Academics | The Hong Kong University of Science and Technology

This project aims to develop and validate a multimodal AI model for predicting the efficacy of various drug therapies in treating Gastroesophageal Reflux (GER) symptoms. GER symptoms arise from the abnormal reflux of stomach and duodenal contents into the esophagus, causing esophageal and extra-esophageal symptoms or complications. Based on the Rome IV criteria, GER symptom patients are categorized into GERD (including erosive esophagitis and non-erosive reflux disease, NERD) and functional esophageal disorders (such as reflux hypersensitivity, RH, and functional heartburn, FH). Given the distinct pathophysiological mechanisms among these subtypes, current treatments like proton pump inhibitors (PPIs) are often suboptimal, with up to 50% of patients showing partial or no response. This project integrates multidimensional data including patient clinical information, endoscopic images, high-resolution esophageal manometry, and dynamic pH-impedance monitoring to build an AI model that accurately predicts the efficacy of various treatment regimens for GER symptoms, providing personalized medication guidance for clinicians.