Spin-off Brailsports

AI platform estimates lactate thresholds and fitness level based on training data

Brailsports, an imec, University of Antwerp and University of Ghent spin-off, is developing a platform that supports athletes and their coaches to determine optimal training volume and adjust training schedules based on data. Using AI models, the platform accurately estimates fitness and fatigue of the individual athlete, improving performance and reducing the risk of injury. For the development of the platform, it is working closely with cycling team Lotto-Dstny.  

For an athlete, the right training volume is key. Both over- and undertraining weakens fitness level. Personal coaches often base training schedules on the lactate test, the classic exercise test in which the concentration of lactic acid in the blood is measured while the athlete makes an increasing effort, until he finally goes all out. And also for popular sports apps, the result of an exercise test is taken as a basis to estimate metabolism and different training zones. 

But there are two main problems with such tests: they are quite invasive, requiring the athlete to go to extreme fatigue. And secondly, they are snapshots. If an athlete falls ill, there is a chance that training sessions have become too hard. Or it may also be the other way around: an athlete can get better faster than expected. They might not make the most of their training sessions, as they are based on threshold values that are too conservative. 

Personalized models

Brailsports has developed a model that can estimate the impact on an individual athlete's lactate thresholds after each workout. To do so, the model uses internal (heart rate) and external (wattage, tempo) parameters. At the same time, Brailsports estimates an athlete's fatigue.

Current apps and smartwatches often already indicate a ‘training load’, based on averages from scientific literature (e.g. a training fatigue remains in the body for an average of 7 days, a fitness effect for 42 days). However, it is obvious that an elite athlete will recover faster from training than a recreational cyclist. By including a large number of data points in the analysis, Brailsports builds personalized models that estimate fatigue after training more accurately.  

Scientific approach

The name Brailsports refers to Dave Brailsford, the performance manager of British Team Sky during Chris Froome’s golden era. Brailsford brought the scientific approach into cycling: pushing the limits to improve performance, based on thorough data analysis.

For its algorithmic model, the spin-off builds on the AI expertise developed within imec and unites it with knowledge about sports physiology, training theory and coaching. Co-founders are Erika Lutin (CEO Brailsports), Bart Nonneman (sports coaching & AI, imec), Tim Verdonck (statistics & AI, UAntwerp), Jan Boone (physiology and sports training, UGent), Steven Latré (fellow AI at imec).

Co-founder Tim Verdonck (UAntwerp/imec): ‘Brailsports won't replace the coach any time soon, but it will make their work even more effective. By smoothly extracting insights from the data and automating part of the work, schedules can be continuously adjusted based on individual data’.

Co-founder Professor Jan Boone (UGent): ‘By supplementing the classic lactate test with a data-driven approach, we are taking a step forward in terms of performance.’ 

Co-founder and CEO Erika Lutin: ‘For now, we're focusing mainly on coaches of professional cycling teams, but the same data-driven approach is also going to have great benefits for recreational runners and cyclists, who often also set themselves quite ambitious goals and don't want to get injured.’

www.brailsports.com