基于统计学纹理特征的阿尔兹海默病MR图像研究

阿尔兹海默病（Alzheimer’s disease，AD）是一种老年神经系统退行性疾病。流行病学研究显示，在过去的50年里，AD的发病率增加了4倍，是目前威胁老年健康的重要疾病。对AD的影像学研究目前主要采用对图像上特征线、面积、体积测量的方法，还没有发现对AD具有特异性的影像学指标。本文尝试采用基于统计学理论的灰度共生矩阵、游程长矩阵的纹理分析方法提取AD患者MR图像上感兴趣区的纹理特征参数，通过筛选得到的参数，对AD患者和健康对照组进行分类识别，并对采用不同分类方法得到的识别结果进行比较。研究结果显示对统计学纹理特征参数使用非线性判别分析的分类方法得到的识别率最高达到90．12％。可以预见，此项研究对AD的早期诊断具有积极作用。

Statistical Texture Features Study on Magnetic Reson ance Image in Patients with Alzheimer's Disease

Alzheimer＇s disease（AD） is a progressive, degenerative disease of the brain, which causes thinking and memory to become seriously impaired. Epidemiological studies show that over the past 50 years, AD incidence increased by 4 times, and was a major threat to the health of older diseases. The imaging study of AD mostly use the image characteristics of the line, area, volume measurement methods, however, there has not been a specific parameter. In this paper, regions of interesting were chosen in MRI as research object. Statistical texture analysis was performed on MR images of AD patients. Texture features were firstly extracted from gray level co-occurrence matrix, run-length matrix, and then extracted by means of Fisher test. Features were used to classify and recognize in normal controls and patients with AD. Finally, there was a comparative study between the different classification results. This study demonstrated that nonlinear discriminant analysis could achieve high classification accuracy （92.19%） , which was valuable in supporting early diagnosis of AD.