Combining wearable sensor signals, machine learning and biomechanics to estimate tibial bone force and damage during running

Abstract

There are tremendous opportunities to advance science, clinical care, sports performance, and societal health if we are able to develop tools for monitoring musculoskeletal loading (e.g., forces on bones or muscles) outside the lab. While wearable sensors enable non-invasive monitoring of human movement in applied situations, current commercial wearables do not estimate tissue-level loading on structures inside the body. Here we explore the feasibility of using wearable sensors to estimate tibial bone force during running. First, we used lab-based data and musculoskeletal modeling to estimate tibial force for ten participants running across a range of speeds and slopes. Next, we converted lab-based data to signals feasibly measured with wearables (inertial measurement units on the foot and shank, and pressure-sensing insoles) and used these data to develop two multi-sensor algorithms for estimating peak tibial force: one physics-based and one machine learning. Additionally, to reflect current running wearables that utilize running impact metrics to infer musculoskeletal loading or injury risk, we estimated tibial force using a commonly measured impact metric, the ground reaction force vertical average loading rate (VALR). Using VALR to estimate peak tibial force resulted in a mean absolute percent error of 9.9%, which was no more accurate than a theoretical step counter that assumed the same peak force for every 2 running stride. Our physics-based algorithm reduced error to 5.2%, and our machine learning algorithm reduced error to 2.6%. Further, to gain insights into how force estimation accuracy relates to overuse injury risk, we computed bone damage expected due to a given loading cycle. We found that modest errors in tibial force translated into large errors in bone damage estimates. For example, a 9.9% error in tibial force using VALR translated into 104% error in estimated bone damage. Encouragingly, the physics-based and machine learning algorithms reduced damage errors to 41% and 18%, respectively. This study highlights the exciting potential to combine wearables, musculoskeletal biomechanics and machine learning to develop more accurate tools for monitoring musculoskeletal loading in applied situations.

Wearables for Biomechanical Performance Optimization and Risk Assessment in Industrial and Sports Applications

“…Regardless of recent applications, health and safety decision makers are looking to expand industry applications for biomechanical wearables [ 9 ] and their implementation effectiveness seen in the sports sector. Athletes and their performance staff were some of the first user groups to adopt this technology [ 47 , 48 ] and are constantly seeking innovative methods to improve athletic performance through biomechanical data collections [ 49 ].…”

“…Running is an activity that most athletes do to train, and there are many wearable solutions—such as instrumented insoles—to measure ground reaction forces (GRF) of each footfall. Matijevic et al (2020) performed a study using IMUs, pressure-sensing insoles, and machine learning to quantify peak tibial force [ 9 ]. One of the leading causes of injuries for runners and athletes is tibial overuse.…”

“…Risk assessment wearables often overlap with performance enhancement devices, and the data collected can be used to enhance the understanding of performance while assessing the risk of injury. Risk assessment wearables include pressure sensors [ 8 ] and IMU’s that host several sensors (e.g., accelerometers, gyroscopes, and magnetometers; [ 9 ]).…”

Experimental recommendations for estimating lower extremity loading based on joint and activity

“…Understanding how measurement errors impact inverse dynamics calculations is critical as the field continues to explore research questions that are best studied outside of traditional biomechanics laboratory. Estimating joint loading using low-cost motion capture techniques or wearable devices has emerged as promising tools to study patients in more natural settings, both in the clinic and in the real world (Hullfish et al, 2020; Matijevich et al, 2020; Renner et al, 2019). These approaches compare favorably to those made using gold-standard techniques across a wide range of clinically relevant activities (Drazan et al, 2021; Hullfish and Baxter, 2020; Martin et al, 2018).…”

Ankle Angle Prediction Using a Footwear Pressure Sensor and a Machine Learning Technique

“…The prediction of falls using a multivariable logistic regression [ 19 ] was explored using observed gait patterns of stroke patients. Center of pressure foot insoles and IMUs used in a linear regression algorithm predicted the force upon the joint with high accuracy [ 20 ]. Fabric-based strain sensors in combination with a linear regression algorithm had a high accuracy when compared with an angle extracted from a high-speed camera [ 21 ].…”