基于多模态超声成像数据的慢性肝病肝纤维化、炎症和脂肪变性的智能分级诊断  

Intelligent grading diagnosis of liver fibrosis,inflammation,and steatosis in chronic liver disease based on multimodal ultrasound imaging data

作　　者：魏星月 王连双 王媛媛 高孟泽 何琼[1,2] 张瑶 罗建文[1,2] Wei Xingyue;Wang Lianshuang;Wang Yuanyuan;Gao Mengze;He Qiong;Zhang Yao;Luo Jianwen(Department of Biomedical Engineering,School of Medicine,Tsinghua University,Beijing 100084,China;Institute of Precision Medicine,Tsinghua University,Beijing 100084,China;Department of Ultrasound,Beijing Ditan Hospital,Capital Medical University,Beijing 100015,China;Beijing Engineering Research Center of Mixed Reality and Advanced Display,School of Optics and Electronics,Beijing Institute of Technology,Beijing 100081,China;Department of Precision Instrument,Tsinghua University,Beijing 100084,China)

机构地区：[1]清华大学医学院生物医学工程系,北京100084 [2]清华大学精准医学研究院,北京100084 [3]首都医科大学附属北京地坛医院超声科,北京100015 [4]北京理工大学光电学院北京市混合现实与新型显示工程技术研究中心,北京100081 [5]清华大学机械工程学院精密仪器系,北京100084

出　　处：《首都医科大学学报》2023年第6期928-935,共8页Journal of Capital Medical University

基　　金：北京市自然科学基金项目(M22018);首都健康专项研究与发展项目(2022-2G-2177);清华精准医学基金会基金项目(2022TS012);清华大学精准医学科研计划项目(2022ZLA005);清华大学春风基金项目(2021Z99CFY025)。

摘　　要：目的开发一套能同时诊断肝纤维化、炎症及脂肪变性的智能系统。方法基于慢性肝病(chronic liver disease,CLD)患者的多模态超声成像数据,包括超声B模式图像、剪切波弹性成像、瞬时弹性成像数据及其原始射频数据,使用定量超声方法和影像组学方法提取这些数据的多模态特征,以超声引导的肝穿刺活检结果为金标准,使用支持向量机以二分类的方式搭建智能分级诊断系统,同时实现肝纤维化、炎症和脂肪变性的智能辅助诊断。结果本方法对于肝纤维化分级的四个二分类任务:≥F1、≥F2、≥F3、≥F4,受试者工作特征(receiver operating characteristic,ROC)曲线下面积(area under the curve,AUC)达到了0.81、0.80、0.89、0.87;对于炎症分级≥A2、≥A3、≥A4,AUC为0.80、0.93、0.93;对于脂肪变性分级≥S1、≥S2,AUC为0.75、0.92。结论本研究提出了基于多模态超声成像数据的CLD肝纤维化、炎症和脂肪变性的智能分级诊断系统,该系统在3种CLD的智能评估中取得了较好的结果,有望推广至临床应用。Objective To develop a non-invasive,accurate,convenient,and widely applicable intelligent diagnostic system to diagnose simultaneously liver fibrosis,inflammation,and steatosis of chronic liver disease(CLD).Methods This study is based on multimodal ultrasound imaging data from CLD patients,including two-dimensional B-mode ultrasound images,two-dimensional shear wave elastography,transient elastography data,and the corresponding original radio-frequency data.Quantitative ultrasound methods were used to extract multimodal features from these multimodal data,and the results of ultrasound-guided liver biopsy were used as the gold standard.Support vector machine(SVM)was used to construct an intelligent grading diagnosis system for CLD in a binary-classification manner.Results The proposed method achieves high the receiver operating characteristic(ROC)area under the curve(AUC)of 0.81,0.80,0.89,0.87 for the classification of fibrosis grade≥F1,≥F2,≥F3≥F4,and 0.80,0.93,0.93 for inflammation≥A2,≥A3,≥A4,and 0.75,0.92 for steatosis≥S1,≥S2.Conclusion The results indicated that the proposed method showed potential expected to be promoted to clinical applications.

关 键 词：慢性肝病 肝纤维化 炎症 脂肪变性 多模态超声 支持向量机 

分 类 号：R445[医药卫生—影像医学与核医学]