a sharper signal • If the contrast is increased the signal will appear to be clearer, but so will the noise • To compensate for all the factors is impossible, therefore the final result has to be a compromise (Fig. 7.1)
Contrast	A radiograph is the product of a transfer of information. During this transfer it is exposed to a number of different influences. Contrast helps to determine the quality of the radiograph There are three principal ‘types’ of contrast • Subject contrast • Image contrast • Radiographic contrast
Subject Contrast	Subject contrast (Fig. 7.2) can be defined as the ratio of the emergent intensities, i.e.: • This is caused by differential attenuation and absorption of the X-ray beam as it passes through the patient (i.e. the subject) • It is responsible for the differing intensities of the emergent X-ray beam, and therefore the exposures that eventually reach the film • Bone attenuates more of the beam than the fatty tissue and therefore the emergent intensity in the area below the bone is less than the surrounding fatty tissue • The film receives less exposure and produces a lower image density when compared to the fatty tissue areas
Factors Affecting Subject Contrast
Different Thicknesses of the Same Tissue Type	Subject contrast is the ratio of the intensity that has passed through the thin part, compared with the thicker part The thicker of the two will: • Attenuate more of the beam • Allow less exposure to reach the film
Different Densities of the Same Tissue with the Same Volume but at a Higher Density	Subject contrast is the ratio of the intensity that has passed through the less dense part, compared with the denser part The higher density will: • Attenuate more of the beam • Reduce the intensity of the emergent beam
Different Atomic Numbers of Different Tissues	The higher the atomic number: • The more the attenuation of the incident X-ray beam • The less the intensity of the emergent beam
Different Atomic Numbers of Different Tissues	Note At the energies used in diagnostic radiography, photoelectric absorption predominates and is the largest contributing factor to subject contrast
Radiation Quality – The kiloVoltage (kV) Set for the Exposure	• Changing kV gives a high contrast variation in bone work • The influence of kV is smaller on soft tissue rendition • In the low kV range, 10 kV has more effect on contrast than in the higher kV range • In the middle kV range, 20 kV is needed to give an appreciable change in contrast For the same subject, increasing the kV: • Decreases the difference in intensities of the emergent beam • Therefore decreases the subject contrast Low kV will produce high subject contrast
	Note The kV must be high enough to adequately penetrate the area being examined
X-ray Equipment	Factors affecting subject contrast include: • The anode material, e.g. mammography • Varying voltage ripple components • Tube filters and supplementary filters
Scattered Radiation	Radiation fog, which increases the overall density of the image Scatter can be limited by: • Lowering the kV In general (below 150 kV), the lower the selected kV, the lower the amount of scatter and the lower the radiation fog and reduction in image latitude The use of: • Grids • Collimators
Use of Contrast Agents	Used to fill a cavity or space in the body that usually has a low subject contrast when compared with surrounding structures Positive agents • Non-ionic iodine compounds or barium which increase the X-ray absorption properties of the structure concerned
Use of Contrast Agents	Negative agents • Carbon dioxide or air which decrease the absorption properties, with an effect on the intensity of the emergent beam These can also be used in combination, e.g. double contrast barium meals and enemas
Radiographic Contrast	• This is the density difference between two adjacent areas on the image • These differences in density may or may not be observable to the naked eye
Subjective Contrast	• The eye analyses the structures in the radiograph and will select details • The success rate depends on the different degrees of brightness: Density detail – Density surroundings • This variable indicates the difference in logarithmic form which suits the physiology of the eye • This measurement is also the measurement of subjective contrast, i.e. Contrast = Density detail = Density surroundings • It is the observer’s opinion of the contrast that is seen on the film • It is a combination of all the other factors listed, plus: Viewing conditions The performance of the eye The observers ability and perhaps their personal opinion

Fig. 7.1 Image quality is a compromise between sharpness, contrast and noise.

Fig. 7.2 Diagrammatic illustration of subject contrast.

