Empowering Cardiologists with Deep Learning YOLOv8 Model for Accurate Coronary Artery Stenosis Detection in Angiography Images

Abstract

Coronary artery stenosis is a common manifestation of cardiovascular disease. For a prompt diagnosis and successful treatment of cardiovascular disease, precise stenosis identification and quantification are essential. Manual interpretation of angiography images is subjective and time-consuming, leading to potential errors and delays in diagnosis. In Present time, deep learning techniques such as You Only Look Once (YOLO) algorithms have shown promising results in image recognition tasks. In this type of study, YOLOv8 algorithm has been used for accurate detection of coronary artery stenosis in angiography images. The collected dataset of angiography images from Mendeley website. There are 236 images that have been utilised in the experiment, with 80% of the images being used for training and the remaining of 20% being used for testing. All of these techniques aid patients who have been given a diagnosis of coronary artery disease and annotate the images to show whether or not there is stenosis. The YOLOv8 algorithm was subjected to transfer learning. Which also includes fine tuning on pre trained model. The experimental results shown that after fine tuning of YOLOv8 algorithm provides better detection of coronary arteries stenosis. The Performance to this approach includes Precision, Recall and Mean Average Precision (mAP) with value of 82%, 58% and 65% respectively. The investigation of results shows that parameter tunned YOLOv8 can help to empower cardiologist for fast and reliable diagnosis of cardiovascular disease and ultimately leading to patient survival rate

Topics & Concepts