Artificial intelligence for diagnosing microvessels of precancerous lesions and superficial esophageal squamous cell carcinomas: a multicenter study

• DOI: 10.1007/s00464-022-09353-0
• Publication Date: 2022-06-15T21:02:37Z
• Authors: Xiang-Lei Yuan; Wei Liu; Yan Liu; Xian-Hui Zeng; Yi Mou; Chun-Cheng Wu; Lian-Song Ye; Yu-Hang Zhang; Long He; Jing Feng; Wan-Hong Zhang; Jun Wang; Xin Chen; Yan-Xing Hu; Kai-Hua Zhang; Bing Hu
• Keywords: Narrow-band imaging

Abstract Background Intrapapillary capillary loop (IPCL) is an important factor for predicting invasion depth of esophageal squamous cell carcinoma (ESCC). The closely related to the selection treatment strategy. However, diagnosis IPCLs complicated and subject interobserver variability. This study aimed develop artificial intelligence (AI) system predict subtypes precancerous lesions superficial ESCC. Methods Images magnifying endoscopy with narrow band imaging from three hospitals were collected retrospectively. annotated on images by expert endoscopists according Japanese Endoscopic Society classification. performance AI was evaluated using internal external validation datasets (IVD EVD) compared that 11 endoscopists. Results A total 7094 685 patients used train validate system. combined accuracy diagnosing in IVD EVD 91.3% 89.8%, respectively. achieved better than depth. ability junior diagnose (combined accuracy: 84.7% vs 78.2%, P < 0.0001) 74.4% 67.9%, significantly improved assistance. Although there no significant differences, senior slightly elevated. Conclusions proposed could improve diagnostic classification