A Transformer-Based Approach to Diagnose Amyotrophic Lateral Sclerosis via Electroencephalogram Analysis

Neurodegenerative disorders are the leading cause of physical disability worldwide. In particular, amyotrophic lateral sclerosis (ALS) is one such condition that significantly impacts the quality of life for millions by impairing nerve cell function in the central nervous system. Despite extensive research, ALS remains difficult to diagnose in its early stages and the exact cause is largely unknown, with contemporary methods taking up to 15 months for a definitive diagnosis. Electroencephalogram (EEG) analysis, a non-invasive method for recording brain electrical activity, has shown promise in identifying subtle neural changes associated with neurodegenerative disorders. Transformers, known for their ability to capture complex data dependencies, offer a novel framework for analyzing EEG signals with high temporal resolution. This study introduces a Transformer-based approach to diagnose ALS by leveraging the EEG and eye-tracking dataset of ALS patients (EEGET-ALS), comprising a total of 1,989 recordings. The model achieved exceptional accuracies of 98.49% in training and 99.33% in both validation and testing. Furthermore, with an area under the curve (AUC) of 0.9963, precision of 100.0%, and recall of 96.36% in testing, the model demonstrates promise in enhancing the accuracy and timeliness of ALS diagnosis with a low rate of false positives and false negatives. Overall, this approach represents a significant advancement in the field of neurodegenerative disease diagnosis, potentially improving patient outcomes and quality of life through a two-minute recording.