IBM Watson Health Imaging Clinical Review Support Tool Study
Research type
Research Study
Full title
IBM Watson Health Imaging Clinical Review Support Tool Study
IRAS ID
278884
Contact name
Bahadar Bhatia
Contact email
Sponsor organisation
IBM Watson Health Imaging
Duration of Study in the UK
0 years, 3 months, 1 days
Research summary
IBM Watson Health has developed a software support tool to identify potential findings after an X-ray or CT scan has been performed, reviewed, and reported by an on-site radiologist (a specialist doctor who reads and interprets these images). The software is called Clinical Review 3.0 (CR3), which uses augmented intelligence (AI) to collect information from the selected images. The collected information is then compared to the radiologist's report.
Any differences between the collected information and the radiologist's report are provided to an on-site radiologist to review. The on-site radiologist then decides if these differences are clinically relevant or significant to require an addendum to the original report. If an addendum is created, the existing hospital clinical workflow is followed.
Under this expected workflow, no physical interaction will occur between the patient and the software support tool. No clinical data or results from this research will be written back to the site’s clinical systems. The use of this software support tool does not directly affect the disposition of the patient. This software tool does not make any care decisions, nor does it directly assist the radiologist in their diagnostic decision-making process.
The CR3 software support tool aids our radiologist(s) to report more efficiently and to train our future specialist doctors. The patients will be provided with research information in their appointment letter, a weblink to the research information, a privacy notice, and an animation describing the research.
We would automatically enroll patients unless they choose to opt-out by completing the tear-off form at the end of their appointment letter and leave it at X-ray reception. If the patient has decided to opt-out, their record will be removed from the research database by a system administrator.
By collecting this information from many patients, we can use statistics to work out how well the AI supports our radiologists in their decisions.
Summary of study results:
This study was a real-world evaluation of artificial intelligence within a busy acute hospital. Yet despite the ever-increasing pressure on the radiologists’ productivity, they generated a low number of missed findings for the specified chest clinical conditions of less than one percent of X-rays and medical scans checked.
We have shown that our real-world evaluation was important to perform, validating the high diagnostic accuracy of the radiologists in the imaging department.
The intended use of AI as a quality tool augments the radiologist’s armoury, maintaining reporting quality.REC name
London - South East Research Ethics Committee
REC reference
20/PR/0091
Date of REC Opinion
16 Jul 2020
REC opinion
Favourable Opinion