RTSense: A Fabric-based Wearable System for Return-To-Sport Assessment of Anterior Cruciate Ligament Injuries

Abstract

Return-to-sport (RTS) assessment is critical for patients recovering from anterior cruciate ligament (ACL) injuries, who face a high risk of re-injury. A key requirement is accurate monitoring of knee flexion during high-dynamic tasks where knee motion is coupled across three anatomical planes. Motion capture offers gold-standard accuracy but requires specialized infrastructure and controlled environments; inertial measurement units are portable but often degrade due to sensor displacement under intense motion. Fabric sensors are comfortable and stable, yet existing single-plane designs cannot reliably separate the main flexion from coupled multi-plane motion. We present RTSense, the first fabric-based system capable of accurately estimating knee flexion during high-dynamic RTS tasks. RTSense places multiple fabric sensors in a cross-plane layout to capture complementary biomechanical cues, addressing two key challenges: fabric data distribution shifts and the bio-structural coupling of multi-plane knee motion. Building on this layout, RTSense combines a fabric-specific data augmentation framework to handle inter-subject and intra-subject variability with a disentangling model that leverages cross-plane features to recover knee flexion angles. To validate RTSense, we collaborate with a medical center and collect data from 35 participants, including ACL patients and healthy controls across athlete and non-athlete groups. RTSense achieves an average RMSE of 3.96° for knee flexion, 3.71° for peak flexion, and yields 100% correct RTS decisions, demonstrating its potential to support reliable RTS decision-making.