Multi-center study based on CT artificial intelligence models for classification of acute appendicitis

ZUO Zongye; ZHANG Qi; HUANG Guoquan; YANG Guang; ZHANG Linjie; ZHANG Tingting; HU Xianlin; YANG Lanying

doi:10.19300/j.2025.L21614

Abstract：Objective To evaluate the diagnostic performance of a deep learning model based on conventional abdominal CT scans for differentiating acute simple appendicitis from acute non-simple appendicitis,and to assess its clinical application value.Methods Clinical and plain CT data from 336 patients with acute appendicitis confirmed by surgical pathology were retrospectively collected from three hospitals.The patients were divided into acute simple appendicitis(n=109)and acute non-simple appendicitis(n=227)based on pathological results.The dataset was randomly split into a training set(n=268)and an independent test set(n=68)at a ratio of 8∶2.The training set was further divided into a training subset(n=214)and a validation subset(n=54)using five-fold cross-validation.A model was constructed using Densenet-121(Dense Convolutional Network 121)and subsequently trained and tested.Receiver operating characteristic(ROC)curves,precision-recall(PR)curves,and confusion matrices were used to evaluate the model's diagnostic performance.Metrics such as the area under the curve(AUC)and PR-AUC were calculated.Results The Densenet-121 deep learning model demonstrated excellent diagnostic performance in classifying acute simple and acute non-simple appendicitis across the training subset,validation subset,and independent test set.The AUC values were 0.999,0.895,and 0.894,respectively,and the PR-AUC values were 0.999,0.831,and 0.813,respectively.The confusion matrix showed high prediction accuracy and effective classification in all three datasets for both acute simple and non-simple appendicitis.Conclusion The deep learning model based on neural networks exhibited robust stability and diagnostic efficiency in the classification of acute appendicitis using CT scans,providing a valuable reference for clinical decision-making.