A Multi-Class Bone Fracture X-Ray Image Dataset for Automatic Fracture Detection and Classification

Dataset Description:
This dataset is a carefully prepared collection of bone X-ray images, categorized by fracture type, including Normal, Simple Fracture, and Major Fracture. To better reflect real-world scenarios, each image includes augmented versions that simulate variations in lighting, orientation, and imaging conditions. It is designed for researchers and developers working on machine learning and deep learning applications such as fracture detection, automated classification, and image segmentation. With its rich and well-structured content, the dataset provides a reliable foundation for developing innovative diagnostic and analytical solutions that support medical imaging, orthopedic diagnostics, and clinical decision-making, ultimately contributing to more accurate and efficient patient care.

Data Content:

Fracture Categories:

Normal:
Source: Collected directly from hospital records.
Number of Original X-ray Images: 509
Number of Augmented Images: 3,054
Total Images in Dataset: 3,563

Simple Fracture:
Source: Collected directly from hospital records.
Number of Original X-ray Images: 616
Number of Augmented Images: 3,696
Total Images in Dataset: 4,312

Major Fracture:
Source: Collected directly from hospital records.
Number of Original X-ray Images: 559
Number of Augmented Images: 3,354
Total Images in Dataset: 3,913

Key Features:
Image Format: PNG
Total Number of Original Images: 1,684
Total Number of Augmented Images: 10,104
Total Dataset Size (Original + Augmented): 11,788 images

Augmentation Techniques:
Rotation: ±30° to simulate variations in bone orientation and patient positioning.
Brightness & Contrast: α 0.85–1.15, β −25 to +25 to mimic lighting variability.
Shift: ±20 pixels to handle minor misalignments.
Zoom: 0.85–1.15 scaling for scale and translation invariance.
Flips: Horizontal/vertical, 50% probability for spatial diversity.

Applications:
Medical Imaging: Train and evaluate Machine Learning models for fracture detection, classification, and segmentation.
Healthcare Technology: Develop automated diagnostic tools and real-time fracture detection systems for clinical and mobile applications.
Medical Research: Study fracture patterns to advance diagnostic methods and treatment strategies.

Data Acquisition:
The X-ray images in this dataset were collected from hospital records following ethical guidelines and anonymized to ensure patient privacy. Experienced radiologists and medical professionals manually annotated the images to accurately classify fracture types. The dataset underwent careful validation to ensure high quality and reliability, making it suitable for research in fracture detection, classification, and medical imaging applications.

Clinical Relevance:
Represents diverse real-world imaging conditions, including different bone orientations, exposure levels, and patient positioning, providing robust data for model training and evaluation.