Health and Behavior: Wearables and Data Analytics for Personalized Therapeutic Management
Led by Behnaz Ghoraani, Ph.D.
This project aims to develop novel machine learning algorithms for converting sensor data from wearable devices and smartphone cameras into clinically actionable insights for managing various neurological conditions, including Parkinson’s disease and Alzheimer’s disease. The objective is to facilitate personalized therapy adjustments in real-world settings, outside of traditional healthcare facilities. The intellectual merit lies in the creation of robust algorithmic and engineering frameworks capable of handling data from multiple sensor types, including inertial measurement units and smartphone camera imagery. These frameworks need to be efficient and scalable to operate under the constraints of battery-powered wearables and limited bandwidth. Identifying diagnostically relevant activity patterns within vast, high-density datasets poses a fundamental challenge that this project seeks to overcome. Additionally, ensuring data privacy, especially with smartphone camera data, and aligning the data-driven insights with clinical standards present substantial hurdles. Our strategy hinges on a dynamic, cloud-based computing architecture that intelligently distributes computational workloads. This infrastructure not only addresses the immediate challenges related to data size and computational demand, but also provides a platform that can incorporate advancements in machine learning and data analytics methodologies. The broader impacts are manifold. The project aligns closely with national and global precision medicine objectives, offering the prospect of fundamentally new, individualized therapeutic strategies that could enhance healthcare delivery and quality of life for millions suffering from neurological disorders.
Under the leadership of Dr. Ghoraani, the project will engage up to two REU participants. One participant will concentrate on optimizing machine learning algorithms for on-device execution, while the second will explore cloud-based computational offloading strategies. Both will acquire hands-on experience in sensor data analysis, computational optimization techniques, and cloud computing architectures. Participants will contribute to a pioneering system capable of analyzing movement and behavioral data in patients’ natural environments, unlocking new avenues for personalized healthcare management.