Mechanics-informed MRI (MRI-MECH) for estimating esophageal health

Dynamic magnetic resonance imaging (MRI) is a technique used extensively for imaging flow through blood vessels, but its application to bolus transport in the esophagus has been limited due to the lack of a framework that quantifies the state and functioning of the esophagus. We present a framework called mechanics-informed MRI (MRI-MECH) that uses MRI measurements of esophageal bolus transport to provide quantitative prediction of esophageal mechanical health, thus enhancing the capability of MRI to diagnose esophageal disorders. MRI-MECH models the esophagus as a flexible 1D tube that follows a 1D tube-law with the flow through it governed by 1D mass and momentum conservation equations. These equations are solved using a physics-informed neural network (PINN) which minimizes the difference between the measurements from the MRI and the solutions of the governing equations. MRI-MECH works well with missing data or poor image resolution, and we demonstrate that by predicting missing information about the lower esophageal sphincter (LES) during the esophageal emptying process. Finally, the MRI-MECH estimates the mechanical health of the esophagus by calculating the esophageal wall stiffness and active relaxation.