A digital twin of the human stomach combining multiphysics and artificial intelligence

We will design a digital twin of the human stomach that combines first¬-principles mathematical modelling (Multiphysics) with Machine Learning (ML). Multiphysics will reproduce the mechanics and the chemistry of the gastric environment, while an Artificial Neural Network (ANN) trained with Reinforcement Learning (RL) will replicate the activity of the Enteric Nervous System (ENS). The model will be validated against in-vivo data from MRI images.

Thanks to its ‘neural’ component, the model will autonomously adapt its motility, emptying, and secretion to the physicochemical properties of the food and the nutrient feedback mechanism from small intestine to control gastric emptying. In collaboration with GSK, different pharmaceutical dosage forms will be simulated in the virtual gastric environment. The resulting disintegration and spatiotemporal dissolution profiles will inform Physiologically Based Pharmacokinetic (PBPK) platforms used by pharmaceutical companies predicting as well as the in vitro dissolution specifications enabling the development of biopredictive methods. Furthermore, the model will provide valuable information on how a specific type of meal affects the motility of the stomach and, therefore, the dissolution and transit of the drug to the small intestine. This is of particular importance for the development of immediate and extended release formulation products.