机器学习模型辅助唇腺活检干燥综合征诊断的初步探索  

A preliminary study on the diagnosis of Sjogren’s syndrome assisted by machine learning model

作　　者：焦德强 梁江 卢云 王晶 陈昱江 陈大琳 罗莉 Jiao Deqiang;Liang Jiang;Lu Yun;Wang Jing;Chen Yujiang;Chen Dalin;Luo Li(Graduate School of Guizhou University of TCM,Guiyang,Guizhou,550005;The first affiliated Hospital of Guizhou University of TCM,Guiyang,Guizhou,550000;School of Information Engineering,Guizhou University of TCM,Guiyang,Guizhou,550025)

机构地区：[1]贵州中医药大学研究生院,贵州贵阳550005 [2]贵州中医药大学第一附属医院,贵州贵阳550000 [3]贵州中医药大学信息工程学院,贵州贵阳550025

出　　处：《中医眼耳鼻喉杂志》2023年第3期152-156,共5页Journal of Chinese Ophthalmology and Otorhinolaryngology

基　　金：国家中医药管理局2022年全国名老中医药专家传承工作室建设项目:吴正石全国名老中医药专家传承工作室;贵州省“十四五”中医药、民族医药重点学科建设规划[重点培育学科:中医痹病学;编号:QZYYZDXK(PY)-2021-04];国家中医药管理局传承创新中心培育单位计划;第四批贵州省中医名医传承指导老师和继承人工作项目;2021年贵州中医药大学大学生创新创业训练计划项目[贵中医大创合自(2021)67号];2016年贵州省“千”层次创新人才培养计划;贵中医大创合自(2020)14号。

摘　　要：目的 探索多种机器学习模型对干燥综合征(SS)唇腺活检病理诊断的预测价值。方法 使用3D Slicer软件对我院既往70例疑似干燥综合征患者的唇腺活检病理图像进行勾画并用Pyradiomics包提取纹理特征参数,建立多个模型并进行AUC值、特异度、灵敏度和准确率的模型诊断性能评估。结果 SVC模型的测试集AUC、准确度、灵敏度、特异度分别为0.989、0.917、0.950、0.800,诊断预测准确率达91.7%,在多个模型中表现稳定可靠。结论 机器学习模型SVC有助于高效率初步识别病理阴性和SS的病理切片,具有较好的应用前景。Objective To explore the predictive value of multiple machine learning models for the pathological diagnosis of labial gland biopsy in patients with Sjogren’s syndrome(SS).Methods The pathological images of labial gland biopsy of 70 patients with suspected Sjogren’s syndrome in our hospital were sketched by 3D Slicer software,and the texture feature parameters were extracted by Pyradiomics package.Several models were established and the diagnostic performance of the model was evaluated by AUC value,specificity,sensitivity and accuracy.Results The test set AUC,accuracy,sensitivity and specificity of SVC model were 0.989,0.917,0.95,0.80 respectively,and the accuracy of diagnosis and prediction is 91.7%.The performance is stable and reliable in many models.Conclusion The machine learning model SVC is helpful to identify pathologically negative and SS pathological sections with high efficiency and has a good application prospect.

关 键 词：干燥综合征 机器学习 纹理特征 唇腺活检 

分 类 号：R78[医药卫生—口腔医学]