Automatic Detection of Obstructive Sleep Apnea Based on Multimodal Imaging System and Binary Code Alignment

There are many patients with obstructive sleep apnea syndrome, which has caused concern. When it occurs, the nasal airflow disappears, and the breathing action of the chest and abdomen still exists. Therefore, we propose a multimodal imaging system in tandem with binary code alignment to remotely and automatically detect the obstructive sleep apnea. The RGB-thermal imaging module is applied to monitor the breathing conditions in the nose or mouth area. The depth camera is utilized to monitor the undulating state of the chest cavity. The obtained respiratory waveform is afterwards subjected to the limiting filter to suppress noisy signals that are not related to the mission objective. The signals derived from thermal camera and depth camera are pooled together for binary code alignment. When these two synchronously acquired bimodal signals do not match well, the obstructive sleep apnea is present. Our method achieves the satisfactory performance with accuracy, precision, recall and F1 of 91.38 %, 96.15 %, 86.21 %, and 90.91 %, respectively. This indicates that the proposed multimodal imaging system coupled with binary code alignment has a potential for non-contact and automatic detection of obstructive sleep apnea.