SVM-Based Classification of Diffusion Tensor Imaging Data for Diagnosing Alzheimer’s Disease and Mild Cognitive Impairment

Abstract

Alzheimer’s disease (AD) is the most common cause of dementia. Early detection of AD is important since treatment is more efficacious if introduced earlier. Mild cognitive impairment (MCI) is often a precursory stage of AD, so is considered to be a good target for early detection of AD. However, MCI is not easy to diagnose due to the subtlety of cognitive impairment. In this study, we developed a method to automate the diagnosis of AD and MCI using support vector machines (SVMs) and diffusion tensor imaging (DTI) data. We implemented two SVM models: one for classifying AD and MCI and another for classifying AD and normal control (NC). In both SVM models, the fractional anisotropy (FA) and the mode of anisotropy (MO) values of DTI were used as features. MO values resulted in a better performance than FA values in both models. In independent testing, the AD-MCI classifier showed a sensitivity of 69.2 %, a specificity of 100 % and an accuracy of 89.7 %, and the AD-NC classifier showed a sensitivity of 84.6 %, a specificity of 90.9 % and an accuracy of 87.5 %. These results are encouraging and suggest that SVM-based classification of DTI data is potentially powerful in early detection of MCI and AD.