We trained high performing open source models on image scans of tissue biopsies to predict endoscopic categories in inflammatory bowel disease. These predictive models can help us better understand the disease pathology and represent a step towards automated clinical recruitment strategies. - View it on GitHub