Although the resting-state networks of patients with disorders of consciousness (DOC) have been widely investigated, the underlying neural mechanisms remain unclear. Here we aimed to explore the static and dynamic alterations in the regional brain activity in patients with DOC and detect the diagnostic ability of each index. Nineteen patients in the vegetative state, 19 in the minimally conscious state (MCS), and 41 healthy controls were recruited for this study. The fractional amplitudes of the low-frequency fluctuation (fALFF) and dynamic fALFF (dfALFF) values were computed, and intergroup differences were detected by analysis of variance. Sub-frequency analysis (slow-4 band and slow-5 band) of fALFF was also performed. Machine learning classifiers were established based on these measures to explore the classification accuracies between patients and controls. The fALFF and dfALFF analyses showed significant intergroup differences in the medial prefrontal gyrus, precuneus, left angular gyrus, and right middle cingulate cortex (MCC), whereas only the dfALFF analysis revealed aberrations in the right inferior frontal gyrus (IFG), right angular gyrus, left supramarginal gyrus (SMG), and left middle occipital gyrus (MOG). Sub-frequency analysis suggested a potential frequency dependent alteration in the default mode network (DMN). The fALFF model exhibited a higher classification accuracy (ACC) (90.50%) than the dfALFF model (86.29%). The combination of fALFF and dfALFF did not improve classification performance. The strength and stability of regional brain activities were disrupted in patients with DOC. Our findings demonstrate that dynamic analysis may reveal more pathological regions and provide a better understanding of the pathophysiologic mechanisms of DOC.
Keywords: disorders of consciousness; fALFF; fMRI; machine learning; temporal dynamics.
Copyright © 2021 IBRO. Published by Elsevier Ltd. All rights reserved.