Classification of MRI and psychological testing data based on support vector machine.

Abstract

Alzheimer's disease (AD) is a progressive, and often fatal, brain disease that causes neurodegeneration, resulting in memory loss as well as other cognitive and behavioral problems. Here, we propose a novel multimodal method combining independent components from MRI measures and clinical assessments to distinguish Alzheimer's patients or mild cognitive impairment (MCI) subjects from healthy elderly controls. 70 AD subjects (mean age: 77.15 ± 6.2 years), 98 MCI subjects (mean age: 76.91 ± 5.7 years), and 150 HC subjects (mean age: 75.69 ± 3.8 years) were analyzed. Our method includes the following steps: pre-processing, estimating the number of independent components from the MR image data, extracting effective voxels for classification, and classification using a support vector machine (SVM)-based classifier. As a result, with regards to classifying AD from healthy controls, we achieved a classification accuracy of 97.7%, sensitivity of 99.2%, and specificity of 96.7%; for differentiating MCI from healthy controls, we achieved a classification accuracy of 87.8%, a sensitivity of 86.0%, and a specificity of 89.6; these results are better than those obtained with clinical measurements alone (accuracy of 79.5%, sensitivity of 74.0%, and specificity of 85.1%). We found that (1) both AD patients and MCI subjects showed brain tissue loss, but the volumes of gray matter loss in MCI subjects was far less, supporting the notion that MCI is a prodromal stage of AD; and (2) combining gray matter features from MRI and three commonly used measures of mental status, cognitive function improved classification accuracy, sensitivity, and specificity compared with classification using only independent components or clinical measurements.