Several recent improvements in magnetic resonance imaging (MR) technology have led to more rapidly acquired, higher resolution images of the pelvis. As a result, MR has become the imaging modality of choice for many pelvic diseases, including diagnosis of congenital uterine anomalies and preoperative assessment of fibroids (table 1). The multiplanar capability of MR has always been useful for determining the organ of origin of various pathologic entities in cases where ultrasound or physical exam findings are ambiguous, e.g. pedunculated fibroid versus solid adnexal mass. In addition, MR's high intrinsic contrast resolution is unmatched in depicting the zonal anatomy of the uterus and alterations thereof.

Contrast enhancement with intravenous gadolinium, while seldom necessary in benign uterine disease, has proven useful in certain situations, including characterization of adnexal masses, detection of endometriosis, and diagnosis of endometrial polyps. As with ultrasound, MR has not been shown to cause harm to the fetus or to a woman's fertility.1

While often viewed as an expensive test, net cost analyses have shown that using MR imaging in the diagnosis of some pelvic diseases often saves health care dollars by obviating surgery or by prompting less-invasive surgery.2 This article will review MR techniques, anatomy, and pathology as they apply to benign diseases of the female pelvis.

Technical considerations

Coils-While using a multi-coil phased array is optimal, the body coil will often produce high quality images for diagnosis. The multi-coil, a receive-only surface coil, does offer several advantages, though. It allows a smaller field of view (FOV) and, even more importantly, a two to threefold higher signal-to-noise ratio (S/N). The multi-coil's four receive-only surface coils, two anterior and two posterior, each sample a small effective volume, thereby maximizing S/N. Each coil has its own amplifier, receiver channel, digitizer, and memory. Four individual data sets are produced and then reconstructed to form a high resolution image.1

The area of coverage of a multi-coil array is confined to a radius of approximately 10 cm from the center of the coils; hence, imaging of large patients can be suboptimal. Also, the presence of a protuberant lower abdomen will not allow for adequate balancing of the coils on the body surface. For these reasons, it is better to use the body coil to examine large patients. In addition, the pelvic coil is highly sensitive to motion, hence prone positioning of the patient may be helpful; however, as many patients cannot lie prone comfortably, most patients are imaged while lying in the supine position.

An abdominal binder (Velcro® wraps) and both anterior and posterior saturation bands over the subcutaneous fat usually suffice to minimize respiratory motion (phase-encoding) artifact.3 Glucagon (1 mg IM) is routinely given just prior to the exam to decrease motion artifact from bowel peristalsis, provided there are no contraindications to its administration, such as a history of glucagon hypersensitivity, pheochromocytoma, insulinoma, or poorly controlled diabetes mellitus. Currently, oral contrast is not used for standard pelvic MR exams. Tampon insertion is only necessary if specific questions about the vagina are to be addressed.

For higher resolution imaging of the cervix and urethra, endoluminal coils may also be used, and can be especially helpful in staging cervical cancer and in imaging urethral diverticula. These coils (endorectal or endovaginal) allow for small FOVs. Examination of the cervix is, therefore, best performed with an intravaginal ring coil which surrounds the cervix, rather than an endorectal coil. An intravaginal coil will more clearly show the anterior lip of the cervix and the fascial planes between the cervix and bladder.4 When an endoluminal coil is used in conjunction with a multi-coil via an external adapter, the full benefits of all coils are realized.

Pulse sequences-T2-weighted imaging (T2WI) is the cornerstone of female pelvic MR imaging, with regard to both uterine zonal anatomy and gynecologic pathology. Fast spin echo (FSE) imaging has revolutionized MR by simultaneously shortening imaging time and improving contrast-to-noise and signal-to-noise ratios over conventional T2WI, secondary to the longer TRs and longer effective TEs that may be used.5 T1-weighted imaging (T1WI) is most useful for gauging general pelvic anatomic landmarks, further characterizing adnexal lesions (detecting fat, hemorrhage), and examining the intrapelvic fat for infiltrative changes and lymphadenopathy.

