CAREER: Personalized and Trustworthy Fall Risk Assessment in Free-Living Environments Through Auto-Calibrated Insole and XAI

About This Grant

Among adults aged 65 years and older, falls are a serious, growing, and costly public health problem. Falling can lead to reduced mobility, functional decline, and loss of independence. Detecting and mitigating fall risks are effective for avoiding the serious consequences of falls. Smart insoles have the potential to improve upon current clinical fall risk assessments by taking less time and by enabling automatic assessments during a daily routine. However, existing smart insole systems face two major hurdles. First, the flexible pressure sensor arrays commonly used in them are prone to errors that slowly change over time, called “sensor drift.” Sensor drift makes it hard to accurately measure critical fall risk assessment values, such as gait, balance, and leg strength. Second, the artificial intelligence models that are widely used to assess fall risk lack clinically meaningful explanations of the results, which are needed to identify fall risk factors, inform effective interventions, and build trust. This project will develop and evaluate a new auto-calibrated insole system along with explainable artificial intelligence (XAI) models. The project will enable accurate, tailored, and reliable fall risk assessment during the normal daily routines of older adults. This research will result in fall risk assessment technologies that promote health, independence, and overall quality of life for older adults. This project also provides education and training for K-12, undergraduate, and graduate students, as well as elderly care workers, on how to develop and apply wearable technologies and XAI for healthcare. The technical aims of this project are divided into three thrusts. The first thrust is to develop a reconfigurable and multimodal insole system that can be auto-calibrated to accurately measure gait, balance, and lower-extremity strength over time. The auto-calibration will be achieved by a synergistic integration of flexible pressure sensor array data, inertial measurement unit data, and gait patterns. The second thrust is to create a context-aware XAI model to detect the risk of fall, identify fall risk factors, and inform personalized interventions. An interactive human-in-the-loop approach is proposed to enable the collection of comprehensive contextual information (for example, location, activities, and medication) with both the insole system and user engagement. Context-aware case-based reasoning is proposed to develop an expert-in-the-loop system that facilitates efficient interaction between medical professionals and the intelligent system, enables explainable fall risk assessment, and supports the development of personalized intervention plans. The final thrust is to establish a simulated free-living environment to comprehensively evaluate the long-term performance of the proposed system in personalized and trustworthy fall risk assessment. 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.