Since the discovery of X-rays by Roentgen in 1895, there has been an explosion in technology for imaging the human body; in addition to ionizing radiation, high-frequency sound (Wade 1999), radiolabelled pharmaceuticals and non-ionizing radiation in strong magnetic fields are used. The parallel expansion in computer technology has enabled the development of cross-sectional imaging using computed tomographic (CT) techniques, and three-dimensional imaging using magnetic resonance imaging (MRI) data.

Methodology

X-ray techniques

Standard radiographs are obtained by placing the patient between an X-ray source and a photographic film. Images are produced by the attenuation of X-ray photons by the patient. As a result, structures of interest are often obscured by the attenuation of photons by other overlying structures. A basic limitation of the standard X-ray technique is the effect of scattered photons adding to the image noise and degrading the quality of the final image. Various manoeuvres have been adopted to reduce this effect, such as scatter grids. The linear tomograph is obtained by the synchronous movement of the X-ray tube and the film, causing a blurring of the overlying and underlying structures while leaving the plane of interest in focus. This procedure increases the detection of lesions such as chest metastases.

Contrast studies

To increase contrast in the soft tissues, an iodine-containing agent may be injected into the lumen of a cavity such as the uterus and fallopian tubes (hysterosalpingogram) or blood vessels (angiogram). Such contrast agents are widely available in ionic and (more expensive) non-ionic forms. The less toxic non-ionic preparations are preferred.

Ultrasound

In Glasgow in 1950, Ian Donald used an industrial flaw detector, designed for testing the integrity of steel boilers, to demonstrate that the massive abdominal swelling of one of his patients was fluid and not solid (Kurjak 2000). A large ovarian cyst that proved to be benign was subsequently removed. Since then, the rapid and continuing evolution of ultrasound technology has led to a proliferation of applications in almost all aspects of medicine and surgery.

Ultrasound waves are generated by a piezoelectric crystal mounted in a transducer housing, and the whole construction is referred to as a ‘probe’. Piezo materials respond to an applied voltage by changing thickness, and produce electrical signals of a few millivolts when compressed. In the final image, the resolution along the beam is determined by the wavelength, while the resolution across it depends on the width of the beam, which is improved by focusing using combinations of lenses and electronic mechanisms.

The time from transmission of the ultrasound pulse to receipt of an echo is translated into depth using the speed of sound in tissue. Echoes arise wherever the pulse of ultrasound encounters an interface between tissues of different acoustic impedance, which may simplistically be regarded as changes in density. Thus, homogeneous materials such as clear fluids are echo-free, while low-level echoes are given by the interfaces between the various components of soft tissue structures. Stronger echoes result from the interface between fibrous tissues (high impedance) and fatty tissues (low impedance), and maximal echoes are given by interfaces between soft tissues and bone or gas, both of which are effectively opaque to ultrasound. To form a real-time image, the beam is electronically swept through a region of tissue and the images are pie-shaped (sector and curved arrays) or rectangular (linear probes).

The choice of frequency is a compromise between the spatial resolution and the depth of penetration required. The higher the frequency, the better the resolution, but the greater the attenuation of the beam within the tissues. Hence, for abdominal scanning, frequencies of between 3.5 and 5 MHz are commonly employed. Probes designed for intracavitary and intraoperative use employ higher frequencies (8–15 MHz) because the organs of interest are closer to the transducer. This enables higher resolution images to be obtained, and also allows study of the movement of structures when pressure is applied to the probe, which gives important diagnostic clues on their relative mobility. The drawbacks of transvaginal scanning are practical (intolerance of the probe by the very young and old, and in the presence of scarring) and diagnostic (the depth of view is limited to approximately 70 mm) (Bennett and Richards 2000). By sweeping an array through the tissues, a three-dimensional scan can be created; while most useful in obstetrics to visualize complex structures such as the fetal face, it also displays uterine anatomy well (Alcazar and Galvan 2009).

The interactivity of ultrasound is an important feature; it is effectively an extension of clinical examination. Ultrasound also provides numerous views during the real-time examination and facilitates guided needle biopsies. However, it has many limitations, notably the variable image quality, that frustrate its application. An important problem is the complete barrier posed by bowel gas, which is why a full bladder is required for transabdominal scanning.

Doppler operates by detecting the change in frequency of ultrasound echoes caused by movement of the target. In the simplest system, this is done continuously with only minimal control of the beam direction. Continuous wave Doppler is well suited to fetal heart monitoring, especially because movements of the fetus do not interrupt the signal. In gynaecology, the vessels of interest are small so precise placement of the Doppler sample is a prerequisite. This can be achieved with pulsed Doppler. The Doppler gate is superimposed on a real-time image that allows the target to be pinpointed (Figure 5.1). The Doppler signal depicts a range of blood velocities occurring in the selected vessel over time. Output can be audio or as a strip chart on which intensity indicates the strength of the signal at each velocity in the spectrum. A variety of measurements can be made from these tracings, the most useful of which are systolic/diastolic velocity ratios. These are measures of the arteriolar resistance to flow, with a low value indicating low vasomotor tone, typical of the placenta and the corpus luteum. This pattern also occurs in inflammation and malignancy. High indices are typical of inactive tissue, such as the resting uterus. Doppler can be used to measure fetal and uteroplacental blood flow, thereby producing physiological information (Brinkman and Wladimiroff 2000).

Figure 5.1 Spectral Doppler. In this scan of an ovarian cancer, the Doppler gate (arrow) has been positioned on the tumour. The spectral tracing is in the lower portion of the figure and shows the typical low-resistance pattern of malignant neovascularization with marked diastolic flow.

In colour Doppler scanning, the same process is applied across an area of interest. The velocity signals are presented as a colour-coded overlay, superimposed on the real-time scan (Figure 5.2). Although the Doppler information is less rich than with spectral Doppler (only the mean velocity or the amount of flow is depicted), the angiogram-like map provides information on the morphological arrangement of the vascular tree, and its sensitivity allows vessels as small as approximately 1 mm in diameter to be detected. Colour Doppler is often used to locate a vessel to guide placement of the spectral Doppler gate for haemodynamic analysis.

Figure 5.2 Colour Doppler of an ovarian carcinoma. On colour Doppler, flow signals are colour coded and presented as an overlay on the grey-scale image in the left-hand panel of this transvaginal scan, while the right-hand panel shows the grey-scale image on its own. The neovascularization in this ovarian carcinoma is shown as knots of colour. Apart from providing a graphic display of vascularity, a useful feature of colour Doppler is to guide placement of the gate for spectral Doppler tracings.

A recent advance has been the introduction of contrast agents for ultrasound in the form of microbubbles for injection. Not only does the signal improvement they produce allow smaller vessels to be detected, but they can be visualized in grey scale to reveal regions of flow in real time. In addition, they can be used as tracers by tracking a bolus as it crosses a tissue of interest, and this may yield valuable functional information.