For our standard MR exam of the female pelvis, which takes approximately 30 to 45 minutes, we image with axial spin echo T1WIs, and sagittal, axial, and coronal FSE T2WIs, using a 16 cm to 20 cm FOV and a matrix of 256 ¥ 192 (table 2). The decision to give intravenous contrast (gadopentetate dimeglumine-Magnevist, Berlex Laboratories, Wayne, NJ) is made on an individual basis and will be described below. The contrast is given as a 10-cc to 20-cc bolus, followed by a saline flush. All contrast-enhanced pulse sequences are performed with fat saturation to prevent misinterpretation of fat as enhancing tissue. The contrast-enhanced pulse sequences include dynamic fast multiplanar spoiled gradient echo (FMPSPGR) scans (1.5 Tesla, Signa system; GE Medical Systems, Milwaukee, Wisconsin) and post-gadolinium T1-weighted imaging, with the typical dose of intravenous gadolinium ranging from 10-cc to 20-cc, depending on the patient's weight. Chemical shift- and fat- and water-saturation imaging can be added to the routine imaging sequences when applicable, as discussed below.

Normal MR anatomy

MR is an excellent noninvasive imaging modality for the demonstration of uterine and parauterine anatomy and for ovary identification.

Uterus-On T1WI, the uterus is usually of homogeneous, low-to-intermediate signal intensity and the uterine contours and supporting structures are well seen. Nevertheless, the uterus tends to be isointense on T1WI with bowel and soft tissues. The uterus is easily identified on T2WI by its characteristic zonal anatomy. On T2-weighted MR images, four major zones can be identified: the endometrium, the junctional zone or innermost one-third of myometrium, the myometrium proper, which begins at the level of the arcuate arteries (stratum vasculare) and extends outward to the serosa (subserosal muscle), and the serosa. The serosa is often identified as a thin line of low T2 signal. The endometrium is of high T2 signal, secondary to its composition of mucosa, glandular tissue, and varied amounts of endocavitary fluid.6 The junctional zone is normally a continuous band of low T2 signal, averaging 4.4 mm in thickness,7 that is well-demarcated on its inner, endometrial side and less well-demarcated on its outer, myometrial side as it blends into the stratum vasculare. Its low signal has been attributed to the dense packing of smooth muscle fibers and the increased number and size of nuclei within this layer.8 The myometrium proper exhibits an intermediate signal (figure 1).

Of note, several changes in the appearance of the uterus on MR take place during the menstrual cycle and with other endogenous, as well as exogenous, alterations in hormone levels. The endometrial stripe, which is best evaluated in the sagittal plane, averages 6 mm to 7 mm during the follicular phase and 10 mm during the luteal or secretory phase.6 During the secretory phase, myometrial T2 signal increases secondary to increased fluid content and vascular engorgement.3 Prior to puberty, postmenopausally, and post-radiation, the uterus is reduced in size, exhibits less distinct zonal architecture (the myometrium decreases in signal and the junctional zone is not identifiable), and possesses a thin endometrial lining, unless the patient is on hormone replacement therapy. Oral contraceptives and GnRH analogs also have the effect of reducing endometrial thickness. Oral contraceptives may reduce junctional zone thickness and may result in a higher signal intensity myometrium as well.9

Cervix-The above-described signal-defined layers of the uterus are continuous with those of the cervix (figure 1A), although they carry different names and are of slightly different composition within the cervix. The cervical mucosa and canal also are of high T2 signal, but the cervical canal is distinguished by the presence of arbor vitae or plicae palmatae, which produce an irregular, branched mucosal pattern. The surrounding low signal band or middle layer represents densely packed muscle and fibrous stroma, and the outer or intermediate signal layer is continuous with the myometrium. Often, Nabothian cysts are found here. These normal structures arise from pinched-off endocervical glands, can be up to 2 cm in size, and share characteristics of other simple cysts; they typically are of low T1 signal, high T2 signal, and are round in shape, with barely perceptible walls. Sagittal T2WI best depicts the supravaginal and intravaginal portions of the cervix by outlining the border-forming vaginal fornices.9

