Problem
The stomach stores, grinds, and propels its contents, and its normal function is essential for food we eat to be adequately processed and prepared for nutrient absorption in the small intestine. These actions require the efficient coordination of gastric muscles. Contractions of gastric muscles deform the stomach wall throughout digestion, and the luminal contents move passively in response to these forces. However, existing in vivo methods cannot fully capture gastric motor function. Recent advances in gastrointestinal magnetic resonance imaging (GI-MRI) have made it possible to visualize the movement of food inside the stomach with high resolution and fast speed, but a comprehensive, 3-D understanding of the movements and shape changes of the stomach itself has been missing. This limits precise diagnosis and treatment of gastrointestinal malfunctions, such as delayed emptying, impaired accommodation, and dysmotility.
Solution
In a recent study led by Xiaokai Wang, a PhD student at the University of Michigan, along with collaborators at Michigan and from the University of Melbourne and the University of Auckland, MRI and deep learning were used to map the movement of the stomach wall that executes digestive functions. A special neural network, known as a neural ordinary differential equation (neural ODE), was used to model the biomechanical process that dynamically deforms the stomach. This involved representing the stomach wall as a surface, then tracking the continuous deformation of this surface during the process of digestion. Next, this model was fit to gastrointestinal MRI data collected from rats during the movement of gastric luminal contents and a 3-D “virtual stomach” was reconstructed. This virtual stomach showed realistic movement at every moment in time throughout digestion and could be tailored to individual rats. Summarizing the motility patterns across ten different rats provided unprecedented details regarding how the stomach organizes and coordinates its functional regions for coherent motor functions.
Impact
This recently published study demonstrated the efficacy of combining a neural ODE with gastrointestinal MRI to model gastric anatomy and function, using a surface morphing approach that effectively characterizes the biomechanical dynamics of the stomach wall that encloses and moves luminal content. This approach is ultimately a more direct assessment of gastric motility than was previously achievable with gastrointestinal MRI data alone, and may provide a more direct route to diagnosis and treatment of GI dysfunction. Furthermore, this approach is non-invasive, performed in vivo, and can be done for both individual and group-level analyses, allowing one to compare gastric motility patterns between individuals or disease states. Ongoing research is further extending this work from animals to humans and linking in vivo observations of gastric motor events to underlying structures, circuits, and mechanisms, moving us one step further towards clinical applications of gastrointestinal MRI.
