Hybrid AI Framework for the Early Detection of Heart Failure: Integrating Traditional Machine Learning and Generative...

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Informatics

Background:

• Cardiovascular disease (CVD) is the leading global cause of mortality.
• Early detection and risk stratification are crucial for managing CVD, particularly heart failure (HF).
• Existing diagnostic methods require enhancement for improved accuracy and efficiency.

Purpose of the Study:

• To develop and evaluate a hybrid AI model for enhanced heart failure prediction.
• To integrate Convolutional Neural Networks (CNNs) and Large Language Models (LLMs) for comprehensive patient data analysis.
• To assess the model's accuracy, interpretability, and clinical relevance.

Main Methods:

• A hybrid AI model was developed, integrating CNNs for structured data and LLMs for unstructured clinical information.
• The model was trained and validated using patient health records.
• Explainable AI (XAI) techniques, including SHAP, were employed for model transparency.

Main Results:

• The hybrid AI model achieved a prediction accuracy of 95.1%.
• Performance metrics included high precision, recall, F1-score, and AUC-ROC, surpassing standalone models.
• Key identified predictors of HF included Chest Pain Type, Maximum Heart Rate (maxHR), and Exercise-Induced Angina.

Conclusions:

• Hybrid AI models offer a promising approach for accurate and interpretable cardiovascular diagnostics.
• The developed model can significantly aid healthcare professionals in early heart failure detection and risk stratification.
• Integrating AI in cardiovascular medicine has the potential to improve patient outcomes and support clinical decision-making.