Performance of MRI for the detection of anterior pelvic endometriotic lesions

To study the performance of MRI for the detection of anterior pelvic endometriotic lesions.

We carried out a retrospective, single site, case-controlled study of patients who underwent surgery for endometriosis between March 2005 and December 2013. Laparoscopy was used to obtain the reference diagnosis of the endometriotic lesions. We age-matched patients with anterior endometriosis with those with isolated posterior endometriosis for reference. All of the pre-therapeutic MRI data were anonymized and blindly reread by two radiologists (junior and senior) twice. They assessed the overall presence of anterior lesions, those of the inter-vesicouterine space, the detrusor, the uterus, and posterior lesions. For each site, we calculated the sensitivity (Se) and specificity (Sp) for each reading compared to the reference, as well as the inter- and intra-operator variability using the Kappa coefficient (K) with its 95% confidence interval (95% CI).

During the study period, 256 patients underwent surgery for endometriosis: 22 presented with anterior endometriosis at surgery, and had had pre-preoperative imagery. We included 22 controls who had an isolated posterior lesion. For the overall detection of anterior lesions, the two radiologists had an identical Se of 77.3% (95% CI; 54.6-92.2). The Sp was 100% (95% CI; 82.4-100) for the junior, and 89.5% (95% CI; 66.9-98.7) for the senior radiologist. The area under the ROC curve was 0.89 (95% CI; 0.80-0.98) for the junior and 0.81 (95% CI; 0.68-0.93) for the senior radiologist. The intra-operator variability was low with almost perfect reproducibility for the overall detection of anterior lesions; k=0.90 (95% CI; 0.77-1) for the junior and k=0.85 (95% CI; 0.70-1) for the senior radiologist. For the various anterior sites, the junior radiologist had k values between 0.60 and 1, whereas those of the senior radiologist were between 0.87 and 1. For inter-operator reproducibility, there was modest agreement between the two radiologists, k=0.46 (95% CI; 0.19-0.73), for the overall detection of anterior lesions; k varied between 0.43 and 0.61, depending on the site.

Our results show that the characterization of anterior lesions by MRI is specific, but not very sensitive, with only moderate inter-operator reproducibility depending on the site. MRI can be used to diagnose anterior lesions, but cannot replace laparoscopy.