Magnetic resonance imaging characteristics of deep endometriosis

Transcription

1 Human Reproduction vol.14 no.4 pp , 1999 Magnetic resonance imaging characteristics of deep endometriosis Karen Kinkel 1,2,4, Charles Chapron 3, Corinne Balleyguier 1, Xavier Fritel 3, Jean-Bernard Dubuisson 3 and Jean-François Moreau 1 1 Department of Radiology, Hopital Necker, 161, Rue de Sevres, Paris, Cedex 15, 2 CIERM, Hopital Bicetre, Le Kremlin- Bicetre and 3 Department of Gynecological Surgery, Clinique Universitaire Baudelocque, CHU Cochin, 123, Blvd. Port Royal, Paris, France 4 To whom correspondence should be addressed at: UCSF, Department of Radiology, Box 0628, M372, 505 Parnassus Avenue, San Francisco, CA , USA The aim of this study was to describe magnetic resonance (MR) imaging findings in histopathologically proven deep endometriosis infiltrating the uterosacral ligaments, the pouch of Douglas, the rectum or the bladder. Twenty patients presenting with a clinical suspicion of deep endometriosis underwent preoperative MR imaging. Sagittal and axial fast T2- and axial T1-weighted spin echo MR sequences were performed. Four patients had post-contrast images. MR results, including morphology and signal intensity of each lesion, were compared to intraoperative gross appearance and histopathology. Histopathology diagnosed 24 lesions of deep endometriosis in the uterosacral ligaments (n 12), the pouch of Douglas (n 2), the rectum (n 3), the bladder (n 7). Uterosacral ligaments with deep endometriosis were statistically different from normal uterosacral ligaments for proximal nodularity (P 0.001). There was no difference in signal intensity between normal and abnormal uterosacral ligaments. Contrast-enhanced SE images in four patients with detrusor invasion showed an interruption of the hypointense detrusor by the enhancing bladder endometriosis. Rectal endometriosis was missed in two of three patients and showed non-specific rectal wall thickening in one patient. It is concluded that MR imaging can diagnose deep endometriosis of uterosacral ligaments, the bladder and the pouch of Douglas, but lacks sensitivity in detecting rectal endometriosis without rectal distension. Key words: endometriosis/intracavitary coils/mri/surface coils/uterosacral ligaments Introduction Deep endometriosis is a particular form of subperitoneal endometriosis, due to ectopic endometrial tissue penetrating the peritoneum 5 mm in depth (Cornillie et al., 1990; Koninckx et al., 1991). Uterosacral ligaments are one of the most frequent locations of subperitoneal or deep endometriosis, followed by the rectum and the bladder (Cornillie et al., 1990). The preoperative diagnosis of deep peritoneal endometriosis is difficult and requires palpation and opening of the subperitoneal space in order to confirm and to evaluate the extent of the lesion (Chapron and Dubuisson, 1996; Koninckx et al., 1996). Preoperative lesion evaluation is necessary in order to plan adequate surgical management consisting of complete surgical excision (Koninckx and Martin, 1995; Chapron and Dubuisson, 1996). Transrectal ultrasonography has been proposed recently to diagnose endometriosis of uterosacral ligaments (Ohba et al., 1996). However, in that study, conclusions were only drawn on different thickness and regularity between normal uterosacral ligaments and uterosacral ligaments infiltrated by deep endometriosis, without any attempt to relate measurements to pathology. Although magnetic resonance (MR) imaging is currently used for the diagnosis of ovarian endometriosis (Togashi et al., 1991; Sugimura et al., 1992), it has not been recommended for the primary evaluation of peritoneal endometriosis. One reason is the low sensitivity of MR imaging for superficial endometrial implants and adhesions (Arrivé et al., 1989; Bis et al., 1997). Because no previous study evaluates the value of MR imaging in diagnosing deep endometriosis, including uterosacral ligaments, the purpose of our study is to describe MR findings of histopathologically proven deep endometriosis. Materials and methods Patients This study was conducted according to the guidelines of the ethical committee of the Necker Hospital (Paris, France) and was approved by the Review Board. All patients gave informed consent orally before MR examination. From January 1996 to December 1997, 