AI-Assisted Fracture Detection in Emergency Radiography

Artificial Intelligence-Assisted Fracture Detection in Emergency Radiography: A Multicentre Pragmatic Randomised Controlled Trial

This study evaluates whether artificial intelligence (AI) can support doctors who interpret X-rays for suspected fractures in emergency care.

In the participating hospitals, X-rays are usually interpreted first by the frontline treating physician, while the formal radiology report is generally available later and not before the patient leaves the emergency department. AI may therefore provide an immediate additional assessment while clinical decisions are being made.

Patients were randomly assigned to one of two groups. In the AI-assisted group, physicians interpreted the X-rays with support from an AI system. In the control group, physicians interpreted the same type of X-rays without AI support. All final diagnoses and treatment decisions remained with the treating physician.

The main question is whether AI assistance affects the time from triage to completion of emergency department treatment. The study also evaluates whether AI influences physician diagnostic confidence, the use of additional imaging, missed fractures, and diagnostic accuracy.

The study includes patients aged 2 years or older presenting after trauma with a suspected fracture requiring X-ray imaging. No additional imaging or treatment was required solely because of study participation.

The FAIR (Fracture detection with AI in emergency Radiography) Trial is a prospective, international, multicentre, pragmatic randomized controlled trial evaluating the clinical impact of AI-assisted interpretation of emergency radiographs.

The study was conducted at three hospitals in Austria and Germany: University Hospital Salzburg, Regional Hospital Hallein, and University Hospital Nuremberg.

Clinical setting

In the participating departments, plain radiographs of patients with suspected fractures are routinely interpreted by frontline orthopaedic trauma or paediatric physicians, who make immediate diagnostic and treatment decisions. Formal radiology reports are generally available later and are usually not available before completion of the emergency department encounter.

The study therefore evaluates AI as an immediate diagnostic decision-support tool during the period in which frontline physicians make clinical decisions, rather than as a replacement for formal radiological interpretation.

Study design

Eligible patient encounters were randomized in a 1:1 ratio to one of two parallel groups:

Control group: radiographs were interpreted by the treating physician without access to AI output.

AI-assisted group: radiographs were interpreted by the treating physician with access to real-time AI output.

The randomization sequence was generated by the trial statistician as one global sequence. Allocation was concealed during recruitment using folded sequential study forms on which only the study number was visible on the front and the treatment allocation was printed on the reverse. Allocation was revealed after radiography, when the triage nurse opened the study form and directed the patient to the corresponding treatment pathway.

All final diagnoses, decisions regarding additional imaging, treatment decisions, and discharge decisions remained the responsibility of the treating physician.

AI intervention

The AI system used was BoneView version 2.3.8 (Gleamer, Paris, France). BoneView analyzes DICOM radiographs and provides visual annotations and classifications for supported musculoskeletal findings.

In addition to fracture-related findings, other BoneView outputs supported by the system, including dislocations, joint effusions, and focal bone lesions, could be visible to physicians in the AI-assisted group. However, the main clinical outcomes of the FAIR Trial focus on fracture-related emergency care.

At University Hospital Salzburg and Regional Hospital Hallein, BoneView was provided through the Aidoc aiOS platform. At University Hospital Nuremberg, BoneView was integrated directly into the local imaging workflow.

Participating physicians received standardized onboarding consisting of a lecture and practical demonstration of the AI system before study implementation.

Study population

Patients were eligible if they were aged 2 years or older, presented after trauma with a suspected fracture requiring plain radiography, and had an injury involving a single anatomical region within the supported scope of the AI system.

Major exclusions included injuries involving multiple anatomical regions, head or cervical spine injuries, previous imaging or medical assessment for the same injury, contraindications to X-ray imaging, and lack of informed consent.

The unit of observation is the patient encounter. The same individual could therefore participate more than once if they presented with separate and unrelated injuries during the study period.

Outcomes

During ongoing recruitment in March 2026, following methodological review and before comparative outcome analysis, the outcome hierarchy was revised to prioritize patient- and physician-centered measures of clinical utility.

The primary outcome is time from triage to completion of emergency department treatment.

Key secondary outcomes include:

physician diagnostic confidence; additional imaging requested during the index emergency department encounter.

Secondary clinical outcomes include:

missed fractures; diagnostic performance compared with an expert-adjudicated reference standard, including sensitivity, specificity, positive predictive value, negative predictive value, and overall accuracy.

The originally registered outcome hierarchy placed greater emphasis on diagnostic performance. The registry record was not updated at the time of the March 2026 methodological revision and is being updated retrospectively to reflect the final analysis plan.

Reference standard

The reference standard for whether a fracture was visible on the index radiograph is established through expert review by a senior radiologist and a senior orthopaedic trauma surgeon. Reviewers can use the available imaging and clinical information when adjudicating each case.

The adjudication specifically determines whether a fracture was visible on the original radiograph. A fracture identified on subsequent CT or other imaging but considered occult on the original radiograph is therefore classified as no fracture visible on the index radiograph.

Disagreements between the two reviewers are resolved by discussion. If consensus cannot be reached, a third expert reviewer determines the final classification.

Study duration

Patient recruitment was conducted during predefined study periods between October 2025 and April 2026. Recruitment ended after completion of the planned site-specific recruitment periods and the available funded period of AI access, rather than because of observed treatment effects.

Inclusion Criteria:

• Age 2 years or older.
• Presentation to the emergency department following trauma requiring plain radiographic imaging.
• Injury involving a single anatomical region within the supported anatomical scope of the AI system.
• Written informed consent provided by the participant or legally authorized representative, with age-appropriate assent where applicable.

Exclusion Criteria:

• Age younger than 2 years.
• Injuries involving multiple anatomical regions.
• Head or cervical spine injuries.
• Previous radiographic imaging or medical assessment for the same injury before the index presentation.
• Contraindication to X-ray imaging, including pregnancy.
• Lack of informed consent.

Reduced image quality was not an exclusion criterion. Multiple fractures within the same anatomical injury region were eligible.