Cross-sectional imaging of acute gynaecologic disorders: CT and MRI findings with differential diagnosis—part I: corpus luteum and haemorrhagic ovarian cysts, genital causes of haemoperitoneum and adnexal torsion

To investigate patients with acute abdominal or pelvic pain, intravenous iodinated contrast agent is generally used, unless contraindicated by allergy or impaired renal function. Alternatively, gynaecologic abnormalities (such as adnexal enlargement) may be encountered in unenhanced abdominal CT studies, such as those requested to investigate suspected urolithiasis and acute renal colic: in these situations, further investigation with contrast-enhanced CT or better MRI should be suggested after gynaecologic consultation [2, 4, 8].

In postmenopausal women, most institutions perform a preliminary unenhanced acquisition followed by optional arterial and mandatory portal venous phase CT scanning. In patients with hip prosthesis or other metallic implants, metal artefact reduction algorithms decrease beam-hardening effects that impair the assessment of the pelvic organs [10].

In reproductive age women with unspecific abdominal complaints, a limited CT protocol that includes a single acquisition in the portal venous phase is beneficial to limit the ionising radiation dose. If available, novel CT techniques such as iterative reconstruction should be used. When a gynaecologic disease is prospectively considered or sonography shows a non-trivial pelvic effusion, we suggest obtaining also a precontrast scanning at least of the pelvis, as it may be helpful to discriminate high attenuation reflecting the presence of blood or debris within a cyst or collection from contrast enhancement. Regardless of patient age, an additional arterial-phase (CT-angiography) acquisition using bolus tracking technique with the region-of-interest in the abdominal aorta is warranted when a concern for blood loss exists, such as in patients with vaginal bloody discharge, dropping haematocrit, hypotension or haemodynamic instability and after sonographic detection of haemoperitoneum [2, 4, 8].

In state-of-the art multidetector scanners, CT generates almost isotropic sub-millimetre voxels that allow reconstruction of high-resolution images along arbitrary planes. Even if a gynaecologic disease is not prospectively suspected, in females, the routine reconstruction of 3- to 5-mm-thick sagittal and coronal CT images of the pelvis is recommended to depict the anatomy of normal and diseased genital organs. Sagittal viewing allows appropriate assessment of size, morphology and enhancement patterns of the uterus, the latter markedly varies according to age, menstrual phase and delay between contrast injection and CT acquisition. Additional MRI-like oblique-axial and oblique-coronal reconstructions (perpendicular and parallel to the main uterine axis, respectively) may be helpful to discriminate extra-uterine structures and to better assess topography of uterine abnormalities such as subserosal or pedunculated leiomyomas [1‐4, 11].

At CT, the fallopian tubes are generally not visible in normal conditions but become identifiable as tubular structures when dilated by intraluminal fluid, secretions or blood. The normal ovaries measure approximately 3 × 3 × 2 cm and show low attenuation with a faintly recognizable follicular structure on contrast-enhanced images. Identification of the ovaries relies on their location close to the external iliac vessels, posterior to the round ligament. Recognition of the suspensory ligament (which connects the ovary to the pelvic sidewall) as a linear or fan-shaped soft-tissue band allows diagnosis of ovarian versus non-ovarian masses. The ovaries are located anterior or antero-medial to the pelvic ureters, conversely iliac lymph nodes lie lateral or posterolaterally [3, 12].

Our comprehensive MRI acquisition protocol of the pelvis for acute pelvic pain or suspected acute gynaecologic disorders is performed on a closed-configuration 1.5-Tesla superconducting MRI scanner (Signa HDxT; GE Healthcare, Milwaukee, Wisconsin, USA) using an eight-channel high-resolution phased-array torso coil with array spatial sensitivity technique (ASSET) parallel acquisition, with the patient lying in the supine position (entry position feet first). An antispasmodic agent such as glucagon or N-butyl-scopolamine, unless contraindicated (e.g. unstable cardiac disease, tachycardia, diabetes, acute angle-closure glaucoma and phaeochromocytoma) can be administered intravenously or intramuscularly to reduce bowel motion artefacts [13, 14].

