Unconstrained Respiratory Event Detection Using a Flexible Tactile Sensor

Sleep apnea-hypopnea syndrome (SAHS) is a highly prevalent and underdiagnosed respiratory sleep disorder. Traditional SAHS diagnostics such as polysomnography (PSG) tend to be complex and costly. The multiple wearable sensors could interfere with the user’s normal sleep, and the manual scoring process is time-consuming and labor-intensive. In this work, we propose a user-friendly and automated monitoring system to detect respiratory events during sleep using a bed-embedded flexible tactile sensor (FTS) in a noncontact mode. A deep learning model based on temporal convolutional network (TCN) is proposed to realize automated respiratory event detection and apnea-hypopnea index (AHI) estimation. The proposed deep learning model provides multi-resolution convolution to the recognition of respiratory patterns, thereby improving the accuracy and robustness of respiratory event detection, especially for irregular respiratory signals. The synchronous PSG result in a clinical setting is used to evaluate the performance of the proposed system. Our method achieves an accuracy of 98.0%, sensitivity of 95.0%, and specificity of 98.5% in detecting respiratory events. Compared to other noncontact approaches, our method provides an accurate and reliable respiratory profile to enable overnight automated respiratory event detection. The 51-subject experiment shows that the predicted AHI is highly correlated with the reference AHI (<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$ R^{{2}}={0.98} $ </tex-math></inline-formula>). The severity classification has shown the potential of the proposed system for SAHS diagnosis.