The application of artificial intelligence in pouchoscopy of patients with restorative proctocolectomy might improve the diagnosis of pouchitis and neoplasms. The aim of this pilot study is to develop a convolutional neural network algorithm for pouchoscopy

Restorative proctocolectomy is the standard procedure for treatment of refractory severe colitis in inflammatory bowel disease as well as the standard procedure for carcinoma preventive treatment of patients with inflammatory bowel disease with colonic neoplasia and patients with familial adenomatous polyposis coli (FAP). Pouchoscopy can be used to monitor the success of therapy and to detect complications such as pouchitis or neoplasia. Artificial Intelligence assisted image recognition programs can support the examiner in finding a diagnosis and train physicians in training, objectify endoscopic findings in the context of studies and might make biopsies unnecessary, thus saving costs. The application of Artificial Intelligence in pouchoscopy has not been demonstrated to date. The aim of this study is to develop, an image recognition algorithm that reliably detects the different graduations of pouch inflammation. This requires training and fine-tuning of the image recognition program PiTorch using the largest possible amount of image data, which will be recruited from the image databases of the UMM and the Theresienkrankenhaus Mannheim. A test run for statistical evaluation will be performed on an independent cohort.

Inclusion Criteria:

• All patients aged ≥ 18 years with inflammatory bowel disease and status after restorative proctocolectomy with ileoanal pouch who had received a pouchoscopy

Exclusion Criteria:

• Very poor endoscopic image quality

Adult patients with inflammatory bowel disease and status after restorative proctocolectomy with ileoanal pouch might develop pouchitis or neoplasia in the pouch.