Computed tomography

The sensitivity of conventional radiography may be improved by using a radiation detector such as a scintillation crystal and a photomultiplier tube. By measuring the attenuation of a finely collimated beam of radiation, passing through the patient at multiple angles, it has been possible to produce images of very high quality. A computer uses the attenuation of each beam passing through the patient to calculate the attenuation coefficient for each area of tissue in the cross-section of interest. The final images are reconstructed using a filtered back-projection technique and displayed in grey scale as a series of attenuation units (with values between +500 and −500) in a matrix of 512 × 512 or 1024 × 1024 elements. The reconstruction of data is limited to the transverse (transaxial) plane. This imaging technique revolutionized modern medicine in the 1970s, but its impact on gynaecology has been less marked.

Magnetic resonance imaging

MRI depends upon the magnetic properties of certain atomic nuclei which, when placed within a magnetic field and stimulated by radio waves of a specific frequency, will absorb and then re-emit some of this energy as a radio signal. This phenomenon of nuclear magnetic resonance (NMR) was first described by Felix Bloch and Edward Purcell in 1946. NMR as a basis for an imaging technique was proposed some 30 years later by Lauterbur.

Nuclei possessing magnetic properties have an odd number of protons or neutrons. The charged particles are spinning, which causes them to behave as tiny bar magnets. If placed within a magnetic field, a majority of protons will line up in the direction of the magnetic field. In addition, their axes are tilted and caused to rotate like small gyroscopes. The frequency of this precessional movement is directly proportional to the strength of the applied field and is called the ‘Larmor frequency’. The hydrogen proton gives a relatively high signal due to its abundance in biological tissues. Other potential NMR isotopes include ¹³C, ²³Na, ¹⁹F and ³¹P.

If a pulse of radio waves is imposed on the nuclei, a strong interaction or ‘resonance’ will occur when the frequency coincides with the precessional frequency of the nuclei. The energy absorbed by the nuclei is then re-emitted as a signal that can be detected in a receiver coil situated around the sample. The initial strength of this signal is proportional to the proton density of the sample. It will then decay in an exponential fashion as the disturbed protons relax back to their original state. The T₁ relaxation time is described as the time taken for the stimulated protons to return to their initial state. The T₂ relaxation time is that taken for the precessing nuclei to get out of step with one another. A variety of sequences of radio wave pulses have been devised so that the resulting signal is weighted to different degrees by the proton density and the T₁ and T₂ relaxation times.

The contrast between different tissues in MRI can be manipulated by altering the pattern of radio waves applied. This is done by changing the time constants associated with the different sequences of radio waves: the repetition time, echo time or inversion time. A further advantage of MRI is the ability to obtain images in multiple planes, such as sagittal, coronal or oblique planes, without moving the patient.

The female pelvis is suitable for MRI examination because of the minimal effect of respiratory motion on the pelvic organs. By placing the receiver coil adjacent to the tissue of interest (e.g. endovaginally to maximize signal from the cervix), very-high-resolution images of an area of interest may be obtained. However, as the data to produce a set of images are accumulated over 5–10 min, patients are required to remain still during this time. If multiple sequences and multiple planes are employed, the total imaging time may exceed 30 min which reduces patient tolerance. However, newer faster pulse sequences with breath-hold techniques and improved computer software are greatly reducing scan times. Recently, open design MR scanners and production of MR-compatible equipment has produced an explosion in the field of ‘interventional MRI’. This has expanded the applications of MRI from being purely diagnostic to its use during treatment, such as for placing brachytherapy implants (Popowski et al 2000) or monitoring thermal ablative therapy.

Radionuclide imaging

Unlike the other imaging modalities, radionuclide imaging provides physiological rather than anatomical detail. Modern gamma cameras are capable of accurately imaging the distribution of administered radiopharmaceuticals, and the use of tomographic systems for single photon emission tomography has improved image resolution. In gynaecological oncology, radiolabelled monoclonal antibodies may be employed in the localization of malignancies (radioimmunoscintigraphy) and in their treatment (radioimmunotherapy).

The purity and specificity of monoclonal antibodies gives them an important role in tumour detection and possibly in the targeting of antitumour agents. Different anticytokeratin antibodies may help in distinguishing a primary ovarian adenocarcinoma from a metastatic adenocarcinoma, especially of colorectal origin (McCluggage 2000). These antibodies have also helped to clarify the origin of the peritoneal disease in most cases of pseudomyxoma peritonei. In recent years, several studies have also investigated the value of a variety of monoclonal antibodies in the diagnosis of ovarian sex cord stromal tumours, and in the distinction between these neoplasms and their histological mimics. Of these, anti-alpha-inhibin and CD99 appear to be of most diagnostic value (Choi et al 2000). Antibodies should always be used as part of a larger panel and not in isolation.

The antibodies or antibody fragments are radiolabelled with a gamma-emitting radionuclide such as ^99mTc, ¹³¹I, ¹²³I or ¹¹¹In. A subcutaneous test dose is no longer given because of the immune response it may generate in the patient. If ¹³¹I or ¹²³I is employed, oral potassium iodide is given to block thyroid uptake of free radioactive iodide. In addition, blood pool subtraction techniques are required for ¹³¹I studies to simulate the non-tumour distribution of labelled antibody. This is not necessary for ¹²³I- or ¹¹¹In-labelled preparations.

Following injection of the antibody, a gamma camera provides serial images of the distribution and uptake of radiolabelled antibody between 18 and 72 h after the antibody has been administered. The radionuclide ¹¹¹In is ideally suited for tomographic studies using images produced by a gamma camera mounted on a gantry which rotates through 360°. The images are reconstructed by computer in the same manner as X-ray CT. ¹¹¹In-labelled antibody is taken up into the liver, spleen, bone marrow and, occasionally, adrenal glands but is a more suitable radiolabel than ¹³¹I. Excretion of ¹¹¹In into the bladder has not been a problem, although non-specific bowel uptake may account for false-positive results. An example of an ¹¹¹In study is shown in Figure 5.3. A central area of increased uptake is apparent at the site of a primary ovarian cancer.

Figure 5.3 Radionuclide scan: primary ovarian carcinoma. Anterior view of the pelvis with 80 MBq ¹¹¹In-labelled OC-125.

¹⁸FDG (¹⁸fluorodeoxyglucose) is a non-specific positron-emitting tracer. It is an analogue of glucose, thereby reflecting metabolism and detecting the increased glycolysis associated with several tumour types. It gains entry into cells by membrane transporter proteins, such as Glut-1, which are expressed in many tumours. It is phosphorylated intracellularly to FDG-6-phosphate and retained within malignant cells. Tumour hypoxia may also increase ¹⁸FDG accumulation via activation of glycolysis. The quantitative parameter used is typically a standardized uptake value (SUV), which represents the tissue activity within a region of interest corrected for the injected activity and the patient’s body weight. A new lesion (Figure 5.4) or a rise in SUV within a lesion indicates disease progression. Whether this can be used as a marker for early therapy remains to be proven.

Figure 5.4 ¹⁸F positron emission tomography of a patient with cervical cancer showing metastatic deposits with increased glucose uptake in left axilla (short arrow) and right scapula (long arrow).

de Souza NM, Brosens JJ, Schwieso JE, Paraschos T, Winston RM. The potential value of magnetic resonance imaging in infertility. Clin Radiol 1995; 50(2):75–79, with permission of Elsevier.

