Integrating EPSOSA-BP neural network algorithm for enhanced accuracy and robustness in optimizing coronary artery...

Area of Science:

• Cardiovascular Medicine
• Artificial Intelligence
• Biomedical Engineering

Background:

• Coronary artery disease (CAD) poses a significant global health challenge, particularly for aging populations.
• Existing heart disease prediction models face limitations in feature selection and predictive precision.
• Early identification of high-risk individuals is crucial for timely intervention and resource optimization.

Purpose of the Study:

• To introduce an advanced BP neural network algorithm, EPSOSA-BP, for enhanced heart disease prediction.
• To improve feature selection and learning capabilities in cardiovascular risk assessment models.
• To validate the efficacy of the proposed algorithm against classical methods and datasets.

Main Methods:

• Development of the EPSOSA-BP algorithm, integrating particle swarm optimization, simulated annealing, and a particle elimination mechanism.
• Implementation of single-hot encoding and Principal Component Analysis for robust feature selection and learning.
• Testing and validation using the UCI and Kaggle datasets, including ablation experiments.

Main Results:

• The EPSOSA-BP model achieved high prediction accuracies of 93.22% (UCI) and 95.20% (Kaggle).
• The algorithm demonstrated superior convergence speed, sensitivity, and specificity compared to classical optimization techniques.
• Ablation studies confirmed the effectiveness of the data preprocessing and feature selection strategies.

Conclusions:

• The EPSOSA-BP algorithm offers a significant advancement in the precision of heart disease prediction.
• The model shows potential for early identification of high-risk patients, improving outcomes and resource allocation.
• Despite implementation challenges, the algorithm is promising for regular cardiac risk monitoring in clinical and community settings.