Optimizing YOLOv8 for Enhanced Abnormality Detection in Abdominal CT Imaging: A Deep Learning Perspective.

Abstract

This research delves into enhancing the capabilities of YOLOv8, a cutting-edge object detection model, for the specific purpose of identifying abnormalities in abdominal computed tomography (CT) scans. With the rising need for swift and precise diagnosis in medical imaging, especially in high-pressure clinical settings, there is a growing reliance on artificial intelligence (AI)-powered systems to support medical decision-making. YOLOv8 stands out due to its advanced architecture, particularly its anchor-free mechanism and streamlined backbone, which collectively offer significant advantages over conventional diagnostic techniques. By training this model on a comprehensive dataset of well-annotated CT images, we were able to evaluate its ability to not only increase detection accuracy but also substantially reduce the time taken to reach diagnostic conclusions. The model’s effectiveness is validated through critical performance metrics such as precision, recall, F1 score, and mean average precision (mAP), all of which indicate strong reliability in identifying diverse pathological features. The outcomes of this study underscore the potential of integrating YOLOv8 into clinical diagnostic workflows, where it can act as a valuable assistant to radiologists—enabling earlier detection of serious conditions and improving overall patient outcomes.