This state-of-the-art MRI protocol (Table 1) relies on T2- and T1-weighted sequences with and without fat suppression and diffusion-weighted imaging (DWI) sequences. Multiplanar T2-weighted sequences are crucial in the assessment of the female pelvis since they provide excellent soft-tissue contrast of the genital organs and surrounding structures. Fat-suppressed T2-weighted images increase the conspicuity of inflammatory lesions and are useful to identify ascites, complex fluid, oedema of tissues and perivisceral fat [15]. Optional gradient-echo T2*-weighted sequences are suitable to detect haemorrhage and air bubbles due to their sensitivity to magnetic susceptibility effects but are particularly prone to susceptibility artefacts from bowel gas. Precontrast fat-suppressed T1-weighted sequences allow differentiating blood from fat, ruling out fat-containing lesions such as teratomas, and highly sensitive for the detection of haemorrhagic foci in endometriosis. DWI sequences are helpful in the characterisation of fluids: hypercellular fluids, such as pus, show restricted diffusion and therefore appear hyperintense on DWI images and hypointense on the corresponding apparent diffusion coefficient (ADC) map. According to the European Society of Urogenital Radiology (ESUR) guidelines, gadolinium-enhanced fat-suppressed T1-weighted sequences are useful (a) to assess a lesion’s contrast enhancement pattern, (b) to depict enhancing mural nodules within adnexal masses, (c) to identify contrast extravasation indicating active bleeding, (d) to differentiate PID from endometriosis, (e) to identify inflammatory enhancement of perivisceral fat and peritoneum and (f) for characterisation of leiomyomas and differential diagnosis from leiomyosarcomas and adnexal masses [13, 16]. Finally, delayed urographic acquisitions allow preoperative depiction of the ureters and their anatomical relationship to adnexal masses [17].

Although very complete, the abovementioned protocol (Table 1) lasts approximately 30 min (40 min including urographic acquisitions) and seems hardly applicable in an emergency setting. When dealing with ill or non-cooperating patients with acute pelvic pain, we suggest adopting a time-efficient non-contrast-enhanced MR imaging protocol that relies on faster sequences, which lasts approximately 15 min (Table 2).

On MRI pelvic examinations, orthopaedic metal implants such as hip prosthesis may determine signal loss, geometric distortion, and failure of fat suppression. Metal artefact reduction sequences (MARS) is a general definition which does not refer to a single specific technique, but rather encompasses a variety of sequences optimised to reduce artefacts from metal. As a general consideration, artefacts are less severe at 1.5 T than at 3 T, since the susceptibility-induced field inhomogeneity is doubled at 3 T MR scanners. As for conventional techniques, turbo spin echo are preferable to gradient echo sequences, and short tau inversion recovery (STIR) performs better than standard spectral fat saturation.

Other tricks to optimise standard sequences, although often at the expense of acquisition time, include increasing receiver bandwidth, increasing matrix size, decreasing echo times and decreasing slice thickness. In recent years, new dedicated metal artefact reduction techniques have been developed including view angle tilting, the multispectral imaging techniques, multiacquisition variable-resonance image combination and slice-encoding for metal artefact correction. In patients with metal implants, these sequences, associated with advanced image acquisition techniques such as parallel imaging, allow to acquire MARS in a clinically feasible scan time [18].

In reproductive age females, the normal ovary shows homogenous soft-tissue T1-weighted signal intensity and is often barely perceptible from adjacent bowel. On the other hand, on T2-weighted images, ovaries are almost always identifiable on the basis of characteristic zonal anatomy including a low signal intensity cortex and high signal intensity medulla with scattered well-circumscribed fluid foci corresponding to follicles. The normal, non-dilated fallopian tubes are usually hardly recognizable at MRI unless outlined by pelvic fluid. In the presence of peritoneal effusion, the fallopian tubes become visible as paired, thin and serpentine juxtauterine structures. When dilated by serous fluid, haemorrhage or pus, fallopian tubes appear as fluid-filled tubular structures that arise from the upper lateral margin of the uterine fundus, showing incomplete parietal plicae or folds that give them the characteristic convoluted appearance [19].

