Artificial Intelligence in Gastroenterology: Future Patient Care Will Be a Human–Computer Partnership

As in many medical specialties, artificial intelligence (AI) is bringing innovations that will profoundly change gastroenterology (the digestive system). Solutions are being developed, particularly for gastrointestinal (GI) diseases, to accelerate diagnostic and treatment decisions and measure disease activity more objectively. Dr. Ryan W. Stidham, a leading gastroenterology and computational medicine expert, conducts AI research at the University of Michigan and is taking important steps toward shaping this future. His efforts aim to turn AI into a usable tool that makes gastroenterologists' decision-making more effective.

Artificial Intelligence and Its Potential in GI Disease

Dr. Stidham's career coincided with the spread of electronic health records (EHRs) and increasing access to medical data. Despite this abundance of data, however, health records did not adequately describe patients' clinical condition. Basic information about disease type, location, and severity was missing, particularly in complex conditions such as inflammatory bowel disease (IBD). Stidham recognized the potential to use machine learning to learn and reproduce expert decision-making, but this required better-quality observational data.

Early Research and Challenges

As the use of objective biomarkers increased in gastroenterology, standardized grading of endoscopic mucosal inflammation became a prominent example of this change. However, differences between observers created a need for centralized readings. Because endoscopic findings were qualitative and images did not fully capture every feature, even experts could disagree about disease severity and type.

To address this gap, Dr. Stidham aimed to improve endoscopic assessment using AI. Convolutional neural networks (CNN = convolutional neural network) had become feasible for medical imaging thanks to increasing computing power. Learning from labels such as disease severity, CNN models offered considerable potential for assessing complex image features.

Digital Endoscopic Assessment: An AI Success Story

Dr. Stidham had a simple but difficult plan: collect patients' endoscopic images, have experts evaluate them, and use those evaluations to train a CNN model. In 2016, however, collecting and labeling enough endoscopic images was not easy. The real challenge was finding tens of thousands of images representing different levels of disease severity and overcoming the technical limitations involved in analyzing them. After years of effort, a working model was developed.

The resulting CNN model assessed ulcerative colitis severity in almost exact agreement with IBD experts. This success provided the first evidence that AI could automatically measure complex diseases.

AI Will Allow Gastroenterologists to Play a More Effective Role

Dr. Stidham says that a major part of AI's benefits in medicine will involve reducing administrative burdens. Applications such as automated documentation, generating notes from clinic audio recordings, and producing procedure notes by analyzing endoscopic videos will soon become available. These applications will give gastroenterologists more time to focus on individual treatment plans and complex decisions.

AI is also being used in health maintenance processes such as vaccination, laboratory screening, and treatment monitoring. These technologies are expected to lighten gastroenterologists' workload and improve access to treatment. However, according to Dr. Stidham, these innovations do not appear able to remedy burnout by immediately reducing tumor burden.

Can AI Match Human Doctors?

Dr. Stidham initially thought AI would leave gastroenterologists astonished, but he has since stepped back from this view. He notes that AI struggles to understand individual patient context and reach the best decisions for each person. An expert gastroenterologist can anticipate gaps in data and adjust the weight given to information according to the patient. Although AI will therefore outperform gastroenterologists in routine and repetitive tasks, it cannot replace clinical experience, creativity, and expertise in the day-to-day management of GI diseases.

AI and Gastroenterology: Moving into Real-World Practice

For Dr. Stidham, a turning point in applying this technology to the real world was a talk given last spring by Microsoft Research president Peter Lee. It demonstrated large language models' (LLMs) ability to understand medical texts and answer questions. Watching applications such as generating a SOAP note (subjective, objective, assessment, and plan) from an audio recording of a clinical visit impressed him.

With these AI-supported advances, it became clear that AI would handle administrative tasks, documentation, billing, and patient follow-up notes. We are moving toward a future in which GI physicians can obtain help from machines and focus more on the creative and clinical decision-making aspects of their work.

How Gastroenterologists Will Adapt: Human–Computer Collaboration

In the coming years, gastroenterologists will explore how to use technologies such as AI prediction models, chatbots, and image analysis. Memory, information retrieval, calculations, and tedious tasks will become the machines' work. We will be able to use our human understanding of context, judgment, and creativity to offer more personalized solutions to more patients. The future points toward an era of strong collaboration between humans and computers in gastroenterology—and, indeed, throughout medicine.