Angiography

Diagnostic and therapeutic angiography have made great strides in recent years with the development of catheters whose tips can be manipulated. This allows considerable control over the guidance of these catheters into selected vessels. In parallel with this, ingenious devices have been devised which can be introduced via these catheters for purposes as diverse as angioplasty and embolization.

In gynaecological practice, the main use of angiography is the control of pelvic bleeding, but both the diagnosis of pelvic venous thrombosis and the prevention of subsequent pulmonary embolism are important roles for this exciting new technique. Pelvic arteriovenous malformations are most readily diagnosed and treated with angiography, and pelvic varices causing chronic pain may be managed in the same way. Selective sampling of gonadal venous blood and steroid hormone assay can be valuable in the preoperative assessment of phenotypic women with chromosomal abnormalities and intra-abdominal gonads of uncertain nature.

Clinical Applications

Normal anatomy and variations

Uterus

The normal uterine cavity is delineated as a triangular structure on hysterosalpingography, with the cornua and the internal cervical os as the corners. Intracavitary abnormalities, both congenital and acquired, such as internal septations and divisions may be defined using this technique (Figure 5.5). Similar information can be acquired with ultrasound following instillation of saline into the uterine cavity (hysterosonography) (van den Brule et al 1998).

Figure 5.5 Hysterosalpingogram: contrast outlining the cavities of a bicornuate uterus. Free spill into the peritoneal cavity can be seen on the right (arrow).

On ultrasound, the uterine cavity is normally represented by a bright line of apposition of the endometrial layers, although occasionally a trace of fluid can be seen at menstruation. Persistent fluid or fluid in a non-menstruating woman is abnormal. In adolescence, it may be caused by vaginal atresia, where the upper vagina dilates more than the uterus so that haematocolpos dominates the picture, while in postmenopausal women, cervical stenosis, either fibrotic or malignant, must be considered. Ultrasound is also useful in detecting the position of an intrauterine device; provided the coil lies in the uterus, it is easy to locate, but a coil that has penetrated through the myometrium is obscured by echoes from bowel gas (Figure 5.6).

Figure 5.6 Two views of an intrauterine device in the uterine cavity (arrows) with a coexisting pregnancy (arrowheads).

The entire uterus can be imaged using ultrasound (Figure 5.7) and its size measured accurately. This is useful in precocious or delayed puberty and in planning brachytherapy. The endometrium is seen as an echogenic layer of uniform texture, often with a fine echo-poor margin. Its thickness varies through the menstrual cycle from 12 mm premenstrually (double-layer thickness) to a thin line after menstruation. The endometrium gradually thins after menopause unless supported by hormones.

Figure 5.7 Transvaginal scan of the uterus. In this scan, taken in the luteal phase of the cycle, the secretory endometrium is seen as an echogenic band in comparison with the relatively echo-poor myometrium.

The morphology of the normal uterus, however, is best defined on MRI where its zonal architecture may be recognized (Hricak et al 1983). On T₂-weighted images, the endometrium is seen as a central high-signal stripe which increases in thickness in the secretory phase of the menstrual cycle. The inner myometrium (junctional zone) is of lower signal intensity than the outer myometrium on T₂ weighting, and correlates histologically with a layer of more densely packed smooth muscle.

Fallopian tubes

The patency of the fallopian tubes may be demonstrated on hysterosalpingography where filling and free spill into the peritoneal cavity may be seen (Figure 5.5). Ultrasound contrast salpingography is currently being evaluated to reduce the radiation dose to the ovaries in patients trying to conceive. Normal uterine tubes cannot be demonstrated on ultrasound, but instillation of microbubble contrast agents into the uterus and tubes permits their visualization and can be used as an initial screening test for tubal patency. Although the anatomical detail is less than that offered by conventional X-ray salpingography and false-positive results occur, this technique of ultrasound salpingography seems likely to find an important role as an initial screen; if the tubes are demonstrated to be patent, no further investigation is necessary and this should result in a reduction of radiation exposure (Van Voorhis 2008).

The normal fallopian tubes are too small in calibre and of too variable a course to be reliably imaged using a cross-sectional technique such as CT or MRI.

Ovaries

The ovaries are best imaged with ultrasound (either transabdominal or, preferably, transvaginal). Changes in their appearance can be correlated with their functional status. Infantile ovaries are small (except in the neonate when hypertrophy and follicles stimulated by maternal hormones may be a surprising finding) and they enlarge before puberty. Follicular development begins before menstruation, but these cycles and those at the menopause are often imperfect so that follicles may persist and continue enlarging for several months. Normally, ovulation occurs at a follicle size of 20–25 mm diameter and the echo-free follicle is replaced by a corpus luteum which can be cystic or solid. Corpora lutea produce a confusing variety of appearances, whose only consistent feature is their transience (Figure 5.8A–C). In doubtful cases, a rescan at 6 weeks to image them at a different phase of the cycle may be needed to resolve their identity. The postmenopausal ovary is usually too small to identify with ultrasound.

Figure 5.8 Transvaginal scans of the ovary. (A) An inactive ovary is indicated by the caliper marks. It is surrounded by a small amount of free fluid and contains minute developing follicles. (B) Two views of an ovary containing a 3-cm simple cystic structure which presumably represents an unruptured follicle. (C) An ovary in the luteal phase containing a solid corpus luteum.

On CT and MRI, the normal ovaries are often difficult to define, but may be recognized on short inversion recovery MRI sequences as very-high-signal foci in the adnexae.

Ovarian varices

The pelvic congestion syndrome is one of many causes of chronic pelvic pain, and is associated with the presence of large varices within the broad ligaments (Liddle and Davies 2007). When pelvic symptoms are severe, there may be gross dilatation of the ovarian veins, with reflux down into the pelvis and often into the legs. These vessels are best demonstrated by selective ovarian venography, although they may also be imaged with Doppler ultrasound (Haag and Manhes 1999). Venography is performed via a femoral or internal jugular venous approach, and the ovarian veins are selectively catheterized with an appropriately shaped angiographic catheter. Satisfactory retrograde opacification of pelvic varices is achieved by injecting contrast medium through the selectively placed catheter with the patient almost upright on a tilting table, while the Valsalva manoeuvre is performed.

Treatment of this condition is primarily surgical and consists of venous ligation. Symptomatic relief has been reported following transcatheter ovarian vein embolization (Ganeshan et al 2007).

Investigation of female infertility

The success of in-vitro fertilization (IVF) has been due, in part, to the correct timing of ovulation and subsequent oocyte recovery that ultrasound can provide. Using transabdominal scanning, ovarian follicles of 3–5 mm diameter can be visualized. They appear as echo-free structures amidst the more echogenic ovarian tissue. Their rate of growth is linear and the mean diameter prior to ovulation is 20 mm (range 18–24 mm). Structures within the follicle, such as the cumulus oophorus, can also be visualized. Following ovulation, internal echoes appear because of bleeding. Free fluid may also be observed in the pouch of Douglas.

