SAMURAI-Reader
Research type
Research Study
Full title
SAMURAI-Reader (Systematic Assessment of the Medical Utility of Radiology and diagnostic Artificial Intelligence: Multicase Multireader Studies)
IRAS ID
363994
Contact name
Alexander Thomas Novak
Contact email
Sponsor organisation
Oxford University Hospital Trust
Duration of Study in the UK
2 years, 11 months, 31 days
Research summary
Artificial intelligence (AI) is being used more often in healthcare to help doctors interpret medical tests such as scans and heart recordings (ECGs). Although many computer tools can automatically analyse these results, their findings still need to be checked by trained clinicians. This research protocol outlines a general plan for studying how AI support might improve doctors' accuracy and confidence when making diagnoses. The design can be easily adapted for future studies that use a similar format involving multiple readers and cases, allowing researchers to test AI tools in different clinical settings.
The first substudy will evaluate a new artificial intelligence (AI) tool called DL Precise, developed by DeepLook Medical. The tool is designed to assist radiologists (doctors who interpret breast imaging) by providing clearer measurements and highlighting suspicious areas on mammograms and ultrasound images. We want to see if DL Precise can improve accuracy, confidence, and consistency when doctors review breast images, especially in women with dense breasts or with cancers that are harder to detect.
The research will take place at Oxford University Hospitals NHS Foundation Trust. Around 500 previously collected breast imaging cases will be reviewed. All cases already have biopsy results confirming whether cancer was present. Radiologists of different experience levels (specialists, general radiologists, and trainees) will each read the same cases twice: once without the AI tool, and once with it, with a time gap between readings to avoid recall.
We will compare performance with and without DL Precise in terms of:
Accuracy of biopsy recommendations (whether they match the final pathology result).
Sensitivity for detecting cancers, including ILC.
Confidence in BI-RADS classification (the standard reporting system).
Consistency between different readers.
Time taken to make decisions.
The results will show whether DL Precise can help radiologists detect cancer more accurately, reduce unnecessary biopsies, and provide greater confidence in breast cancer diagnosis.
REC name
London - Brighton & Sussex Research Ethics Committee
REC reference
26/PR/0376
Date of REC Opinion
30 Apr 2026
REC opinion
Further Information Favourable Opinion