A Machine Learning Approach to Developing an Accurate Prediction of Maximal Heart Rate During Exercise Testing in...

Area of Science:

• Exercise Physiology
• Cardiovascular Health
• Biomedical Data Science

Background:

• Maximal heart rate (HR max) is a key indicator of exercise intensity and effort.
• Current prediction formulas for HR max have limitations in accuracy.
• Improving HR max prediction is crucial for effective exercise prescription and clinical assessment.

Purpose of the Study:

• To enhance the accuracy of maximal heart rate (HR max) prediction.
• To evaluate the efficacy of machine learning (ML) algorithms for HR max estimation.
• To compare ML model performance against established HR max prediction formulas.

Main Methods:

• Utilized a large dataset (n=17,325) from the Fitness Registry of the Importance of Exercise National Database.
• Applied various ML algorithms including lasso regression, neural networks, support vector machine, and random forests.
• Evaluated models using cross-validation, RMSE, RRMSE, Pearson correlation, and Bland-Altman analysis.

Main Results:

• All ML models demonstrated improved HR max prediction accuracy, reducing RMSE and RRMSE compared to standard formulas.
• Random forests (RF) model showed the highest improvement in prediction accuracy.
• ML predictions exhibited significant correlations with measured HR max and lower bias than traditional equations.

Conclusions:

• Machine learning, especially the RF model, offers a superior method for predicting HR max.
• Readily available clinical measures can be effectively used with ML for HR max prediction.
• This ML-driven approach warrants consideration for clinical applications to refine HR max assessments.