Vagina-Tampon insertion and, preferably, endorectal coil use enables MR evaluation of the vagina. Vaginal secretions yield high T2 signal in the canal, though the mucosa itself is low in signal, and the muscular vaginal wall exhibits signal intensity that is comparable to that of the myometrium. Axial and coronal T2WI best depicts the high signalpara-vaginal and paracervical venous plexa.9

Ovaries-The premenopausal ovary exhibits intermediate signal on T1WI and low-to-intermediate signal on T2WI, and contains numerous small, primarily peripheral physiologic or functional cysts (figure 1B). The low-to-intermediate signal of the ovary is attributed to increased cellularity of the cortical tissue and dense connective tissue. The occasional higher signal observed within the ovarian medulla of some premenopausal women has been attributed to less-dense vascular and connective tissue in this location. Most cysts that are imaged as less than 1 cm to 1.5 cm represent follicles which are filled with serous fluid and, therefore, are of low T1 signal and high T2 signal. Their featureless, thin wall enhances upon gadolinium administration. Most cysts greater than 1 cm represent corpus luteum cysts, which exhibit wall enhancement upon contrast administration and more varied internal signal intensity secondary to the presence of hemorrhage in its various stages-corpus luteal cysts are more apt to be of high T1 signal and slightly lower T2 signal on heavily T2WI than their follicular counterparts.1 The postmenopausal ovary is smaller and devoid of follicles, unless it is stimulated by HRT.

Uterine pathology

Congenital uterine anomalies-The clinical presentation of congenital uterine anomalies varies with the type of anomaly. Some anomalies, such as the arcuate uterus, are asymptomatic and inconsequential. Others may present with infertility (e.g. septate uterus), pelvic pain, and hematometrocolpos (e.g. uterus didelphys with partial vaginal septum). The septate uterus is the most common müllerian anomaly, and can result in a 90% abortion rate.9 A bicornuate uterus offers a lower, but still real risk of infertility. Uterus didelphys is usually asymptomatic, unless there is a horizontal septation of the upper vagina that results in hematometrocolpos. When a unicornuate uterus possesses a rudimentary horn, this horn usually does not communicate with the vagina and may contain functioning endometrium, leading to unilateral hematometra.

The American Fertility Society has recently reclassified müllerian anomalies into seven major categories (table 3).10 These categories are hypoplasia/agenesis (Class I), unicornuate (Class II), didelphus (Class III), bicornuate (Class IV), septate (Class V), arcuate (Class VI), and DES drug-related (Class VII). With the exclusion of DES-related anomalies, congenital uterine anomalies can be viewed conceptually as failures of fusion or resorption. The müllerian (paramesonephric) ducts form the entire uterus and the upper one-third of the vagina; the lower two-thirds of the vagina are formed by the urogenital sinus. During normal uterine development, the two paramesonephric ducts (each with a uterine tube and canal) fuse in the midline, forming the uterine septum. Shortly thereafter (around 9 weeks of gestation), the septum is resorbed. When there is incomplete fusion, didelphys (most extreme), bicornuate, and arcuate (least severe) anomalies result. When there is no resorption or only partial septal resorption, a complete septate or partial septate uterus results. A unicornuate uterus with rudimentary horn results from partial atresia of one of the paramesonephric ducts.11

Discerning a septate uterus from a bicornuate uterus has always been a challenge in both hysterosalpingography and ultrasound, and is a critical distinction to make, as the treatment for each is very different. A septate uterus can be repaired hysteroscopically with obliteration of the midline uterine septum (hysteroscopic metroplasty), whereas a bicornuate uterus can only be reconstructed with open surgery (abdominal metroplasty). MR offers the best chance at discerning between the two. In a comparative study, MR demonstrated an accuracy of 100% in classifying all müllerian anomalies, compared to 92% with endovaginal ultrasound, and much poorer accuracy with hysterosalpingography.12

Uterine hypoplasia, a subcategory of Class I-hypoplasia/agenesis, is best appreciated in the sagittal plane. It manifests itself as a small uterus with a small endometrial canal, poor zonal definition, and abnormal T2-hypointense myometrium.13 Whenever hypoplasia of the female genital tract is observed, it is important to describe those portions of the genital tract that are present and those that are absent, indicating the level at which the agenesis has occurred.

