fracture.ai

What It Is

fracture.ai is a web-based AI fracture detection tool for uploading and analyzing X-ray images. Users can drag and drop DICOM, BMP, JPG, PNG, or TIF files, or import images via the clipboard. A single upload supports up to 4 views from the same exam, with a maximum file size of 50MB per file. The page clearly states that “this is not a medical product,” so it is better suited for research, teaching, or informal auxiliary reference rather than as a basis for clinical diagnosis.

AI Capabilities and Output

Its core capability is fracture detection using computer vision and neural networks, covering tasks such as image classification and object detection. The site says the model was trained on hundreds of thousands of radiographs and uses EfficientNet for classification; there may be specialized models for different body parts and age groups. When no suitable model is available, a general algorithm is used as a fallback. Outputs typically include the probability that a fracture is present in a given image, along with a colored heatmap. However, the official text also notes that the heatmap may, but does not necessarily, indicate the true fracture location, and accuracy varies by body part and age group.

Pricing, Ease of Use, and Integration

The website describes it as a free online X-ray fracture detection service, with no disclosed paid plans, subscriptions, commercial licensing, or payment methods. In terms of usability, it works directly through browser upload and supports modern versions of Edge, Chrome, Firefox, and Safari. DICOM input is considered preferable because it preserves full resolution and more grayscale information. The site does not mention an API, PACS/RIS integration, SDK, or batch processing capabilities, so there is limited information on how it could fit into institutional workflows.

Data Privacy and Compliance Risks

fracture.ai provides fairly detailed privacy explanations: DICOM metadata is stripped in the browser, and only image pixels are uploaded. File names are generated using a hashed patient ID, hashed timestamp, gender, and rounded age. The platform reviews for potential personal information and deletes suspicious images. However, users must ensure they have the right to upload the images and must not upload files containing personal information within the image itself. More importantly, uploaded content grants the platform a worldwide, perpetual, free, irrevocable, and sublicensable license to use it, and anonymized X-rays may even involve a CC0 license. This is a significant limitation for medical institutions or users subject to privacy and compliance requirements.

Pros, Cons, and Who It’s For

Its strengths are that it is free, easy to use, supports DICOM and multiple image formats, and provides probabilities plus heatmaps. Its limitations are that it is not a medical product, lacks verifiable regulatory certification information, has unclear API and service support, and imposes broad licensing terms on uploaded images. It is suitable for medical imaging AI learning, research exploration, demonstrations, and users who want a preliminary AI reference for individual X-rays. It is not suitable for direct clinical decision-making or for institutions with strict requirements around data ownership, compliance, and auditability.

Access from China

The site does not provide information about access from mainland China, network nodes, or payment options, so its accessibility status is unknown. For medical use cases in China, priority should be given to alternative medical imaging AI solutions that have local compliance qualifications, can integrate with hospital systems, and can sign data processing agreements.