Automated Bone Fracture Detection with AlexNet on Multi-Region Radiographic Analysis

Bone fractures are a significant medical issue that requires accurate diagnosis to ensure appropriate treatment and quick recovery. Early detection of fractures is crucial in minimising complications and improving patient outcomes. This study utilises a dataset comprising 10,581 radiographic images, covering a broad spectrum of anatomical regions such as the lower limb, upper limb, lumbar spine, hips, and knees. The dataset is divided into 4,606 images labelled as fractured and 4,640 images labelled as Not fractured for training, 337 fractured and 492 non-fractured images for validation, and 238 fractured and 268 non-fractured images for testing. The study employs the AlexNet deep learning architecture, leveraging advanced data augmentation and preprocessing techniques to enhance model performance. The model demonstrated a robust capacity to distinguish between fractured and non-fractured cases, achieving an impressive accuracy of 96%. These results underscore the model’s potential as a valuable diagnostic tool, supporting healthcare professionals in early and accurate fracture detection across various body regions and patient conditions.