Deep Learning Based Early Cancer Detection in Barrett’s Esophagus
Research type
Research Study
Full title
Early Detection of Cancer in cases of Barrett’s Esophagus using Endoscopy Image Analysis through Deep Image Retrieval
IRAS ID
281513
Contact name
Lyndon Smith
Contact email
Sponsor organisation
University of the West of England
Duration of Study in the UK
1 years, 0 months, 30 days
Research summary
Barrett's oesophagus is a condition where the epithelium (lining) of the lower oesophagus (food pipe) is replaced from normal to abnormal. This abnormal epithelium is known as metaplasia. Metaplasia increases the lifetime risk of cancer (oesophageal adenocarcinoma) and is associated with acid reflux. If pre-cancerous changes (known as dysplasia) could be seen within the metaplasia, this could direct biopsy targets and improve diagnosis and management, or identify early cancer that could be removed en bloc, and cured.
We wish to develop a machine-learning algorithm using artificial intelligence from images of known normal epithelium and known pathologies. Therefore we wish to ask patient attending for routine OGD (upper gastrointestinal endoscopy or flexible camera into the mouth down into the stomach) if extra photographs can be taken for these purposes.
The project aims to assist clinicians by offering informed biopsy process, in which the system presents the operator with clinical outcomes of patients with visually similar GI tracts. The project goal is to assess the use of artificial intelligence-based similarity detection systems to better inform biopsy placement, increasing the reliability of bi-yearly inspections.
The system aims to identify low grade dysplasia, high grade dysplasia or indeterminate for dysplasia within Barrett’s to provided targeted biopsy of that spot. Additionally, we aim for it to identify small, early adenocarcinomas or nodules that could be removed en bloc through the gastroscope and ‘curing’ the patient.
REC name
London - Surrey Research Ethics Committee
REC reference
20/HRA/2372
Date of REC Opinion
28 Jul 2020
REC opinion
Further Information Favourable Opinion