Predicting neurocognitive function with hippocampal volumes and DTI metrics in patients with Alzheimer's dementia and mild cognitive impairment

INTRODUCTION Cognitive performance in patients with Alzheimer's dementia (AD) and mild cognitive impairment (MCI) has been reported to be related to hippocampal atrophy and microstructural changes in white matter (WM). We aimed to predict the neurocognitive functions of patients with MCI or AD using hippocampal volumes and diffusion tensor imaging (DTI) metrics via partial least squares regression (PLSR). METHODS A total of 148 elderly female subjects were included: AD (n = 49), MCI (n = 66), and healthy controls (n = 33). Twenty-four hippocampal subfield volumes and the average values for fractional anisotropy (FA) and mean diffusivity (MD) of 48 WM tracts were used as predictors, CERAD-K total scores, scores of CERAD-K 7 cognitive subdomains and K-GDS were used as dependent variables in PLSR. RESULTS Regarding MCI patients, DTI metrics such as the MD values of the left retrolenticular part of the internal capsule and left fornix (cres)/stria terminalis were significant predictors, while hippocampal subfield volumes, like the left CA1 and hippocampal tail, were main contributors to cognitive function in AD patients, although global FA/MD values were also strong predictors. The 10-fold cross-validation and stricter 300-iteration tests proved that global cognition measured by the CERAD-K total scores and the scores of several CERAD-K subdomains can be reliably predicted using the PLSR models. CONCLUSIONS Our findings indicate different structural contributions to cognitive function in MCI and AD patients, implying that diffuse WM microstructural changes may precede hippocampal atrophy during the AD neurodegenerative process.