Transvaginal sonography has largely replaced the transabdominal approach in infertility practice because of the superior anatomical display and because it allows precisely guided aspiration of follicles and fluid in the pouch of Douglas. More precise measurement of follicles is possible, and the corpus luteum is easily recognized. In the midluteal phase, it appears as an oval structure 30–35 mm long and 20–25 mm wide with a wide variety of sonographic appearances (Baerwald et al 2005). Endometrial reflectivity patterns have been proposed as another way to assess ovulation (Randall et al 1989). Ultrasound can only provide presumptive evidence of ovulation/pregnancy. The collection of secondary oocytes constitutes definitive evidence of ovulation. The appearance of internal echoes before the follicle reaches 18 mm or continuous enlargement to 30–40 mm indicates follicular failure.

Ultrasound is also useful for accurate timing of artificial insemination, while a postcoital test can help to differentiate between inadequate sperm penetration and poor mucus production in the presence of immature follicles. Ultrasound scanning is also employed to monitor patients on clomiphene therapy, and there is good correlation between follicular diameter and plasma oestradiol concentration. Ovarian hyperstimulation syndrome is uncommon if gonadotrophin therapy is monitored by ultrasound in conjunction with measurements of plasma oestradiol levels.

The waveforms of blood flow in vessels supplying the ovaries of women undergoing IVF have been studied using transvaginal power Doppler ultrasound (Palomba et al 2008). The observation of blood flow patterns did not improve any reproductive outcomes and is currently not advocated in IVF programmes.

MRI is valuable in the investigation of infertility where uterine pathology is suspected, and provides a particularly high diagnostic yield in patients with dysmenorrhoea and menorrhagia (deSouza et al 1995). It should be part of the investigation of patients with persistent unexplained infertility awaiting costly procedures such as gamete intrafallopian transfer and IVF. MRI should also be used before myomectomy in order to differentiate between leiomyoma and adenomyoma, as attempts to perform a myomectomy on a localized adenomyoma often result in extensive uterine damage.

Adenomyosis and endometriosis

The diagnosis of adenomyosis is often suggested by symptoms of hypermenorrhoea and dysmenorrhoea, but similar symptoms are also produced by leiomyomas. Hysterosalpingography may show multiple small tracks of contrast extending into the myometrium but is now obsolete. Ultrasound in skilled hands may be useful (Dueholm and Lundorf 2007).

Adenomyosis is best demonstrated on T₂-weighted MRI, with which two patterns (focal and diffuse) can be recognized (Hricak et al 1992, Byun et al 1999). In focal adenomyosis, there is a localized ill-defined mixed-signal-intensity mass (adenomyoma) within the myometrium (Figure 5.9). Diffuse adenomyosis presents with diffuse or irregular thickening of the junctional zone, often with underlying high-signal foci (Figure 5.10). Pathologically, this represents smooth muscle hypertrophy and hyperplasia surrounding a focus of basal endometrium. It is the smooth muscle changes that are easily recognized by MRI rather than the foci of heterotopic glandular epithelium. Values for junctional zone thickness from >5 to >12 mm have been suggested for the diagnosis of adenomyosis. In the authors’ experience, however, the thickening of the junctional zone that is described in adenomyosis (Reinhold et al 1999) is not diagnostic of adenomyosis on its own; it may be seen in hypermenorrhoea where no pathological evidence for adenomyosis has subsequently been shown. In these instances, absolute values for junctional zone widths are unhelpful (Hauth et al 2007), and ratios to the width of the outer myometrium or the cervical stroma may be more meaningful.

Figure 5.9 Focal adenomyoma. T₂-weighted spin-echo (SE 2500/80) midline sagittal section through the uterus, demonstrating a large ill-defined mixed-signal intensity mass characteristic of an adenomyoma (arrows).

Figure 5.10 Diffuse adenomyosis. T₂-weighted spin-echo (SE 2500/80) midline sagittal section through the uterus, showing thickening and irregularity of the low-signal band of the junctional zone (short arrows). Diffuse infiltration of a mixed-signal lesion (long arrow) is seen in the posterior wall.

The lesions of endometriosis are difficult to detect with imaging techniques, mainly because of their small size. Only larger endometriomas can be detected on ultrasound and are seen as well-defined spaces, most commonly in the adnexal region (Figure 5.11). The role of MRI in the diagnosis of endometriosis also depends on the site of the endometriotic implants; deposits on the ovarian cortex are much more readily identified than small peritoneal deposits. The latter are recognized at laparoscopy but may be missed on MRI (Brosens et al 2004). Endoscopic ultrasound has been advocated to improve the diagnosis of endometriotic infiltration into the digestive tract (Dumontier et al 2000). MRI is particularly valuable in detecting deep enclosed endometriosis typically found in the uterosacral ligaments and rectovaginal septum (Figure 5.12) commonly associated with pelvic pain. Deep pelvic endometriosis consists mainly of fibromuscular rather than endometrial tissue (Del Frate et al 2006), and often goes undetected on visual inspection of the pelvic cavity because of its position. However, MRI lacks sensitivity in detecting rectal endometriosis if the rectum is not distended (Kinkel et al 1999). More recently, there have been isolated reports correlating an increased junctional zone thickness with endometriosis in patients with infertility (Kunz et al 2000).

Figure 5.11 Endometriotic deposits. T₂-weighted spin-echo (SE 2500/80) midline sagittal image; fluid levels are seen in cystic lesions with mixed-signal intensity suggestive of haemorrhage (arrows).

Figure 5.12 Rectovaginal endometriosis. T₂-weighted spin-echo (SE 2500/80) sagittal section demonstrating a rectovaginal mass (arrows) in a patient with laparoscopically proven endometriosis. This was not detected at laparoscopy because it is retroperitoneal.

Leiomyomas

The most common cause of disturbance of the normal uniform echo texture of the myometrium is fibroids. Generally, a focal thickening occurs and their position in the uterus can be determined. Submucosal fibroids, or the polyps they may produce, can be mistaken for endometrial thickening, although occasionally the way they move with normal uterine contractions is diagnostic. Instillation of saline into the uterine cavity, a form of ultrasound contrast, provides much more information about intracavitary lesions. Pedunculated or broad ligament fibroids frequently masquerade as extrauterine masses. The echo texture of fibroids is very variable, in accordance with their histological spectrum. A characteristic is calcification which is easily recognized on ultrasound as intensely reflective foci with accompanying acoustic shadowing.

MRI has been reported to be superior to ultrasound for the diagnosis of leiomyomas (Ascher et al 1994). Characteristic well-defined low-intensity masses are seen within the myometrium on T₂-weighted scans, often with high-signal foci within them. There are no completely reliable features that distinguish fibroids from leiomyosarcomas on ultrasound or MRI, and even Doppler has proved disappointing here. In those patients presenting with more advanced disease, alterations in the lesion margins may suggest malignant change.

