Multi-step validation of a deep learning-based system with visual explanations for optical diagnosis of polyps with advanced features
Qing-Wei Zhang,
Zhengjie Zhang,
Jianwei Xu,
Zi-Hao Dai,
Ran Zhao,
Jian Huang,
Hong Qiu,
Zhao-Rong Tang,
Bo Niu,
Xun-Bing Zhang,
Peng-Fei Wang,
Mei Yang,
Wan-Yin Deng,
Yan-Sheng Lin,
Suncheng Xiang,
Zhi-Zheng Ge,
Dahong Qian,
Xiao-Bo Li
Affiliations
Qing-Wei Zhang
Division of Gastroenterology and Hepatology, Key Laboratory of Gastroenterology and Hepatology, Ministry of Health, Renji Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai Institute of Digestive Disease, Shanghai, China
Zhengjie Zhang
School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, China
Jianwei Xu
School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, China
Zi-Hao Dai
Division of Gastroenterology and Hepatology, Key Laboratory of Gastroenterology and Hepatology, Ministry of Health, Renji Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai Institute of Digestive Disease, Shanghai, China
Ran Zhao
Department of Gastroenterology, Guangdong Provincial People’s Hospital, Guangdong Academy of Medical Sciences, Guangzhou 510080, China
Jian Huang
Department of Gastroenterology, Yuyao People’s Hospital, Medical School of Ningbo University, Ningbo, Zhejiang Province, China
Hong Qiu
Department of Gastroenterology and Hepatology, Chongqing Hospital of Traditional Chinese Medicine, Chongqing, China
Zhao-Rong Tang
Department of Gastroenterology and Hepatology, Chongqing Hospital of Traditional Chinese Medicine, Chongqing, China
Bo Niu
Department of Digestive Endoscopy Center, Yuncheng First Hospital, Yuncheng, Shanxi Province, China
Xun-Bing Zhang
Department of Digestive Endoscopy Center, Shuguang Hospital Affiliated to Shanghai University of Traditional Chinese Medicine, Shanghai 201203, China
Peng-Fei Wang
First Division of Gastroenterology Department, Lanzhou University Second Hospital, Lanzhou, Gansu Province, China
Mei Yang
Department of Gastroenterology and Hepatology, The Third People’s Hospital of Chengdu, Chengdu, Sichuan Province, China
Wan-Yin Deng
Department of Digestive Endoscopy Center, Fujian Provincial Hospital, Shengli Clinical Medical College of Fujian Medical University, Fuzhou, Fujian, China
Yan-Sheng Lin
Department of Medicine, The University of Hong Kong-Shenzhen Hospital, Shenzhen, China
Suncheng Xiang
School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, China
Zhi-Zheng Ge
Division of Gastroenterology and Hepatology, Key Laboratory of Gastroenterology and Hepatology, Ministry of Health, Renji Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai Institute of Digestive Disease, Shanghai, China; Corresponding author
Dahong Qian
School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, China; Corresponding author
Xiao-Bo Li
Division of Gastroenterology and Hepatology, Key Laboratory of Gastroenterology and Hepatology, Ministry of Health, Renji Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai Institute of Digestive Disease, Shanghai, China; Corresponding author
Summary: Artificial intelligence (AI) has been found to assist in optical differentiation of hyperplastic and adenomatous colorectal polyps. We investigated whether AI can improve the accuracy of endoscopists’ optical diagnosis of polyps with advanced features. We introduced our AI system distinguishing polyps with advanced features with more than 0.870 of accuracy in the internal and external validation datasets. All 19 endoscopists with different levels showed significantly lower diagnostic accuracy (0.410–0.580) than the AI. Prospective randomized controlled study involving 120 endoscopists into optical diagnosis of polyps with advanced features with or without AI demonstration identified that AI improved endoscopists’ proportion of polyps with advanced features correctly sent for histological examination (0.960 versus 0.840, p < 0.001), and the proportion of polyps without advanced features resected and discarded (0.490 versus 0.380, p = 0.007). We thus developed an AI technique that significantly increases the accuracy of colorectal polyps with advanced features.
