Improving ALS Diagnosis Based on Human EEG Signal Analysis with Machine Learning

Amyotrophic lateral sclerosis (ALS) is a progressive neurodegenerative disease of the motor neurons in the central nervous system, which causes considerable morbidity and mortality. The basis for effective management and intervention rests on early and accurate diagnosis. This paper seeks to propose a method to improve diagnostic precision for ALS using an approach with machine learning techniques coupled with Electroencephalography (EEG) signals. EEG data from 10 ALS patients and 7 healthy controls are broadly classified based on state-of-the-art EEG-based brain signal processing and advanced algorithms of classification. Using the K-Nearest Neighbors and Support Vector Machine classifiers results in an accuracy of $\mathbf{9 4. 1 1 \%}$ and $100 \%$, respectively. We further shared common EEG features across ALS, Parkinson, and Multiple Sclerosis, which we termed as common mode features (CMF), suggesting the possibility of a shared biomarker for these neurodegenerative conditions. Our results emphasize the capability of EEG-based Machine Learning models in distinguishing ALS from healthy conditions, hence opening an avenue for its early diagnosis and distinction from other neurological disorders. This underlines the great potential that machine learning, when combined with EEG analysis, may have for transformational ALS diagnostics.