40 consecutive patients underwent preoperative MR examination for clinical suspicion of deep endometriosis. The inclusion criteria were based on the association of severe pelvic pain (deep dyspareunia, premenstrual cystalgia or tenesmus) and/or an abnormal physical examination of uterosacral ligaments, the rectum or the pouch of Douglas and/or history of endometriosis. Among the 40 patients, 20 patients underwent surgery at the Clinique Universitaire Baudelocque, CHU Cochin (Paris, France), and were considered to be the study population. The remaining 20 patients were excluded because of the patient s refusal to undergo surgery (n 3), insufficient surgical and pathological information on the patients operated on at other institutions (n 6), absent insurance coverage (n 3) or absent follow-up (n 8). The women ranged in age from 21 to 40 years (with a mean of 32.3 years). History of pelvic surgery was found in 11 of 20 patients, nine for endometriosis, one for myomectomy and one for Caesarean section. The clinical symptoms, sometimes 1080 European Society of Human Reproduction and Embryology

2 MRI of deep endometriosis associated, were dyspareunia (n 11), dysmenorrhoea (n 8), sterility (n 4), urinary burning (n 5), tenesma (n 2) or indeterminate severe pelvic pain (n 5). The physical examination demonstrated either the presence of a nodule on the uterosacral ligament (n 8), a painful taut uterosacral ligament (n 2), a nodule in the pouch of Douglas (n 2) or was normal (n 8). MR imaging MR imaging was performed with a 1.5 T unit (Signa 4.5; GE Medical Systems, Milwaukee, WI, USA) using first either a body coil (n 12) or a torso coil (n 8), followed by image acquisition with an endorectal cervix receiver coil (Medrad, Pittsburgh, PA, USA). All image acquisition was monitored by a radiologist aware of the clinical and other imaging modality findings. For closer proximity to the lesions, the endorectal coil was positioned in the vagina in 17 patients and in the rectum in three patients. Axial and sagittal T2-weighted fast spin echo images [repetition time, 4000 ms; echo time, 104 ms (4000/104)] and axial T1-weighted (500/10) spin echo images were obtained with 5 mm thick contiguous sections. For images obtained with the body coil or the torso coil, two signals were acquired with a matrix and a cm field of view. For images obtained with the endorectal coil, four signals were acquired with a matrix and a cm field of view. Superior saturation pulse and no-phase wrap options were used for both body coil imaging and intracavitary coil imaging. Patients with T1 hyperintense lesions underwent a fat suppressed T1-weighted spin echo sequence. Four patients with nodules in the vesico-uterine pouch underwent contrastenhanced T1- and T2-weighted fat suppressed spin echo images. The MR images were analysed prospectively by a board certified independent reader, who was aware of the clinical suspicion of deep endometriosis, but blinded to the results of physical examination, other imaging modality results or histopathology. The reader was asked to identify any lesion of the uterosacral ligaments, the pouch of Douglas, the rectum or the vesicouterine pouch. The following parameters were recorded for each uterosacral ligament: (a) visibility on axial and/or sagittal T2-weighted and/or axial T1-weighted images; (b) morphology: presence of a nodule, size and interface to adjacent organs; and (c) signal intensity on T2-weighted fast spin echo and T1-weighted spin echo images in reference to the myometrium. Morphology (shape, size and margins), signal intensity on unenhanced T1- and T2-weighted images and homogeneity were recorded for all other lesions. Possible extension of the lesion into the bladder detrusor, the ureters or the muscularis propria of the rectum was noted and compared to results of surgery and histopathology. The presence of endometriomas was noted for each patient. No attempt was made to diagnose superficial implants or adhesions. MR imaging findings were compared to clinical results, findings of other imaging modalities, such as transvaginal ultrasound, rectal echoendoscopy, cystoscopy and intravenous pyelography (IVP), when available, to the video of intraoperative gross appearance and the results of histopathology. Lesion classification Uterosacral ligaments were considered normal