CCSS: Discovery of Individualized Disease Features for Personalized Health Monitoring

Florida Atlantic University, Boca Raton FL

Investigators

Abstract

Parkinson's disease (PD) affects approximately six million people globally, which is predicted to double by 2040. Therapeutic interventions such as medication and Deep Brain Stimulation need to be progressively adjusted, from three times daily in early stages to as often as every two hours in advance cases. These therapeutic adjustments are based on the patients' response-to-medication, which is gathered from patient interviews. However, the patient interview can be unreliable and suffer from recall bias, resulting in over- or under-treatment of the patients. Therefore, this project proposes to develop methodologies and algorithms to be used along with wearable sensors (battery-powered inertial measurement units) to monitor the response-to-medication of Parkinson's disease patients in their natural environment. This proposal is a significant contribution to Precision Medicine national efforts. Success of this project may result in fundamentally new individualized therapy adjustment strategies, thereby providing considerable improvement in both healthcare delivery and quality of life for the millions of patients afflicted by PD. In addition, an integrated education and outreach program is designed to promote the research-informed education, dissemination, and engagement activities at the Florida Atlantic University, a Hispanic-serving Institute, to increase a pipeline of underrepresented minority students in electrical and computer engineering and K-12 students. This project develops novel methodologies and algorithms to translate the wearable sensors data into clinically actionable information about the response-to-medication of Parkinson's disease patients enabling personalized therapeutic adjustments outside hospital settings. The proposed approach is a major departure from the current efforts as it will develop an integrated sensing and computational framework to explore raw sensor data for identifying patient-specific disease features. The proposed framework uses novel algorithms, combined with wearable sensors, to update the disease features according to each patient's disease severity as the patient is being monitored. The transformative design addresses current technical obstacles. First, the project develops an innovative sensing and computation framework to explore the multiset fused raw sensor data for individualized disease features that could significantly improve response-to-medication detection accuracy. This new approach integrates tensor decomposition to explore raw, multiset sensor data for data-driven disease features and reinforcement learning to derive decisions on when to update the individualized disease features. Second, the project will extend the developed innovative sensing and computation framework to support the implementation of the algorithms with battery-powered wearable sensors. The new framework is based on a distributed design (to off-load computationally expensive components of the algorithms to a server) and a lightweight architecture (to reduce on-device computation load) and enables the wearable sensor device to transmit the raw sensors' data to a server only when the disease features are being updated. The key transformative aspect of the proposed research is the advancement of data analysis tools for personalized monitoring of PD patients while translating the research into a clinically applicable system. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.