Staging by the Book: Automatic Sleep Stage Classification Using Scoring Rules
Researchers have developed a new method for automatic sleep stage classification that strictly adheres to clinical scoring rules, offering a transparent alternative to opaque deep learning models. This rule-based approach operationalizes the American Academy of Sleep Medicine's scoring logic and provides natural-language justifications for its decisions. While its agreement with human scorers is lower than current deep learning methods, it serves as a valuable tool for auditing and governing AI-driven sleep staging. AI
IMPACT Provides a transparent, rule-based alternative for sleep staging, aiding in the auditing and governance of AI models in clinical settings.