MobileViT and YOLOv8: Improving Bone Fracture Detection and Classification Through Deep Learning

An excessive force can cause a bone fracture and result in problems such as bleeding, infection, hypoxia, interfragmentary strain, and loss of support. The manual procedure of identifying fractures from X-rays or CT scans mostly depends on orthopedic surgeons’ expertise, which might be difficult in remote regions. Medical diagnoses in real time are becoming feasible with deep learning and machine learning development. The datasets used in this study were obtained from Kaggle. The first dataset, titled “Bone Fracture Detection: Computer Vision Project,” included 4,148 labeled images and 10,581 X-ray images segmented into “fractured” and “not fractured” categories for bone fracture classification. This research aims to use deep learning models like CNN, ConvNeXt, ViT, MobileViT, VGG16, and VGG19 to detect bone fractures and characterize their severity effectively. YOLOV8 is utilized for fracture diagnosis, and image processing techniques improve noisy X-ray images obtained 0.995 with maPA50 and 0.991 with maPA50-95. In terms of classification accuracy, MobileViT showed 99%. This research aims to develop a reliable, automated system for detecting bone fractures facilitating early diagnosis and treatment.