Establishment and Prospective Validation of a Classification System for Surface Microvessels in Peritoneal Lesions...

Background: Peritoneal nodules require abdominal exploration and biopsy. However, blind biopsies may yield false-negative results. We developed a magnifying endoscopy-based microvascular classification for peritoneal lesions, prospectively validated its pathological predictive accuracy, and explored a precision-targeted biopsy strategy. Methods: This two-phase study was conducted at three hospitals from June 2016 to May 2022 (retrospective classification development) and four hospitals from June 2022 to February 2023 (prospective validation). Two experienced endoscopists first extracted features independently under blinded conditions and then constructed the microvascular classification (Types 1-3D) through an unblinded iterative process. Retrospective patients undergoing white-light endoscopic exploration served as non-randomized controls. Diagnostic accuracy, biopsy number, and histological diagnostic yield were compared. Results: This retrospective cohort included 56 patients with 77 peritoneal lesions. Of the 29 prospectively enrolled patients, 25 patients with 76 peritoneal lesions were included in the final validation. Three out-of-scope cases were reported separately. At the patient level, magnifying endoscopy showed significantly higher accuracy in predicting lesion pathology than white-light endoscopy (83.93% vs. 46.00%, p < 0.001), with fewer biopsies (p = 0.002) and a higher histological diagnostic rate (p < 0.001). Prospectively, the classification accuracy was 96.05% for tuberculosis (Type 2C), 96.05% for tumor-like lesions (Type 3A), 96.05% for mesothelioma (Type 3B), 97.37% for Müllerian epithelial tumors (Type 3C), and 90.79% for metastatic carcinoma (Type 3D). Conclusions: The microvascular classification showed promising preliminary accuracy for predicting peritoneal lesion pathology and may support precision-targeted biopsy. External validation and formal reproducibility assessment are required before clinical implementation.