The Effect of Gray Matter ICA and Coefficient of Variation Mapping of BOLD Data on the Detection of Functional Connectivity Changes in Alzheimer's Disease and bvFTD

Resting-state fMRI results in neurodegenerative diseases have been somewhat conflicting. This may be due to complex partial volume effects of CSF in BOLD signal in patients with brain atrophy. To encounter this problem, we used a coefficient of variation (CV) map to highlight artifacts in the data, followed by analysis of gray matter voxels in order to minimize brain volume effects between groups. The effects of these measures were compared to whole brain ICA dual regression results in Alzheimer's disease (AD) and behavioral variant frontotemporal dementia (bvFTD). 23 AD patients, 21 bvFTD patients and 25 healthy controls were included. The quality of the data was controlled by CV mapping. For detecting functional connectivity (FC) differences whole brain ICA (wbICA) and also segmented gray matter ICA (gmICA) followed by dual regression were conducted, both of which were performed both before and after data quality control. Decreased FC was detected in posterior DMN in the AD group and in the Salience network in the bvFTD group after combining CV quality control with gmICA. Before CV quality control, the decreased connectivity finding was not detectable in gmICA in neither of the groups. Same finding recurred when exclusion was based on randomization. The subjects excluded due to artifacts noticed in the CV maps had significantly lower temporal signal-to-noise ratio than the included subjects. Data quality measure CV is an effective tool in detecting artifacts from resting state analysis. CV reflects temporal dispersion of the BOLD signal stability and may thus be most helpful for spatial ICA, which has a blind spot in spatially correlating widespread artifacts. CV mapping in conjunction with gmICA yields results suiting previous findings both in AD and bvFTD.