Revisiting the subtyping of bowel-based disorders using unsupervised machine learning

Abstract

The Rome IV criteria, a widely used symptom-based diagnostic tool for disorders of gut-brain interaction (DGBI), have enabled clinicians to provide diagnoses for patients whose gastrointestinal (GI) symptoms lack organic explanations. However, challenges remain in identifying robust diagnostic biomarkers, predicting treatment outcomes, and ensuring diagnostic stability. Unsupervised machine learning can be used to step beyond existing clinical definitions to uncover intrinsic patient subtypes possibly unburdened by these limitations. In this talk, Jarrah will provide an overview of bowel-based DGBI subtyping, present the findings from his recent application of unsupervised machine learning to revisit this problem, and introduce his next steps in improving DGBI diagnosis and treatment strategies.