Long-Term Sleep Respiratory Monitoring by Dual-Channel Flexible Wearable System and Deep Learning-Aided Analysis

The long-term sleep respiratory monitoring has been implemented by a dual-channel flexible wearable system, which consists of ultrathin flexible respiratory sensors and a packaged system box. Subjects can directly wear the system and sleep at home all night, collecting the respiratory signals from two nasal cavities independently and simultaneously. One night (8 h) sleep respiratory signals are completely collected from a suspected obstructive sleep apnea-hypopnea syndrome (OSAHS) patient. The results demonstrate that the apnea and hypopnea signals can be accurately distinguished and easily extracted from night breathing signals. For the first time, the different types of nasal cycle are clearly detected from four subjects (aged 27-42) via a flexible wearable monitoring system, which include the classic, in-concert, and mixed type. The 1-D convolutional neural network (1DCNN) model is constructed and utilized for the disease classification and identity recognition with an accuracy of 96.67% and 93.67%, respectively. Therefore, a high accuracy, long-term, and dual-channel sleep respiratory monitoring has been achieved. The proposed wearable respiratory monitoring system shows promising applications (APPs) in early screening of OSAHS, further research in the nasal cycle, and big data interconnection.