In reproductive age females, the majority of ovarian cysts are physiological or functional in nature and encompass dominant follicles, follicular cysts and luteal cysts. During the follicular phase of menstrual cycle, under oestrogen stimulation, ovarian follicles grow until only one in each cycle reaches a size of approximately 20–25 mm, becoming the dominant follicle. Sometimes a dominant follicle fails to ovulate, but remains hormonally sensitive and continues to grow becoming a follicular cyst. During ovulation, the dominant follicle ruptures, releases the oocyte and then collapses; at this point, the luteinising hormone causes proliferation of granulosa cells in the inner lining of the remaining follicular bed, resulting in the formation of the corpus luteum.

A functional structure which regresses spontaneously when pregnancy does not occur, the corpus luteum is frequently seen when cross-sectional imaging is performed in the second part (luteal phase) of the menstrual cycle. The CT hallmark of a corpus luteum is a well-circumscribed, unilocular (rarely bilocular) cystic structure located within the ovary, which measures up to 3 cm and is demarcated by a crenulated, briskly enhancing rim that reflects prominent blood flow within thickened walls (Fig. 1) and corresponds to the colour Doppler US ‘ring of fire’ sign. MRI shows a similar appearance with internal fluid-like hypointense and hyperintense signal on T1- and T2-weighted images, respectively, and intense wall enhancement after gadolinium administration (Fig. 2) [8, 9, 20].

When a corpus luteum fails to regress, a luteal cyst develops and may be filled with either fluid or blood, so the terms corpus luteum cyst and haemorrhagic corpus luteum are often used as synonyms [21]. At CT, both follicular and corpus luteum cysts present as well-marginated unilocular lesions, usually measuring less than 5 cm, with thin walls and homogeneous fluid-like attenuation (Figs. 3 and 4). A 3-cm threshold has been suggested to differentiate follicles and corpus luteum (below 3 cm) from functional (follicular/luteal) cysts (over 3 cm). At MRI, ovarian cysts containing simple fluid have prolonged T1 and T2 relaxation times and therefore show homogeneously low T1-weighted and very high T2-weighted signal intensity (Figs. 4 and 5) [22].

Unilateral acute pelvic pain and tenderness leading to ED admission may result from cyst enlargement, haemorrhage or rupture. Bleeding within luteal cyst or other functional cyst results in the formation of a haemorrhagic ovarian cyst. In the former case, bleeding is caused by the hypervascularity and fragile vessels of the granulosa layer within the wall of the cyst. At CT, the presence of blood leads to the appearance of increased attenuation (> 40 Hounsfield units, HU) within the cyst. Haemorrhagic ovarian cysts are unilateral, show mixed (in the range of 25–100 HU) internal attenuation with common fluid-debris or fluid-fluid levels (Fig. 6) and often appear as complex masses. Considerable overlap exists between imaging appearances of haemorrhagic follicular and luteal cysts, and differentiation among the two entities should rely on the current phase of the menstrual cycle. After intravenous contrast, the walls of luteal cysts appear thicker than those of follicular cysts and highly vascularised. The MRI pattern of haemorrhagic cysts depends on the age of the blood, the amount of serous fluid and clot formation and retraction (Fig. 7). The most usual appearance is that of a unilocular, unilateral cyst with relatively high signal intensity on T1-weighted images and variable, heterogeneous T2-weighted signal intensity; intracystic fluid-fluid levels are commonly seen [8, 23].

Table 3 summarises the CT and MR imaging features of ovarian cysts. Additionally, a flowchart summarising the steps of an algorithm for MRI differential diagnosis of ovarian cystic masses is proposed in Fig. 8.

The endometrioma represents the key differential diagnosis of haemorrhagic ovarian cysts. Affecting 10% of women of reproductive age, endometriosis is defined by growth of ectopic endometrial tissue outside the uterus, which bleeds synchronously with the endometrium. Clinically, endometriosis is suggested by cyclic perimenstrual pain and by other manifestations such as dysmenorrhoea, dyspareunia, infertility and pain during defecation. Although dysmenorrhoea and chronic pelvic pain represent the classic clinical presentation, a sudden onset with acute pelvic pain may also occur [24]. Because of the great clinical variability of the disease, delayed diagnosis up to 7–10 years after initial symptoms is common [25].