A unicornuate uterus, Class II, is banana-shaped. When rudimentary horns are present, they may or may not contain endometrium and may or may not communicate with the main endometrial cavity. Those containing endometrium that do not communicate with the main endometrial cavity may bleed and present as cryptomenorrhea secondary to hematometra (figure 2). Identification of a rudimentary horn and reporting its presence is important, as it is a possible site for ectopic implantation.

Uterus didelphys, Class III, when complete, appears as two separate normal-sized uteri and cervices (figure 3). Look for a horizontal vaginal septum (usually on the same side as renal agenesis, when present) in this context. When a vaginal septum is obstructive, "reflux" endometriosis may ensue. A renal evaluation should be performed in this and all congenital uterine anomalies.

A bicornuate uterus, Class IV, exhibits an abnormal, double convex fundal contour with a central cleft (>2 cm deep),* a widened intercornual distance (greater than 4 cm)** with an angle of cornual divergence greater than 75 degrees,14 and a divider between cornua comprised of either myometrium alone, or myometrium and fibrous tissue combined. In either case, you may find some soft tissue that is centrally isointense to myometrium. You will also see low T1- and T2 signal tissue if fibrous tissue is present (figures 4A and 4B).

A septate uterus, Class V, exhibits normal uterine shape and size. It has a flat or convex external uterine contour, but it possesses a fibrous septum that is of low signal on both T1- and T2WI. The septum is considered complete only if it extends down to the internal os; it is otherwise considered partial. Of note, the fundal aspect of the septum may contain some myometrium.15

Our congenital anomaly MR imaging protocol includes a coronal FMPSPGR (large FOV) localizer that includes the kidneys, axial T1WI, and sagittal, axial oblique (uterine short axis), and coronal oblique (uterine long axis) FSE T2WI. This protocol requires no intravenous contrast administration (table 2). The coronal localizer serves to screen for associated urinary tract anomalies (e.g. renal agenesis) and vertebral anomalies (of segmentation and formation).

Fibroids-Uterine leiomyomas, or fibroids, are benign, estrogen-dependent, neoplastic proliferations of smooth muscle (figure 5). They occur in approximately one in four women of childbearing age and are 3 to 9 times more common in black women, with a prevalence of approximately 50% in black women in their fifth decade.16 While often asymptomatic, fibroids are most often a cause of morbidity to women during their reproductive years; they usually regress after menopause. Common presentations include pelvic pain or pressure, menorrhagia, menometrorrhagia, and infertility. Fibroids may cause tubal narrowing, if ostial, or difficulty with regard to implantation of the fertilized ovum, if submucosal (figures 6A and 6B).

When located in the lower uterine segment, they can also interfere with normal labor and delivery. While pelvic ultrasound is often sufficient to make the diagnosis, it occasionally fails to detect very small submucosal fibroids. In addition, the technical limitations of ultrasound, with regard to its small field of view and bowel gas obscuration, often make it inadequate to evaluate the extent and location of large fibroids. MR does not share these disadvantages and also is better at distinguishing a pedunculated fibroid from a solid adnexal mass, and at defining the precise location of the fibroids as they relate to the endometrium and to the uterus in general, assets which are critical for surgical planning (especially when selective myomectomy is being considered). For these same reasons, MR is also useful for monitoring the efficacy of medical therapy with GnRH analogs regarding fibroid shrinkage.9

We use our standard pelvic protocol without gadolinium to evaluate fibroids (table 2). Fibroids are categorized as submucosal, intramural, or subserosal, based on their predominant location. The descriptor "pedunculated" should be reserved for those fibroids on a stalk. Fibroids should be characterized in reports as such, to help the referring physician determine whether the fibroid represent the true cause of the patients symptoms and, again, to assist in surgical planning; e.g. submucosal fibroids can often be removed hysteroscopically.

