Pelvic MRI in the diagnosis and staging of pelvic endometriosis: added value of structured reporting and expertise

Purpose To compare the diagnostic characteristics of routine-read (RR), structured-reported read (SR), and structured expert-read pelvic (SER) MRI for staging of pelvic endometriosis in a tertiary care academic center. Methods Of 530 patients with endometriosis (2013-2018), 59/530 (11.1%) were staged surgically and underwent pelvic MRI. Radiology reports were considered RR; MRI studies were independently reassessed by SR and SER. Involvement was recorded by compartment [anterior (AC), middle (MC), posterior (PC), adnexal (AX), and other (OC)]. Diagnostic discrepancy between review methods was assessed with McNemar's test. Interobserver agreement was assessed using Cohen's unweighted kappa. Results Of 295 compartments in 59 women (mean age = 38.8 years; range 20-69), 147/295 (49.8%) had confirmed endometriosis. Overall sensitivity: RR = 42.9%; SR = 86.4%; SER = 74.2%. SR's increased sensitivity was significant for PC (p < 0.001), MC (p < 0.001), AC (p = 0.001), AX (p = 0.038). Higher sensitivity by SER was significant for PC (p < 0.001), MC (p = 0.004) and AC (p < 0.001), but not AX (p > 0.05). Overall specificity: RR = 95.3%; SR = 45.9%; SER = 81.8%. SER specificity was no different than RR for PC or AX (p > 0.5). RR sensitivity relied heavily on detection of AX involvement, whereas SR and SER showed additional sites of disease while maintaining comparable specificity for SER. Overall agreement between SR and SER was fair [k = 0.342 (95% CI 0.25, 0.44)]. Conclusions Even at a tertiary care academic center, SER outperforms both SR and RR in the assessment of pelvic endometriosis. Although lack of expertise may negatively impact specificity, use of structured reporting is significantly more sensitive than RR. Therefore, its use can be of assistance in surgical planning and patient counseling.