Coordinative Learning with Ordinal and Relational Priors for Volumetric Medical Image Segmentation

TL;DR

This paper introduces CORAL, a novel framework for volumetric medical image segmentation that leverages continuous anatomical similarities and global directional consistency to improve representation learning and segmentation accuracy.

Contribution

CORAL is the first method to integrate both ordinal and relational priors for volumetric segmentation, capturing local and global anatomical structures.

Findings

Achieves state-of-the-art results with limited annotations

Learns anatomically meaningful representations

Outperforms existing contrastive methods

Abstract

Volumetric medical image segmentation presents unique challenges due to the inherent anatomical structure and limited availability of annotations. While recent methods have shown promise by contrasting spatial relationships between slices, they rely on hard binary thresholds to define positive and negative samples, thereby discarding valuable continuous information about anatomical similarity. Moreover, these methods overlook the global directional consistency of anatomical progression, resulting in distorted feature spaces that fail to capture the canonical anatomical manifold shared across patients. To address these limitations, we propose Coordinative Ordinal-Relational Anatomical Learning (CORAL) to capture both local and global structure in volumetric images. First, CORAL employs a contrastive ranking objective to leverage continuous anatomical similarity, ensuring relational feature distances between slices are proportional to their anatomical position differences. In addition, CORAL incorporates an ordinal objective to enforce global directional consistency, aligning the learned feature distribution with the canonical anatomical progression across patients. Learning these inter-slice relationships produces anatomically informed representations that benefit the downstream segmentation task. Through this coordinative learning framework, CORAL achieves state-of-the-art performance on benchmark datasets under limited-annotation settings while learning representations with meaningful anatomical structure. Code is available at https://github.com/haoyiwang25/CORAL.