Search for: All records

Amid the COVID-19 pandemic, it has been reported that greater than 35% of patients with confirmed or suspected COVID-19 develop postacute sequelae of SARS CoV-2 (PASC). PASC is still a disease for which preliminary medical data are being collected—mostly measurements collected during hospital or clinical visits—and pathophysiological understanding is yet in its infancy. The disease is notable for its prevalence and its variable symptom presentation, and as such, management plans could be more holistically made if health care providers had access to unobtrusive home-based wearable and contactless continuous physiologic and physical sensor data. Such between-hospital or between-clinic data can quantitatively elucidate a majority of the temporal evolution of PASC symptoms. Although not universally of comparable accuracy to gold standard medical devices, home-deployed sensors offer great insights into the development and progression of PASC. Suitable sensors include those providing vital signs and activity measurements that correlate directly or by proxy to documented PASC symptoms. Such continuous, home-based data can give care providers contextualized information from which symptom exacerbation or relieving factors may be classified. Such data can also improve the collective academic understanding of PASC by providing temporally and activity-associated symptom cataloging. In this viewpoint, we make a case for the utilization of home-based continuous sensing that can serve as a foundation from which medical professionals and engineers may develop and pursue long-term mitigation strategies for PASC. 

This demonstration presents a working prototype of VitalHub, a practical solution for longitudinal in-home vital signs monitoring. To balance the trade-offs between the challenges related to an individual’s efforts thus compliance, and robustness with vital signs monitoring, we introduce a passive monitoring solution, which is free of any on-body device or cooperative efforts from the user. By fusing the inputs from a pair of co-located UWB and depth sensors, VitalHub achieves robust, passive, context-aware and privacy-preserving sensing. We use a COTS UWB sensor to detect chest wall displacement due to the respiration and heartbeat for vital signs extraction. We use the depth information from Microsoft Kinect to detect and locate the users in the field of view and recognize the activities of the respective users for further analysis. We have tested the prototype extensively in engineering and medical lab environments. We will demonstrate the features and performance of VitalHub using real-world data in comparison with an FDA approved medical device. 

Vital signs (e.g., heart and respiratory rate) are indicative for health status assessment. Efforts have been made to extract vital signs using radio frequency (RF) techniques (e.g., Wi-Fi, FMCW, UWB), which offer a non-touch solution for continuous and ubiquitous monitoring without users’ cooperative efforts. While RF-based vital signs monitoring is user-friendly, its robustness faces two challenges. On the one hand, the RF signal is modulated by the periodic chest wall displacement due to heartbeat and breathing in a nonlinear manner. It is inherently hard to identify the fundamental heart and respiratory rates (HR and RR) in the presence of higher order harmonics of them and intermodulation between HR and RR, especially when they have overlapping frequency bands. On the other hand, the inadvertent body movements may disturb and distort the RF signal, overwhelming the vital signals, thus inhibiting the parameter estimation of the physiological movement (i.e., heartbeat and breathing). In this paper, we propose DeepVS, a deep learning approach that addresses the aforementioned challenges from the non-linearity and inadvertent movements for robust RF-based vital signs sensing in a unified manner. DeepVS combines 1D CNN and attention models to exploit local features and temporal correlations. Moreover, it leverages a two-stream scheme to integrate features from both time and frequency domains. Additionally, DeepVS unifies the estimation of HR and RR with a multi-head structure, which only adds limited extra overhead (<1%) to the existing model, compared to doubling the overhead using two separate models for HR and RR respectively. Our experiments demonstrate that DeepVS achieves 80-percentile HR/RR errors of 7.4/4.9 beat/breaths per minute (bpm) on a challenging dataset, as compared to 11.8/7.3 bpm of a non-learning solution. Besides, an ablation study has been conducted to quantify the effectiveness of DeepVS.