automated detection of coronary artery stenosis and stent size using convolutional neural networks

Introduction: coronary artery stenosis is a key marker of coronary heart disease, making accurate detection vital for timely diagnosis and effective treatment. precise stent sizing is also essential for successful angioplasty procedures. this study aims to automate the identification and measurement of coronary artery stenosis and stent size using convolutional neural networks (cnns). methods and materials: the dataset included 400 angiography clips, with 200 clips showing angioplasty procedures and 200 normal angiography clips, sourced from both the left and right coronary arteries. specifically, 100 images corresponded to the left anterior descending (lad) artery, 66 to the right coronary artery (rca), and 34 to the left circumflex artery (lcx). participant ages ranged from 35 to 87 years, with stenosis severity ranging from 65% to 99% obstruction. to automate stenosis detection and stent sizing, we employed several cnn architectures, including inception v3, vgg16, and a simple cnn model, all trained on annotated angiography images. performance was evaluated using metrics such as mean squared error (mse) and accuracy. results: the proposed cnn-based approach showed high accuracy and low error rates in detecting coronary artery stenosis and estimating stent dimensions. key findings indicate that these models were able to effectively recognize stenosis regions and provide precise stent size estimates, showing potential for real-world diagnostic applications. conclusion and discussion: this automated diagnostic method reduces reliance on manual assessment, enhancing patient outcomes in coronary artery disease management. the findings represent a significant step forward in cardiovascular diagnostics, suggesting that future research could improve model architectures or expand datasets for enhanced generalization and accuracy.