基于骨振信号和深度网络的膝关节退行性病变早期筛查方法  

Early screening methods for knee osteoarthritis based on vibroarthrographic signals and deep network

作　　者：郑田田 周海天 宋江玲 张瑞[1] ZHENG Tiantian;ZHOU Haitian;SONG Jiangling;ZHANG Rui(Medical Big Data Research Center,Northwest University,Xi′an 710127,China;School of Electronic and Information Engineering,South China University of Technology,Guangzhou 510641,China)

机构地区：[1]西北大学医学大数据研究中心,陕西西安710127 [2]华南理工大学电子与信息学院,广东广州510641

出　　处：《西北大学学报（自然科学版）》2021年第4期549-557,共9页Journal of Northwest University（Natural Science Edition）

基　　金：国家自然科学基金面上项目(12071369);陕西省重点研发计划资助项目(2019ZDLSF02-09-02);国家自然科学基金青年项目(62006189)。

摘　　要：膝关节退行性病变(knee osteoarthritis,KOA)是一种由关节软骨病变导致膝关节发生不可逆损伤的慢性疾病。若在KOA发病初期及时诊断并采取有效措施,则可有效延缓病程进展。由于在发病初期关节病变范围小、程度轻、患者症状不明显,传统检查手段(MRI,CT等)几乎无法得到显著的病理信息,这使得早期筛查变得十分困难。而髌骨关节摩擦音(又称为骨振信号,vibroarthrographic signal,VAG)有可能为实现早期筛查提供一种新的途径。基于此,文中以VAG信号为数据源,提出了一种基于卷积神经网络的KOA早期筛查方法。首先,结合自动裁剪、补零处理、白化处理等方法对VAG信号进行信号对齐与去相关等预处理;其次,从VAG信号频域角度出发,采用卷积神经网络实现KOA的早期筛查;最后,采用西安市某医院临床采集的772条VAG信号数据集验证所提方法的可行性与有效性。数值实验结果表明,本文所提KOA早期筛查方法准确率、灵敏度、特异性分别可达86.2%、88.20%、83.3%。Knee osteoarthritis(KOA)is a common chronic joint disease caused by articular cartilage lesions.If the timely diagnosis can be taken at the early stage of KOA,the progression of KOA can be effectively delayed.Due to the small range and mild severity of lesions,and unobvious symptoms,the traditional examination methods(MRI,CT,etc.)can hardly get significant pathological information,which makes the early screening of KOA difficult.The vibroarthrographic signal(VAG)may provide a new way for early screening of KOA.According to it,a new automated KOA early screening method based on VAG signals by using convolutional neural networks(CNNs)is proposed.Firstly,the methods of cutting,zero filling,whitening are used to preprocess the VAG signals;secondly,the convolutional neural network is applied to realize the automated early screening of KOA;finally,simulation results on 772 VAG signals clinically collected from an hospital in Xi′an are conducted to verify the feasibility and effectiveness of the proposed method.Experimental results show that the accuracy,sensitivity,and specificity of the proposed method are 86.2%,88.20%,and 83.3%.

关 键 词：膝关节退行性病变 骨振信号 早期筛查 卷积神经网络 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]