when physical examination and laparoscopic inspection and palpation were normal. Histopathological criteria of deep endometriosis were the presence of both endometriotic glands and stromal tissue infiltrating the peritoneum 5 mm in depth. Lesions in the vesicouterine pouch extending into the bladder detrusor were classified as endometriosis of the bladder. Statistical analysis Morphology and signal intensity on T2-weighted images were compared between normal uterosacral ligaments and uterosacral ligaments infiltrated by deep endometriosis. The statistical analysis used the χ 2 two-tailed test in order to evaluate if abnormal uterosacral ligaments were different from normal uterosacral ligaments. A P value of 0.05 was considered to indicate a statistically significant difference. Results Surgery and histopathology diagnosed unilateral (n 8) and bilateral (n 2) lesions of deep endometriosis of uterosacral ligaments (n 12), of the pouch of Douglas (n 2), the bladder (n 7) or the rectum (n 3) in 20 patients. Other lesions of endometriosis included ovarian endometriomas (n 10), superficial peritoneal implants (n 11) and adhesions (n 15). During laparoscopy, endometriosis was scored according to the revised American Fertility Society classification (American Fertility Society, 1985). Patients were distributed according to the classification as: stage 1 (minimal, n 3), stage 2 (mild, n 2), stage 3 (moderate, n 6) and stage 4 (severe, n 9). Surgery consisted of resection of endometriomas (n 10), uterosacral ligaments (n 12), subperitoneal nodules of the pouch of Douglas (n 2), in partial cystectomy (n 7) and in partial colo-rectal resection (n 5). Lesions of uterosacral ligaments (n 12) Laparoscopic palpation diagnosed 28 normal and 12 abnormal uterosacral ligaments. Histopathology confirmed infiltration by deep endometriosis in all 12 removed uterosacral ligaments. T2-weighted fast spin echo images identified 86% (24/ 28) of normal uterosacral ligaments and 100% (12/12) of uterosacral ligaments with deep endometriosis. A markedly lateral positioned cervix (n 2) or indistinct margins between the rectosigmoid and the cervix (n 2) prevented us from identifying four normal uterosacral ligaments. T1-weighted SE images identified 25% (7/28) of normal uterosacral ligaments and 50% (6/12) of uterosacral ligaments with deep endometriosis; all were isointense to myometrium. Non-visualization of uterosacral ligaments on T1-weighted images was due to the presence of surrounding pelvic organs (small bowel, rectum) isointense to the uterus. Uterosacral ligaments with deep endometriosis were irregular due to the presence of a proximal nodule measuring between 10 and 18 mm in largest diameter (mean 13.4 mm) (Figures 1 3). T2-signal intensity of the nodule was iso- (6/12), hypo- (5/12) or hyperintense to the myometrium (1/12); the nodule was isointense (6/6) when identified on T1-weighted SE imaging. Pathology showed abundant fibrosis with small clusters of endometriotic glandular tissue in uterosacral ligaments with T2 hypo- or isointensity. The T2 hyperintense uterosacral ligaments correlated with multiple clusters of endometriotic glandular tissue and minimal stromal reaction on histopathology. Normal uterosacral ligaments were regular (24/24) and iso- (14/24) or hypointense (10/24) to the myometrium. None of the normal uterosacral ligaments was irregular or presented with a nodule (Figures 1A, B and 4A). Comparison between uterosacral ligaments with and without deep endometriosis showed significant differences in nodularity 1081

3 K.Kinkel et al. Figure 1. Deep endometriosis of the right uterosacral ligament (USL) extending to the rectal wall muscles in a 32 year old woman. (A) Axial T2-weighted image obtained with the body coil shows a hypointense nodule on the proximal part of the right USL (arrow) and a normal regular left USL (arrowheads). (B) Axial T2- weighted image obtained with an endorectal cervix coil demonstrates an enlarged right uterosacral ligament (arrow) with a proximal nodule (arrowhead) extending up to the rectum (small arrow). Figure 2. Deep endometriosis of the right (USL) in a 33 year old woman. (A) Axial T2-weighted image obtained with the body coil and (B) axial T2-weighted image obtained with an endorectal cervix coil in endovaginal position (small arrow) show a hypointense