Radio Frequency Sensing with Genetic Manifestations for Assessing Circadian Rhythm Dysregulation from Mouth Breathing

The conventional approaches for mouth breathing monitoring are uncomfortable with invasive devices and prone to missed detections due to brief clinical observations. A novel radio frequency (RF) sensing technique is proposed for contactless monitoring of mouth breathing, and the results are consistent with genetic dysregulation manifestations. Compared to camera-based systems, RF sensing offers a lighting-independent, privacy-preserving, and cost-effective solution for long-term circadian rhythm monitoring. The RF signals contain macro-level physiological information on activity and vital signs. The technique holds promise for addressing the challenge of preventing mouth breathing in children, with its feasibility validated by experiments on rats. Latent features extracted via a variational autoencoder (VAE) reveal RF feature discriminability comparable to gene features. In support vector machine (SVM) classification tests, the accuracy of distinguishing mouth breathing samples by activity level is 0.853, and by breathing rate variability (BRV) is 0.867. Combining these two macro-level features enables complete classification with an accuracy of 1, equivalent to the gene-based features. Moreover, dynamic time warping (DTW) is introduced to link and align gene expression sequences with activity level sequences, facilitating the localization of gene regulatory events. The minimal DTW match points in the four groups of mouth breathing (MB), control (C), mouth breathing + rhythm (MBR), and control + rhythm (CR) are 11, 8, 7, and 5, respectively, indicating that mouth breathing is associated with more pronounced micro-gene macro-activity joint dysregulation.