More common (80% of patients) compared to superficial peritoneal implants and deep pelvic endometriosis, endometrioma results from ectopic endometrial glands in the ovarian cortex that form haemorrhagic ‘pseudocysts’ which become detectable by imaging. Endometriomas have highly variable, often heterogeneous CT appearance and often mimic mixed solid/cystic or predominantly solid masses (Fig. 9a, b). Compared to MRI, CT has poor specificity and often does not allow reliable differentiation from other complex cystic adnexal masses, either benign or malignant (Fig. 9). Suggestive ancillary findings include multiplicity, bilateral adnexal involvement (‘kissing ovaries’ sign, Fig. 9a, b), stability over time and associated lesions on the peritoneum and bowel serosa [8, 17].

MRI is arguably the best imaging technique for the diagnosis of endometrioma, relying on high signal intensity on T1-weighted images and the pathognomonic T2 ‘shading’ sign (gradual loss of signal till complete signal void in the most dependent portion) reflecting cyclic repeated bleeding over time (Figs. 10 and 11). Haemorrhagic ovarian cyst is suggested over endometrioma when the cyst is solitary and unilocular, without the ‘shading sign’, and disappears at follow-up (6–12 weeks). Other T1-hyperintense lesions which may be confused with endometriomas include dermoids and mucinous cystic tumours: precontrast fat-suppressed T1-weighted images help distinguish lipid-containing dermoids from the blood. Mucinous cystic tumours show signal intensity lower than the blood on T1-weighted images. Finally, the identification of vascularised projections and nodular septa should suggest malignancy [17, 22, 26, 27].

In this regard, it is important to remember the importance of post-processing images when facing a cystic mass that is hyperintense on unenhanced T1-weighted fat-suppressed sequences. Subtraction techniques that suppress the T1 hyperintense background of the cystic content allow to detect true gadolinium enhancement in suspicious solid mural nodules (a finding suggesting tumour arising from endometrioma), which may be hardly identifiable at visual assessment alone [14, 28].

In women of reproductive age, the genital tract is the most common source of non-traumatic haemoperitoneum and bleeding corpus luteum represents the leading cause, secondary to the high vascularity of the luteal walls [29, 30]. This potentially life-threatening condition is rare in adolescence and occurs at a median age 26 years, often (nearly 60% of cases) after sexual intercourse or physical exercise. The right ovary is more frequently involved because the contralateral one is protected by the sigmoid colon ‘cushion’. The usual presentation includes a sudden onset of pelvic pain which gradually spreads to the entire abdomen, ultimately leading to peritoneal irritation and even life-threatening shock. Amenorrhoea, vaginal bleeding, early pregnancy signs and elevated beta human chorionic gonadotropin (β-hCG) are absent [31, 32].

At CT, the ruptured corpus luteum is seen as a shrunken ovarian cyst with discontinuity of the thickened wall, often surrounded by hyperattenuating ‘sentinel clot’. Occasionally, active bleeding may be detected as serpiginous or jet-like contrast extravasation adjacent to the cyst, most usually in the portal phase reflecting intermittent or venous haemorrhage (Figs. 12 and 13). Haemoperitoneum appears as hyperdense (> 30 HU) peritoneal-free fluid, with higher attenuation typically detected within the dependent portion of peritoneal cul-de-sac (Figs. 12 and 13) [3, 8, 20, 29, 30].

At MRI, haemoperitoneum shows variable signal intensity depending on its age (haemoglobin degradation stages), the extent of the bleeding and blood clot formation. Within 48 h from bleeding, pelvic intraperitoneal haemorrhage generally shows intermediate signal intensity (higher than that of urine) on T1-weighted images and intermediate T2-weighted signal intensity (Fig. 14). Later, a fluid-fluid level can be seen, with the dependent portion of the bloody ascites being hyperintense compared with the supernatant on T1-weighted images; on T2-weighted images, the signal intensity relationship is reversed. This appearance corresponds to the CT ‘haematocrit effect’, where sedimented erythrocytes produce a dependent layer of high attenuation. Intraperitoneal haemorrhage may contain localised clot showing higher signal intensity than the free intraperitoneal blood on T1-weighted images and low signal intensity on T2-weighted images. These clots may be useful in identifying the bleeding source since they generally form close to the organ from which the haemorrhage originated (‘sentinel clot’ sign) (Fig. 15). Compared to CT, active haemorrhage is occasionally identified on MRI as hyperintense blush on fat-suppressed T1-weighted images after gadolinium administration [8, 9].

