Bone Fracture Detection And Localisation Using Enhanced YOLOv8 Model

Abstract The use of Machine Learning and Artificial Intelligence in multiple areas including medical imaging sector has become the center of attention, specifically for medical purposes. The potential to utilize AI-powered tools offers many benefits, such as: enhancing the decision support for diagnosis, as AI can aid clinicians in identifying fractures with high accuracy. It also helps reducing discrepancies in diagnosis by assisting in subtle or complex cases, AI helps minimize diagnostic inconsistencies. These AI tools are particularly beneficial in emergency radiology, where speed and precision are essential for optimal patient outcomes. This research aims to contribute towards novel deep machine learning and AI based methods for bone fracture detection and localisation based on enhancing the YOLOv8 architecture. By leveraging an innovative algorithm pipeline comprising data augmentation and iterative model optimization based on cosine-annealing strategy, it was possible to achieve significant improvements in fracture detection and localisation system performance. Availability of such objective imaging tools, can ultimately lead to improving patient outcomes and facilitating wider adoption of intelligent systems in clinical settings.