Prediction of Coronary Artery Calcium with CT-Images using Deep Learning

The coronary-artery-calcium(CAC)-score of CT is a popular tool for recognition of coronary-artery disease. Nevertheless, the present diagnostic approach utilizing CAC-score CT is laborious, since the radiologist needs to meticulously examine each CT image individually and verify the precise range. This research applies three convolutional neural network (CNN) models to 1200 CT scans of the cardiovascular system, including both normal and calcium-containing scans. As part of our experimental test, we categorize the CT-image data into three different sets: original CAC-score CT, cardiac-segmented, and cardiac-cropped images. The former set includes the entire rib cage, while the latter two sets exclude all but the heart region. The investigational test for determining the incidence of calcium in a CT-picture utilizing the Inception ResnetV2, VGG, and Resnet50 are yielded the best results with a 98.52% accuracy rate when the Resnet 50 model was applied to cardiac cropped image data. As a result, the authors of this work hope that future studies will pave the way for automated calcium study scores for each CAC-score CT as well as for the detection of calcium alone.