A proximal policy optimisation algorithm-based algorithm for cardiovascular disorders detection.

Area of Science:

• Cardiology
• Artificial Intelligence
• Sports Medicine

Background:

• Cardiovascular diseases (CVDs) pose significant risks to athletes.
• Accurate detection of CVDs in athletes is crucial for preventing adverse events.
• Existing diagnostic methods may face challenges with imbalanced datasets.

Purpose of the Study:

• To develop and evaluate a novel artificial neural network (ANN) model for assessing CVD risk in athletes.
• To address the challenge of imbalanced classification inherent in medical datasets.
• To improve the reliability and accuracy of cardiovascular disorder detection in athletic populations.

Main Methods:

• Utilized a mutual learning-based artificial bee colony (ML-ABC) algorithm for initial weight setting.
• Employed proximal policy optimization (PPO) to ensure stable and efficient ANN updates.
• Formulated classification as a decision-making process with rewards for minority class identification to handle imbalance.

Main Results:

• The proposed ANN model demonstrated superior performance across multiple datasets.
• Achieved high accuracies: 0.88 (Polyclinic dataset), 0.86 (NCAA dataset), and 0.82 (NHANES dataset).
• Outperformed existing models in cardiovascular disorder detection for athletes.

Conclusions:

• The novel ANN approach effectively detects cardiovascular diseases in athletes.
• The ML-ABC and PPO integration enhances model reliability and handles class imbalance.
• This methodology advances cardiovascular disorder detection and clinical applications in sports medicine.