摘要
目的利用同轴相衬成像(in-line phase contrast imaging,IL-PCI)技术与支持向量机(support vector machine,SVM)算法对正常与早期骨性关节炎(osteoarthritis,OA)软骨组织建立分类模型。方法研究样本分别来自接受人工膝关节置换手术及创伤性关节损伤患者的离体膝关节软骨组织。实验对象分为正常组与早期OA组,每组18例。利用IL-PCI技术分别获取两组样本的电子计算机断层扫描(computed tomography,CT)图像,运用行程长度矩阵、分形维度及小波变换三种分析方法对两组样本的CT图像提取纹理参数26个,经主成分分析法降维,利用纹理参数构建支持向量机分类诊断模型,对正常软骨与早期OA软骨进行自动分类。结果利用IL-PCI技术可以清晰观察到正常与早期OA软骨中微结构的变化。两组软骨的纹理特征参数比较显示,大部分纹理特征参数差异有统计学意义。利用SVM算法建立的分类模型能较好地区分正常与早期OA软骨组织,分类准确性可达86.1%。结论利用支持向量机算法与同轴相衬成像方法,能够为早期OA的鉴别诊断提供一种量化判别的手段。
Objective To establish the classification model for normal and early osteoarthritis(OA)of cartilage tissue by in-line phase contrast imaging(IL-PCI)technique and support vector machine(SVM)algorithm.Methods The study samples were in vitro knee cartilage samples from patients undergoing arthroplasty.The research subjects were divided into normal group and early OA group,18 cases in each group.Computed tomography(CT)images of cartilage tissues were obtained by in-line phase contrast imaging(IL-PCI)technology.Totally 26 texture parameters were extracted from the two groups of CT images by using three analysis methods of stroke length matrix,fractal dimension and discrete wavelet transform.Dimensionality of texture parameters was reduced by principal component analysis.A SVM classification model was established based on texture parameters to classify normal cartilage and early OA cartilage automatically.Results The changes of microstructure in normal and early OA cartilage could be clearly observed by IL-PCI technique.The cartilage texture parameters of the two groups were compared,and the results showed that most of the texture parameters were significantly different.The classification accuracy of the two groups can reach 86.1%by support vector machine algorithm.Conclusion The support vector machine algorithm and IL-PCI method can provide an effective auxiliary method for the diagnosis of early osteoarthritis.
作者
李君
程程
董立男
吴明树
张璐
Li Jun;Cheng Cheng;Dong Linan;Wu Mingshu;Zhang Lu(Department of Radiation Oncology, Peking University Third Hospital, Beijing 100191, China;Department of Imaging, The General Hospital of the People's Liberation Army, Beijing 100039, China;School of Biomedical Engineering, Capital Medical University, Beijing 100069, China)
出处
《首都医科大学学报》
CAS
北大核心
2022年第1期120-126,共7页
Journal of Capital Medical University
基金
国家自然科学基金(81401549)。
关键词
同轴相衬成像
支持向量机
骨关节炎
纹理特征
in-line phase contrast imaging(IL-PCI)
support vector machine
osteoarthritis
texture feature
