| Background and Objective:Background:The detection rate of colorectal polyps is getting higher and higher.Different pathological properties of polyps affect the formulation of colorectal cancer screening strategies.However,due to the lag of pathological results and the lack of pathologists in clinical practice,follow-up strategy formulation is not timely and individualized.Objective:1.Explore the accuracy of Kudo pit pattern classification for judging the pathological nature of colorectal polyps under the white light endoscopy(WLE);2.To explore which factor of polyp size or polyp shape has more influence on the selection of endoscopic treatment for polyps;3.To explore the feasibility of optical coherence tomography(OCT)combined with artificial intelligence(AI)in real-time diagnosis of polyp properties.Subjects and methods:Part Ⅰ:The sensitivity,specificity,Youden index and Kappa value of the diagnostic test were calculated by retrospectively analyzing 192 polyps classified by Kudo pit pattern.According to the years of use of endoscopy,the examining doctors were divided into three groups:high seniority group,medium seniority group and low seniority group,and the above indexes of each group were calculated respectively.Part Ⅱ:Retrospective analysis was conducted on 386 polyps treated under endoscopic treatment.The size and shape of polyps were taken as independent variables,and endoscopic treatment method was taken as dependent variable.After the assignment,disordered multi-classification Logistics regression analysis was conducted to explore the influencing factors of endoscopic treatment method selection for polyps.Part Ⅲ:Collect colorectal polyps of different pathological types in a certain proportion,and conduct 3D reconstruction after OCT imaging of in vitro polyps.After reconstruction,3D data 7:3 was divided into training set and verification set to build AI model.Results:Part Ⅰ:The overall sensitivity,specificity,Youden index,positive predictive value,negative predictive value and Kappa value of Kudo Pit Pattern were 0.6389,0.7297,0.3681,0.8762,0.4023 and 0.290.The results of Kudo pit pattern typing of endoscopic physicians with high,middle,and low senors were as follows:in the high senors group,negative predictive value was 0.5862,Kappa value was 0.595;The middle seniority group:negative predictive value 0.4137,Kappa value 0.239;Low seniority group:negative predictive value 0.2068,Kappa value 0.058.Part Ⅱ:The size and morphology of polyps treated by cauterization and argon ion coagulation(APC)were similar,and the smaller the polyps were,the more inclined they were to choose these two methods.The morphology of polyps had no effect on the choice of forceps.The size of polyps treated by forceps was larger than that treated by cautery and APC.The size and morphology of polyps are affected by the choice of cold resection,electrocoagulation resection,and endoscopic mucosal resection(EMR).Compared with electroresection and EMR,polyps treated with cold resection tend to choose electrocoagulation resection,while those with large and unpedicled polyps tend to choose EMR.Part Ⅲ:The in vitro OCT imaging system was successfully established,and OCT imaging of pig large intestine was performed in vitro.Twenty cases of adenomatous polyps and 60 cases of non-adenomatous polyps were collected.OCT scanning of isolated polyps was completed to construct the AI model.The accuracy of the AI model was>90%in the final verification set test.Conclusions:1.The overall efficacy of Kudo Pit Pattern typing in the diagnosis of polyp properties under WLE was generally evaluated,and its accuracy was related to the seniority of endoscopic physicians.2.The size and shape of polyps will influence the selection of endoscopic treatment methods,and the size of polyps is a more important factor.3.The AI-enabled intelligent analysis of OCT images preliminarily verified the clinical feasibility of automatic polyp classification and diagnosis in vitro. |
PDF Full Text Request