基于人工智能深度学习算法辅助诊断早期ESCC的研究  

Deep learning algorithm based on artificial intelligence to assist the diagnosis of early ESCC

作　　者：王娜 温静[3] 冯佳 卢娜利 刘翠华[3] 智佳 王子阳 黄锦 WANG Na;WEN Jing;FENG Jia;LU Nali;LIU Cuihua;ZHI Jia;WANG Ziyang;HUANG Jin(Graduate School of Xinxiang Medical University,Xinxiang 453003;Department of Gastroenterology,the 988th Hospital of PLA Joint Logistics Support Force;Department of Gastroenterology and Hepatology,the 984th Hospital of PLA Joint Logistics Support Force;Department of Gastroenterology,the 980th Hospital of PLA Joint Logistics Support Force,China)

机构地区：[1]新乡医学院研究生院,河南新乡453003 [2]中国人民解放军联勤保障部队第九八八医院消化内科 [3]中国人民解放军联勤保障部队第九八四医院消化内科 [4]中国人民解放军联勤保障部队第九八〇医院消化内科

出　　处：《胃肠病学和肝病学杂志》2024年第2期156-161,共6页Chinese Journal of Gastroenterology and Hepatology

基　　金：河南省医学科技攻关计划科研项目(LHGJ20190881)。

摘　　要：目的探讨基于人工智能(artificial intelligence,AI)深度学习算法的内镜识别系统在胃镜诊疗过程中对早期ESCC检出率的研究。方法选取中国人民解放军联勤保障部队第九八八医院、中国人民解放军联勤保障部队第九八四医院及中国人民解放军联勤保障部队第九八〇医院三个消化内镜中心2018年6月至2020年6月早期ESCC、ESCC、食管隆起性病变以及食管憩室的白光图像、碘染色图像。通过训练和验证不同的目标检测模型和实例分割模型,最终选取表现最优的目标检测模型Yolov 5和实例分割模型Yolact++共同构建AI“嵌合模型”,评估该模型诊断早期ESCC的性能。结果AI“嵌合模型”对早期ESCC诊断的敏感度为95.60%,特异度为91.60%,准确率为90.70%,均优于单模型。结论本研究构建的AI“嵌合模型”可显著提高早期ESCC的检出率。Objective To establish an artificial intelligence-aided diagnosis model to improve the detection rate of early ESCC.[WTHZ]Methods White light images,iodine staining images and complete videos of early ESCC,ESCC,esophageal protuberant lesions and esophageal diverticulum were selected from 3 digestive endoscopic centers of the 988th Hospital of PLA Joint Logistics Support Force,the 984th Hospital of PLA Joint Logistics Support Force,and the 980th Hospital of PLA Joint Logistics Support Force from Jun.2018 to Jun.2020.The lesions in the picture were marked with rectangles and polygons,which were divided into training set,verification set and test set.Through training and verifying different target detection models and case segmentation models,the best target detection model Yolov 5 and case segmentation model Yolact++were selected to construct AI″chimera model″.Finally,the performance of AI″chimeric model″in the diagnosis of early ESCC was evaluated.[WTHZ]Results The sensitivity,specificity and accuracy of AI″chimeric model″in the diagnosis of early ESCC were 95.60%,91.60%and 90.70%,respectively.[WTHZ]Conclusion The AI″chimeric model″constructed in this study can significantly improve the detection rate of early ESCC.

关 键 词：人工智能 深度学习 实例分割 食管鳞状细胞癌 碘染色 

分 类 号：R735.1[医药卫生—肿瘤]