Whereas in the past, the majority of cases were treated surgically, nowadays, in haemodynamically stable patients, the management approach is increasingly conservative with close observation and roughly 20% of patients ultimately require laparoscopic surgery [31, 32]. Some authors proposed the use of CT to predict the clinical outcome, stating that in their experience, active bleeding and massive pelvic haemoperitoneum (anteroposterior diameter > 6 cm) dictate the need for operative treatment [20, 33].

The second most frequent gynaecologic cause of haemoperitoneum, ruptured ectopic pregnancy (EP) accounts for 6% of pregnancy-related mortality. Mostly encountered in the fourth decade of life, EP occurs in approximately 1.3–2.4% of all pregnancies and refers to the implantation of a blastocyst at a different location from the endometrium of uterine cavity, most usually in the fallopian tube (about 95% of all ectopic pregnancies, involving the ampulla, isthmus and fimbria in descending order of incidence). The other uncommon implantation sites are listed in Table 4 [34, 35]. Predisposing factors include prior EP, history of tubal surgery, scarring related to previous PID and infertility treatments such as in vitro fertilisation and embryo transfer. The presentation includes acute abdominal pain closely similar to that produced by ruptured corpus luteum, but associated with amenorrhoea, vaginal bleeding and positive beta human chorionic gonadotropin (β-hCG) [34].

The key distinction between pregnant and non-pregnant patients is determined by β-hCG levels in correlation with menstrual history. Albeit acute appendicitis is the most frequent cause of acute abdominal or pelvic pain in pregnancy, EP represents a leading consideration in patients with positive β-hCG test and non-visualisation of intrauterine pregnancy at transvaginal US. Unfortunately, US has a high specificity (near to 100%) but low sensitivity (about 15–20%) [34, 36]. Occasionally, patients with EP may undergo CT under the clinical diagnosis of severe acute abdomen or haemoperitoneum, sometimes without prior US. Alternatively, in our experience, females may be investigated after false-negative early urine pregnancy tests and are afterwards found pregnant by in-hospital β-hCG blood testing. The CT appearance of EP is an extrauterine (most usually adnexal) cystic-like structure with some degree of peripheral enhancement, representing a gestational sac, with or without haemoperitoneum (Fig. 16) [15, 37].

In patients with equivocal clinical and sonographic findings, MRI represents an appropriate tool to make the correct diagnosis of EP. MRI findings include (a) haemoperitoneum in the pelvic cul-de-sac; (b) a heterogeneous, partly haemorrhagic, adnexal mass; and (c) tubal dilatation (haematosalpinx) with wall enhancement. The adnexal mass corresponding to the gestational sac appears as a thick-walled cystic structure which typically has high signal on T2-weighted sequences and frequently contains foci of acute blood (with variable T1-weighted signal intensity according to haemoglobin degradation stages). Unfortunately, the cystic component may not be visible, and in these cases, EP appears as a heterogeneous, mostly T2-hyperintense mass. Secondary to the invasion of the tubal wall by the trophoblasts, haematosalpinx appears as hyperintense fluid on T1-weighted sequences within the dilated fallopian tube (Fig. 17). The blood in different stages of evolution appears as iso- to hyperintense fluid on T1-weighted MR images. Additional gradient-echo T2*- and susceptibility-weighted imaging may be helpful in the identification of EP-related bloody adnexal masses. At CT and MRI, the cystic mass may show variable peripheral enhancement, corresponding to the sonographic ‘ring of fire’ sign; similarly, tubal wall enhancement has been reported, albeit the use of gadolinium contrast is controversial [9, 15, 38, 39].

Ruptured EP is still considered the primary cause of death in the first trimester of pregnancy. It is due to the invasive growth of trophoblast into the wall of the fallopian tube; the risk of rupture increases with the enlargement of the EP [40].

