Efficient Anatomical Labeling of Pulmonary Tree Structures via Deep Point-Graph Representation-based Implicit Fields

TL;DR

This work introduces graph learning on skeletonized structures, incorporating differentiable feature fusion for improved topology and long-distance context capture and employs an implicit function for efficient conversion of sparse representations into dense reconstructions end-to-end.

Abstract

Pulmonary diseases rank prominently among the principal causes of death worldwide. Curing them will require, among other things, a better understanding of the complex 3D tree-shaped structures within the pulmonary system, such as airways, arteries, and veins. Traditional approaches using high-resolution image stacks and standard CNNs on dense voxel grids face challenges in computational efficiency, limited resolution, local context, and inadequate preservation of shape topology. Our method addresses these issues by shifting from dense voxel to sparse point representation, offering better memory efficiency and global context utilization. However, the inherent sparsity in point representation can lead to a loss of crucial connectivity in tree-shaped structures. To mitigate this, we introduce graph learning on skeletonized structures, incorporating differentiable feature fusion for improved topology and long-distance context capture. Furthermore, we employ an implicit function for efficient conversion of sparse representations into dense reconstructions end-to-end. The proposed method not only delivers state-of-the-art performance in labeling accuracy, both overall and at key locations, but also enables efficient inference and the generation of closed surface shapes. Addressing data scarcity in this field, we have also curated a comprehensive dataset to validate our approach. Data and code are available at \url{https://github.com/M3DV/pulmonary-tree-labeling}.

Figures (10)

• Figure 1: Pulmonary Tree Labeling. (a) The pulmonary tree consists of three anatomic structures (airway, artery and vein). (b) Given a binary volume representing a tree structure as input, we label each voxel into one of 19 classes (18 pulmonary segments + 1 background out of lung) based on branching regions.
• Figure 2: A Comparison of Data Representation for Pulmonary Tree Labeling. The CNN-based voxel methods are either low-resolution (down-sampled) or local (sliding-window). The standard sparse representation like point and graph is global but it is not trivial to reconstruct high-quality dense volume. Our method that combines point, graph, and implicit fields produces high-quality dense reconstruction efficiently.
• Figure 3: Anatomy of Pulmonary Trees and Pulmonary Segments. Each pulmonary tree branch corresponds to a pulmonary segment. The intersegmental vein is highlighted in red, which lies along the pulmonary segment border.
• Figure 4: Sample Illustration with Voxel-level and Graph-level Metrics. Due to the large number of voxels, voxel-level metric often overlooks labeling at key regions. Given two test samples with similar voxel performance, the sample with better graph-level performance ( Case A and B) tends to perform better at key points than its counterpart ( Case A' and B').
• Figure 5: Skeleton Graphs Directly from vesselvio and Two Types of Imperfections. (1) The centerline (CL) points, which represent the path of the tree branch as discrete and connected coordinates, might contain cliques. (2) The coordinates of leaf graph nodes might fall outside the foreground volume.