BoneFract: A Bone Fracture Dataset
Description
This dataset was created to support research on automated bone fracture detection and anatomical classification using X-ray images. The dataset allows researchers to use deep learning techniques to detect bone fractures and classify the anatomical region from X-ray images. The underlying research hypothesis is that deep learning models can learn meaningful visual patterns from X-ray images and can be used to accurately distinguish between fracture-positive and fracture-negative cases while identifying the anatomical region affected by a fracture. The X-ray images in this dataset were individually collected from Peoples University of Medical & Health Sciences for Women (PUMHSW), Nawabshah, Shaheed Benazirabad, Sindh, Pakistan. The images are provided in PNG format and were organized according to anatomical region and fracture status for bone fracture detection and classification research. The dataset consists of 47,931 X-ray images categorized into seven anatomical groups: Arm, Wrist, Hand, Pelvis, Thigh, Lower Leg, Foot. The dataset includes both fracture-positive and fracture-negative cases. The dataset is divided into 3 categories for deep learning procedures. These categories contain trained 38,344, valid 4,793, and test 4,794 images, respectively. The dataset contains X-ray images from various anatomical regions, including both fracture-positive and fracture-negative cases, enabling the investigation of the ability of deep learning algorithms to detect patterns in images associated with fractures. The dataset is suitable for research and teaching in medical imaging, computer vision, artificial intelligence, deep learning, and computer-aided diagnosis.
Files
Steps to reproduce
The X-ray images were individually collected from Peoples University of Medical & Health Sciences for Women (PUMHSW), Nawabshah, Shaheed Benazirabad, Sindh, Pakistan. The images were obtained from X-ray examinations covering seven anatomical regions: Arm, Wrist, Hand, Pelvis, Thigh, Lower Leg, and Foot. The collected images included both fracture-positive and fracture-negative cases. The collected X-ray images were reviewed and organized according to their anatomical region and fracture status. The images were converted and standardized into PNG format to provide a consistent format for subsequent image-processing and deep learning procedures. The dataset was organized into appropriate anatomical and fracture-status categories and subsequently divided into training, validation, and testing subsets containing 38,344, 4,793, and 4,794 images, respectively. The final dataset contains 47,931 X-ray images. The prepared dataset was used for automated bone fracture detection and anatomical classification using deep learning techniques. The workflow includes data collection, image organization, format standardization, dataset splitting, and subsequent model development and evaluation.
Institutions
- People's University of Medical and Health Sciences for WomenSindh, Nawabshah