Ectopic pregnancy

Traditionally, ultrasound has been used to exclude an ectopic pregnancy by demonstrating an intrauterine pregnancy, but this sign is unreliable in assisted pregnancy because the likelihood of coexistent intra- and extrauterine pregnancies is not negligible. The positive diagnosis of ectopic pregnancy requires the demonstration of an extrauterine fetal heart. Secondary features that may be recognized are an adnexal mass with peritrophoblastic Doppler signals, free fluid in the pelvis and increased endometrial thickness. With a combination of these features, the negative predictive value of an ultrasound examination may be 96%. However, significant problems can arise in distinguishing an ectopic pregnancy with a pseudogestational sac from an early normal or failed intrauterine pregnancy.

The role of ultrasound in the diagnosis of ectopic gestations has been strengthened by the use of transvaginal probes and the introduction of Doppler. A weakness of transabdominal scanning was the difficulty in demonstrating the sac itself, because of gas in overlying or adherent bowel. With transvaginal scanning, the adnexal sac can be demonstrated and women at high risk can be monitored through the first few weeks of pregnancy (Figure 5.13). Early ectopic sacs can be treated by injection of potassium chloride or methotrexate under ultrasound guidance. This approach is especially useful in IVF programmes where sparing the uterine tubes is important.

Figure 5.13 Ectopic gestation. In this transvaginal scan, colour signals are seen around the ectopic sac using power Doppler.

Infections

Ultrasound is useful in the diagnosis and management of ovarian abscesses, where shaggy walled cavities can be demonstrated in the adnexal region (Figure 5.14). They usually have a vascular pseudocapsule that can be demonstrated on colour Doppler. However, in pelvic inflammatory disease without abscess formation, the ultrasonic changes are too subtle to make this a useful technique. Monitoring the effects of antibiotic treatment on tubo-ovarian abscesses may be useful, although, in some cases, sterile cavities persist for many weeks after signs of infection have been controlled. The hydrosalpinx that complicates pelvic inflammatory disease is seen as a funnel-shaped cavity in the adnexal region with thin walls and no internal echoes.

Figure 5.14 Tubo-ovarian abscess. The irregular walls of this cavity are vascularized as part of the inflammatory response.

Venous thrombosis

Diagnosis

A relatively common complication of pregnancy and gynaecological malignancies, deep venous thrombosis is most easily diagnosed using colour Doppler ultrasound and this should be the investigation of first choice. Visualization of the iliac veins may be difficult and some patients will have to proceed to contrast venography. If colour Doppler ultrasound has shown the femoral veins to be clear of thrombus, this is most easily performed by puncturing these vessels directly, which allows excellent opacification of the iliac veins and thus a confident diagnosis or exclusion of thrombosis, which can be difficult when contrast is injected into foot veins.

Inferior vena cava filter insertion

The indications for insertion of an inferior vena cava filter are listed in Box 5.1. The procedure is easily performed via femoral or jugular venous punctures, and some devices may even be inserted through a sheath introduced into an antecubital fossa vein. The majority of these filters are inserted permanently; short-term placement is possible, however, and is most commonly indicated in the late stages of pregnancy (when free-floating thrombus is demonstrated or when a recent pulmonary embolus has occurred), or prior to surgical removal of a large pelvic tumour which has caused compression and thrombosis of one or both iliac veins. Two types of short-term filter may be used: temporary and retrievable.

• Temporary filters are mounted on a long catheter and are placed via an internal jugular or antecubital fossa venous approach into the infrarenal inferior vena cava above the thrombus (Figure 5.15). The device is then sutured in place. They can only be left in situ for approximately 10 days and should then be removed because of the risk of infection.

• Retrievable filters are designed for permanent placement and are inserted via an internal jugular venous approach. One such device is the Günther Tulip filter (William Cook Europe, Bjaeverskov, Denmark) which has a hook on its superior aspect. This may be grasped with a loop snare and withdrawn into a sheath prior to its removal. These devices can be removed after 10 days, but this is associated with more potential complications as some endothelialization of the filter is likely to have occurred at this time and the filter limbs may be partially fixed to the caval wall.

Box 5.1 Indications for inferior vena cava filter placement

• Recurrent pulmonary emboli despite anticoagulation

• Pulmonary embolic disease in a patient in whom anticoagulation is contraindicated

• Free-floating pelvic or inferior vena caval thrombus

• Prophylaxis during surgery (see text)

• Previous life-threatening pulmonary embolism

Figure 5.15 Temporary vena caval filter insertion above free-floating thrombus. (A) Control film. (B) Inferior vena cavogram using digital subtraction demonstrates a large free-floating thrombus. (C, D) A temporary filter has been placed on top of the thrombus using a right internal jugular approach. The plastic catheter on which the filter is mounted is barely visible.

Haemorrhage

Haemorrhage from the genital tract may complicate parturition or may be associated with benign (e.g. uterine arteriovenous malformation) or malignant (e.g. cervical carcinoma) disease. Traditional surgical therapy of severe pelvic haemorrhage consists most commonly of unilateral or bilateral internal iliac artery ligation, although in the case of primary postpartum haemorrhage, hysterectomy is more often performed.

There is a strong argument in favour of angiography and embolization prior to arterial ligation or hysterectomy in hospitals having the necessary expertise. Bleeding sites are often impossible to localize at surgery (hence the necessity for non-specific ligation of the internal iliac artery), but are usually easily identified at angiography; selective occlusion of the haemorrhaging vessel can then be performed with a much higher likelihood of controlling bleeding (Figure 5.16). Previous internal iliac artery ligation does not necessarily prevent successful arterial embolization but may make the procedure technically more difficult (Collins and Jackson 1995).

Figure 5.16 Severe postpartum haemorrhage in a patient who had been explored twice vaginally and who continued to bleed in spite of vaginal packing. (A) Control film showing a large pack in pelvis. (B) Selective internal iliac arteriogram demonstrates brisk extravasation of contrast from a branch of the anterior division. This was selectively catheterized and occluded with polyvinyl alcohol particles. (C) Postembolization arteriogram demonstrates occlusion of the bleeding vessel with no further extravasation of contrast.

Embolization is very successful in a variety of gynaecological disorders complicated by bleeding. For certain conditions (e.g. uterine arteriovenous malformations, bleeding from recurrent cervical carcinoma), this should be considered as the procedure of first choice (Figure 5.17). In others, it may be life-saving when more conventional treatment has failed.

Figure 5.17 Embolization of an area of neovascularity in the right side of the pelvis due to recurrent carcinoma of the cervix in a patient with persistent bleeding. (A) Selective right internal iliac arteriogram using digital subtraction shows encasement of the origin of the right uterine artery which supplies an extensive area of neovascularity. (B) The right uterine artery has been selectively catheterized prior to embolization. Note the marked irregularity of this vessel due to involvement by tumour. Embolization performed with polyvinyl alcohol particles gave excellent control of bleeding.