nodule (arrow) on the right USL close to the rectal wall (arrowhead). (P 0.001) and no differences in T2 signal intensity. Irregularity with presence of a proximal nodule on a uterosacral ligament was highly indicative of deep endometriosis. Two lesions of uterosacral ligaments extended to the rectal muscularis propria. The normal hyposignal of the rectal muscularis propria was visible only in patients presenting a distended rectum, as it occurred incidentally by the intrarectal position of the receiver coil. Deep rectal wall invasion corresponded to an interruption of the hyposignal of the rectal muscular propria on T2-weighted spin echo images (Figure 1b). Extension to the rectum could not be analysed in patients with an endovaginal position of the coil due to absent rectal distension. Comparison between MR results and physical examination showed a lower detection rate of abnormal uterosacral ligaments on physical examination (10/12) than on T2-weighted 1082 Figure 3. Deep endometriosis of the left (USL) in a 39 year old woman. Axial T2-weighted image shows proximal thickening of the left USL (arrowheads) associated with a hypointense nodule on the left side of the ligament (arrow). imaging (12/12). Two abnormal uterosacral ligaments were missed clinically in two patients presenting with bilateral nodules of deep endometriosis in each uterosacral ligament. None of the normal uterosacral ligaments presented with a nodule either at MR imaging or at physical examination.

4 MRI of deep endometriosis Lesions of the rectum (n 3) Two patients complaining about tenesma presented with three nodules of deep endometriosis of the rectal wall at pathology. Preoperative diagnoses had been performed by transrectal ultrasound and confirmed by surgery and histopathology. MR imaging prospectively identified one of the three lesions. This lesion was positioned 8 cm above the anus and measured 15 mm in largest diameter. MR images showed asymmetric T2 isointense thickening of the rectal wall at 8 cm from the anus. In one patient with two rectal lesions, measuring 10 and 20 mm and located in the rectosigmoid segment, at 12 and 15 cm respectively from the anus, motility artefacts impaired accurate assessment of the rectal wall. MR images obtained with the endorectal coil could not provide interpretable images above 8 cm from the anus in both patients. Positioning of the intrarectal coil higher in the rectum, above the level of the cervix, was limited by the patient s symptoms. Lesions of the pouch of Douglas (n 2) Two patients presented with a nodule in the pouch of Douglas, measuring 15 and 18 mm in largest diameter and extending Figure 4. Deep endometriosis of the pouch of Douglas in a 30 year old woman. Sagittal T2-weighted image shows a hypointense nodule on the posterior wall of the uterus extending up to the rectal wall (arrow). Table I. Clinical and imaging findings in seven patients with histopathologically proven bladder endometriosis either into the rectal or into the rectovaginal wall. On T2- weighted fast spin echo images the signal intensity of deep endometriosis of the pouch of Douglas was isointense to myometrium and homogeneous in one patient (Figure 4), and isointense and heterogeneous due to hyperintense spots in the other patient. Signal intensity on T1-weighted images was isointense and homogenous in both patients. Both lesions were suspected at physical examination. Lesions of the bladder (n 7) The size of the seven lesions ranged from 10 to 40 mm (mean 26 mm) (Table I). The margins were either irregular (n 6) or smooth (n 1). All lesions involved the posterior bladder wall; three were located on the left, two in the middle and one on the right of the vesico-uterine pouch. In a full bladder, the posterior limit of the lesion was projected at the level of the isthmus uteri (n 5) or the cervix (n 2). In five of seven patients the lesion protruded into the bladder lumen, mimicking bladder cancer (Figure 5). In two other patients, localized thickening of the posterior bladder wall was the sole morphological abnormality (Figure 6). On T2-weighted fast SE images the signal intensity of bladder endometriosis was isointense and heterogeneous due to hyperintense spots in five lesions and isointense and homogeneous in two lesions. On T1-weighted non-enhanced SE images, the signal