Contrast Enhancement in Digital Imaging

Histogram Production	For a given image, a histogram can be produced, plotting: • Intensity level of the image pixels (horizontal axis) • Number of pixels (vertical axis)
Histogram Equalisation Software	• Alters a computer image to equally distribute the brightness levels over the whole histogram • This effect increases the contrast of the image
Histogram Equalisation Software	Note If the intensity range of the original histogram is small the changed image will have a lot of noise • Used to lighten a dark image
Contrast Stretching Software	• If the image has a narrow range of intensity values the histogram can be ‘stretched’ to cover a specified upper and lower intensity range • This improves the contrast of the image without changing the range of grey levels

Noise Reduction

Density Slicing or Thresholding	The separating (segmenting) of an area of interest and removing unwanted information. Works best if there is a clear peak on the histogram that can be selected • Either by selecting pixels of similar values (intensity threshold) The image is scanned Each pixel is compared with the threshold Those above the threshold are turned white Those below are turned black (or vice versa) • Or by selecting specific colours or intensities in the image background, e.g. The background pixels could be selected and turned black The remaining pixels are unchanged
Image Smoothing	Neighbourhood averaging (Gaussian smoothing) • Either the average value of the pixels is calculated • Or an area of the image is selected, e.g. 3 pixels by 3 pixels • This would result in 9 pixels being selected • The average value of the pixels is calculated by: Multiplying by 1/9, and adding the results together to find the average pixel value The resultant pixels form the mask kernel or spatial filter The mask is applied to the whole image • This removes the higher frequencies from the image • The resultant image has lower contrast
Image Smoothing	Median filtering • The median pixel value is selected for the mask This preserves the edges of the image The pixels with the extreme values are reduced • The resultant image has less noise
Hough Transform	A method of highlighting areas of a specific shape in an image • As curved lines are usually used, the technique is useful for highlighting specific anatomical areas • The transform identifies a number of pixels that form a curved line • The technique looks for pixels of similar value and: Calculates the distance from the origin (the coordinates of the centre of a circle) Calculates the angle from the origin using coordinates (or the radius in a circle) Puts the calculated values into an equation • A mesh is created from the equation and overlaid on the image • At each point on the image a measurement is made and compared with the mesh • Mesh points with higher values than the image are overlaid on the image accentuating the area of interest
Hough Transform	Note This process does not work if the image has a lot of noise as clear edges of the object have to be identified
Unsharpness	Unsharpness on a radiograph – total image unsharpness is caused by the following three factors: • Movement of the object • The geometry of the radiation beam • The imaging system
Movement of the Object	Calculating movement • The object speed (v ) in any direction other than perpendicular to the film plane, in mm • Multiplied by the exposure time (t ) in seconds • Multiplied by the enlargement ratio, i.e. the focus film distance (ffd) divided by the focus object distance (fod). (about 10–20%, providing macro techniques are not used)
Movement of the Object
Movement Unsharpness	Movement unsharpness can be either voluntary or involuntary
	Voluntary movement • Controlled by asking the patient to hold their breath • By the use of immobilisation devices
	Involuntary movement • The movement of organs • The fluid in vessels which cannot be controlled by the patient
	Example: • If organ movement is about 5 mm and an exposure time of 0.2 s The unsharpness = 1 mm • If the exposure time was reduced to 0.02 second The unsharpness = 0.1 mm Therefore unsharpness would be improved by a factor of 10
Geometric Unsharpness	This is caused because the focal spot in the tube is not a point source • For a point source, the edges of an object will be recorded as sharp edges on the film • For an extended source, each edge of the object receives a ray as if it were two separate points, causing the penumbra effect, which produces unsharpness • If the focus–object distance is reduced, then the unsharpness is increased Fig. 7.3 Effect of (A) point source, (B) extended source, (C) focus object distance and (D) object film distance on geometric unsharpness (Ug). Note • If the focus–object distance needs to be increased, to reduce geometric unsharpness, this will mean increasing the mAs and the exposure time, which increases movement unsharpness • If the object film distance (ofd) is increased geometric unsharpness is increased, and an increase in exposure will be required, which may increase movement unsharpness
