MRI Visualization of Bowel Endometriosis: A Pilot Study

Abstract

<h3>Study Objective</h3> To develop and validate a magnetic resonance imaging (MRI)-based nomogram to optimize the preoperative evaluation of bowel endometriosis (BE). <h3>Design</h3> A prospective cohort study. <h3>Setting</h3> University-based hospital. <h3>Patients or Participants</h3> Between 2018 and 2020, patients diagnosed as deep endometriosis were enrolled prospectively and distributed to the development and validation cohort equally. <h3>Interventions</h3> MRI parameters were extracted from T1-weighted and T2-fat saturated images. A Nomogram was developed based on potential predictors for BE by LASSO regression and validated in an external cohort to evaluate the accuracy of predicting BE by generating ROC curve and calibrating curve. <h3>Measurements and Main Results</h3> Among 220 deep endometriosis patients, 154 (70%) were diagnosed as BE. By using logistic regression analysis, thickness of rectal wall (OR=2.45, 95%CI: 1.60-3.73, <i>P</i><0.01) and traction sign (OR=5.62, 95%CI: 1.09-29.00, <i>P</i>=0.04) were integrated in the Nomogram to predict the presence of BE. Predictive accuracy with AUC value 0.94 (95%CI: 0.90-0.98, <i>P</i><0.01) was shown in the development cohort, and AUC=0.94 (95%CI: 0.89-0.98, <i>P</i><0.01) in the validation cohort. A cut-off value of 6.0mm for the thickness of rectal wall resulted in the highest predictive accuracy of BE (sensitivity: 78.6%, specificity: 90.9%, <i>P</i><0.01). <h3>Conclusion</h3> This pilot study indicates the MRI-based Nomogram is efficient for visualization and assessment of BE. The probability of BE tended to be greater in deep endometriosis patients with thickness of rectal wall over 6.0mm