intensity was isointense and homogeneous in six lesions and isointense and heterogeneous due to hyperintense spots in one lesion. Among the four patients with contrast-enhanced images, the degree of contrast-enhancement of bladder endometriosis was higher than the normal detrusor. Lesion extension into the detrusor could not be assessed on precontrast images because of isointensity of the lesion to the detrusor. Contrast-enhanced SE images, obtained with an endovaginal coil, were obtained in four patients with detrusor invasion and showed an interruption of the hypointense detrusor by the enhancing nodule of endometriosis (Figures 7B, 8). In two patients the lesion extended to the distal ureter. Invasion of the lower ureteral junction was suspected by unilateral dilatation of the ureter in one of two patients. Patient A B C D E F G Urinary symptoms Yes Yes Yes No No Yes Yes Lesion size Other lesions Ovary USL Tubes Adhesions Implants Large ligament Ovary Ultrasound BT Myoma BT BT Myoma BT Not performed Cystoscopy Negative Not performed Positive Negative Not performed Positive Negative MRI Lesion location Mid-VIP Left VUP Left p BW Left VUP Mid p BW Right p BW Mid p BW Lesion margin Irregular Irregular Irregular Irregular Irregular Irregular Smooth Lesion T2 signal Heterogeneous Heterogeneous Heterogeneous Heterogeneous Heterogeneous Heterogeneous Homogenous Lesion T1 signal Homogeneous Homogeneous Homogeneous Homogeneous Heterogeneous Homogeneous Homogeneous Lesion enhancement after i.v. contrast Yes * * Yes Yes * Yes Detrusor invasion Yes * * Yes Yes * Yes Ureteral invasion No No No Yes No No Not seen IVP Ureteral invasion Not performed None None Yes None None Yes MRI magnetic resonance imaging; USL uterosacral ligament; BT bladder tumour; VUP vesico-uterine pouch; p post; BW bladder wall; IVP intravenous pyelography; * unable to assess due to absent contrast-enhanced images. 1083

5 K.Kinkel et al. Figure 5. Bladder endometriosis in a 33 year old woman. Sagittal T2-weighted image shows a heterogeneous mass in the posterior bladder wall (arrow) protruding into the bladder lumen and containing hyperintense spots. A thin hyperintense fat plan (white open arrow) separates the mass from the uterus. Figure 6. Bladder endometriosis in a 33 year old woman. Sagittal T2-weighted image shows thickening of the posterior bladder wall obliterating the vesicouterine pouch. The mass contains some small hyperintense spots (arrow). Urinary symptoms were present in five of seven patients and consisted of permanent (n 1) or periodic (n 4) urinary burning during menses. Two patients had no urinary complaints and presented with a history of endometriosis and amenorrhoea. Six of the seven patients underwent preoperative transvaginal ultrasound misdiagnosing a subserosal anterior fibroid of the uterus (n 2) or a bladder tumour (n 4). Preoperative percystoscopic biopsy was performed during menses in five patients; results were negative in three patients and endometriosis was diagnosed in two patients. Intravenous pyelography correctly diagnosed ureteral involvement in both patients with non-specific distal ureteral obstruction, whereas MR imaging missed one of the two ureteral obstructions. Discussion Deep endometriosis is a poorly known form of subperitoneal endometriosis responsible for severe pelvic pain (Koninckx et al., 1991). Uterosacral ligaments are frequently involved (Cornillie et al., 1990). The natural history of such an infiltration is still unknown, but the depth of the lesion increases with time and is correlated to the intensity of pain (Koninckx 1084 Figure 7. Bladder endometriosis in a 40 year old woman. (A) Sagittal T2-weighted image obtained with the body coil shows a heterogeneous mass between the cervix and the posterior bladder wall (arrow). (B) The T2-weighted fat suppressed axial image, obtained with an endovaginal coil (open arrow), demonstrates a mass between the anterior vaginal fornix and the bladder wall (large arrow). The hypointense line in the bladder wall corresponds to the detrusor (small arrow) and is interrupted by the mass. Figure 8. Bladder endometriosis in a 25 year old woman. T1- weighted contrast-enhanced fat suppressed image, obtained with an endovaginal coil shows a small mass in the posterior bladder wall (arrow). The hypointense line corresponding to the detrusor (arrowheads) is interrupted. et al., 1991). The treatment of deep endometriosis is complete surgical excision (Koninckx et al., 1995; Chapron and Dubuisson, 1996). Indeed interruption of the medical treatment is associated with a high risk of recurrence (Waller and Shaw,

