Estimating regional cerebral blood flow using resting-state functional MRI via machine learning

Abstract

Perfusion MRI is an important modality in many brain imaging protocols, since it probes cerebrovascular changes in aging and many diseases; however, it may not be always available. Here we introduce a method that seeks to estimate regional perfusion properties using spectral information of resting-state functional MRI (rsfMRI) via machine learning. We used pairs of rsfMRI and arterial spin labeling (ASL) images from the same elderly individuals with normal cognition (NC; n = 45) and mild cognitive impairment (MCI; n = 26), and built support vector machine models aiming to estimate regional cerebral blood flow (CBF) from the rsfMRI signal alone. This method demonstrated higher associations between the estimated CBF and actual CBF (ASL-CBF) at the total lobar gray matter (r = 0.40; FDR-p = 1.9e-03), parietal lobe (r = 0.46, FDR-p = 8e-04), and occipital lobe (r = 0.35; FDR-p = 0.01) using rsfMRI signals of frequencies [0.01-0.15] Hertz compared to frequencies [0.01-0.10] Hertz and [0.01-0.20] Hertz, respectively. We further observed significant associations between the estimated CBF and actual CBF in 24 regions of interest (p < 0.05), with the highest association observed in the superior parietal lobule (r = 0.50, FDR-p = 0.002). Moreover, the estimated CBF at superior parietal lobule showed significant correlation with the mini-mental state exam (MMSE) score (r = 0.27; FDR-p = 0.04) and decreased in MCI with lower MMSE score compared to NC group (FDR-p = 0.04). Overall, these results suggest that the proposed framework can obtain estimates of regional perfusion from rsfMRI, which can serve as surrogate perfusion measures in the absence of ASL.

Neuro-anatomical and neuro-molecular markers in salience network and their integration in machine learning-based prediction of cognitive dysfunction in mild cognitive impairment

“…SVM regression modeling was performed using the libSVM package (Chang and Lin, 2011) using 10-folds cross-validation. SVM cross-validation was consistent with our previous study (Chand et al, 2020b). In this approach, participants were divided into 10-folds, SVM model was trained in 9-folds data, and then the trained model was tested on the remaining 1-fold (test) data.…”

“…This procedure was repeated for all combinations of 10-folds data. Gaussian kernel function was used and SVM model parameters were optimized within each training set using 10-folds nested cross-validation grid search method (Chand et al, 2020b). SVM regression modeling was also performed using SN volumes or SN DVRs alone as input features.…”

Salience network anatomical and molecular markers are linked with cognitive dysfunction in mild cognitive impairment

“…SVM cross-validation was consistent with our previous study. 37 In this approach, participants were divided into 10-fold data, SVM model was trained in 9-folds data, and then the trained model was tested on the remaining 1-fold (test) data. This procedure was repeated for all combinations of 10-fold data.…”

Stress Changes the Resting-State Cortical Flow of Information from Distributed to Frontally Directed Patterns

“…However, further analytical studies are necessary to allow for harnessing the full utility of TE for quantification of the effect of various psychological and mental disorders on brain function. This is in particular crucial to enable the use of TE as a useful feature for real-time data-driven approaches to decoding of the brain activity [57][58][59][60].…”

“…From a broader perspective, these results posit the use of TE as a potential diagnostic/prognostic tool in identification of the effect of stress on distributed brain networks that are involved in the brain responses to stress. This observation becomes more intriguing considering the recent surge in application of machine learning and statistical frameworks to decoding of the brain activity [57][58][59][60].…”