Deep Learning of Multimodal Features Derived from Raw Polysomnography for Cognitive Impairment Prediction

Predicting Future Cognitive Impairment Using Whole-Night Sleep-Derived Outcomes

It is suggested that increased representational complexity does not necessarily improve cross-site generalization and that compact, clinically interpretable sleep features may help predict future cognitive impairment.

Frozen SleepFM and engineered polysomnographic features for prediction of subsequent cognitive impairment diagnoses

Multimodal Sleep Feature Extraction and Stratified Feature Selection for Dementia Prediction

Fusion of Age-Residualized Engineered Sleep Features with Raw EEG for Predicting Future Cognitive Impairment

A hybrid system to predict a diagnosis of cognitive impairment recorded one to six years after polysomnography and an engineered branch combined demographics, stage-specific electroencephalogram (EEG) features, and automated-annotation descriptors were developed.

Annotation-Guided Temporal Attention with a Sleep Foundation Model for Predicting Future Cognitive Impairment from Polysomnography

A complete prediction pipeline was developed that used a frozen SleepFM encoder for five-minute PSG segments, a bidirectional LSTM for temporal modelling, and annotation-guided attention based on algorithmic sleep stages and events for future cognitive-impairment prediction.

Predicting Future Cognitive Impairment from Polysomnography with Stage-Aware Temporal Selection