Analysis and classification on multiple cortical features of patients with mild cognitive impairment

作者：Guo, Shengwen; Wu, Congling; Lai, Chunren; Wu, Yupeng; Jiang, Xingjun; Zhao, Di

来源：Journal of Southeast University(Natural Science Edition), 2017, 47(3): 483-489.

DOI：10.3969/j.issn.1001-0505.2017.03.012

• Automatic classification
• Cortical feature
• Dimensionality reduction
• Linear Support Vector Machines
• Longitudinal change
• Mild cognitive impairments
• Morphological features
• Recursive feature elimination

The difference of the brain structures and morphological changes among the patients with stable mild cognitive impairment (sMCI), the patients with converted mild cognitive impairment (cMCI) and the normal control (NC) was revealed and three groups were discriminated. First, 73 NC, 46 sMCI and 40 cMCI were selected, and the baseline, 1-year and 2-year longitudinal follow-up magnetic resonance (MR) images were acquired. Secondly, the FreeSurfer software was used to calculate the cortical morphological features including the cortical thickness, the gray matter volume, the surface area, and the mean curvature. The T-test method, the sparsity-constrained dimensionality reduction (SCDR) method and the recursive feature elimination (RFE) method were adopted to extract the salient features in discrimination. Finally, the linear support vector machine (LSVM) was applied to classify these three groups, and the brain regions with strong capability in the classification and their distributions were analyzed. The experimental results show that the RFE method exhibits the best performance in classification, followed by the SCDR method, and the T-test method is least. The combination of four types of cortical features, especially the combination of the baseline feature with the longitudinal change feature, can improve the performance of the classifier. Therefore, the cortical morphological features and their changes with time can be applied for automatic classification between the patients with sMCI and the patients with cMCI.

• 出版日期2017-5-20
• 单位华南理工大学