Remote Kinesiology for Improving Human Health with Auto-locating Compliant Motion Tracking Stickers and Artificial Intelligence

Carnegie-Mellon University, Pittsburgh PA

Investigators

Abstract

PROJECT SUMMARY/ABSTRACT Body mounted inertial measurement units (IMUs) enable human motion tracking and kinematic analysis beyond the traditional healthcare setting of a hospital or lab. This is a powerful platform for studying how physical conditions and diseases affect the kinematics of daily activities. Clinical adoption of these body mounted IMUs can lead to new treatment methods and more efficient utilization of healthcare resources [1]– [3]. Bulky and uncomfortable form factors of body mounted IMUs, however, compromise patient adherence and limit clinical adoption [4], [5]. Furthermore, the inability of body mounted IMUs to be removed/replaced on the body without corrupting most kinematic models and activity classification algorithms, creates the possibility of significant error during unsupervised use or extended durations of wear [3]. Such challenges with wearability and usability interfere with the ability to fully realize the potential of body mounted IMU motion systems for clinical applications. In an effort to address this need, I will apply and evaluate the rapid fabrication techniques, developed by my lab, to manufacture a fully flexible, extensible, and soft IMU sticker [6]. I will also investigate the use of artificial intelligence/machine learning to detect the anatomical placement location of body mounted IMUs through common physical therapy exercises. To accomplish this, I will leverage the natural kinematic constraints of the body such as axes of rotation, range of motion, and angular velocity, which are unique to specific regions of the body, to classify the IMU locations with supervised learning [7]. I have already manufactured a network of unobtrusive body mounted IMUs stickers and developed an approach of using globally referenced quaternions to track, visualize, and analyze human motion in real-time. I have empirically determined that the accuracy of this system is comparable to other IMU-based motion tracking systems, and I have already begun collecting motion tracking data of common physical therapy exercises. I hypothesize that a more compliant IMU sticker developed using rapid fabrication methods and a K-nearest neighbor (KNN) classifier trained with motion tracking data (unaltered or decomposed using Principal Component Analysis (PCA)) will detect the anatomical location of randomly placed body mounted IMU stickers through common physical therapy exercises. The result of this work will allow me to improve the health outcomes of people in the future through kinematics, IMUs, and AI.