Alzheimer's disease (AD) is a prevalent degenerative disease that is both idiopathic and highly debilitating, calling for early detection and intervention before extensive irreversible brain damage occurs. Here, a novel functional complexity measure based on information theory was applied to mark a disease-related reduction in stochastic low frequency (0.01-0.08 Hz) BOLD signal fluctuations at rest. As AD progresses, BOLD signal complexity is expected to decrease in affected regions, alongside an overall increase in brain amyloid burden.
In this study, a retrospective analysis was performed on 18F-AV-45 PET and rs-fMRI data from 65 subjects (30 males, 35 females; mean age ± SD: 74 ± 7.4 years), across four clinical groups (Control, Early Mild Cognitive Impairment (EMCI), MCI, and AD). The regional BOLD signal complexity measures, or transient information (TI) values, were determined from the change in uncertainty of BOLD pattern prediction over time (i.e. block entropy growth rate).
The main findings in this study included an expected regional reduction in functional complexity with increasing amyloid burden, as well as an unexpected, albeit non-significant, global increase in functional complexity with increased amyloid burden and disease progression. One third (27 out of 82) of the cortical and subcortical grey matter regions presented a significant (>90% CI) effect of brain amyloid load on BOLD signal complexity, including regions associated with disease-related dysfunction in memory, language processing, attention, behavior, somatosensory functions, and motor functions. With disease diagnosis taken into account, only the EMCI group indicated a decrease in global BOLD signal complexity with increased brain amyloid load.
Analysis of rs-BOLD signal complexity has the potential to provide a more accurate representation of disease state than current amyloid plaque or structural measures, as well as identify the regions altered by disease pathology. Based on the reduced signal complexity reported in regions previously linked to disease pathology (precuneus/posterior cingulate, lateral temporal lobe, and frontal regions), further study of this novel metric is advised. With validation, the BOLD complexity analysis metric can be attuned to use as a cognitive biomarker in the clinical setting, potentially improving disease diagnosis, treatment monitoring, and evaluation of future treatment options.