Amyotrophic lateral sclerosis (ALS) is notoriously difficult to diagnose early because its symptoms vary widely and current methods depend heavily on clinical judgment. Researchers have proposed Siam-ALS, a deep learning framework that uses EEG and eye-tracking signals to detect the disease in a more objective, data-driven way.

The framework is built on a Siamese network architecture, which learns to compare pairs of patient-generated signals and identify patterns indicative of ALS. Unlike many black-box models, Siam-ALS is designed to be explainable, offering insights into which features drive its decisions—a key requirement for clinical adoption.

By tapping into non-invasive EEG and eye-tracking data, the approach could complement existing assessments and potentially catch ALS earlier, when interventions might be more effective. The authors emphasize that this is a step toward reducing reliance on subjective clinical evaluation, though further validation is needed before it can be used in practice.