Progression from mild cognitive impairment to dementia in Alzheimer's disease: Whole cortex voxelwise functional connectivity analysis with multivariate distance matrix regression

Background Differentiating individuals with mild cognitive impairment who convert to dementia due to Alzheimer's disease (MCI-C) from those who do not convert (MCI-NC) is increasingly important. Functional connectivity (FC) derived from resting state functional MRI (rs-fMRI) has been investigated as a potential biomarker. However, improved data analysis strategies are needed. One underexplored approach is pairwise voxel-to-voxel analysis. Objective To describe differences in FC between amyloid positive MCI-C and MCI-NC using a whole-cortex voxel-to-voxel pairwise approach. Methods Baseline rs-fMRI from the Alzheimer's Disease Neuroimaging Initiative was retrieved for amyloid positive MCI participants. Voxel-to-voxel, pairwise, cortical FC was computed. Multivariate distance matrix regression was used to identify voxels presenting FC patterns that were significantly different between MCI-C and MCI-NC. Results 21 MCI-C and 28 MCI-NC were included. The primary analysis with voxel-level threshold at p < 0.001 combined with cluster-level p < 0.05 yielded no significant results. At voxel-level p < 0.01 and the same cluster-level threshold, three significant clusters on the right visual cortex were found. These clusters, however, were not robust to head motion, fMRI protocol and additionally clinical or biological severity. Conclusions Progression from Alzheimer-related MCI to dementia was not significantly associated with FC in the primary analysis. However, a less stringent threshold yielded FC differences in the occipital lobe in alignment with previous studies but were not robust to methodological and biological covariates between groups. Our findings highlight the need for larger samples and careful control of covariates to identify robust FC alterations related to dementia conversion.