[Paper Review] Enhancing the Detection of Coronary Artery Disease Using Machine Learning

Karan Kumar Singh, Nikita Gajbhiye+1 more|arXiv (Cornell University)|Mar 6, 2026

Artificial Intelligence in Healthcare0 citations

TL;DR

The study compares Bi-LSTM, GRU, and a Bi-LSTM+GRU hybrid for non-invasive CAD detection from 3D CCTA data, with the hybrid model achieving the highest accuracy (97.07%).

ABSTRACT

Coronary Artery Disease (CAD) remains a leading cause of morbidity and mortality worldwide. Early detection is critical to recover patient outcomes and decrease healthcare costs. In recent years, machine learning (ML) advancements have shown significant potential in enhancing the accuracy of CAD diagnosis. This study investigates the application of ML algorithms to improve the detection of CAD by analyzing patient data, including clinical features, imaging, and biomarker profiles. Bi-directional Long Short-Term Memory (Bi-LSTM), Gated Recurrent Units (GRU), and a hybrid of Bi-LSTM+GRU were trained on large datasets to predict the presence of CAD. Results demonstrated that these ML models outperformed traditional diagnostic methods in sensitivity and specificity, offering a robust tool for clinicians to make more informed decisions. The experimental results show that the hybrid model achieved an accuracy of 97.07%. By integrating advanced data preprocessing techniques and feature selection, this study ensures optimal learning and model performance, setting a benchmark for the application of ML in CAD diagnosis. The integration of ML into CAD detection presents a promising avenue for personalized healthcare and could play a pivotal role in the future of cardiovascular disease management.

Motivation & Objective

• Develop an ML model to predict CAD using demographic, clinical, and imaging data.
• Compare the performance of different ML algorithms (e.g., DT, RF, SVM, Neural Networks) for CAD prediction.
• Identify significant predictors via feature selection to improve model accuracy.
• Enhance data quality with preprocessing steps (handling missing data, normalization, outlier handling) to boost learning outcomes.

Proposed method

• Assemble a dataset of 1000 3D CCTA images from a Siemens 128-slice scanner.
• Apply data cleaning, normalization, and standardization (z-score) to prepare features for modeling.
• Use Pearson correlation matrix for feature selection to reduce dimensionality.
• Implement Bi-LSTM, GRU, and a Bi-LSTM+GRU hybrid classifier for CAD prediction.
• Evaluate models using accuracy, precision, recall, F1-score, and ROC-related metrics.

Figure 1: Block diagram of proposed methodology

Experimental results

Research questions

• RQ1 Can ML models accurately predict CAD from non-invasive imaging data combined with clinical features?
• RQ2 Which ML architecture (Bi-LSTM, GRU, or a hybrid) yields the best CAD detection performance on the dataset?
• RQ3 How do feature extraction and preprocessing affect model performance and generalization?
• RQ4 How do the proposed models compare to existing CAD detection methods in terms of accuracy and other metrics?

Key findings

• Bi-LSTM achieved 92.7% accuracy, 92.9% precision, 92.7% recall, and 92.7% F1-score.
• GRU achieved 93.9% accuracy, 94% precision, 93.9% recall, and 93.8% F1-score.
• Hybrid Bi-LSTM+GRU achieved 97.07% accuracy, 94.13% precision, 94.07% recall, and 94% F1-score.
• The hybrid model outperformed individual Bi-LSTM and GRU models and prior methods reported in the literature.
• The study highlights non-invasive data integration as a means to improve clinical decision-making and CAD diagnosis.

Figure 2: Training and validation loss of the Bi-LSTM model across epochs