Cervical carcinoma

The cervix

In cervical carcinoma, precise staging of the primary disease provides a prognosis, allows the institution of correct treatment and permits comparison of different treatment protocols. Clinical staging, applied according to the system of the International Federation of Gynecology and Obstetrics, is subjective and is still widely used. Imaging with ultrasound does not improve the accuracy of clinical staging (Fotopoulou et al 2008), with poor image quality and difficulty in interpretation as the major problems. Although transrectal ultrasound produces clearer views of the cervix, definition of tumour from normal cervix is still poor. The same difficulty limits the value of CT because the normal cervix and cervical carcinomas have similar attenuation values, so the tumour can only be recognized if it alters the contour or size of the cervix (Subak et al 1995). MRI is now widely recognized as the staging modality of choice.

Several studies confirm the accuracy of MRI in the staging of early cervical cancer in comparison to surgical staging (Whitten and deSouza 2006). The primary tumour is best assessed using T₂-weighted MRI, which is superior to CT in detection (sensitivity 75% vs 51%, P < 0.005) and staging (accuracy 75–77% vs 32–69%, P < 0.025) (Kim et al 1993, Ozsarlak et al 2003).

The resolution of the primary tumour on MRI may be further improved by using intracavitary receiver coils. Endorectal coils give high-resolution images of the posterior cervix, but the anterior margin of the tumour and its relation to the bladder base is often difficult to define because of drop-off of signal. Endovaginal coils give very-high-resolution images of the cervix and adjacent parametrium (deSouza et al 1996). The distortion of the low-signal ring of inner stroma may be apparent, and any breaks in the ring representing tumour extension may be identified (Figure 5.18A,B).

Figure 5.18 Carcinoma of the cervix. (A) T₁-weighted (SE 780/20) and (B) T₂-weighted (SE 2500/80) transverse spin echo through the cervix showing a large intermediate-signal-intensity mass mainly on the left. There is distortion and displacement of the normal low-signal band of inner stroma (arrows). A break in this stromal ring indicating parametrial extension is seen on the T₂-weighted images (arrowhead in B).

With MRI, the invasion of cervical cancer may be assessed in three planes: the coronal and axial planes are used for determining parametrial invasion, the axial plane is used for determining extension into the bladder and rectum, and the sagittal plane is used for extension into the uterine body, bladder and rectum (Sala et al 2007). Fast spin-echo T₂-weighted sequences provide the best image contrast (Figure 5.19). MRI is also accurate in the prediction of myometrial tumour involvement and in showing the relationship of the tumour to the internal os, and hence the patient’s suitability for trachelectomy (Peppercorn et al 1999). Volume (three-dimensional) imaging also provides the information necessary to calculate tumour volumes, which is of prognostic significance. Volumes obtained with MRI correlate well with those obtained by histomorphometric methods, but only weakly with clinical stage (Burghardt et al 1989). The volumetry of the tumour also gives a more accurate prediction of parametrial invasion and lymph node involvement (Burghardt et al 1992). Measuring tumour volume on MRI is of paramount importance; it has been shown that it is the size of tumour burden that determines the outcome rather than invasion beyond the anatomical margins of the uterus (Soutter et al 2004). Dynamic contrast-enhanced studies have been used for assessment of tumour angiogenesis (Hawighorst et al 1999), and the rate of contrast uptake has been shown to correlate with microvessel density in the tumour. However, no correlation with tumour aggressiveness has been demonstrated (Postema et al 1999) and these images are not used routinely. Recently, diffusion-weighted MRI has shown higher accuracy than T₂-weighted imaging alone for tumour detection (Charles-Edwards et al 2008).

Figure 5.19 Carcinoma of the cervix: a sagittal magnetic resonance image using a newly developed double inversion recovery sequence through the pelvis. A large carcinoma is seen (arrows) abutting but not invading the bladder base anteriorly and the rectal wall posteriorly. The uterus lies immediately cranial (large arrow).

Parametrium

CT and MRI have both been used to assess parametrial spread. False-positive diagnoses may arise from misinterpreting inflammatory parametrial soft tissue strands associated with a tumour as actual tumour invasion on both CT and MRI. A comparison of the assessment by these modalities with histological findings after radical hysterectomy showed an accuracy rate for parametrial involvement of 87–90% for MRI, 55–80% for CT and 82.5% for examination under anaesthesia (Kim et al 1993). Extracervical extension on MRI is best defined using T₂-weighted fast spin-echo sequences transverse to the cervix (deSouza et al 1998). Fat-suppression techniques do not provide additional benefits (Lam et al 2000). Contrast enhancement has not proved beneficial; in a series of 73 patients, fast spin-echo T₂-weighted images had an accuracy of 83% for determining parametrial extension (compared with 65% for T₁-weighted gadolinium-enhanced images and 72% for T₁-weighted gadolinium-enhanced fat-suppressed images). The high negative predictive value (95%) for the exclusion of parametrial tumour invasion was the principal contributor to the staging accuracy obtained with fast spin-echo T₂-weighted imaging (Sironi et al 2002). MRI therefore yields valuable information for treatment planning and should be used routinely in conjunction with clinical staging to determine the appropriate therapy in patients with cervical carcinoma.

An intravenous urogram is now obsolete as part of the staging process as it is has been superseded by MRI.

Nodal involvement

The role of lymphangiography in the staging of cervical cancer is obsolete, with up to 71% false-positive results and 16% false-negative results. Percutaneous fine-needle aspiration biopsy of nodes had been used to improve the results obtained with pelvic lymphangiography, but this has been replaced by CT. The major drawback for epithelial tumours is that nodes must be enlarged to be detectable. Thus, metastases less than 2 cm in diameter will not be identified.

Some studies have shown an improvement in pelvic lymph node evaluation with MRI compared with CT (accuracy 88% vs 83%, P < 0.01) (Kim et al 1993). However, like CT, this relies on changes in the size of the lymph nodes (Figure 5.20) since the tumour deposits themselves are not highlighted.

Figure 5.20 Computed tomography scan of stage IIIb cervical carcinoma.

A lymph-node-specific MR contrast agent has been developed that allows the identification of malignant nodal infiltration independent of the lymph node size. This novel MR contrast agent is classified as a nanoparticle (mean diameter 30 nm), and is composed of an iron oxide core, coated with low-molecular-weight dextran (ultrasmall particles of iron oxide, USPIO). The particles, administered intravenously, are taken up by macrophages in the reticuloendothelial system, predominantly within the lymph nodes. Uptake of USPIO results in marked loss of signal intensity (darkening) of the node on T₂– and T₂*-weighted sequences because of a susceptibility artifact caused by the iron. Metastatic tissue within a node displaces the normal macrophages, thus preventing uptake of contrast agent, and the node continues to remain high in signal intensity. A significant increase in sensitivity with no loss of specificity has been demonstrated for the detection of malignant lymph nodes with USPIOs in patients with cervical and endometrial cancer (Rockall et al 2005). On a node-by-node basis, the sensitivity increased from 29% using standard size criteria to 93% using USPIO criteria. On a patient-by-patient basis, sensitivity increased from 27% using standard size criteria to 100% using USPIO criteria.

