In sports physiology, the key concepts of aerobic and anaerobic threshold are necessary to assess an athlete’s performance and individual exercise strain, as well as to optimize the training load. Traditionally, these thresholds for a three-zone exercise model are determined either by invasive blood lactate samples taken from the fingertip or by monitoring changes in respiratory gas variables in laboratory conditions. A combination of these methods is also possible. The development of computational methods and the proliferation of electronics available to consumers, such as sports watches and heart rate monitors, creates an interesting alternative to these traditional threshold determination methods.

It has long been known that heart rate variability (HRV), i.e. the time between successive beats of the heart, varies in a complex way. In this thesis, a recently developed method, dynamical detrended fluctuation analysis (DDFA) is used to analyze HRV. This method examines the dependence of correlations in a time series as functions of scale and time. The thesis studies and compares the scalability results provided by DDFA with the thresholds measured using conventional methods including lactate and ventilation measurements. In addition, the effect of raw data filtering on the operation of the DDFA method is tested. The aim of this thesis is to demonstrate the qualitative link between HRV and thresholds.

The material for the experimental part of the thesis consists of 15 direct tests of maximum oxygen uptake on bicycle ergometers and 9 tests on a treadmill. The tests were carried out under laboratory conditions, in which both respiratory gas variables and blood lactate concentrations were used to determine individual thresholds for each volunteer subject. The intervals between consecutive heartbeats (RR intervals, RRIs) during the test were also recorded and this RRI time series was analyzed using the DDFA.

As indicated by previous studies, the results show clear changes in HRV during exercise. Despite individual differences, these changes are systematic, allowing them to be utilized in the assessment of thresholds as well. In addition, adequate filtering of data prior to the calculation of DDFA is shown to be necessary. On the other hand, even excessive filtering does not significantly reduce the quality of the results. In conclusion, it is possible to use the results of DDFA to define predictions of thresholds without tedious lactate measurements and/or gas tests. Forecasts and indicators will be developed in further research.