Imaging findings of ruptured EP include a poorly defined complex adnexal mass with the presence of pelvic fluid; the identification of bloody effusion (measuring 30–45 HU) in the Morrison pouch and in the cul-de-sac should raise concern for a ruptured EP. On contrast-enhanced images, active bleeding may be detected, especially in the case of a massive rupture. A specific CT finding is the sentinel clot sign: the presence of a hyperdense clot (> 60 HU) within the adnexum, thus confirming the source of bleeding [35, 40].

At MRI, diagnostic features of tubal rupture include the disruption of tubal wall enhancement and the occurrence of acute haematoma (low T2 signal intensity outside the implantation site). However, in pregnant patients, the detection of blood in the pelvic cul-de-sac is not specific for EP; the differential diagnosis includes haemorrhagic corpus luteum cyst and spontaneous abortion; however, the association of haemorrhagic fluid without an intrauterine pregnancy should raise suspicion for EP [9, 15, 38, 39]. Some authors reported that the size may help in the differentiation between ruptured EP (median 3.76 cm) and corpus luteum (2 cm) [15, 37].

Overall, due to its higher contrast resolution, MRI may detect EP more accurately than CT. The value of MRI in diagnosing EP has been evaluated in retrospective studies that included the presence of a direct sign (ectopic gestational sac) and indirect signs (haematosalpinx, adnexal haematoma and bloody effusion). The diagnostic accuracy of the direct sign was 92%, with a sensitivity of 91.3% and specificity of 100%, more accurate than that of any single indirect sign; accuracy increased up to 100% when diagnostic criteria required the presence of a direct sign or at least two indirect signs [41]. Furthermore, using T2*-weighted sequences in diagnosing EP sensitivity, specificity and accuracy were, respectively, 95%, 100% and 96%, due to the capability of T2*-weighted images to identify fresh haematoma [39].

Laparoscopic surgery is the gold standard treatment for EP. Indications include rupture, pain and haemodynamic instability. However, in selected patients, expectant management or pharmacotherapy (methotrexate) is being increasingly used [34, 36].

Isolated fallopian tube torsion without ovarian torsion is exceptionally rare and requires surgical intervention too. Predisposing factors are hydrosalpinx, tubal ligation, tubal or paratubal cystic and solid masses [9].

CT, better if with multiplanar reformatted images, may show an adnexal structure distinct from the ovary. MRI usually shows a distended dilated fallopian tube with thickened walls; the tube may demonstrate a vortex-like appearance (due to more twists) distant from the ipsilateral ovary, and the latter appears normal. Thick and twisted pedicle can also be observed (whirlpool sign) [19, 50].

Paraovarian or paratubal cysts are unilocular cystic structures located in the broad ligament, between the fallopian tube and the ovary. Isolated torsion of paraovarian/paratubal cysts rarely occurs and its prevalence is higher in children than in adults. MR imaging generally demonstrates a unilocular cyst distinct from the ipsilateral ovary, hyperintense on T2-weighted images and hypointense on T1-weighted images. The signal intensity of the cystic content may be slightly high on T1-weighted images due to haemorrhagic changes related to torsion [9].

The differential diagnosis of AT includes massive ovarian oedema and ovarian hyperstimulation syndrome (OHSS). The former entity is a rare, benign condition thought to result from intermittent or partial torsion of the mesovary compromising the venous and lymphatic drainage but with preserved arterial supply [55].

The imaging appearance of massive ovarian oedema is similar to that of AT: an enlarged ovary with oedematous stroma and multiple nonovulatory follicles pressed towards the periphery of the cortex (Fig. 20) [56]. However, patients’ clinical history generally records self-limiting episodes of abdominal pain of different durations. The features of pain vary (acute pain or progressive and profound diffuse pain) depending on the rapidity of the torsion; menstrual irregularities, infertility, abdominal distension, signs of virilisation and precocious puberty are also reported [57, 58].