	To calculate geometric unsharpness: • The ratio of the object film distance to the distance between the focal spot and the object is used • Therefore the larger the focal spot, the greater the distance between the object and the film, and the closer the focal spot is to the object, the more geometric unsharpness will increase
	Example If the focus film distance (ffd) = 100 cm and the focus object distance (fod) = 80 cm calculate the geometric unsharpness if a 1.2 mm focal spot was used • ofd = 100 − 80 = 20 cm geometric unsharpness:focal spot size = 20 : 80 geometric unsharpness = of the focal spot size • If focal spot = 1.2 mm, then geometric unsharpness = = 0.3 mm
To Produce a Sharp Image	Theoretically use: • The shortest possible exposure time • A well collimated beam • A grid • The smallest possible focal spot • A large focus film distance • A small object film distance • A cooperative, thin patient (!) In reality, a number of compromises must be made
Measurement of Resolution (Sharpness)	• The line pair phantom is made up of strips of lead separated by an air gap equal to the width of the preceding lead strip • Starting at low frequencies (perhaps 1 line pair per millimetre, or lp mm⁻¹), the phantom then progressively decreases the width of the lead strip and the following air gap until high frequencies, perhaps as high as 10 or 14 lp mm⁻¹, are reached Fig. 7.4 Example of one type of line pair phantom.
Resolution	The size of the smallest object or distance between two objects that must exist before the imaging system will record that object or objects as separate entities. It gives no indication of how the system will record objects of larger dimensions • Or whether they will be visible to an observer • Usually measured in line pairs per millimetre (lp mm⁻¹) • Resolution has a limited value in assessing system performance
Resolution	Note Definition • A subjective impression of the details that can be seen in the radiograph • Is difficult to quantify • Should not be confused with resolution
Modulation Transfer Function (MTF)	Allows assessment of system performance at different spatial frequencies (i.e. ‘object sizes’) • It would be expected that the differences in density of the line pair phantom would be recorded in exactly the same way as the original phantom with the sharp edges clearly delineated • Because of transition density within the image, distinct transition edges are formed which produce a density profile when scanned with a microdensitometer • At low frequencies the transfer of the signal through the system is quite good • As the frequencies get higher the contrast begins to drop rapidly and the pairs of lines become impossible to see • Figures can be obtained for the percentage of accuracy of the transfer of the signal through the system • Therefore the modulation, or altering, of the signal through the system is measured and the modulation transfer function (MTF) for the system is obtained Fig. 7.5 Reading from an image of a perfectly transferred line pair phantom (100% transfer at all frequencies). Fig. 7.6 Example of an actual density profile obtained from a line pair phantom showing decreasing information transfer at higher lp mm⁻¹, due to transition densities. Fig. 7.7 Example of MTF curve, showing loss of information transfer at high spatial frequencies, e.g. at high lp mm⁻¹.
Application	• Figure 7.8 shows four different films, each with different MTF characteristics • Film A appears to be most suited to a general radiographic department which requires about 2 lp mm⁻¹ • Film A transfers most information at low and high frequencies • Film D appears suited to mammography where the demands are greater, and in the order of 4 lp mm⁻¹ are required Fig. 7.9 Suggested limiting resolutions shown on an MTF curve (curve D from Fig. 7.8).
	Note
	• The minimum size of object which is seen is the signal • Everything smaller than this size is information that is not required and is noise

Fig. 7.8 Example of four different MTFs.

Sharpening Digital Image

Edge Enhancement	The use of filters to highlight the boundaries between objects • A filter multiplies all the pixel values by a constant • The values are then added together • The results are then divided by a second figure • The results are then applied to a range of grey scales • Makes the boundaries more conspicuous
	Application • Can aid the identification of small lesions in the body
	Note • Changes the subjective appearance of the image and may result in a loss of information
Frequency Domain Method	• Take the Fourier transform of the image • Multiply the transform by a filter • Invert the filtered transform to get the enhanced image • By increasing the size of the high frequency areas of the image the image becomes sharper