Human Bone Fractures
Description
A medical imaging dataset for automated bone fracture detection and classification. It contains 1,539 annotated X-ray images covering 10 fracture categories and healthy bone cases. Dataset Overview The dataset includes X-ray images with bounding box annotations for fracture detection and classification. Each image is labeled with fracture type and location, supporting computer vision research in medical imaging. Fracture Types The dataset covers 10 categories: Comminuted — Bone broken into multiple fragments Greenstick — Incomplete fracture common in children Healthy — Normal bone structure without fractures Linear — Simple straight-line fracture Oblique Displaced — Angled fracture with bone displacement Oblique — Angled fracture without displacement Segmental — Fracture with isolated bone segment Spiral — Twisting fracture pattern Transverse Displaced — Horizontal fracture with displacement Transverse — Horizontal fracture without displacement Dataset Statistics Total Images: 1,539 Training Set: 1,347 images (87.5%) Validation Set: 128 images (8.3%) Test Set: 64 images (4.2%) Data Characteristics Image Format: JPEG Annotation Format: Bounding box coordinates with class labels Image Content: X-ray images showing various bone structures and fracture patterns Coverage: Multiple fracture types with varying severity and presentation Applications Medical image analysis and computer-aided diagnosis Fracture detection and classification Research in medical AI and radiology Training and evaluation of detection algorithms Educational purposes for medical imaging analysis Annotation Details Each image includes precise bounding box annotations indicating fracture locations and corresponding class labels. Annotations enable object detection and classification tasks for automated fracture identification in X-ray images. This dataset supports research and development of AI systems for medical image analysis, with potential applications in clinical decision support and radiological workflow automation.
Files
Steps to reproduce
Configuration File: data.yaml — Dataset configuration with class names, paths, and metadata Dataset Information: dataset_info.csv — CSV with image paths, labels, class counts, and split assignments Data Splits: train/ — Training set images/ — 1,347 X-ray images (JPEG) labels/ — 1,347 annotation files (YOLO format) valid/ — Validation set images/ — 128 X-ray images (JPEG) labels/ — 128 annotation files (YOLO format) test/ — Test set images/ — 64 X-ray images (JPEG)