Comparative Analysis of YOLOv8 Variants for Prostate Cancer Detection

Abstract

Prostate cancer (PCa) is still a big health problem all over the world, and it's important to find it early and correctly so that it can be treated well. Traditional ways of diagnosing don't always have the right level of sensitivity and specificity. This thesis talks about how to use of YOLOv8, a new framework for finding things in real time, to automatically find prostate cancer inhistopathology pictures. We used a dataset made just for prostate cancer to test three differen versions of YOLOv8: YOLOv8n, YOLOv8s, and YOLOv8m. The results of the experiment show that all three models work well. YOLOv8n has competitive accuracy and better efficiency, so it looks like a great option for clinical settings with few resources. This study adds to what we already know about using AI to help with medical diagnoses. It shows how useful YOLOv8 can be for making workflows for finding prostate cancer better. It also suggests areas for future research, like hyperparameter optimization, dataset expansion, and using explainable AI methods.