Machine Learning-Based Risk Stratification Model to Guide Cardiopulmonary Exercise Training in Cardiovascular...

Area of Science:

• Cardiology
• Exercise Physiology
• Artificial Intelligence in Medicine

Background:

• Precise risk stratification is vital for cardiovascular patients undergoing cardiac rehabilitation.
• Current risk classification methods vary, leading to inconsistencies.
• Standardization is needed for safe and effective cardiopulmonary exercise training.

Purpose of the Study:

• To develop and validate machine-learning (ML) models for standardizing risk classification in cardiopulmonary exercise training.
• To assist clinicians in minimizing exercise-related risks for cardiovascular patients.
• To optimize treatment efficacy during cardiac rehabilitation programs.

Main Methods:

• Retrospective data from 1163 patients across three institutions were analyzed.
• Machine learning models including logistic regression, SVM, XGBoost, and random forests were employed.
• Feature selection using LASSO and hyperparameter optimization via grid search were utilized.

Main Results:

• The XGBoost model demonstrated strong performance in replicating clinician-assigned risk stratification.
• LASSO feature selection enhanced model accuracy to 79.31% with an AUC of 89.40%.
• Key predictors identified include ventilatory efficiency, maximal systolic blood pressure, maximal oxygen consumption, and weight.

Conclusions:

• An ML model was successfully developed to categorize exercise-related risk types in cardiopulmonary rehabilitation.
• This AI tool can support clinical decision-making for safe and effective patient training.
• The model aids in optimizing cardiac rehabilitation strategies for cardiovascular patients.