Graph Strength for Identification of Pre-training Desynchronization

Motor processing can result in coordinated changes of ongoing/decreasing brain neural activity or event-related de/synchronization (ERD/S) that raises brain responses over either contralateral brain hemisphere. Because of the affordability and time resolution provided, Electroencephalographic (EEG) signals are commonly used to acquire motor imagery paradigms. However, the widely-known condition of low-noise signals makes detection and spatial localization of ERD/S challenging. Here, to deal with the high variability between subjects, we propose to perform group analysis of graph representations extracted from the weighted Phase lock Index. Statistical thresholding of the functional connectivity estimates is also accomplished to improve the assessments of phase synchronization between electrodes. The obtained results on a real-world database with 50 individuals show that the proposed methodology improves interpretation of ERD/S, allowing better prediction of motor imagery ability in subjects having low skills for practicing this paradigm.