English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Finite element analysis (FEA) forms the cornerstone of modeling cardiacbiomechanics but is computationally expensive, limiting its clinicalapplication for digital twin creation, which often requires tens to hundreds ofsimulations to estimate tissue parameters. We developed an attention-enhancedgraph neural network (GNN)-based FEA emulator, HeartSimSage, to rapidly predictpassive biventricular myocardial displacements from patient-specificgeometries, chamber pressures, and material properties. HeartSimSage addressesthe limitations of current emulators by effectively handling diversethree-dimensional (3D) biventricular geometries, mesh topologies, fiberdirections, structurally based constitutive models, and physiological boundaryconditions. It supports flexible mesh structures with variable node counts,orderings, and element connectivity. To optimize information propagation, wedesigned a neighboring connection strategy inspired by Graph Sample andAggregate (GraphSAGE) that prioritizes local interactions while maintainingmid-to-long-range dependencies. We further incorporated Laplace-Dirichletsolutions for enhanced spatial encoding and employed subset-based training forimproved efficiency. By integrating an attention mechanism, HeartSimSageadaptively weighs neighbor contributions and filters irrelevant information,enhancing prediction accuracy. HeartSimSage achieves approximately 13,000xspeedup on GPU and 190x on CPU compared to traditional FEA, while maintaining anominal averaged error of 0.13% +- 0.12% in predicting biventriculardisplacements. We validated our model using a published left ventricle datasetand conducted sensitivity analyses on hyperparameters, neighboring strategies,and the attention mechanism.

HeartSimSage: Attention-Enhanced Graph Neural Networks for Accelerating Cardiac Mechanics Modeling