Tailored compartment-based comparison of transvaginal ultrasound and MRI for suspected deep pelvic endometriosis: a prospective concordance study.

objectiveTo assess the intermodality agreement between transvaginal ultrasound (TVUS) and magnetic resonance imaging (MRI) in the evaluation of deep pelvic endometriosis (DE), using a compartment-based framework that includes both classical and extended pelvic structures. This study was not designed to assess diagnostic accuracy, as no surgical or histological reference standard was used.

methodsIn this prospective study, 79 symptomatic women with clinical suspicion of DE underwent standardized TVUS and MRI protocols within a 20-day interval. Agreement was analyzed by anatomical compartment using Cohen's Kappa coefficient. Discordant findings (TVUS-only vs. MRI-only) were quantified and tested using McNemar's test. Detection frequencies were also compared using chi-squared tests for paired proportions. In addition to anterior and posterior compartments, we included clinically relevant structures not encompassed by standard classification models. Based on agreement and discordance patterns, each structure was categorized into a descriptive suggestion tier.

resultsIntermodality agreement varied across compartments. Almost perfect agreement was observed for the rectosigmoid (κ = 0.83), and substantial agreement for the rectum (κ = 0.65), rectovaginal septum (κ = 0.68), and ovarian endometriomas (κ ≥ 0.71). Moderate agreement was seen for the uterine serosa (κ = 0.49), vagina (κ = 0.57), bladder (κ = 0.42), and vesicouterine recess (κ = 0.54). TVUS detected significantly more lesions in key regions, including the rectosigmoid (84.8% vs. 64.6%, p = 0.006) and posterior cul-de-sac (53.2% vs. 31.6%, p = 0.010), while MRI showed relative advantages for evaluating the round ligaments and uterine serosa. An exploratory suggestion model (Fig. 1) summarizes these findings.

conclusionTVUS and MRI demonstrated substantial concordance across multiple pelvic compartments, particularly when interpreted by expert operators. By extending the compartment-based framework and applying a structured analysis of agreement and detection bias, this study provides insights into the complementary roles of TVUS and MRI in non-invasive DE diagnosis. In the absence of pathological confirmation, these findings should be interpreted as hypothesis-generating and not as measures of diagnostic accuracy.