ObjectivesTo evaluate the diagnostic accuracy of 2-dimensional (2D) and 3-dimensional (3D) transvaginal ultrasonography (US) in comparison with magnetic resonance imaging (MRI) for identification of deep infiltrating endometriosis. MethodsIn this prospective observational study, 159 premenopausal women who underwent surgery for a clinical suspicion of deep infiltrating endometriosis were prospectively enrolled. All women underwent 2DUS, 3DUS, and MRI. The following 3 locations of deep endometriosis were considered: (1) intestinal; (2) other posterior lesions (retrocervical septum, rectovaginal septum, uterosacral ligaments, and vaginal fornix); and (3) anterior. The sensitivity, specificity, positive predictive value, and negative predictive value of 2D and 3D transvaginal US in comparison with MRI were determined. ResultsIntestinal deep infiltrating endometriosis was identified by 2DUS in 56 of 66 patients, by 3DUS in 59 of 66, and by MRI in 61 of 66. A receiver operating characteristic curve analysis showed optimal results for 2DUS, 3DUS, and MRI (areas under the curve, 0.86, 0.915, and 0.935, respectively) with a statistically significant difference between 2DUS and MRI (P=.0103), even when the 95% confidence interval showed an overlap. Other posterior deep infiltrating endometriosis was identified by 2DUS in 55 of 75 patients, by 3DUS in 65 of 75, and by MRI in 66 of 75. A receiver operating characteristic curve analysis showed very good results for 2DUS, 3DUS, and MRI (areas under the curve, 0.801, 0.838, and 0.857) with no statistically significant differences. In the 12 women with deep infiltrating endometriosis in the anterior location, the nodules were correctly identified by 2DUS in 3 of 12 patients, by 3DUS in 5 of 12, and by MRI in 6 of 12. ConclusionsOur results seem to suggest that there is a statistically significant difference between 2DUS and MRI for the intestinal location of deep infiltrating endometriosis, whereas no differences were found among the techniques for the other locations.