To develop and train a convolutional neural network to detect and characterize disease severity of inflammatory bowel disease during endoscopy

To develop and train a Convolutional Neural Network to detect and characterize disease severity in inflammatory bowel disease during endoscopy. This initiative will inevitably establish a high-quality large image database. Our secondary study aims are therefore to use the images we collect to advance the field of deep learning and computer aided diagnosis in inflammatory bowel disease by establishing an image database. This will involve developing a framework combining deep learning and computer vision algorithms. The ultimate aim is to use the image database to produce high impact research outcomes and training resources leading to an improvement in the quality of endoscopy performed, reduce inter-observer variability in disease assessment and a reduction in missed bowel cancer rates and associated mortality.

Inclusion Criteria:

• • Any adult patient aged 16 years or older who has consented to undergo endoscopic investigation where images are captured as part of routine clinical care.

Exclusion Criteria:

• • Any patient under the age of 16

  • Patients who are unable to give informed consent to undergo endoscopic investigation or those who do not wish their pseudo-anonymised images to be used

Inflammatory Bowel Disease (IBD) affects at least one in 250 people of the UK population and the prevalence is rising.