DREAM

Explainable Neural Network for Sleep Apnea Detection from ECG

DREAM (Deep Residual Explainable Apnea Model) is an explainable neural network for detecting sleep apnea using single-lead ECG signals. The system provides clinically interpretable predictions.

Key Contributions

  • Novel architecture combining residual blocks with attention for single-lead ECG analysis
  • Explainability layer for clinical decision support, highlighting ECG segments driving predictions
  • Evaluated on the PhysioNet Apnea-ECG benchmark dataset
  • Published in Biomedical Signal Processing and Control, Elsevier (2026)

References