The classic MR appearance of fibroids is one of a well-circumscribed, round mass with low-to-intermediate T1 signal and low T2 signal (figure 5). Nevertheless, fibroids often exhibit evidence of degeneration (internal high T2 signal areas representing hyaline or cystic degeneration) and occasionally are homogeneously high in T2 signal (highly cellular fibroids).17 Carneous degeneration yields high T1 signal and low T2 signal, representing the presence of blood products. In more than one-third of cases, a high T2 signal rim encircles the fibroid. This rim has been found to represent dilated lymphatics, veins, and/or edema, all of which are likely related to vascular congestion.18 When contrast is administered, fibroids exhibit a variety of signal intensities. Most fibroids will enhance, but usually later than normal myometrium. In one study, contrast administration caused 65% of leiomyomas to be hypointense, 23% to be isointense, and 12% to be hyperintense to normal myometrium. High T2 signal (cellular) leiomyomas typically were isointense-to-hyperintense to normal myometrium upon gadolinium administration.17

Adenomyosis-Adenomyosis is a disease characterized by implantation of endometrial basalis glands and stroma in the myometrium, and is hypothesized to occur secondary to uterine trauma, which may be caused by parturition, myomectomy, or curettage.19,*** When these ectopic glands remain in continuity with the endometrial surface, they result in the classic hysterosalpingographic appearance of adenomyosis. However, some of these glands eventually are cut off from their source secondary to surrounding muscle hypertrophy and fibrosis, resulting in true ectopic islands of endometrial tissue within the myometrium. Adenomyosis can occur in a diffuse or a focal manner (figures 7A and 7B). The more focal form is referred to as an adenomyoma because the encasement of the ectopic glands by hypertrophied smooth muscle causes the adenomyosis to appear more discrete; however, an adenomyoma does not bear a pseudocapsule and, therefore, should not be as well-circumscribed as a leiomyoma.16

Adenomyosis occurs most commonly in multiparous women and is often asymptomatic (reported incidence of 5% to 70% );20 therefore, many cases are discovered incidentally. When symptomatic, it typically presents after age 30 with menorrhagia, dysmenorrhea, and pelvic pain, and abates after menopause.16 Because adenomyosis lacks the pseudocapsule of leiomyomas, it is not readily resectable, and its only definitive cure is hysterectomy. Leiomyoma therapy is currently geared toward uterine preservation. Therefore, it is incumbent upon those imaging the patient to distinguish between adenomyosis and leiomyomas.

Clinical findings of uterine enlargement and tenderness (rather than lobulated contour) may tilt the differential diagnosis towards adenomyosis; however, these are not always present. While endovaginal sonography can discern between these two entities, even in the best of hands, its sensitivity and specificity for detection of adenomyosis only reaches 87% and 98%, respectively, for focal lesions, and 80% and 74% for diffuse disease.21 MR, on the other hand, can make the distinction between adenomyosis and leiomyomas more reliably, with up to 99% accuracy.22

The MR appearance of adenomyosis is one of an abnormally thickened junctional zone containing foci of high T2 signal. The lesions typically appear ill-defined. Focal adenomyomas should exhibit an indistinct border (figure 7B), unlike small leiomyomas. The entire uterus may be enlarged. The low T2 signal of adenomyosis may represent hypertrophied smooth muscle and the focal areas of high T2 signal have been found histologically to represent embedded endometrial tissue. Those foci that are both of high T1 and T2 signal represent hemorrhage.22 When contrast is administered, adenomyosis, whether focal or diffuse, always enhances less than the adjacent myometrium.17

