A non-exercise based V02max prediction using FRIEND dataset with a neural network.

Area of Science:

• Cardiology
• Exercise Physiology
• Computational Intelligence

Background:

• Maximum oxygen consumption (VO2 max) is a vital measure of cardiorespiratory fitness.
• VO2 max also predicts risks for adverse health outcomes, including hypertension, obesity, and diabetes.
• Current methods for assessing VO2 max, such as cardiopulmonary exercise tests, can be complex and less accessible for routine clinical practice.

Purpose of the Study:

• To develop and compare computational intelligence models, specifically neural networks, for predicting maximum oxygen consumption.
• To establish simpler and accurate models as an alternative to standard cardiopulmonary exercise tests.
• To facilitate the use of VO2 max prediction in daily clinical practice for population health stratification.

Main Methods:

• Implementation and comparison of three predictive models: Wasserman/Hansen equation, linear regression, and non-linear neural networks.
• Evaluation of model performance using a large dataset (12,262 individuals) from the FRIEND database.
• Assessment of model accuracy through sensitivity and specificity calculations.

Main Results:

• Neural network models demonstrated superior performance in predicting maximum oxygen consumption compared to traditional equations and linear regression.
• The study confirmed the potential of computational intelligence for accurate VO2 max estimation.
• High accuracy metrics were achieved, indicating the reliability of the developed neural network models.

Conclusions:

• Neural networks offer a highly accurate and efficient method for predicting maximum oxygen consumption.
• These models provide a valuable alternative to traditional cardiopulmonary exercise tests, enhancing clinical utility.
• The developed models can aid in the risk stratification of the general population, improving preventative healthcare strategies.