Optimized hybrid RNN-GRU model for predictive diagnosis of cardiovascular disease

Area of Science:

• Cardiology and Artificial Intelligence
• Biomedical Engineering
• Data Science in Healthcare

Background:

• Cardiovascular disease (CVD) is a leading global cause of death, with India facing a significant burden.
• Accurate and efficient heart disease risk prediction is crucial for timely intervention and improved patient outcomes.

Purpose of the Study:

• To develop and evaluate a hybrid deep learning model for enhanced heart disease risk prediction.
• To improve the accuracy and efficiency of identifying individuals at risk of cardiovascular disease.

Main Methods:

• Utilized a hybrid Recurrent Neural Network (RNN)-Gated Recurrent Unit (GRU) deep learning model on a dataset of 918 samples.
• Applied data preprocessing including outlier correction (IQR), normalization, and Synthetic Minority Over Sampling Technique (SMOTE) for dataset balancing.
• Employed GridSearchCV with 10-fold cross-validation for model fine-tuning.

Main Results:

• The hybrid RNN-GRU model achieved superior performance, outperforming individual RNN and GRU models.
• Achieved high accuracy (99.6%), F1 score (99.6%), precision (99.6%), and recall (99%).
• Demonstrated performance significantly higher than previously reported accuracies of 87% and 97%.

Conclusions:

• The hybrid RNN-GRU model effectively extracts temporal features from cardiac signals, crucial for accurate risk prediction.
• The model shows significant potential for enhancing clinical decision-making in early and precise cardiac disease detection.
• Highlights the importance of robust data preprocessing techniques in developing effective predictive models.