肺结核病人工智能辅助诊断系统微信小程序的开发和应用

Development and application of an artificial intelligence-assisted diagnosis system for tu-berculosis embedded in a WeChat applet

目的探索构建基于智能手机微信小程序的肺结核病在线智能辅助诊断系统,提高肺结核病诊断的准确率和效率。方法采用C5.0决策树算法,从汕头市结核病防治所2018年1月至2020年12月3174份结核门诊病案资料中,挖掘高价值的肺结核病诊断决策规则,建立诊断预测模型,进而开发肺结核病人工智能辅助诊断系统,并对其在肺结核病诊断中的准确性进行评估。结果2559例训练样本中通过数据挖掘出对肺结核病诊断具有显著价值的决策规则8条,置信度在77.49%~99.81%之间;对3174个研究对象中随机抽取的20%验证样本(肺结核确诊病例403例和非结核其他病例212例)进行验证,以原诊断为金标准,系统诊断的敏感度为94.79%(382/403)、特异性为96.23%(204/212)、准确性为95.28%(586/615)。结论基于智能手机微信小程序的肺结核病智能辅助诊断系统可作为肺结核病的在线辅助诊断工具,提高了肺结核病诊断的质量和效率。

Objective To explore the construction of an online intelligent auxiliary diagnosis system for pulmonary tuberculosis(PTB)based on the WeChat applet installed in smartphones,so as to improve the accuracy and efficiency of PTB diagnosis.Methods Using the data of 3174 PTB outpatient medical records in Shantou Tuberculosis Prevention and Control Institute from January 2018 to December 2020,the decision tree model based on C5.0 algorithm was adopted to establish a diagnostic prediction model,and the optimal decision rules in the diagnosis and decision-making of PTB were found out through data mining.Then an artificial intelligence-assisted diagnosis system for PTB was developed and its accuracy in PTB diagnosis was evaluated.Results Among 2559 training samples,8 decision rules with significant value for the diagnosis of tuberculosis were found through data mining,with confidence between 77.49%and 99.81%.20%of the validation samples randomly selected from 3174 research subjects(403 confirmed cases of pulmonary tuberculosis and 212 non tuberculosis cases)were validated.Using the original diagnosis as the gold standard,the sensitivity of the system diagnosis was 94.79%(382/403),specificity was 96.23%(204/212),and accuracy was 95.28%(586/615).Conclusion The intelligent auxiliary diagnosis system for PTB based on the WeChat applet on smartphones can be used as an online auxiliary diagnosis tool for PTB,which improves the quality and efficiency of the disease diagnosis.