Running: how to prevent injuries with sensors and AI

The most common injury that affects runners stems from overuse, not falls. Early warning signs include changes in motion. Successful injury prevention could pick up on this aspect by detecting and responding to these deviations at an early stage. It is the focus of the "Smart Injury Prevention" project.

In this MEDICA-tradefair.com interview, Dr. Lina Rahlf talks about "Smart Injury Prevention" and explains how sensors and AI can help prevent running injuries.

Dr. Rahlf, what is the goal of the "Smart Injury Prevention" project?

Dr. Lina Rahlf: The project aims to develop artificial intelligence software for the personalized prediction of running injuries using innovative applications of machine learning.

Running is one of Germany's most popular sports, although it comes with a great risk of injury. There are many causes for these injuries. However, running biomechanics are increasingly a topic of conversation in this setting. By attaching inertial sensors to the body, our goal is to assess specific biomechanical factors related to the running action using machine learning algorithms. Using inertial sensor technology to measure running biomechanics is a common method and has proven beneficial in previous research. By applying artificial intelligence to analyze the data, we expect to achieve innovative and successful software.

The process works like this: The inertial sensor attaches to the runner's waist with a waistband and continuously measures biomechanical running parameters such as stride frequency or ground contact times and detects the athlete’s personal running pattern. Thanks to artificial intelligence software we can identify deviations in the running pattern. Another goal is to use a large amount of data on individual running patterns and additional injury data to create a risk assessment that facilitates injury prevention.

What causes these deviations?

Rahlf: Deviations in running patterns can have different causes. For example, it might be due to runner's fatigue, which may increase injury risk. We know that overuse injuries are far more common than traumatic injuries in the running world. An injury that has not yet fully healed can also prompt deviations in the running pattern.

What types of sensors do you use?

Rahlf: We use inertial sensors that are commonly used in sports science. The sensors can measure acceleration and angular velocity of movements. This data is meant to detect and derive biomechanical parameters such as stride frequency, stride length or ground contact times.

Who are your project partners?

Rahlf: Humotion GmbH is responsible for the sensor and software development. PI Probaligence GmbH is likewise involved as a software development specialist and also in charge of the project's machine learning processes. We at the Department of Human Movement Science at the University of Hamburg evaluate the measurement system.

The project has been ongoing since October 2020. What is the status of the project?

Rahlf: We are in the first year of the project and thus in its early stage. The overall project duration is three years. We are about to implement our large-scale empirical study in which we will equip all participating runners with a sensor and collect data from training sessions and subsequent injuries over an extended period. The machine learning processes are designed to use the initial data to detect any changes in running patterns and analyze correlations between the changes and any resulting injuries.

Thanks to the innovative analysis of running parameters and injuries, the project provides an important approach on how to design and improve preventive measures to reduce the risk of sports injuries and – after successful project completion – has the potential to take root beyond the sport of running.