【作者】 杨磊；

【导师】 王洪瑞；

【作者基本信息】 河北大学 ， 模式识别与智能系统， 2019， 硕士

【摘要】 随着老龄化日益严重,心血管疾病已成为我国发病率和死亡率占据首位的疾病,动脉粥样硬化斑块是心血管疾病的主要原因。斑块组织准确分类与识别对辅助医生进行心血管疾病的诊断及治疗方式选择具有极大意义。本文针对OCT影像和真实斑块组织,基于影像特征与物理特征对斑块组织类型的识别与分类进行研究。主要研究内容如下:一、提出改进区域生长算法用于提高OCT影像冠脉斑块的分割精度,实现冠状动脉斑块组织的半自动化分割,同时结合图像特征(颜色特征、纹理特征),使用支持向量机算法,进行钙化斑块组织和非钙化斑块组织的识别分类,准确率为82.5%。二、引入动态流变学对离体真实斑块组织的储能模量、损耗模量以及粘度进行测定,对斑块组织进行频率扫描实验、恒温扫描实验与剪切速率扫描实验。实验结果分析表明,斑块组织呈剪切变稀的假塑性流体特征,具有一定的弹性和黏度;频率和剪切速率对斑块组织稳定性影响较大;不同类型斑块组织的存在明显差异。三、通过流变学实验获得真实斑块组织的流变学特征曲线,采用幂律模型建立了斑块组织物理模型,提出了粘度指数K与非牛顿指数N作为特征参数对钙化斑块组织和非钙化斑块组织进行分类识别,准确率达到了86.67%。更多还原

【Abstract】 With the aging becoming more and more serious,cardiovascular disease has become the leading disease of morbidity and mortality in China.Atherosclerotic plaque is the main cause of cardiovascular disease.Accurate classification and recognition of plaque tissue is of great significance to assist doctors in the diagnosis and treatment of cardiovascular diseases.This paper studies the recognition and classification of plaque tissue types from image features and physical features.The main contents are following:Firstly,improved region growing algorithm is proposed to improve the segmentation accuracy and realize the semi-automatic segmentation of coronary artery plaque tissue.At the same time,combined with image features(color features,texture features)and support vector machine(SVM)is used to recognize and classify different types of plaque tissue.The accuracy rate is 82.5%.Secondly,designing dynamic rheology in vitro experiment to measure the storage modulus,loss modulus and viscosity of real plaque tissue,and carrying out frequency scanning experiment,constant temperature scanning experiment and shear rate scanning experiment on plaque tissue.The experimental results show that the plaque structure is a shear thinning pseudoplastic fluid with certain elasticity and viscosity;frequency and shear rate have great influence on the stability of plaque structure;there are obvious differences between different types of plaque structure.Thirdly,the rheological characteristic curves of real plaque tissues were obtained by rheological experiments,and the physical model of plaque tissues was established by power law model.Viscosity index K and non-Newtonian index N(mean error 0.1631 and 0.0187,respectively)were proposed as characteristic parameters to classify and identify different types of plaque tissues.The accuracy rate was 86.67%.更多还原