Identifying novel neuromarkers for detection of Amyotrophic Lateral Sclerosis from resting state magnetoencephalography: a functional connectivity study

In this study, functional connectivity (FC) analyses were carried out on resting state magnetoencephalography (MEG) data from 26 healthy and 26 Amyotrophic lateral sclerosis (ALS) patients during eyes-closed (RS-EC) and eyes-open (RS-EO) conditions for five minutes each to identify functional differences in brain connectivity between the two groups. After bandpass filtration (0.5-100 Hz) and artefact correction, the MEG data were segmented to 0.4 seconds of non-overlapping time windows, then source transformed using linearly constrained minimum variance beamformer. A total of 126 sources were reconstructed and FC on the source level time-series has been computed using phase synchrony measure for alpha and beta rhythms, as these are associated with awake-resting state and focus respectively. Significant alterations in FC were identified using Wilcoxon’s signed rank test with 5% threshold. Connections exhibiting weaker (hypo-connectivity) and stronger (hyper-connectivity) FC in ALS compared to healthy has been identified for potential neuromarker detection. It has been found, areas with cognitive control, muscle control, vision, concentration, emotional and sensory association exhibits hypo-connectivity while primary sensorimotor cortex and somatosensory area shows hyper-connectivity in ALS cohort. Previous study encompassing 4-30 Hz frequency band has reported hyper-connectivity in posterior cingulate cortex or the right visual and occipital regions as potential biomarkers for ALS during RS-EO condition, but to the best of authors’ knowledge, no hypo-connectivity study has been reported yet for source-level MEG and thus, hypo-connectivity in aforesaid regions unfold new scope for further research, especially during RS-EC condition which has not been explored much. Node strength of both hypo and hyper connected networks have been used to classify ALS and Healthy in Alpha and Beta bands and it has been observed for both states, hypo-connectivity exhibit at least 2.3% higher classification accuracy compared to hyper-connected regions when using SVM-RBF as the classifier.