Automatic classification method of Alzheimer's disease by voxel-based morphometry on MR images

作者：Guo, Shengwen^*; Chi, Minyue; Cen, Guiyin; Kuang, Cuili; Niu, Chuanxiao; Lai, Chunren; Wu, Xiaoming; The Alzheimer's Disease Neuroimaging Initiative

来源：Journal of Southeast University(Natural Science Edition), 2015, 45(2): 260-265.

DOI：10.3969/j.issn.1001-0505.2015.02.012

• Alzheimer's disease
• Automatic classification
• Classification accuracy
• Linear Support Vector Machines
• Mild cognitive impairments
• Mild cognitive impairments (MCI)
• Recursive feature elimination
• Voxel-based morphometry

To determine the atrophy in important brain regions from magnetic resonance(MR) images of patients with Alzheimer's disease (AD) and mild cognitive impairment (MCI) and classify the normal control(NC), MCI and AD groups, the MR images of 178 subjects were selected for analysis. The voxel-based morphometry (VBM) and analysis of variance (ANOVA) were adopted to investigate the grey matter(GM) volume differences of brain structure in MR images from NC group, MCI group and AD group. Then, the dimension of the detected features was reduced by recursive feature elimination (RFE) method. Finally, the linear support vector machine (LSVM) was applied to classify these three groups. The experimental results show that the average classification accuracies of MCI group and NC group, MCI group and AD group, AD group and NC group are (90.2±1.3)%, (74.7±0.9)% and 100%, respectively. The dominant regions sensitive to classification include hippocampus, parahippocampal gyrus, amygdala, olfactory cortex, fusiform gyrus and so on. The proposed method not only can reveal differences in brain gray matter among NC group, MCI group and AD group effectively, illustrate the shrinking process and characteristics in brain regions of MCI patients and AD patients, but also can exhibit great potentials to accurately distinguish these three groups in clinical application.

• 出版日期2015-3-20
• 单位华南理工大学