LETI: LETI Scientists Suggested Increasing the Endurance of Athletes Using Machine Learning

The method created at LETI will allow calculating the anaerobic threshold individually for each athlete with high accuracy, which is the most important indicator for monitoring physical fitness.

The anaerobic threshold is the highest level of intensity that a person can withstand for a long time without accumulating a significant amount of lactate, which reduces the overall physical condition of the body.

Therefore, one of the tasks of professional athletes during the training process is to constantly increase the anaerobic threshold to enhance the overall endurance of the body. However, the exact determination of the anaerobic threshold is a difficult task since it depends on many factors, including the physiological characteristics of a particular athlete and the system of training methods used by the coaches.

“Using machine learning methods, we have developed a model that can improve the accuracy of predicting the anaerobic threshold, which is one of the main criteria for monitoring the training of professional athletes. This development will increase the effectiveness of the training process.”

Dmitry Kaplun, Senior Lecturer of the Department of Automation and Control Processes at LETI
Before the creation of the model, researchers of the Research Institute of Hygiene, Occupational Pathology and Human Ecology and the Mechnikov Northwestern State Medical University collected data. Scientists tested athletes simulating the training process and physiological state when reaching the anaerobic threshold. Data on heart rate, blood oxygen saturation, and other information were collected using sensors. More than 1.2 thousand tests were conducted to collect data.

After that, the obtained data were used by LETI scientists to train a predictive model. The researchers used four different machine learning methods to achieve the highest possible accuracy of data analysis. The resulting model can determine the physiological parameters (in quantitative terms) that limit the increase in the anaerobic threshold during training. To do this, scientists used a special explanatory algorithm LIME (Local Interpretable Model-Agnostic Explanations). The results are published in the journal Biomedical Signal Processing and Control.

“The developed model for determining the anaerobic threshold allows us to identify patterns that affect the test result and therefore predict the course of the training process so that the athlete acts effectively without flaws or overwork and enters the competition at the peak of form,” explains Dmitry Kaplun.

Now scientists are working to improve the accuracy of the created model by applying other more complex machine learning algorithms.