6 MRI of deep endometriosis 1993). The results of the surgical treatment depend on the quality of complete surgical excision and require therefore a preoperative diagnosis of the extent of the lesion. The diagnosis of endometriosis infiltrating uterosacral ligaments is difficult due to the subperitoneal position of the ligaments. Physical examination can demonstrate a painful uterosacral ligament during menses (Koninckx et al., 1995; Chapron and Dubuisson, 1996). Our study showed that a negative physical examination cannot exclude deep endometriosis as shown in two of 10 patients with infiltrated uterosacral ligaments, in five of seven patients with bladder endometriosis and in one of two patients with rectal endometriosis. Deep endometriosis is located predominantly in the subperitoneal space and might not be visible in the peritoneum during laparoscopy. Although an experienced surgeon can detect nodules during laparoscopic palpation of the ligaments (Chapron and Dubuisson, 1996), the lesion can be hidden by peritoneal adhesions of the pouch of Douglas, which is frequently involved by endometriosis. Our study demonstrated that MR imaging was able to detect abnormal uterosacral ligaments. The sensitivity of lesion detection was significantly higher with T2-weighted images (100%) than with T1-weighted images (50%). Although the sensitivity for normal uterosacral ligament detection was low, 25 and 86% respectively for T1- and T2-weighted imaging, the purpose of our study was not to detect all uterosacral ligaments, but specifically those that will require surgical excision. Our findings do not agree with the study of Arrivé et al. (1989), who reported that endometrial implants on uterosacral ligaments were not visible on MR imaging. These differences may be due to more severe endometriosis in our study populations and a thinner section thickness of 5 versus 10 mm on T2-weighted spin echo MR images. Low T2- signal intensity correlated with abundant stromal fibrosis on pathology. Fibrosis is known to be hypointense to muscle on T2-weighted images (Ebner et al., 1988) and might be due to a long evolution of endometriosis. Our results confirm the study of Siegelman et al. (1994) describing eight solid pelvic masses caused by endometriosis as being predominantly hypointense on T2-weighted images and of intermediate signal intensity on T1-weighted images. In order to differentiate those nodules from the surrounding isointense cervical stroma or myometrium, morphological features of abnormality become a more important diagnostic argument. Our results further demonstrated that MR imaging could reliably differentiate normal from abnormal uterosacral ligaments. The comparative analysis between normal and infiltrated uterosacral ligaments showed a significant difference in the proximal morphology of the ligaments. Deep endometriosis infiltrating uterosacral ligaments differed from normal uterosacral ligaments by the presence of a proximal nodule 9 mm. Signal intensity was of no use in differentiating normal from infiltrated uterosacral ligaments as iso- or hypointense signal intensity was found in both groups. If we had applied the diagnostic criteria of nodularity of uterosacral ligaments to our study population, we could have yielded a maximum sensitivity and specificity of 100% for deep endometriosis of uterosacral ligaments. Our results need to be interpreted in the context of clinical suspicion of deep endometriosis occurring more frequently in patients with previously diagnosed and treated endometriosis (Chapron and Dubuisson, 1996). Indeed our high sensitivity of 100% for depiction of deep endometriosis in uterosacral ligaments is most likely overestimated due to the 30% prevalence of this type of lesion in our study population. Nevertheless, the lower sensitivity of physical examination, the difficulty of prelaparoscopic palpation and the fact that results of surgery depend on total excision of the lesion, emphasize the usefulness of MR imaging before surgery. Further prospective studies are necessary in