Automatic Detection Of Respiratory Effort Related Arousals From Polysomnographic Recordings

A reliable automatic categorization of respiratory effort is paramount for sleep-disordered breathing characterization from polysomnography. A respiratory effort related arousal (RERA) is a subtle breathing obstruction associated with an arousal. For identification of RERAs we focused on: chest and abdomen EMGs, airflow, and EEG; monitoring changes in ECG and SaO(2). The quality of signals was assessed to overcome sensor associated problems and sporadic individual signal losses. We evaluated an ensemble learning and a deep learning approach using the engineered feature set trained on the 994 available records. The initial ensemble model was officially scored achieving a 0.081 area under the precision-recall curve (AUPRC) on a test set, whereas for the recurrent neural network model the average AUPRC was 0.295, obtained using 10-fold cross-validation.