DeepCAD: A Medical Image Analysis Approach for Coronary Artery Disease Detection in CTA

Abstract

Coronary Artery Disease (CAD) continues to be a leading cause of cardiovascular mortality worldwide. Early and accurate diagnosis is crucial for effective treatment planning and improved patient outcomes. While Computed Tomography Angiography (CTA) serves as a widely adopted non-invasive imaging modality for evaluating CAD, the manual interpretation of CTA scans is time-consuming and subject to significant inter-observer variability, often affecting diagnostic consistency and timeliness. To address these challenges, our research introduces an advanced deep learning-based framework aimed at automating the detection of CAD using CTA images. The proposed methodology leverages the power of Convolutional Neural Networks (CNNs) integrated with attention mechanisms to ensure robust and precise feature extraction and classification. These attention modules help the model focus on diagnostically relevant regions, thereby improving interpretability and accuracy.A comprehensive image preprocessing pipeline has been implemented to enhance the input data quality. This includes vessel segmentation to isolate coronary arteries, contrast enhancement for clearer visualization, and noise reduction techniques to suppress irrelevant artifacts. Such preprocessing significantly boosts the model’s performance by ensuring that critical features are preserved and highlighted during training. The model is trained and validated using large-scale, annotated datasets to ensure generalizability and statistical reliability. Quantitative evaluation demonstrates that our approach achieves superior accuracy, sensitivity, and specificity compared to traditional diagnostic methods and baseline deep learning models. These results highlight the potential of AI-driven systems to reduce diagnostic delays, improve consistency, and assist clinicians in making informed decisions.