Medical utility of artificial intelligence for fracture detection in the emergency department
| ISRCTN | ISRCTN23087950 |
|---|---|
| DOI | https://doi.org/10.1186/ISRCTN23087950 |
| ClinicalTrials.gov (NCT) | Nil known |
| Clinical Trials Information System (CTIS) | Nil known |
| Integrated Research Application System (IRAS) | 357391-A |
| REC reference | 25/SC/0252 |
| Sponsor | Oxford University Hospitals NHS Trust |
| Funder | Radiobotics |
- Submission date
- 07/11/2025
- Registration date
- 08/01/2026
- Last edited
- 08/01/2026
- Recruitment status
- Recruiting
- Overall study status
- Ongoing
- Condition category
- Injury, Occupational Diseases, Poisoning
Plain English summary of protocol
Background and study aims
Broken bones (fractures) are commonly misdiagnosed in emergency departments, which can lead to patients experiencing delayed recovery, prolonged pain, and unnecessary follow-up visits to the NHS. In the UK, around 3.7 to 8.7% of suspected fractures are misdiagnosed, primarily because busy emergency staff incorrectly interpret X-rays. This can also lead to patients without fractures being incorrectly told they have a broken bone, resulting in unnecessary treatments.
This study aims to test whether using artificial intelligence (AI) software to help clinicians identify fractures on X-rays can reduce these diagnostic errors and improve patient care. The AI tool automatically analyses X-ray images and highlights areas where fractures might be present, helping doctors and nurses make more accurate diagnoses.
Who can participate?
Anyone aged 2 years or older who attends an emergency department or minor injuries unit and has an X-ray taken for a suspected fracture can be included in this study. This includes both children and adults.
Patients cannot participate if they are under 2 years old, need X-rays for suspected child abuse, or require imaging of the skull, facial bones, teeth, or neck spine. Children under 18 with suspected lower back fractures are also excluded. Patients can opt out by using the national NHS data opt-out system or by completing an opt-out form displayed in the emergency department.
What does the study involve?
The study will run across several NHS hospitals and minor injuries units in the Thames Valley region. During the 6-month study period, AI software will be installed at each site. The AI tool will alternate between being switched on and off each month, so doctors sometimes have access to the AI assistance and sometimes do not.
Most patients will simply receive their normal emergency care. The study team will collect information from medical records to see whether the AI reduces misdiagnoses and unnecessary follow-up appointments. No additional X-rays or tests are needed.
A small number of patients (30 at each hospital) will be asked if they are willing to be contacted one month after their visit to complete a short questionnaire about their experience. Staff at participating hospitals will also be invited to complete questionnaires about their experience using the AI tool.
What are the possible benefits and risks of participating?
The AI tool may help clinicians detect fractures more accurately, which could mean patients receive the right treatment more quickly and have fewer unnecessary follow-up visits. Patients might also experience less pain and recover more quickly if their fracture is correctly identified early on. For patients without fractures, there may be less chance of being incorrectly diagnosed and receiving unnecessary treatment.
The study involves minimal risk to patients. The AI software is a support tool only and does not replace the doctor's judgement. Clinicians will still make the final decision about diagnosis and treatment. The main risk is that the AI might occasionally make errors, but the study team will carefully monitor the tool's performance to ensure it is working safely. There are no physical risks as patients receive standard care with no additional procedures.
Where is the study run from?
The study is run from Oxford University Hospitals NHS Foundation Trust, with the main coordination happening at Headley Way, Oxford OX3 9DU.
The study will take place across four hospital trusts in the Thames Valley region:
1. Oxford University Hospitals (John Radcliffe Hospital and Horton General Hospital emergency departments)
2. Oxford Health NHS Foundation Trust (minor injuries units)
3. Royal Berkshire NHS Foundation Trust (Royal Berkshire Hospital emergency department and associated minor injuries units)
4. Buckinghamshire Healthcare NHS Trust (Stoke Mandeville Hospital emergency department and associated minor injuries units)
When is the study starting and how long is it expected to run for?
The study is expected to start in December 2025 and run until December 2026, lasting approximately 12 months in total. Patient recruitment will take place over a six-month period from December 2025 to July 2026.
Who is funding the study?
The study is funded by Radiobotics, the company that manufactures the AI fracture detection software being tested.
Who is the main contact?
