Injuries are the worst nightmare for athletes. An anterior cruciate ligament tear in the knee can mean 6 months of downtime and even the end of a sports career in football, rugby, or handball, among others. However, injuries also pose a serious sporting and economic problem, especially for professional sports. A study conducted by the company ProFootballDB indicated that 73% of the stars in major football leagues were injured during the 2020-2021 season.
It's not hard to imagine the impact of those injuries on teams like Real Madrid or FC Barcelona.
Indeed, it's no wonder that injury prevention has become a priority for sports clubs. This has also driven research and development, as well as innovation in various fields such as monitoring techniques and medical imaging, among others.
One of the most promising lines of work for this purpose is the use of big data and artificial intelligence to develop models that detect when an athlete is at risk of injury.
However, the various existing initiatives in the market have yet to take off and remain fertile ground for R&D and innovation. Why is that? Perhaps we can find the answer if we momentarily imagine ourselves in the lead-up to a hypothetical World Cup final between Argentina and Portugal in Qatar.
In both cases, during the preparation meeting for the match, the medical team, summoning up their courage, raises their hands.
They don't know how to say, hint, or suggest, if anything, that the artificial intelligence system tells them that Messi in one case and Cristiano Ronaldo in the other should stay on the bench as they have a high probability of getting injured. "According to the algorithm..." they start to say, but they don't finish the sentence, the coach cuts them off with a sharp gesture.
Obviously, this is a highly unlikely scenario, as medical teams, even if they use artificial intelligence systems, don't rely solely on them to make decisions.
That is one of the main challenges facing artificial intelligence today and is driving a global movement at the scientific, technical, and political levels: to be reliable.
Within that movement, explainable artificial intelligence (XAI) is included. It is a tremendously active field of research and development in which techniques and tools are being developed for machine learning algorithms to explain the results they provide. To put it simply, machine and deep learning models, such as neural networks, often function as black boxes: you train them with data, they learn, and when presented with new data, they give you a result. What happens inside the network is opaque. Why these results occur is unknown. Is it something the model has detected in an ultrasound of Messi's thigh? Does CR have an excess of accumulated internal load?
What XAI aims to do is answer that question. Its application to the world of
injury prevention in sports opens up numerous opportunities for innovation.
12/01/2023
Enrique Alcántara/ HUB 4ICVESPORT