| 以下仅列出文章摘要、提纲简介，如需获取全文阅读权限，或原创定制、长期合作，请随时联系。 微信QQ:312050216 |
 |
 |
| 基于声发射的膝关节骨性关节炎的动态检测与诊断识别
摘 要
膝关节骨性关节炎（Knee (D:/Dict/8.9.6.0/resultui/html/index.html" \l "/javascript:;) Osteoarthritis (D:/Dict/8.9.6.0/resultui/html/index.html" \l "/javascript:;), KOA）是一种复杂的退行性疾病，具有较高的发病率和致残率。并且随着人口老龄化趋势的加快其患病率呈上升趋势，KOA不仅是我国所面临的的一项重要社会卫生问题，也是危害全人类健康的一项重大医学问题。由于KOA在临床上早期表现的并不是特别明显，当症状出现时往往已经到了晚期，所以前期的诊断尤为重要。X射线、超声检测、核磁共振检测和关节镜检测等传统的诊断方法无法对膝关节提供完整的动态分析，具有一定的局限性。
本文通过分析膝关节在运动过程中产生的声发射（Acoustic Emission, AE）信号，对KOA进行了动态检测与诊断识别。主要研究内容为：
（1）KOA动态检测系统的设计，通过试验得到在坐-立-坐运动过程中膝关节的角度变化和与之对应的声发射信号。
（2）分别在波形和特征参数两个方面分析健康膝关节和患病膝关节声发射信号的差异性。根据膝关节角度的变化，将坐-立-坐运动分为上升和下降两个运动阶段。
（3）在诊断识别方面，为了提高运算速度而不降低信号所包含的信息量，运用主成分分析的方法对膝关节声发射信号进行降维。采用曼-惠特尼U检验的方式对降维后的信号进行了差异性分析。
（4）对比了BP(Back Propagation)神经网络、支持向量机（Support Vector Machine, SVM）和小生境粒子群优化的支持向量机这三种方法对膝关节声发射信号进行诊断识别的效果。
经过研究得到以下结果：
检测平台可同步获取膝关节在坐-立-坐运动过程中的声发射信号的膝关节角度变化信号，通过对膝关节声发射时域信号、频域信号和时频信号的对比发现患病组的声发射信号中包含了更多高频率的信号和更高的幅值，高频信号出现的时间主要集中在坐-立-坐运动的下半段时间。
通过对两个运动阶段健康组和患病组声发射特征参数的特点对比发现而下降阶段的膝关节声发射信号具有更高的活动性和更强的信号水平。
声发射特征参数经过线性变化提取为2个主成分，能够包含原始信号74.03%的信息量。曼-惠特尼U检验结果表明两个运动阶段的声发射信号存在差异，差异性在患病膝关节上表现的更为显著。
仿真测试发现：三种模式识别方法均能对患有骨性关节炎的膝关节声发射信号做出准确识别，其中小生境粒子群优化的支持向量机的诊断识别率最高。三种方法均显示出对下降阶段的膝关节声发射信号具有更高的分类准确率。对下降阶段的声发射信号进行识别更易对KOA做出早期诊断。
Abstract
Knee osteoarthritis is a complex degenerative disease with high morbidity and disability rate.And with the acceleration of the aging trend of the population, the prevalence of knee osteoarthritis is on the rise. Knee osteoarthritis is not only an important social health problem facing ourcountry, but also a major medical problem endangering the health of all mankind.Since the clinical manifestations of knee osteoarthritis are not particularly obvious in the earlystage, when the symptoms appear, it is often late, so the early diagnosis is particularly important.The traditional diagnostic methods, such as X-ray, ultrasound,nuclear magnetic resonance and arthroscopy,cannot provide a complete dynamic analysis of the knee joint and have certain limitations.
In this paper, the dynamic detection,diagnosis and recognition of knee osteoarthritis were carried out by analyzing the acoustic emission signals generated during the movement of knee joint.The main research contents are as follows:
(1) Design of the dynamic detection system for knee osteoarthritis.Through the test, the Angle change of knee joint and the corresponding acousticemission signal were obtained in the process of sit-rise-sitting movement.
(2) The difference of acoustic emission signals between healthy knee joint and diseased knee joint was analyzed in terms of waveform and characteristic parameters.According to the change of knee joint Angle,seated - upright - seated movement can be divided into two stages: ascending and descending.
(3) In terms of diagnosis and recognition,in order to improve the speed of operation without reducing the information content contained in the signals, principal component analysis was used to reduce the dimension of knee acoustic emission signals.Man-Whitney U test was used to analyze the differenceof the signals after dimensionality reduction.
(4) The effects of BP(Back Propagation) neural network,support vector machine and support vector machine based on niche particle swarm optimization were compared in the diagnosis and recognition of knee joint acoustic emission signals.
The research results are as follows:
Testing platform can get knee joint synchronization in - state - sitting movement in the process of acoustic emission signals of knee joint Angle changes,through to the knee acoustic emission signals,frequency domain and time-frequency signal contrast group discovered in acoustic emission signal contains more high frequency signal and higher amplitude, high frequency signal of time focused on by - stand - by movement of the second half of the period of time.
By comparing the characteristics of acoustic emission characteristic parameters of the healthygroup and the sick group in the two movement stages,it was found that the knee joint acoustic emission signal in the descending stage had higher activity and stronger signal level.
The acoustic emission characteristic parameters are extracted intotwo principal components after linear changes, which can contain 74.03% of the information of the original signal.Mann-Whitney U test results showed that there were differences in acoustic emission signals between the two movement stages,and the difference was more significant in the disaffected knee joint.
The simulation results show that all the three pattern recognition methods can accurately identify the acoustic emission signals of knee joint with osteoarthritis,and the support vector machine based on niche particle swarm optimization has the highest diagnostic recognition rate.All the three methods showed higher classification accuracy ofknee joint acoustic emission signals in the descending stage.Recognition of acoustic emission signals in the descent stage iseasier to make an early diagnosis of knee osteoarthritis.
Keywords Knee joint Acoustic emission Principal component analysisBP neural network Support vector machine Niche particle swarm
|
