Artificial intelligence or AI are computer algorithms that have the ability to learn, solve problems and make decisions. In recent years AI has become more advanced and several types of AI such as Chat GPT2 (an advanced chat bot) have made the news and become well known and easy to access. Currently, AI is in use in medicine and has been approved for the detection of polyps in colonoscopy where it assists doctors by highlighting suspicious lesions on the screen. This process involves the AI analysing large amounts of data so that it can ‘learn’ complex pattern recognition, similar to how doctors learn how to make medical decisions.
Three scientific review articles published in the last two years by gastroenterologists from around the world have shown that AI can assist with the diagnosis, treatment, and prediction of Inflammatory Bowel Disease (IBD). Currently, the cause of IBD is not well understood but it is believed that genes and gut microbiome are contributing factors. AI can analyse large amounts of genomic data to discover new genes that may increase the risk of IBD. AI can also analyse the many types of bacteria present in the gut to see if there are any types associated with IBD.
The diagnosis and assessment of IBD requires tissue to be taken and analysed using a microscope but there are challenges surrounding the interobserver variability where two different pathologists may score the same sample differently. AI has shown promise in this area as it can detect certain types of cells in tissue with a similar ability to a pathologist.
AI can be trained to analyse videos in endoscopy in order to grade the severity of IBD and can prevent errors made due to human fatigue. An example is Video Capsule Endoscopy where the patient swallows a small camera that has the ability to record videos of the entire small bowel. Normally it takes 96.6 minutes for a doctor to review the entire video but with the assistance of AI this time is reduced to 5.9 minutes without any loss in performance.
AI can also make IBD research faster and more efficient. One aspect is that it can analyse patient data bases much faster than a human and find which patients are suitable to be recruited for certain IBD studies.2 AI can also discover new drugs in the treatment of IBD. Based on a database that record the types of proteins found in patients with IBD, AI can detect which ones can be a target for future immune therapies.
AI also has limitations and thus cannot completely replace a doctor in the treatment and diagnosis of IBD. AI currently struggles to find the difference between ulcerative colitis and Crohn’s disease based on endoscopy and will need to undergo training through analysis of IBD video databases. AI also needs to be trained to ignore minor scratches to the bowel caused by the scope in endoscopy and sites where biopsies are taken as they are currently detected as IBD due to the similar appearance. AI is only as good as the data it is trained on. AI can struggle to differentiate between IBD and other rarer diseases with similar appearances because there are limited data on the rarer diseases available. Finally, AI is unlikely to replace the rapport and interpersonal relationship between a doctor and a patient.