The degree of junctional zone thickness required for diagnosis of adenomyosis is of some controversy. Previously, a junctional zone thickness of greater than 5 mm was thought to be abnormal.23 This thickness value has recently been challenged by a report by Kang et al7 that many asymptomatic, control patients have been shown to have junctional zones greater than 5 mm, with a mean thickness of 4.4 mm. This paper states that in order to minimize false positive diagnoses, a cut-off of 7 mm to 8 mm would be more appropriate. In addition, during the MR exam, Kang et al observed many instances of transient increased thickness of the junctional zone (up to 13 mm), which they attribute to myometrial contractions. Currently, we consider a junctional zone of 12 mm or less to be abnormal, but the detection of focal high T2 signal is very helpful in confirming our suspicion.

We use our standard pelvic protocol without gadolinium to image patients with suspected adenomyosis.

Endometrial polyps-While the diagnosis of endometrial polyps is usually made by hysteroscopy or transvaginal sonohysterography, the MR appearance of an endometrial polyp should be known, as it is an important cause of postmenopausal bleeding. Menometrorrhagia is a common premenopausal presentation. Endometrial polyps have been found to be most prevalent in women in their fourth through sixth decades, particularly after age 50, and are found to be present in 10% of the autopsies performed on women.16 Polyps range in size from millimeters to centimeters and may be single or multiple.

While gadolinium has not been shown to be useful in the evaluation of most benign uterine diseases, often decreasing the conspicuity of small fibroids and adenomyosis, it is required when diagnosis of endometrial polyps is at issue because it greatly increases the conspicuity of these lesions. Without contrast, all that is generally observed when a polyp is present is endometrial thickening or lobulation. With contrast, differential enhancement of the polyp relative to endometrium and myometrium is seen, although the signal intensity of the polyp itself is extremely variable. The signal intensity may be homogenous or heterogenous, and either hypointense to endometrium and hyperintense to myometrium, or hypointense to both endometrium and myometrium (these latter polyps typically contain a dense fibrotic stroma). Both smooth and spiculated contours have been observed.17 Of note, endometrial polyps cannot be distinguished from endometrial carcinoma by any imaging modality or by hysteroscopic or gross pathologic appearance; histologic analysis is required to make this discernment.

Adnexal pathology

Polycystic ovarian disease-While a presumptive diagnosis of polycystic ovarian disease, or Stein-Leventhal syndrome, can be made by the clinical history of oligomenorrhea or amenorrhea and infertility in conjunction with bilaterally enlarged ovaries upon physical exam, there are tests that can confirm the diagnosis, including elevation of LH levels with no LH surge, normal FSH and estrogen levels, and imaging or laparoscopic evidence of ovarian enlargement. Obesity and hirsutism, while common, are not required for diagnosis,24 as there is a spectrum of clinical and radiologic findings for this disease.

While the classic sonographic appearance of bilaterally enlarged ovaries with numerous small peripheral cysts suffices to confirm the diagnosis when the above clinical findings are present, one may incidentally observe the ovarian findings by MR, which include bilateral ovarian enlargement with numerous small peripheral cysts, and broad central areas of low T1- and T2 signal representing abundant medullary stroma.25 Of note, this classic appearance is seen in under one-half of patients with the syndrome; nearly one-third of patients may have normal-sized ovaries.26 Additionally, this classic appearance has been observed in women without the syndrome, so one cannot make this diagnosis based on radiologic findings alone.

Endometriosis-A diagnosis of endometriosis is made when endometrial tissue is found external to the uterus. While often asymptomatic, it most commonly presents in women of reproductive age with chronic pelvic pain and infertility, and is rarely seen postmenopausally.27 Endometriotic cysts, or endometriomas, are usually located on the ovary.1 Endometrial implants, however, can occur throughout the abdomen and thorax.27 To date, no specific sonographic or MR imaging criteria exist to differentiate endometriomas from hemorrhagic cysts, as they are both classically of high T1 signal, complex in appearance, and without internal enhancement; however, in the proper clinical setting, the combination of lesion multiplicity and low T2-signal of the lesions, sometimes referred to as "shading," can lead to the specific diagnosis of endometriosis. Fat-saturation imaging and gadolinium administration can assist in the diagnosis of small endometriomas.

