Cognition-related white matter integrity dysfunction in Alzheimer's disease with diffusion tensor image

•Cognition-related white matter (WM) microstructure was disrupted in AD patients.•FA and MD values of cognition related WM tracts were correlation with MMSE scores.•Accuracy of SVM trained by white matter tract features could reach 91.67%.•MD value of cingulum was the most important feature. Alzheimer’s disease (AD) is a worldwide progressive neurodegenerative disorder in the elderly. Previous research has indicated that Alzheimer’s disease impairs white matter (WM) tracts. Anatomical and neuroimaging studies have indicated that WM tracts are associated with cognitive function. Whether the abnormal WM integrity in AD is associated with cognitive impairments and the clinical symptoms is still not clear. To this end, we investigated the relationship between the impairments in WM tracts and the decline of cognitive ability in AD. Diffusion tensor imaging (DTI) data were collected from 38 AD patients and 30 normal, cognitively healthy volunteers. The tract-based spatial statistics (TBSS) approach was used to compare the fractional anisotropy (FA) and mean diffusivity (MD) values between the two groups. WM tracts (cingulum, superior longitudinal fasciculus (SLF), uncinate fasciculus (UF), and inferior longitudinal fasciculus (ILF)) associated with cognition function were extracted for region of interest (ROI)-based analysis. Significantly decreased FA values and increased MD values of the cognition-related WM tracts were observed in the AD group compared with the normal cognition (NC) group. In addition, we further demonstrated that the decreased FA values and increased MD values of the cognition-related WM tracts were significantly correlated with MMSE scores. These results indicated that abnormal changes in WM integrity are observed following AD. Finally, we used support vector machine (SVM) with a repeated, stratified 10-fold cross-validated classifier to evaluate the ability of FA and MD values to discriminate disease. The accuracy of the SVM using cognition-related WM as classified features was higher than that using non-cognition-related tracts. Most importantly, our results showed the relationship between abnormal WM tracts and cognitive ability in AD. These findings further suggested that AD-related impairments in cognition-related WM tracts may influence the cognitive ability of AD patients.