The role of positron emission tomography (PET) for detecting metastatic pelvic lymph nodes remains unproven. Small single-centre studies claim 91–96% sensitivity and 96–100% specificity for detecting nodes involved with tumour (Reinhardt et al 2001, Loft et al 2007), while results from other studies are much poorer (Williams et al 2001). Urinary residues of ¹⁸FDG in the ureters remain a source of false-positive results. Quantitation suggests that a SUV_max ≥3.3 for lymph nodes is a significant adverse factor in those patients with nodal enlargement (Yen et al 2008).

Distant spread

Even with advanced pelvic spread, involvement of distant sites is the exception rather than the rule. Patients with stage I and II cervical cancer do not need to have bone scans.

A chest X-ray is a routine pretreatment investigation, but the yield of lung metastases at first presentation is likely to be no more than 2%.

Assessing response and recurrence

In conjunction with clinical examination, MRI can provide an objective assessment of the effects of surgery or radiotherapy. Serial MRI may be used before and after primary radiation therapy to assess tumour response (Akin et al 2007), and is increasingly important with the increased use of cytotoxic regimes. CT and ultrasound have limited ability to differentiate between fibrosis and tumour. While some say that infiltration of the parametrium may be easily recognizable, it is difficult to differentiate fibrosis after treatment from tumour recurrence. The use of contrast agents in detecting recurrent disease has proved disappointing; overall, diagnosis is best with unenhanced T₂-weighted images, but particularly in patients with treatment complications (e.g. fistula formation), contrast enhancement does help (Hricak et al 1993).

PET-CT is increasingly used in identifying pelvic recurrence with sensitivity, specificity and accuracy of 92.0%, 92.6% and 92.3%, respectively (Kitajima et al 2008). PET-CT is also able to detect lung metastasis, local recurrence, peritoneal dissemination, para-aortic lymph node metastasis and pelvic lymph node metastasis, and is a useful complementary tool for obtaining good anatomical and functional localization of sites of recurrence during follow-up of patients with cervical cancer.

Endometrial carcinoma

Primary tumour

Thickening of the endometrium is characteristic of endometrial carcinoma and is well demonstrated on transvaginal ultrasound (Figure 5.21). However, specificity is low, although Doppler may help to distinguish hormonal causes of thickening (weak signals) and malignancy which gives marked colour and spectral Doppler signals (Weber et al 1998). Three-dimensional power Doppler has been reported to improve detection of endometrial carcinoma in patients with postmenopausal bleeding (Alcazar and Galvan 2009).

Figure 5.21 Transvaginal ultrasound image of cystic hyperplasia of the endometrium. The grossly thickened endometrium is clearly seen, together with the cystic spaces, in this patient on tamoxifen.

In patients already proven to have an endometrial cancer, myometrial extension has important prognostic and therapeutic implications. Myometrial invasion may be classified as absent, superficial or deep, and this can be assessed with high-resolution ultrasound probes. In a study of 75 patients, ultrasound had a diagnostic accuracy of 73% (Berretta et al 2008). Errors occurred when the tumour was exophytic and had significant extension into the uterine cavity. False-positive results also occur if the uterine cavity is distended with pus or blood when a subendometrial hypoechoic halo may be seen. Ultrasound assessment of the integrity of the echo-poor layer can be improved by the use of transvaginal or transrectal sonography. However, ultrasound has been superseded by MRI.

Changes in uterine blood flow have been used to detect endometrial cancer using endometrial thickness (including tumour), and the pulsatility index (PI) derived from flow velocity waveforms recorded from both uterine arteries and from within the tumour (Bourne et al 1991) found an overlap in endometrial thickness between women with endometrial cancer and those without, but the PI was invariably lower in women with postmenopausal bleeding caused by endometrial cancer than in those with other reasons for bleeding. Blood flow impedance is inversely related to the stage of the cancer. However, although PI values in healthy women increase slightly with age, they decrease with oestrogen replacement therapy, so although Doppler studies may be helpful in the detection of endometrial cancer, allowance must be made for oestrogen replacement therapy.

CT has been used to stage endometrial cancer which is seen as a hypodense lesion in the uterine parenchyma (Hardesty et al 2001), or as a fluid-filled uterus due to tumour obstruction of the endocervical canal or vagina. These findings, however, are non-specific and are easily confused with leiomyomata, intrauterine fluid collections and extension of a cervical carcinoma into the uterine body. In addition, a central lucency may occur in normal postmenopausal women.

MRI represents the best method of assessing a patient with an endometrial cancer, with advantages over other radiological techniques. Figure 5.22 is a series of MRI scans of patients with endometrial cancer. On the T₂-weighted pulse sequence, the tumour has a signal intensity similar to that of normal endometrium, but showing some degree of variability. The high signal makes the tumour quite distinct from the surrounding myometrium, which possesses an intermediate-signal intensity. In a premenopausal uterus, however, it may be difficult to differentiate tumour from adenomatous hyperplasia or indeed from the high signal of normal endometrium. Contrast-enhanced T₁-weighted MRI improves the ability to assess the depth of myometrial invasion by endometrial tumour (Koyama et al 2007, Sala et al 2009). MRI has been found to have a sensitivity of 57% and a specificity of 96% for tumour confined to the endometrium, a sensitivity and specificity of 74% for superficial invasion, and a sensitivity and specificity between 83–88% and 72–85%, respectively, for deep penetration (Sironi et al 1992, Cabrita et al 2008). However, the degree of invasiveness may be overestimated for exophytic polypoid tumours with significant extraluminal extension (Chung et al 2007, Sanjuan et al 2008). Powell et al (1986) first described the low-signal band of inner myometrium to be thinned or absent in those patients with deeply invasive tumours, and this correlated well with the pathological measurement of myometrial invasion.

Figure 5.22 Magnetic resonance imaging scans (T₂) of two patients with endometrial cancer. The sagittal (A) and transverse (B) views in the first patient show tumour filling the endometrial cavity (arrows), but the surrounding low-signal-intensity junctional zone is intact. In the second patient, the sagittal (C) and transverse (D) views show tumour filling the endometrial cavity (arrows) with breach of the low-intensity junctional zone (arrowheads) and extension of tumour into the myometrium.

The sagittal plane is the most appropriate for examination of a patient with primary endometrial cancer, as this provides a longitudinal view of the uterus which will include both corpus and cervix. Using the sagittal plane also provides the opportunity to assess anterior invasion of the tumour into the bladder and posteriorly to the rectum. Critical review and expert consensus of MRI protocols by the European Society of Urogenital Radiology, from 10 European institutions, and published literature between 1999 and 2008 indicated that high-field MRI should include at least two T₂-weighted sequences in sagittal, axial oblique or coronal oblique orientation. High-resolution post-contrast images acquired at 2 min ± 30 s after intravenous contrast injection are optimal for the diagnosis of myometrial invasion. If cervical invasion is suspected, additional slice orientation perpendicular to the axis of the endocervical channel is recommended. Due to the limited sensitivity of MRI to detect lymph node metastasis without lymph-node-specific contrast agents, retroperitoneal lymph node screening with precontrast sequences up to the level of the kidneys is optional. The likelihood of lymph node invasion and the need for staging lymphadenectomy are indicated by high-grade histology at endometrial tissue sampling and by deep myometrial or cervical invasion detected by MRI (Kinkel et al 2009).

