RAPID: Collaborative Research: Data Analytics for Mechano-acoustic and Physiological Monitoring of COVID19 Symptoms

On March 11, 2020, the World Health Organization (WHO) declared Covid-19 a global pandemic. The novelty of the Covid-19 pathogen, diversity of its transmission modes, lack of universal testing capability, absence of a vaccine, lack of medical supplies and personal in hospitals needed for effective treatment, prevent us from predicting the duration and recurrence of the pandemic. Our proposed RAPID project addresses a key issue with pandemics in general and Covid-19 in particular – the limited capacity of any health-care system – whereby hospitals and health-care providers struggle to accommodate the huge surge in patients needing treatment. We propose to address this challenge by developing low-cost sensing and in-situ data analytics platform technologies to enable individualized, distributed and continuous health monitoring of individuals and thereby provide early disease detection capabilities in-residence, minimize the number of unnecessary hospital visits, and act as an early warning system to enable preventive measures to be taken early on especially for high-risk individuals such as seniors and elderly individuals who are most vulnerable to Covid-19. Our proposed technology will acquire mechano-acoustic signatures of the underlying physiological processes (such as those measured by a stethoscope) and precision kinematics of core-body motions using individualized skin-mounted soft electronics compute platform (“The Patch”) from individuals tested for Covid-19, develop low-complexity data analytic algorithms using a hybrid of digital signal processing (DSP) and machine learning (ML) to detect the presence of infection with high accuracy, and deploy these algorithms on such resource-constrained compute platforms for rapid diagnosis. The Rogers Research Lab (http://rogersgroup.northwestern.edu/) at Northwestern University has currently deployed the Patch at the Shirley Ryan AbilityLab (SRAL) with plans to expand to Northwestern Memorial and Lurie Children's Hospital. However, the Patch is severely constrained in terms of computational resources and energy, and lacks sophisticated data analytics. The Shanbhag Research Group at the University of Illinois at Urbana-Champaign (http://shanbhag.ece.illinois.edu/) will leverage their recent work on designing low-complexity fixed-point DSP and ML algorithms for vision applications and energy-efficient circuit architectures to: 1) develop low-complexity fixed-point ML algorithms for Covid-19 specific analytics using patient data acquired by the current deployment of the Patch; 2) develop methods for energy-efficient embedding of such algorithms on to the Patch and associated hardware; 3) and work with the Rogers Research Group to deploy the ML-based Covid-19 specific data analytics in the field with real-life patients. This project qualifies for NSF RAPID given the urgency of monitoring and mitigating the spread of the Covid-19 and the urgent need to process massive volumes of data currently being acquired by the Patch, e.g., close to 0.5 TB of high quality signals acquired over more than 1000 hours of continuous monitoring data just in the first 1.5 weeks of deployment. As we scale and expand, these numbers will increase very rapidly and will need to be processed efficiently to enable rapid and accurate diagnosis, and thereby contribute significantly to limiting the impact of the Covid-19 pandemic. Intellectual Merit: Our RAPID proposal brings together innovations in flexible materials, mechano-acoustic sensing devices, energy-efficient inference architectures, and low-complexity