Inference-Time Orthogonal Seeding Enables Geometry-Aligned 3D Organ Segmentation for Slice-Propagation Methods

2026-08-12

AI Summary

This work addresses the geometric error accumulation in existing single-slice propagation methods, which propagate annotations along only one axis and neglect information from coronal and sagittal planes, severely degrading surface distance metrics far from the seed slice. To overcome this limitation, the authors propose an orthogonal seed configuration strategy at inference time that leverages seed slices from all three orthogonal anatomical planes—axial, coronal, and sagittal—without requiring multi-axis training. By integrating unlabeled multi-planar registration and a label fusion rule, the method effectively exploits full 3D geometric structure. Experiments on multi-organ CT datasets demonstrate substantial improvements over single-axis baselines: Dice coefficient increases by 21.9%, normalized surface Dice by 25.5%, and mean Hausdorff distance decreases by 53.5%, confirming that orthogonal geometric layout is more critical than the number of seed slices.

Abstract

Dense voxel-level annotation remains a major bottleneck in 3D medical image segmentation. Single-slice propagation methods such as Sli2Vol reduce this burden by propagating one annotated seed slice through a volume using label-free registration. However, axial-only propagation accumulates errors with distance from the seed, especially in surface-distance metrics, because it ignores coronal and sagittal evidence and therefore underuses the 3D information available in CT/MRI volumes. To better leverage volumetric geometry, we study how key training and inference choices affect slice-propagation models, including single-axis versus multi-axis label-free registration, single-seed versus multi-seed propagation, and orthogonal seed configurations. Instead of propagating from a single axial seed, we use three orthogonal seeds---one axial, one coronal, and one sagittal---and fuse their propagated labels with a simple label-free rule. Our results show that the training paradigm has limited impact: an axially trained network applied to off-axis seeds captures nearly all the improvement, while explicit three-axis training adds little. Instead, performance is driven by inference-time seed geometry, especially orthogonality rather than the number of annotated slices, as a budget-matched three-axial control provides no benefit and can even degrade performance. On a multi-organ CT cohort, orthogonal seeding with the axial Sli2Vol backbone improves Dice by 21.9%, Normalized Surface Dice by 25.5%, and reduces Average Hausdorff Distance by 53.5% over the single-axis baseline.

Problem

• 3D medical image segmentation
• slice propagation
• orthogonal seeding
• volumetric geometry
• annotation bottleneck

Innovation

• orthogonal seeding
• slice propagation
• 3D medical segmentation
• inference-time fusion
• geometry-aligned segmentation