1D Convolutional Neural Network for Obstructive Sleep Apnea Detection

Obstructive sleep apnea (OSA) is a condition that influences many people and is determined by events of reduced respiratory airflow during sleep. However, electrocardiogram (ECG)-based detection of OSA is more suitable for noninvasive requirements and instrument limitations of wearable portable devices. As compared to earlier electrocardiogram (ECG) based OSA detection systems, deep learning approaches demonstrate substantial promises and advantages. This research presents a model for the detection of OSA from a single-lead ECG using a 1D convolutional neural network (1D-CNN). The performance of the proposed model is estimated on the well-established PhysioNet Apnea-ECG dataset. This dataset includes seventy ECG recordings, but only thirty-five released datasets are used in this research. The accuracy, precision, sensitivity, specificity, and F1 scores of the proposed model were evaluated as 94.77±1.35%, 93.80±2.253%, 92.55±4.57%, 96.14±1.66%, and 93.07±2.03%, respectively. The accuracy of the proposed model can be further improved for large datasets. Moreover, the proposed method can be implemented in wearable devices, which could monitor/detect OSA in the home setting and assist the medical expert.