AI-based prediction of VO₂ max from 24-h Holter ECG recording

Area of Science:

• Cardiovascular physiology
• Space medicine
• Biomedical engineering

Background:

• Spaceflight induces cardiovascular (CV) deconditioning, affecting heart and blood vessels due to microgravity.
• This deconditioning leads to reduced exercise capacity, specifically impacting maximum aerobic power (VO2 max).
• Current methods for monitoring CV health in space are limited.

Purpose of the Study:

• To investigate if non-exercise 24-hour Holter ECG recordings can predict microgravity-induced CV deconditioning.
• To develop machine learning models for forecasting VO2 max deterioration.
• To establish a potential method for regular CV monitoring during space missions.

Main Methods:

• Utilized pooled data from baseline and post-bed rest periods of Head-Down Tilt campaigns.
• Developed and trained Machine Learning models using multiple features from Holter ECG data.
• Validated models to predict VO2 max after deconditioning.

Main Results:

• Machine learning models demonstrated good prediction accuracy for VO2 max (RMSE < 5 mL/kg/min).
• The approach successfully identified VO2 max deterioration during simulated microgravity (bed rest).
• The findings highlight the potential of ECG-derived features for assessing CV health.

Conclusions:

• 24-hour Holter ECG recordings can be leveraged to predict cardiovascular deconditioning.
• Machine learning analysis of ECG data offers a viable, non-exercise method for monitoring VO2 max.
• This approach has significant implications for astronaut health management in space and clinical applications on Earth.