FDG-PET has been shown to display a higher sensitivity (96.7%) than CT/MRI (83.3%) for identifying primary and extrauterine (83.3% vs 66.7%) lesions, although no difference in specificity emerged between the two modalities (both 100%) (Suzuki et al 2007). For lymph node evaluation, MRI and FDG-PET are equivalent — sensitivity (91.5% vs 89.4%), specificity (33.3% vs 50.5%), accuracy (84.9% vs 84.9%), positive predictive value (91.5% vs 93.3%) or negative predictive value (33.3% vs 37.5%) (Park et al 2008) — which does not justify the use of PET-CT in disease staging.

Recurrent disease

Transrectal sonography has been investigated in cases where recurrent endometrial cancer is suspected (Savelli et al 2004). Physical examination may be difficult because of postsurgical fibrosis. Infiltration into the surrounding connective tissues and organs such as the rectum and bladder can be identified, and transrectal ultrasound can be used to guide transvaginal or transperineal fine-needle biopsy. As with recurrent cervical cancer, CT, MRI or PET-CT may be used to detect recurrent disease and metastatic carcinoma to omentum or lymph nodes.

Choriocarcinoma

Choriocarcinomas (gestational trophoblastic tumours) are characteristically hypervascular and this can be detected with colour Doppler (Figure 5.23) (Sebire and Seckl 2008). On spectral Doppler, they have a very low resistance index and this is useful for monitoring response to chemotherapy, although ultrasound is not very sensitive in their initial diagnosis since the characteristic snow-storm appearance of the multiple cystic spaces is only present in 40% of cases.

Figure 5.23 Choriocarcinoma. A bulky uterus of non-specific pattern was seen on grey-scale ultrasound of this choriocarcinoma (A). (B) The colour Doppler study shows the marked tumour vascularity. The spectral Doppler sample volume has been placed over the left uterine artery and reveals fast flow with minimal pulsatility, indicating lack of normal resistance to flow. This results in a low pulsatility index, changes in which can be used to monitor response to treatment.

Ovarian carcinoma

In patients with ovarian cancer, the dramatic difference in cure between patients with local disease (80–90%) and distant disease (15–25%) means that imaging to detect ovarian cancer early is desirable.

The ovary

Ultrasound has been explored as a screening method for carcinoma of the ovary in an attempt to detect early disease (Menon and Jacobs 2000, Sato et al 2000). Early studies demonstrated a poor yield: out of 5479 women screened, five women with primary stage I cancer and four women with metastatic disease of the ovary were diagnosed (Campbell et al 1989). The overall rate of false-positive results was 2.3% and the likelihood of a positive scan being a primary ovarian cancer was one in 67. A scoring system has been evolved which reduces the number of false-positive results which otherwise occur due to confusion with unusual-appearing corpora lutea and benign tumours (Timmerman et al 1999). The improved resolution of transvaginal scanning permits a more detailed morphological examination (Figure 5.24A,B). This has been tested in a study of 25,000 asymptomatic postmenopausal women, of whom 740 had raised CA125 levels and were referred for ultrasonography; an ovarian volume of 8.8 ml was used as a cut-off for normality. All 20 cancers had abnormal morphology consisting of complex cysts, although there were numerous false-positive results. The positive predictive value was approximately 20% (Menon et al 2000).

Figure 5.24 Ultrasound scan of ovarian tumours. The smooth walls and echo-free contents of the benign cystic mass in (A) (arrowhead) are very different from the complex appearance of the 10.5 cm carcinoma in (B). The second cyst in (A) contains uniform low-level echoes, probably representing debris.

Benign lesions are unilocular or multilocular with thin septae and no nodules, whereas malignant lesions are often multilocular with thick septae and nodules. The use of transvaginal colour flow imaging has been advocated, but these studies have been variously reported as disappointing or useful (Varras 2004). Initially, a low resistance index (less than 0.4) was advocated as the best discriminator, but this has proved unreliable and a high peak velocity may be a better feature. At present, the role of ultrasound in screening, although attractive, remains unvalidated (Partridge et al 2009) and the results of larger trials are awaited. Additional specificity might be gained by adding contrast agents.

In ovarian cancer staging, MRI and CT are more sensitive than ultrasound (95%, 92% and 69%, respectively) for detecting peritoneal metastases, especially in the subdiaphragmatic spaces and hepatic surfaces (Tempany et al 2000). Either modality therefore can be used to stage advanced ovarian cancer. The potential of PET to identify metabolically active tumour that does not appear on morphological studies is promising. In one study, the PET-CT scan indicated a different disease distribution compared with CT alone in 55% of patients, and changed therapeutic management in 34% (Soussan et al 2008). Therefore, although current evidence for its utility is limited, PET-CT may be of major importance in staging ovarian cancer in the future.

As a tumour arising from a non-essential organ that is initially confined to the peritoneal cavity, ovarian cancer makes an attractive target for monoclonal antibodies. An antibody to CA125 has been assessed in monitoring the course of the disease (Sok et al 2009). However, antibodies are often limited by problems that include antibody specificity, stability and immunoreactivity, as well as patient reaction to the antibodies used. Antibodies coupled to drugs or biological toxins are also under investigation. Some antibodies may have direct antitumour effects through binding to biologically active receptors or through immune receptor functions. The use of antibody fragments, chimeric antibodies and genetically engineered antibodies is also under active investigation (Badgwell and Bast 2007).

Pelvic and abdominal spread

CT is the most useful imaging modality for demonstrating macroscopic recurrence, and can spare patients a second-look laparotomy. A variety of manifestations can be seen with CT, each of which can have a spectrum of appearances including ascites, peritoneal seeding, and visceral and nodal metastases (Forstner 2007). Ultrasound may also be used to detect ascites and liver metastases. MRI should be performed in women with questionable macroscopic recurrent tumour and negative CT examination (Figure 5.25), although neither modality can exclude microscopic disease. Newer whole-body diffusion-weighted MRI measurements are looking extremely promising in identifying peritoneal spread (Kyriazi et al 2009). PET-CT may also prove to be of benefit, although quoted accuracy reflects local expertise at image interpretation: lesion-based sensitivity and accuracy of FDG-PET/CT vs MRI has been recorded as 43% and 86%, and 75% and 94%, respectively (P < 0.05) (Kim et al 2007), indicating that FDG-PET/CT is not as useful as MRI for detecting local pelvic recurrence and peritoneal lesions in ovarian cancer recurrence. In contrast, in a recent meta-analysis of 34 studies comparing CA125, PET alone, PET-CT, CT and MRI, PET-CT had the highest pooled sensitivity (0.91, 95% confidence interval 0.88–0.94). The area under the curve (AUC) was 0.9219, 0.9297, 0.9555, 0.8845 and 0.7955, respectively. Results of pairwise comparison between each modality demonstrated that AUC of PET, whether interpreted with or without the use of CT, was higher than that of CT or MRI (P < 0.05) (Gu et al 2009).