CESCAIL

Name and title

CESCAIL Study - Capsule Endoscopy delivery at sCale through enhanced AI analysis

Recruitment status

Open

Chief investigator

Professor Ramesh Arasaradnam

Sponsor

University Hospitals Coventry & Warwickshire NHS Trust

Funder

NIHR AI in Health and Care Award

Aim

To compare the accuracy of prototype machine learning tool’s artificial intelligence against clinician reporting and measure time taken to complete this task.

Study design

Diagnostic Accuracy Study

Speciality

Gastroenterology

Sample size

674

Summary

Bowel cancer is the area being studied, as the disease is a target for innovation in early detection and diagnosis. The CESCAIL study tests the use of Artificial Intelligence (AI) on a video taken from a minimally-invasive imaging device, to improve efficiency and accuracy of the detection of polyps, which are little outgrowths within the lining of the bowel.

The study will be conducted across multiple sites within the UK, recruiting 674 patients who are having Colon Capsule Endoscopy (CCE) as a part of their standard care pathway. CCE is a swallowable capsule the size of a large vitamin pill with two tiny cameras inside – and it has already proven to be a viable, efficient alternative to traditional colonoscopy (a thin flexible tube with a camera on the end which goes round the large bowel), normally used to check the large bowel.

However, checking CCE videos for problems or signs of disease can be time consuming. A trained clinician can take 20-120 minutes to assess up to 400,000 images from a maximum 12-hour video. AI can reduce the time needed to check CCE videos. Early versions of the system called ‘AiSPEED’ allow clinicians to achieve the same accuracy when checking videos in less than 20% of the time.

As part of this study participants will continue with their standard care pathway, with their capsule video analysed by a clinician to determine their final diagnosis. In addition to this, the patient’s capsule video will be further analysed by the AI tool to support a second clinician’s analysis of the images. The AI-supported analysis report will then be compared with the standard care report, comparing the time taken to analyse and report, and to measure the productivity and accuracy of the AI in detecting polyps.