老年嵌顿疝肠坏死风险预测神经网络模型:融合生物学指标Treg/Th17比值与临床指标模型的验证与开发OA

Objective To establish a neural network model integrating the biological marker regulatory T cell/T helper 17 cell(Treg/Th17 ratio)and clinical indicators,thereby providing individualized prediction for elderly patients at high risk of intestinal necrosis due to incarcerated hernia.Methods A retrospective analysis was conducted on 140 elderly patients with incarcerated inguinal hernia admitted to the Second People's Hospital of Yibin City from January 2023 to January 2025,including 95 males and 45 females,with an average age of(74.6±8.2)years(range,65-83 years).According to the intraoperative presence or absence of intestinal necrosis,all patients were divided into non-intestinal necrosis group(n=88)and intestinal necrosis group(n=52).General clinical data of the two groups were collected and compared.The least absolute shrinkage and selection operator(LASSO)regression method was used to screen the predictive variables for intestinal necrosis in elderly patients with incarcerated hernia.Logistic regression analysis was employed to identify the influencing factors of intestinal necrosis in such patients.The total sample(n=140)was randomly divided into training set(n=98)and internal validation set(n=42)in a ratio of 7:3 by computer-generated random numbers.A back propagation(BP)neural network model was constructed using Python 3.6 based on training set.R software(v3.5.3)combined with the rms package was used to draw calibration curves to evaluate the consistency between the model-predicted probability and actual incidence probability.The receiver operating characteristic(ROC)curve and area under the curve(AUC)were applied to assess the predictive efficacy of the model.Decision curve analysis(DCA)was adopted to evaluate the clinical net benefit of the model at different threshold probabilities.Results(1)History of oral anticoagulant use,abdominal surgery history,history of incarcerated hernia,incarceration type,incarceration duration,D-dimer,neutrophil count(NEUT),and Treg/Th17 ratio were all independent influencing factors for intestinal necrosis in elderly patients with incarcerated hernia(P<0.05).(2)The Treg/Th17 ratio contributed the most to model classification,followed by D-dimer,incarceration duration,incarceration type,NEUT,history of incarcerated hernia,oral anticoagulant history,and abdominal surgery history.The predicted probability of training set and validation set was in good consistency with the actual probability,with Brier scores of 0.073 and 0.075 respectively,indicating favorable model calibration.The prediction accuracy of the BP neural network model in training set was 94.21%,with an AUC of 0.913(95%CI 0.821-0.965,P<0.001).The prediction accuracy in validation set was 90.45%,with an AUC of 0.896(95%CI 0.815-0.957,P<0.001),which verified the high accuracy of this model.DCA confirmed that the model possessed high clinical application value.Conclusion The neural network model integrating Treg/Th17 ratio and clinical indicators exhibits high predictive value for the risk of intestinal necrosis in elderly patients with incarcerated hernia.