Bone Fracture Detection

• Tech Stack: Python, scikit-learn, Huggingface, PyTorch, pandas, Open-CV
• Github URL: Project Link

Developed two pipelines to detect bone fractures in X-ray images using deep learning models. The first pipeline classifies whether the X-ray contains fractures, utilizing ResNet-18, VGG, and ViT architectures on the MURA dataset. The second pipeline pinpoints the locations of fractures with bounding boxes using object detection-based model architectures such as Faster R-CNN, YOLOv8, and DETR.

Achieved the highest F1 score of 82% with ViT on the classification task, demonstrating the advantage of self-attention in modeling long-range dependencies and understanding global structures in X-ray images. For fracture detection, the results were more mixed, with DETR performing well for larger fractures, while Faster R-CNN and YOLO significantly outperformed DETR for small and medium-sized fractures.