基于生物阻抗和深度学习技术的甲状腺组织分类模型研究  

Early thyroid cancer detection and differentiation by using electrical impedance spectroscopy and deep learning:a preliminary study

作　　者：黄奥玲 黄文雯 董鹏伟 鞠仙莉 严丹丹 袁静萍[1] Huang Aoling;Huang Wenwen;Dong Pengwei;Ju Xianli;Yan Dandan;Yuan Jingping(Department of Pathology,Renmin Hospital of Wuhan University,Wuhan 430060,China;School of Remote Sensing and Information Engineering,Wuhan University,Wuhan 430072,China)

机构地区：[1]武汉大学人民医院病理科,武汉430060 [2]武汉大学遥感信息工程学院,武汉430060

出　　处：《中华内分泌外科杂志（中英文）》2024年第4期484-488,共5页Chinese Journal of Endocrine Surgery

基　　金：国家药物监督管理局医疗器械监管科学研究基地(武汉大学)开放项目(2023JDKF005)。

摘　　要：目的近年来甲状腺癌发病率显著提高,超声和细针穿刺活检等术前检查不断完善,但仍存在诊断性手术及过度诊疗的现象。本研究旨在探讨利用生物阻抗及深度学习技术对甲状腺组织进行分类的可行性。方法为了提高分类的准确性,我们设计一套适合甲状腺组织的电阻抗测量系统,采集来自321例患者的331个离体甲状腺标本共1340项数据集。随后,建立一个基于多特征的一维卷积神经网络(one dimensional convolution neural,1D-CNN)结合长短期记忆网络(long short-term memory,LSTM)混合模型对甲状腺组织进行分类。其中80%(1072/1340)的数据用于训练,另外20%(268/1340)用于测试。采用受试者工作特征曲线(receiver operating characteristic curve,ROC)、宏平均ROC、微平均ROC及曲线下面积(area under the curve,AUC)等指标对甲状腺组织分类模型进行评价。此外,用敏感性、特异性、阳性预测值、阴性预测值和约登指数比较该分类模型与超声的诊断价值。结果甲状腺组织两分类(恶性/非恶性组织)模型的ROC-AUC=0.94,总体准确率达到91.4%。进一步建立三分类(恶性/良性/正常组织)模型,其中正常、良性和恶性组的ROC-AUC分别为0.91、0.85和0.92,微平均ROC-AUC和宏平均ROC-AUC分别为0.91和0.90。且与超声相比,甲状腺组织分类模型具有更高的特异性。结论本研究基于生物阻抗及CNN-LSTM建立的甲状腺组织分类模型具有较高的准确率和较好的稳定性,该技术有望在未来的临床应用中提供有用的补充信息,从而有助于甲状腺疾病的治疗决策和管理。ObjectiveTo aid in the detection of thyroid cancer by using deep learning to differentiate the unique bioimpedance parameter patterns of different thyroid tissues.MethodsAn electrical impedance system was designed to measure 331 ex-vivo thyroid specimens from 321 patients during surgery.The impedance data was then analyzed with one dimensional convolution neural(1D-CNN)combining with long short-term memory(LSTM)network models of deep learning.In the process of analysis,we assigned 80%of the data to training set(1072/1340)and the remaining 20%data to the test set(268/1340).The performance of final model was assessed using receiver operating characteristic(ROC)curves.In addition,sensitivity,specificity,positive predictive value,negative predictive value,Youden index were applied to compare impedance model with ultrasound results.ResultsThe ROC curve of the two-classification(malignant/non-malignant tissue)model showed a good performance(area-under-the-curve AUC=0.94),with an overall accuracy of 91.4%.To better fit clinical practice,we further performed a three-classification(malignant/benign/normal tissue)model,of which the areas under ROC curve were 0.91,0.85,0.92 for normal,benign,and malignant group,respectively.The results indicated that the area under micro-average ROC curve and the macro-average ROC curve were 0.91 and 0.90,respectively.Moreover,compared with ultrasound,the impedance model exhibited higher specificity.ConclusionsA deep learning model(CNN-LSTM)trained by thyroid electrical impedance spectroscopy(EIS)parameters shows an excellent performance in distinguishing among different in-vitro thyroid tissues,which is promising for applications.In future clinical utility,our study does not replace existing tests,but rather complements others,thus contributing to therapeutic decision-making and management of thyroid disease.

关 键 词：甲状腺结节 甲状腺癌 生物阻抗谱 电极 

分 类 号：R736.1[医药卫生—肿瘤]