Prof. Alex Novak, Alex.Novak@ouh.nhs.uk
Contact information
Public, Scientific, Principal investigator
Headley Way
Headington
Oxford
OX3 9DU
United Kingdom
 ORCID ID |
0000-0002-5880-8235 |
|---|---|
| Phone | +44 (0)1865 231550 |
| Alex.Novak@ouh.nhs.uk |
Study information
| Primary study design | Interventional |
|---|---|
| Study design | Prospective cluster randomized cross over trial (ABAB/BABA) |
| Secondary study design | Cluster randomised trial |
| Scientific title | Systematic Assessment of the Medical Utility of Radiology Artificial Intelligence (SAMURAI) in Fracture Detection |
| Study acronym | SAMURAI-Fracture |
| Study objectives | Primary Objective: To evaluate whether implementation of an AI fracture detection tool reduces the proportion of patients having unnecessary NHS healthcare contacts as a result of incorrect diagnoses, i.e. false positive/negative conclusions. Secondary Objectives: To assess the technical performance of an AI fracture detection tool, and its impact on clinician experience, overall patient care experience and service provision. |
| Ethics approval(s) |
1. Approved 23/10/2025, UK Health Research Authority (2 Redman Place, Stratford, London, E20 1JQ, United Kingdom; +44 (0)20 7104 8000; approvals@hra.nhs.uk), ref: 357391 2. Approved 23/10/2025, Joint Research Office, Oxford University Hospitals NHS Foundation Trust (Second Floor, OUH Cowley, Unipart House Business Centre, Garsington Road, Oxford, OX4 2PG, United Kingdom; -; shahista.hussain@ouh.nhs.uk), ref: PID18979-SI001 3. Approved 23/10/2025, South Central – Oxford C Research Ethics Committee (Oxford University Hospitals NHS Foundation Trust, Unipart House Business Centre, Garsington Road, Oxford, OX4 2PG, United Kingdom; -; oxfordc.rec@hra.nhs.uk), ref: 25/SC/0252 |
| Health condition(s) or problem(s) studied | Patients undergoing X-ray for suspected fracture in the emergency department or minor injuries unit |
| Intervention | The trial will involve installing a MHRA- and CE-approved AI fracture detection software at each site for 6 months. Clinicians will be able to view AI annotated images to aid in diagnosis when reviewing X-rays for suspected fracture. Sites will be randomly assigned to begin the trial with the AI algorithm either active ("On") or inactive ("Off") during the first month. Thereafter, the algorithm status will alternate each month (“AI On” and “AI Off”) for the remaining 5 months. This crossover design ensures that each site experiences both conditions multiple times, allowing within-site comparisons of outcomes under AI-assisted versus standard practice. This study will be conducted across four sites, divided into three clusters: Cluster 1: Oxford University Hospitals NHS Foundation Trust: John Radcliffe Hospital ED (Level 1 ED) Horton General Hospital ED (Level 1 ED) Oxford Health Trust: MIUs (Level 3 EDs) Cluster 2: Royal Berkshire NHS Foundation Trust: Royal Berkshire Hospital ED (Level 1 ED) Associated MIUs (Level 3 ED) Cluster 3: Buckinghamshire Healthcare NHS Trust: Stoke Mandeville Hospital ED (Level 1 ED) Associated MIUs (Level 3 ED) All patients undergoing an X-ray for suspected fracture will be eligible for the study. Dedicated research teams at each site will extract data from Electronic Patient Records and share the anonymised data with the central trial team for evaluation of primary and secondary outcomes. We will gather feedback from patients and clinical staff through electronic surveys at each site. These sites will allow for sufficient patient heterogeneity (socioeconomic, geographical, population, ethnicity) as per the INCLUDE guidance to ensure that our results are generalisable to the wider UK population. |
| Intervention type | Device |
| Phase | Not Applicable |
| Drug / device / biological / vaccine name(s) | RBFracture |
| Primary outcome measure(s) |
Unnecessary NHS contacts measured via electronic patient record audit of composite endpoints (phone consultations, repeat ED attendances, additional imaging, GP consultations, late fracture clinic referrals, inappropriate early clinical use, unnecessary imaging, immobilization, and sick leave) compared between periods when AI system is active ("On") versus inactive ("Off") throughout the entire 6-month study period (December 2025 – May 2026), with comparison occurring monthly during the alternating On/Off periods. |
| Key secondary outcome measure(s) |
Patient Impact: |
| Completion date | 31/12/2026 |
Eligibility
| Participant type(s) | Patient |
|---|---|
| Age group | Mixed |
| Lower age limit | 2 Years |
| Upper age limit | 100 Years |
| Sex | All |
| Target sample size at registration | 45000 |
| Key inclusion criteria | All patients undergoing X-Rays in ED or MIU for a suspected fracture will be eligible for inclusion |
| Key exclusion criteria | 1. Patients under 2 years old 2. Skeletal survey conducted to assess for non-accidental injury 3. Skull, facial bone, dental, and cervical spine X-rays 4. Thoracolumbar spine X-rays in patients under 18 years old 5. Patient has opted out of data sharing on the National Data Opt-Out 6. Patient completes the opt out questionnaire displayed on posters in the waiting room |
| Date of first enrolment | 28/10/2025 |
| Date of final enrolment | 30/07/2026 |
Locations
Countries of recruitment
- United Kingdom
- England
Study participating centres
Headley Way
Headington
Oxford
OX3 9DU
England
Sandford Road
Littlemore
Oxford
OX4 4XN
England
London Road
Reading
RG1 5AN
England
Whielden Street
Amersham
HP7 0JD
England
Results and Publications
| Individual participant data (IPD) Intention to share | Yes |
|---|---|
| IPD sharing plan summary | Available on request |
| IPD sharing plan | The datasets generated during and/or analysed during the current study will be available upon request to Prof. Alex Novak, Chief Investigator, alex.novak@ouh.nhs.uk. Type of data to be shared: Anonymised imaging data (X-ray images processed by RBFracture™ algorithm) and associated anonymised demographic, clinical, and outcome metadata Timing of availability: Data will become available upon publication of the primary results manuscript, anticipated within 12 months of study completion Duration of availability: Data will be made available for a minimum of 5 years post-publication Access criteria and sharing mechanism: Access will be restricted to bona fide researchers undertaking research aligned with the study objectives. Requests will be reviewed by the study steering group. Data will be shared via secure transfer mechanisms compliant with UK GDPR and NHS information governance requirements Participants' consent for data sharing: Participants were not explicitly consented to data sharing, however, the use of routine clinical data via EPR/RIS and an opt-out consent model supports secondary use for research aligned with the original study purpose Data anonymisation: All data will be fully anonymised with the removal of identifiable information; re-identification will not be possible Ethical and legal restrictions: Data sharing will comply with UK GDPR, NHS information governance protocols, and ethical approval conditions. No ethical restrictions are anticipated beyond these standard requirements Additional comments: This dataset of ~45,000 fracture cases will support future post-market surveillance, algorithm validation studies, comparative evaluations of fracture-detection AI tools, and health economic analyses. |
Editorial Notes
07/11/2025: Study's existence confirmed by the HRA.
