Unraveling motor imagery brain patterns using explainable artificial intelligence based on Shapley values

• DOI: 10.1016/j.cmpb.2024.108048
• Publication Date: 2024-01-30T17:42:17Z
• Authors: Sergio Pérez-Velasco; Diego Marcos-Mart...; Eduardo Santamarí...; Víctor Martínez-C...; Selene Moreno-Cal...; Roberto Hornero
• Keywords: Motor Imagery

Motor imagery (MI) based brain-computer interfaces (BCIs) are widely used in rehabilitation due to the close relationship that exists between MI and motor execution (ME). However, underlying brain mechanisms of remain not well understood. Most MI-BCIs use sensorimotor rhythms elicited primary cortex (M1) somatosensory (S1), which consist an event-related desynchronization followed by synchronization. Consequently, this has resulted systems only record signals around M1 S1. could involve a more complex network including sensory, association, areas. In study, we hypothesize superior accuracies achieved new deep learning (DL) models applied decoding rely on focusing broader activation brain. Parallel success DL, field explainable artificial intelligence (XAI) seen continuous development provide explanations for DL networks success. The goal study is XAI combination with extract information about patterns from non-invasive electroencephalography (EEG) signals.