Multi-view Hybrid Graph Convolutional Network for Volume-to-mesh Reconstruction in Cardiovascular MRI

2023-11-22

arXiv.org

AI Summary

To address error accumulation and low efficiency in conventional multi-stage pipelines for 3D cardiac mesh reconstruction from cardiovascular MRI, this work proposes an end-to-end paradigm that directly generates high-fidelity surface and volumetric hexahedral meshes from 3D volumetric images. Methodologically, we introduce the first multi-view HybridVNet architecture, seamlessly integrating CNNs with graph convolutional networks; further, we pioneer the synergistic incorporation of variational graph autoencoders, deep supervision, and topology-aware mesh regularization into cardiac modeling. Evaluated on the UK Biobank dataset, our method achieves significant improvements over state-of-the-art approaches: mean contour distance for left ventricular myocardium decreases by 27% (1.86 → 1.35 mm), endocardial Hausdorff distance improves by 18% (4.74 → 3.89 mm), and Dice coefficient increases by 8% (0.78 → 0.84). These results demonstrate superior accuracy, robustness, and clinical applicability.

Abstract

Cardiovascular magnetic resonance imaging is emerging as a crucial tool to examine cardiac morphology and function. Essential to this endeavour are anatomical 3D surface and volumetric meshes derived from CMR images, which facilitate computational anatomy studies, biomarker discovery, and in-silico simulations. Traditional approaches typically follow complex multi-step pipelines, first segmenting images and then reconstructing meshes, making them time-consuming and prone to error propagation. In response, we introduce HybridVNet, a novel architecture for direct image-to-mesh extraction seamlessly integrating standard convolutional neural networks with graph convolutions, which we prove can efficiently handle surface and volumetric meshes by encoding them as graph structures. To further enhance accuracy, we propose a multi-view HybridVNet architecture which processes both long axis and short axis CMR, showing that it can increase the performance of cardiac MR mesh generation. Our model combines traditional convolutional networks with variational graph generative models, deep supervision and mesh-specific regularisation. Experiments on a comprehensive dataset from the UK Biobank confirm the potential of HybridVNet to significantly advance cardiac imaging and computational cardiology by efficiently generating high-fidelity meshes from CMR images. Multi-view HybridVNet outperforms the state-of-the-art, achieving improvements of up to $sim$27% reduction in Mean Contour Distance (from 1.86 mm to 1.35 mm for the LV Myocardium), up to $sim$18% improvement in Hausdorff distance (from 4.74 mm to 3.89mm, for the LV Endocardium), and up to $sim$8% in Dice Coefficient (from 0.78 to 0.84, for the LV Myocardium), highlighting its superior accuracy.

Problem

• Direct image-to-mesh reconstruction for cardiac MRI
• Multi-view integration of long and short axis CMR
• Hybrid CNN-graph model for high-fidelity mesh generation

Innovation

• HybridVNet integrates CNNs with graph convolutions
• Multi-view architecture processes long and short axis CMR
• Combines variational graph models with deep supervision