A Robust Framework for Data Generative and Heart Disease Prediction Based on Efficient Deep Learning Models.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Cardiovascular Diagnostics

Background:

• Accurate heart disease diagnosis relies on biomarkers like ECG and blood pressure.
• Deep learning models show promise for heart disease diagnosis but struggle with scarce or imbalanced data.
• Existing diagnostic models are prone to prediction bias due to limited datasets.

Purpose of the Study:

• To propose two deep learning frameworks, GAN-1D-CNN and GAN-Bi-LSTM, for enhanced heart disease diagnosis.
• To address data scarcity and imbalance issues in heart disease datasets.
• To improve the accuracy and reliability of automated heart disease prediction.

Main Methods:

• Utilized a generative adversarial network (GAN) to augment the Cleveland heart disease dataset.
• Developed two deep learning architectures: GAN-1D-CNN and GAN-Bi-LSTM.
• Trained and evaluated the models on the augmented dataset for heart disease prediction.

Main Results:

• The GAN-1D-CNN model achieved 99.1% accuracy, sensitivity, specificity, F1-score, and 100% AUC.
• The GAN-Bi-LSTM model achieved 99.3% accuracy, 99.3% sensitivity, 99.2% specificity, 99.2% F1-score, and 100% AUC.
• Principal Component Analysis (PCA) significantly reduced prediction times.

Conclusions:

• The proposed GAN-based frameworks effectively augment limited datasets for improved heart disease diagnosis.
• Both GAN-1D-CNN and GAN-Bi-LSTM demonstrate high diagnostic performance and reliability.
• These frameworks offer a robust solution for heart disease prediction in data-constrained scenarios.