order to confirm the excellent diagnostic results of MR imaging in deep endometriosis of uterosacral ligaments. In the patients with rectal endometriosis or with deep endometriosis in uterosacral ligaments extending to the rectal wall, a more aggressive surgery with partial resection of the rectum is required (Coronado et al., 1990; Nezhat et al., 1992). Staging of deep rectal involvement is therefore helpful for patient management. Our results show a low sensitivity of 33% for rectal lesion invasion by our MR imaging technique. In women with rectal endometriosis, the mobility of the endorectal coil is limited by the clinical symptoms, therefore preventing imaging levels 8 cm from the anal border. The endovaginal position of the coil was shown a higher tolerance by the patient, but did not distend the rectal wall and could not image levels above the cervix. Rectal contrast administration distending the rectal wall and use of combined pelvic phase array and endovaginal coils might represent a better technical option for detailed anatomic information on the rectal wall. Nevertheless, endorectal ultrasound combines good results, high feasibility and lower costs than MR imaging, and seems to represent a more available technique for rectal wall extension than MR imaging (Ohba et al., 1996; Schroder et al., 1997). Our one-third rate of bladder endometriosis in a population with suspicion of deep endometriosis was higher than that reported in women with all types of endometriosis. Our results demonstrated that bladder endometriosis can be diagnosed on MR imaging by morphologic abnormalities, including localized or diffuse bladder wall thickening and signal intensity abnormalities. On T2-weighted images, 71% of the patients with bladder endometriosis presented with spots of high signal intensity in an abnormal thickening of the bladder wall. Our study did not address whether these signs allow differentiation from bladder cancer which usually does not occur in young women. Contrast-enhanced MR images, obtained with an endovaginal coil, allowed precise staging of bladder endometriosis by demonstrating invasion of the detrusor by an extrinsic bladder mass. Indirect diagnostic arguments in favour of bladder endometriosis were the association with ovarian endometriomas and/or other lesions of endometriosis in 71% of our patients. MR imaging of the bladder was abnormal even in the patients with normal cystoscopy results or without urinary symptoms. The high false negative rate of cystoscopy in bladder endometriosis has been reported in other studies (Savoca et al., 1996; Vercellini et al., 1996). As endometriosis seldom invades the mucosa, lesion identification remains difficult but might be optimized by performing cystoscopy during menstruation. Indeed in two of our patients with bladder endometriosis and negative initial cystoscopy, results were 1085

7 K.Kinkel et al. positive at a subsequent cystoscopy during menstruation, performed after the MR diagnosis and only 2 3 weeks after the first cystoscopy. Compared to transvaginal ultrasound, MR imaging allowed more accurate localization of the nodule in the bladder wall and ruled out the diagnosis of subserosal uterine fibroid. Our results did not confirm recently reported superiority of transvaginal ultrasound over MR imaging in assessing bladder endometriosis (Fedele et al., 1997). Technical differences, such as intravenous injection of contrast media and endovaginal receiver coils, might explain differences in MR results between studies. Intravenous pyelography was more sensitive than MR imaging in diagnosing extension of the bladder mass to the distal ureter. The different surgical treatment options for bladder endometriosis include laparoscopic or laparotomic partial bladder resection (Dubuisson et al., 1994; Nezhat et al., 1996) or urological endoresection during cystoscopy. Preoperative knowledge of the accurate extent of bladder endometriosis guides patient orientation to a trained referral centre. Although MR imaging is not specific for bladder endometriosis, it can detect morphological and signal abnormality of the bladder highly suggestive of endometriosis in patients that might present with other sites of extraperitoneal involvement. The advantage of MR imaging combined with intravenous pyelography is the accurate evaluation of