Cracking the Code to Personalize Care for Children with IBD

Artificial intelligence (AI) has been personalizing your Netflix queue and Facebook feed for years. Now, physicians and scientists at Cincinnati Children’s are harnessing AI technology to bring precision medicine to the bedside for children with inflammatory bowel disease (IBD). If successful, their methods could lead to tailored treatment strategies that achieve and maintain optimal outcomes, with less exposure to corticosteroids.

“IBD is inherently complex with a heterogeneous disease course, and physicians need better methods for predicting which treatments will benefit a given patient,” says Jasbir Dhaliwal, MBBS, MSc, a pediatric gastroenterologist in the Division of Gastroenterology, Hepatology and Nutrition. Dhaliwal has secondary appointments with the hospital’s Division of Biomedical Informatics and the James M. Anderson Center for Health Systems Excellence.

“IBD is fertile ground for AI research since diagnosis and disease monitoring are driven in large part by different imaging modalities, including magnetic resonance and endoscopy, and histology information from biopsies. This enables creation of multimodal data sets that we can analyze with deep learning approaches to create predictive algorithms that deliver individualized IBD care approaches,” Dhaliwal says.

Dhaliwal’s interest in AI began at the Hospital for Sick Children (SickKids) in Toronto, where she completed an advanced inflammatory bowel disease fellowship in 2020. There, she collaborated with fellow researchers and derived a machine learning classifier to differentiate types of colonic IBD, with the view of potentially implementing the classifier in the clinical setting.

“These projects opened my eyes to understanding what machine learning is and some of the novel ways we can use it to analyze data sets,” she says.

Mining the Data from PROTECT

Much of Dhaliwal’s research involves using data from the landmark PROTECT (Predicting Response to Standardized Colitis Therapy) study. PROTECT established a clinical model to help physicians predict which individual patients were likely to achieve clinical remission on mesalamine monotherapy, based on pretreatment clinical, genetic and microbial factors. Lee (Ted) Denson, MD, director of the Schubert-Martin Inflammatory Bowel Disease Center at Cincinnati Children’s, was co-principal investigator of the PROTECT study.

The large, multicenter study of more than 400 patients and 29 centers in North America contains a gold mine of information. As Dhaliwal dives into the data sets with her colleagues at Cincinnati Children’s, she is mindful of the potential pitfalls inherent to AI research.

“To fully realize precision medicine and assign personalized therapy in IBD, we need to integrate multimodal patient data using machine learning algorithms that interface with clinical decision support tools embedded in the electronic health record,” she says. “AI methodologies are sophisticated and exciting, and it’s tempting to get caught up in the possibilities. Developing an accurate, high-quality machine learning platform depends on many factors, and it is important to be cognitive of the limitations and pitfalls of these approaches. We’re in the discovery phase now, with the long-term goal of introducing algorithms into the clinical setting that can help kids with IBD experience the best outcomes possible.”

Dhaliwal and a gastroenterology clinical fellow Ruben Colman, MD, recently published a review in Frontiers in Pediatrics related to this research. It concludes with a prediction of how machine learning will integrate multimodal patient data to bring precision medicine to the bedside of children with ulcerative colitis in the future. Read the paper.