Development of a User-Friendly Pipeline for Constructing Atrial Models at Scale: Importance of the End-User for Clinical Uptake

Laura Bevis1, Semhar Biniam Misghina1, Elisa Rauseo1, Carlos Edgar Lopez Barrera1, Gernot Plank2, Edward Vigmond3, Axel Loewe4, Elias Karabelas5, Steffen E Petersen1, Gregory Slabaugh1, Anthony Mathur1, Caroline H Roney1
1Queen Mary University of London, 2Medical University of Graz, 3LIRYC - University of Bordeaux, 4Karlsruhe Institute of Technology (KIT), 5University of Graz

Abstract

It is critical to consider reproducibility and user-friendliness when developing models to ensure their accuracy for large-scale clinical trials and realistic uptake in clinical practice. Here we present a pipeline for constructing atrial models at scale, developed with the end-user in mind.

The atrial modelling toolkit (atrialmtk: https://github.com/pcmlab/atrialmtk) allows the user to produce simulation grade atrial meshes that incorporate atrial regions, fibres, and transmural variations across the atrial wall, which are used as input for electrophysiological simulation. Several workflows allow the input of various data types that may be available to the user, including MRI images, electroanatomical mapping data, and artificial geometries such as those produced by statistical shape models. Different fibre distributions and bilayer or volumetric meshes can be chosen by the user, and the simulation results used to investigate the effect of fibres and fibrosis on fibrillatory dynamics. Clear instructions, computational requirements, and the required expertise of the user are detailed, and relevant training materials identified at each step of the model pipeline to ensure its correct use and implementation. The pipeline was tested for compatibility, implementation time and user-friendliness, across computer operating systems and by multiple users with differing levels of expertise, computational skill and level of clinical understanding, and improvements made to maximise its compatibility, reproducibility and clarity of instruction.

Overall, we have developed a pipeline for the reproducible construction of atrial models. Its success at scale has been demonstrated in a previous study of 1000 atrial geometries, and its compatibility and ease of use by the reproduction of results by users external to the field of electrophysiology. By extending our testing to a larger range of clinical end-users, we hope to increase the likelihood of the uptake of models in clinical practice and increase the engagement of clinicians with other digital twin initiatives.