Artificial Intelligence-Based Diagnostic Support System for Functional Dyspepsia Based on Brain Activity and Food Preference.

Abstract

Background and aim Disorders of gut-brain interaction (DGBI) aredisorders where no organic clinical abnormalities are detected such as functional dyspepsia (FD) and irritable bowel syndrome (IBS). The brain activity of individuals with FD and IBS differs from that of healthy controls. Artificial intelligence can distinguish healthy controls from individuals with DGBI using several biomarkers. This study aimed to establish an artificial intelligence-based diagnostic support system using food preferences and brain activity in patients with DGBI. Methods ROME IV criteria were used to diagnose patients with FD and IBS. Their food preference was scored using a visual analog scale, and brain activity in the prefrontal cortex was investigated using functional near-infrared spectroscopy (fNIRS). The diagnostic model was developed based on the brain activity and visual analog scale scores for food using an artificial neural network model. Results Forty-one participants, including 25 patients with DGBI were enrolled in the study. The accuracy of the artificial intelligence-based diagnostic model using an artificial neural network in differentiating between healthy controls and patients with DGBI and between healthy controls and those with FD were 72.3% and 77.1%, respectively. Conclusions The artificial intelligence-based diagnostic model using brain activity and preference to food images showed sufficiently high accuracy in distinguishing patients with DGBI from healthy controls, and those with FD from healthy controls. Therefore, the fNIRS system provides objective evidence for diagnosing DGBI.