MV-GNN: Generation of continuous geometric representations of mitral valve motion from 3D+t echocardiography.

We present a geometric deep-learning method for reconstructing a temporally continuous mitral valve surface mesh from 3D transesophageal echocardiography sequences. Our approach features a supervised end-to-end deep learning architecture that combines a convolutional neural network-based voxel encoder and decoder with a graph neural network-based multi-resolution mesh decoder, all trained on sparse landmark annotations. Key elements of our methodology include a tube-shaped prototype mesh with labeled vertices, a specialized loss function to preserve the known inlet and outlet, and a rigid alignment system for anatomical landmarks. A custom term in the loss function prevents self-intersecting geometries within the valve mesh, promoting point correspondence and facilitating a continuous representation of valve anatomy over time. An ablation study evaluates the impact of different loss term configurations on model performance, highlighting the effectiveness of each individual loss term. Our Mitral Valve Graph Neural Network (MV-GNN) outperforms existing deep-learning methods on most distance metrics for the annulus and leaflets. The continuous valve motion representations generated by our approach (3D+t) exhibit distance measures comparable to our 3D solution, demonstrating its potential for analyzing mitral valve dynamics and enhancing personalized simulations for hemodynamic assessment and therapy planning.

Learning three-dimensional aortic root assessment based on sparse annotations.

Purpose: Analyzing the anatomy of the aorta and left ventricular outflow tract (LVOT) is crucial for risk assessment and planning of transcatheter aortic valve implantation (TAVI). A comprehensive analysis of the aortic root and LVOT requires the extraction of the patient-individual anatomy via segmentation. Deep learning has shown good performance on various segmentation tasks. If this is formulated as a supervised problem, large amounts of annotated data are required for training. Therefore, minimizing the annotation complexity is desirable. Approach: We propose two-dimensional (2D) cross-sectional annotation and point cloud-based surface reconstruction to train a fully automatic 3D segmentation network for the aortic root and the LVOT. Our sparse annotation scheme enables easy and fast training data generation for tubular structures such as the aortic root. From the segmentation results, we derive clinically relevant parameters for TAVI planning. Results: The proposed 2D cross-sectional annotation results in high inter-observer agreement [Dice similarity coefficient (DSC): 0.94]. The segmentation model achieves a DSC of 0.90 and an average surface distance of 0.96 mm. Our approach achieves an aortic annulus maximum diameter difference between prediction and annotation of 0.45 mm (inter-observer variance: 0.25 mm). Conclusions: The presented approach facilitates reproducible annotations. The annotations allow for training accurate segmentation models of the aortic root and LVOT. The segmentation results facilitate reproducible and quantifiable measurements for TAVI planning.