The thick, low-intensity, early enhancing wall of endometriomas results from its composition of fibrosis and hemosiderin-laden macrophages, and its vascularity (figure 8). Whether the walls of hemorrhagic cysts also enhance has not been studied. Endometriotic implants are often very difficult to identify, and may occur (as listed here in descending order of frequency) on the ovaries, cul-de-sac, posterior uterine wall, uterosacral ligaments, anterior uterine wall, and bladder dome. They are typically low T2 signal, enhancing solid masses because of the reactive fibrosis surrounding the small amount of endometrial tissue, although foci of high signal may be present within the implants secondary to hemorrhage and/or the presence of more substantial endometrial tissue.1

Mature cystic teratomas, or dermoids-Dermoids represent 10% to 15% of ovarian neoplasms26 and are comprised of all three germ layers, although the ectodermal layer tends to predominate. Malignant conversion is rare, most typically occurring in larger tumors (greater than 10 cm) and in postmenopausal women.25 Dermoids may present with pelvic pain or pressure, ovarian torsion, or as an incidentally discovered pelvic mass, and appear at any age, though they are found most commonly during the childbearing years. The sonographic and MR appearance of dermoids is extremely variable, as it is dependent on the equally variable composition of the given lesion which can contain sebum or liquid fat, hemorrhage, solid fat, hair, and teeth. The three latter contents can form a nodular mass projecting from the cyst wall, known as a dermoid plug or Rokitansky nodule. The large lipid content of these lesions can cause them to camouflage with bowel gas at sonography, and with pelvic fat at MR. When the lipid is identified, however, a specific diagnosis of mature cystic teratoma can usually be made (with the rare exception of fat-containing liposarcomas).

To aid in identification of the fat, several tricks of MR technique and interpretation can be used. First, one can look for high T1 signal within the mass (figure 9A), though this can be due to either the presence of lipids or hemorrhage. Diagnosis is aided by the frequent presence of chemical shift artifact at fat-fluid interfaces (alternate bright and dark bands in the frequency-encoding direction) that infers the presence of lipid within the lesion (figure 9B). When doubt remains, the use of fat- and/or water-saturation pulse sequences on T1WI can be confirmatory (figure 9C).25 Occasionally, however, no fat is seen within the lesion. When this occurs, careful scrutiny of the cyst wall with opposed-phase and in-phase gradient echo images may lead to identification of focal chemical shift artifacts within the cyst.28 Fluid-fluid levels and fat-fluid levels may also be seen. Because 10% to 15% of dermoids are bilateral,26 when one dermoid is found, close examination of the contralateral adnexa is mandatory.

Benign ovarian neoplasms-Because fibromas and fibrothecomas share a composition similar to leiomyomas (that of collagen-forming spindle cells), their signal characteristics are identical (intermediate T1- and low T2-signal), and they have occasionally been misinterpreted as pedunculated or broad ligament leiomyomas.1 It is also difficult to differentiate between benign cystadenomas and cystadenocarcinoma.

Differentiating between benign and malignant pelvic neoplasms-While differentiating between benign and malignant pelvic neoplasms is often impossible, several MR imaging features may help in this discernment. Stevens et al29 statistically derived five primary criteria of malignancy with the aid of dynamic gadolinium enhancement, and were correct in 84% of the time in characterizing the lesion as benign or malignant: 1) size greater than 4 cm, 2) solid mass or large solid component, 3) wall thickness greater than 3 mm, 4) septa greater than 3 mm thick and/or presence of vegetations or nodularity, and 5) necrosis. They also derived four ancillary criteria of malignancy, which, when used in combination with the five primary criteria, led to 95% of th