Exploring EEG signal complexity as a novel diagnostic tool for neuropsychiatric and neurodegenerative disorders

Abstract

In the current landscape of diagnostic tools, neuroimaging devices are considered the apex, however these techniques have significant limitations in cost and temporal resolution. Scalp electroencephalography (EEG), while lacking in spatial resolution, is relatively low-cost with a high temporal resolution, capable of taking measurements in the temporal range of post-synaptic potentials. Many EEG researchers utilise amplitude based measures to assess abnormalities, and while useful information is present in the domain of signal intensity, these techniques may not capture subtle signal variations irrespective of frequency and power. Complexity can be captured with a variety of techniques such as fractal measures including Higuchi fractal dimension (HFD), box counting, and detrended fluctuation analysis or through multiscale entropy. The current study utilised two open source datasets. The first dataset contains resting state recordings from Alzheimer’s Disease and frontotemporal dementia (FTD) patients. The second dataset consists of resting state recordings from major depressive disorder patients. We found that spectral features in AD and FTD displayed a theta and alpha dominance with a reduction in alpha power, whereas MDD displayed increased beta power compared to healthy controls in each respective dataset. The contrast between dementias and MDD was further displayed with AD and FTD showing reductions in complexity, with MDD showing increased complexity. Further, differentiation between dementias was achievable with observing rostrocaudal asymmetry with box counting. With a combination of techniques, the development of an EEG profile for a given disorder can be attained with commonly used and novel metrics.