Bone Fracture Detection using Deep Learning:
Musculoskeletal injuries, particularly bone fractures, represent a significant portion of medical emergencies worldwide. Quick and precise identification is vital for effective clinical intervention. This project introduces an automated diagnostic tool that utilizes Deep Learning to detect fractures in X-ray images. Built on a modern technical stack, the application uses Python for the core logic, Flask as the web framework, and a responsive HTML/CSS/JS interface to provide medical practitioners with an efficient, user-friendly screening platform.
The technical backbone of this system is the YOLOv8 (You Only Look Once) object detection model, selected for its superior speed and localized accuracy. We trained this model on a comprehensive dataset of over 3,700 images, enabling it to categorize seven distinct types of upper-limb conditions, including fractures in the humerus, shoulder, forearm, and wrist, as well as positive indicators for elbows and fingers.
With a achieved mean Average Precision (mAP50) of 86%, this system offers a reliable secondary opinion for radiologists and clinicians. By automating the preliminary screening process, this software aims to bridge the gap in healthcare accessibility, ensuring that accurate fracture detection is available even in resource-limited settings.