Neural Signals Classification through Machine Learning: Distinguishing Voluntary and Involuntary Movements EEG Correlates

This thesis is focused on the development of a Machine Learning (ML) model with the primary goal of accurately classifying voluntary and involuntary movements by means of the EEG signal, contributing to the larger project of diagnosing the level of consciousness in non-responsive patients. Some tens of thousands of patients across the globe have been incorrectly labeled as "awake but unaware" for long periods, when they have actually remained conscious the entire time [1]. The high prevalence of misdiagnosis highlights the need for improved diagnostic methods in this field. The goal is to minimize misdiagnoses, leading to improved patient outcomes and care in the field of Brain Injuries. This project is based on the Readiness Potentials (RPs): slow negative EEG potentials found in the seconds preceding voluntary actions, as indicators of volition and consciousness. In addition to the classification task, this work also undertakes the evaluation of the performance of the improved EEGLAB plug-in (Matlab toolbox). Proper signal preparation is crucial before classification: the signals go through pre-processing to remove artifacts and noise, followed by feature extraction. These features are selected based on their statistical significance or chosen from existing literature. The effectiveness of the optimal pre-processing chain is assessed in the classification task, ensuring the reliability and validity of the classification results. Subsequently, feature selection methods are applied to refine the feature set. The machine learning algorithms employed are: Decision Tree (DT), Support Vector Machines (SVM) and K-Nearest Neighbors (k-NN). These models are evaluated and compared based on their performance. The developed ML model can be part of the BCI framework: it offers opportunities for improving the diagnosis in non-responsive patients and gains a more reliable understanding of their state of consciousness and cognitive state. [1] Owen, Adrian M. “Improving diagnosis and prognosis in disorders of consciousness.” Brain : a journal of neurology vol. 143,4 (2020): 1050-1053. doi:10.1093/brain/awaa056