Cardiac resynchronization therapy (CRT) is a common intervention for patients with dyssynchronous heart failure, yet approximately one-third of recipients fail to respond due to multiple contributing factors, including suboptimal lead placement. Identifying optimal pacing sites remains challenging, largely due to patient-specific anatomical variability and the limitations of current individualized planning strategies. In a step towards constructing an in-silico approach to help address this issue, we develop two geometric deep learning (DL) models, based on graph neural network (GNN) and geometry-informed neural operator (GINO), to predict cardiac activation time maps in real time for CRT planning and optimization. Both models are trained on a large dataset generated from finite-element (FE) simulations over a wide range of synthetic left ventricular (LV) geometries, pacing site configurations, and tissue conductivities. In testing, the GINO model outperforms the GNN model on synthetic test data, with lower prediction errors (1.38% vs 2.44%), while both demonstrate comparable performance on real-world LV geometries (GINO: 4.79% vs GNN: 4.07%). Using the trained models, we also develop a workflow for optimizing the pacing site in CRT from a given activation time map and LV geometry. The trained DL models were capable of recovering the ground truth subject-specific parameters from the noisy activation time map with small errors. In conjunction with an interactive web-based graphical user interface (GUI) available at https://dcsim.egr.msu.edu/, this study shows promising potential as a clinical decision-support tool for personalized pre-procedural CRT optimization.

翻译：心脏再同步化治疗（CRT）是治疗不同步性心力衰竭患者的常见干预手段，但约三分之一的受治者因多种因素（包括电极放置位置欠佳）而未能获得良好疗效。确定最佳起搏位点仍具挑战性，这主要源于患者个体间的解剖结构差异以及现有个体化规划策略的局限性。为构建一种计算模拟方法以帮助解决此问题，我们基于图神经网络（GNN）和几何信息神经算子（GINO）开发了两种几何深度学习（DL）模型，用于实时预测心脏激动时间图，以支持CRT规划与优化。两种模型均在通过有限元（FE）仿真生成的大规模数据集上进行训练，该数据集涵盖了广泛的合成左心室（LV）几何结构、起搏位点配置及组织电导率。在测试中，GINO模型在合成测试数据上表现优于GNN模型，预测误差更低（1.38%对比2.44%），而两者在真实左心室几何结构上表现出相近的性能（GINO：4.79%对比GNN：4.07%）。利用训练好的模型，我们还开发了一个工作流程，用于根据给定的激动时间图和左心室几何结构优化CRT中的起搏位点。训练后的深度学习模型能够以较小误差从含噪声的激动时间图中恢复出真实受试者特异性参数。结合可通过https://dcsim.egr.msu.edu/访问的交互式网络图形用户界面（GUI），本研究显示出作为个体化术前CRT优化临床决策支持工具的潜在应用前景。