most extraperitoneal sites of involvement, contents of a pelvic mass and lesions hidden by dense adhesions. If bladder or rectal endometriosis is suspected, the endovaginal MR imaging protocol should include intravenous contrast for lesions in the bladder and rectal contrast for lesions in the rectum. Technically, optimal and individually adapted pelvic MR imaging might contribute to better patient management. In conclusion, MR imaging can diagnose deep endometriosis of uterosacral ligaments by demonstrating a proximal nodule. Staging of possible urinary or rectal endometriosis requires the use of intravenous or intrarectal contrast media for optimal MR imaging results. In those circumstances, preoperative MR imaging can guide laparoscopic treatment planning and might improve patient management. References American Fertility Society (1985) Revised American Fertility Society classification of endometriosis: Fertil. Steril., 43, Arrivé, L., Hricak, H. and Martin, M.C. (1989) Pelvic endometriosis: MR imaging. Radiology, 171, Bis, K.G., Vrachliotis, T.G., Agrawal, R. et al. (1997) Pelvic endometriosis: MR imaging spectrum with laparoscopic correlation and diagnostic pitfalls. Radiographics, 17, Chapron, C. and Dubuisson, J.B. (1996) Laparoscopic treatment of deep endometriosis located on the uterosacral ligaments. Hum. Reprod., 11, Cornillie, F.J., Oosterlynck, D., Lauweryns, J.M. and Koninckx, P.R. (1990) Deeply infiltrating pelvic endometriosis: histology and clinical significance. Fertil. Steril., 53, Coronado, C., Franklin, R.R., Lotze, E.C. et al. (1990) Surgical treatment of symptomatic colorectal endometriosis. Fertil. Steril., 53, Dubuisson, J.B., Chapron, C., Aubriot, F.X. et al. (1994) Pregnancy after laparoscopic partial cystectomy for bladder endometriosis. Hum. Reprod., 9, Ebner, F., Kressel, H.Y. and Mintz, M.C. (1988) Tumor recurrence versus fibrosis in the female pelvis: differentiation with MR imaging at 1.5 T. Radiology, 166, Fedele, L., Bianchi, S., Raffaelli, R. and Portuese, A. (1997) Pre-operative assessment of bladder endometriosis. Hum. Reprod., 12, Koninckx, P.R., Meuleman, C., Demeyere, S. et al. (1991) Suggestive evidence that pelvic endometriosis is a progressive disease, whereas deeply infiltrating endometriosis is associated with pelvic pain [see comments]. Fertil. Steril., 55, Koninckx, P.R., Martin D.C., and Shaw, R.D. (1995) Surgical treatment of deeply infiltrating endometriosis. In Shaw, R.D., Endometriosis. Current understanding and management. 1st edn, Blackwell Science Ltd, Oxford, pp Koninckx, P.R., Meuleman, C., Oosterlynck, D. and Cornillie, F.J. (1996) Diagnosis of deep endometriosis by clinical examination during menstruation and plasma CA-125 concentration. Fertil. Steril., 65, Nezhat, C., Nezhat, F., Nezhat, C.H. et al. (1996) Urinary tract endometriosis treated by laparoscopy. Fertil. Steril., 66, Nezhat, F., Nezhat, C., Pennington, E. and Ambroze, W. Jr (1992) Laparoscopic segmental resection for infiltrating endometriosis of the rectosigmoid colon: a preliminary report. Surg. Laparosc. Endosc., 2, Ohba, T., Mizutani, H., Maeda, T. et al. (1996) Evaluation of endometriosis in uterosacral ligaments by transrectal ultrasonography. Hum. Reprod., 11, Savoca, G., Trombetta, C., Troiano, L. et al. (1996) [Echographic, MRI and CT features in a case of bladder endometriosis]. Arch. Ital. Urol. Androl., 68, Schroder, J., Lohnert, M., Doniec, J.M. and Dohrmann, P. (1997) Endoluminal ultrasound diagnosis and operative management of rectal endometriosis. Dis. Colon Rectum, 40, Siegelman, E.S., Outwater, E., Wang, T. and Mitchell, D.G. (1994) Solid pelvic masses caused by endometriosis: MR imaging features. Am. J. Roentgenol., 163, Sugimura, K., Takemori, M., Sugiura, M. et al. (1992) The value of magnetic resonance relaxation time in staging ovarian endometrial cysts. Br. J. Radiol., 65, Togashi, K., Nishimura, K., Kimura, I. et al. (1991) Endometrial cysts: diagnosis with MR imaging. Radiology, 180, Vercellini, P., Meschia, M., De Giorgi, O. et al. (1996) Bladder detrusor endometriosis: clinical and pathogenetic implications. J. Urol., 155, Waller, K.G. and Shaw, R.W. (1993) Gonadotropin-releasing hormone analogues for the treatment of endometriosis: long-term follow-up. Fertil. Steril., 59, Received on May 12, 1998; accepted on December 21, 1998