Endoscopic Image-Based Prediction of Esophageal Stenosis after ESD Using Mucosal Defect Metrics

Stricture is a major adverse event following esophageal endoscopic submucosal dissection (ESD). Although postoperative esophageal stenosis after ESD is clearly associated with large mucosal defects, there are no quantitative criteria for the degree of the defect. We aimed to examine the predictive factors associated with the development of esophageal stenosis after ESD, and to explore quantitative indicators for predicting postoperative stenosis. A retrospective analysis of endoscopic data from patients with esophageal ESD was conducted. The area and perimeter of the resected mucosa defect were measured by Image Pro Plus 6.0. Logistic regression, receiver operating characteristic (ROC) curve, and nomogram model were adopted for analysis. The median area of resected mucosal defect was 527.74 mm2 (IQR 322.80-823.48), the median perimeter was 91.75 mm (IQR 71.69-118.32), and the median circumferential ratio was 33% (IQR 20-40). Owing to collinearity between the perimeter and area of the resected mucosal defect, these variables were analyzed individually in the multivariate analysis (r = 0.958, P < 0.001). All three factors mentioned above were identified as independent risk factors for postoperative esophageal stenosis (P < 0.05), while intraoperative muscularis propria injury was not (P > 0.05). On the basis of the above results, prediction models were constructed and validated internally. The concordance statistics (C-statistics) of the training and validation sets including perimeter or area were 0.875 and 0.830 and 0.747 and 0.835, respectively. The Hosmer-Lemeshow goodness-of-fit test results were not significant (P > 0.05). Accurate quantification of the area and perimeter of resected mucosa by image analysis technology can accurately predict esophageal stenosis after ESD.