OHSS is an iatrogenic complication of assisted reproduction techniques, characterised by bilaterally enlarged ovaries with several, peripherally located cysts accompanied by fluid shift from the intravascular to third space from increased capillary permeability that may cause the development of ascites, pleural effusions and oliguria in severe forms [59]. Typical imaging findings at CT and MRI include bilateral symmetric enlargement of ovaries due to the presence of multiple cysts representing enlarged follicles or corpus luteum cysts. Cystic content can be fluid-like or haemorrhagic; in the latter case, cysts demonstrate higher attenuation at CT and high T1 signal intensity at MRI. The characteristic peripheral location of the follicles, separated by thin septa and surrounding a central core of ovarian stroma, has been referred to as ‘wheel spoke’ appearance [60]. Differently from cystic neoplasms, OHSS lacks abnormal enhancing solid components within the cystic lesions; in addition, follow-up imaging should demonstrate resolution. It should also be remembered that enlarged hyperstimulated ovaries are themselves at risk for torsion [61]. If torsion occurs in hyperstimulated ovaries, the latter may demonstrate asymmetrical augmentation, abnormal enhancement, twisted pedicle and eventually haemorrhage [50].

Other features of OHSS include free fluid in the peritoneal, pleural and pericardial cavity; the free fluid seen in OHSS is generally simple ascites, although it may display slightly higher attenuation due to ruptured haemorrhagic cysts. The modified Golan classification categorises OHSS as mild, moderate or severe on the basis of symptoms severity and radiologic findings [62].

Congested pelvic veins without AT are the hallmark of the pelvic congestion syndrome (PCS), an underdiagnosed condition that affects 9% of premenopausal women and may be exacerbated by intra-abdominal pressure. The multifactorial pathogenesis involves a family history of varicose veins, previous pelvic surgery, multiple pregnancies, hormonal influences and retroverted uterus. Although the key symptom is chronic dull pelvic pain that worsens in the premenstrual phase, acute and severe pain may also occur. Flow reduction, thrombosis and mass effect on nerves may lead to worsening pelvic pain. Other clinical manifestations may include dyspareunia, postcoital ache, dysmenorrhoea and perineal pain [63, 64].

Similarly to male varicocele, dilatation of the gonadal veins and pelvic venous plexus more commonly occurs on the left side because of left ovarian vein draining into left renal vein before reaching the inferior vena cava. Mechanical extrinsic compression of drainage veins such as in retroaortic left renal vein, in May-Thurner syndrome (left common iliac vein compression against the lumbar spine by the right common iliac artery) or in nutcracker syndrome (left renal vein compression between superior mesenteric artery and aorta), may lead to venous congestion and engorgement of the ovarian vein [65]. In this latter syndrome, the abrupt narrowing, with a triangular shape, of the proximally dilated left renal vein between the abdominal aorta and the superior mesenteric artery is referred to as the ‘beak sign’. It is one of the most useful CT findings for the diagnosis of nutcracker syndrome, with a sensitivity of 91.7% and a specificity of 88.9% according to the literature [66, 67]. MR angiography provides findings similar to those of CT with the advantage of being less invasive and without radiation exposure in comparison with retrograde venography [68].

At CT, tortuous and dilated (> 4 mm) tubular structures nearby the ovary and uterus show slow blood flow and enhancement synchronous with veins. The diagnostic criteria for PCS require at least four ipsilateral parauterine veins of varying calibre, at least one measuring more than 4 mm in diameter, or an ovarian vein diameter greater than 8 mm (Fig. 21). Similarly, MRI depicts pelvic varices as multiple dilated vessels lying on uterus (and within the myometrium such as arcuate veins), ovaries and pelvic sidewall that can extend inferiorly to communicate with paravaginal and thigh venous plexus [64, 69‐71]. Dilated veins are hyperintense on gradient-echo T1-weighted images; on T2-weighted MR images, they may show no signal intensity due to flow-void artefacts or mixed low and high signal intensity because of the relatively slow flow inside the vessels. Gradient-echo fat-suppressed T1-weighted sequences after intravenous administration of contrast material demonstrate vascular enhancement of pelvic varices, thus confirming the diagnosis (Figs. 22 and 23) [64]. Asciutto et al. reported MRI to have levels of sensitivity and specificity for PGS respectively of 88% and 67% for ovarian veins, 100% and 38% for iliac veins and 91 and 42% for pelvic plexus [72].

Additionally, up to 50% of women with PVI present cystic formations in the ovary, ranging from scant cysts to classic polycystic ovary aspect (Fig. 24) [64]. The mechanism underlying these cystic changes in the ovary is not completely known, although a relationship with increased oestrogen levels has been reported [73].