Coordinative Learning with Ordinal and Relational Priors for Volumetric Medical Image Segmentation
Haoyi Wang
TL;DR
This work tackles limited-annotation volumetric medical image segmentation by introducing Coordinative Ordinal-Relational Anatomical Learning (CORAL), which learns anatomically meaningful representations through two complementary objectives: Relational Anatomical Learning (RAL) and Ordinal Anatomical Learning (OAL). RAL replaces hard thresholding with a listwise relational ranking to capture continuous anatomical similarity between slices, while OAL enforces global directional consistency by aligning pairwise slice directions with a learnable canonical anatomical progression. The combined loss $\mathcal{L}_{CORAL} = \mathcal{L}_{RAL} + \mathcal{L}_{OAL}$ yields finer local structure and globally aligned trajectories, improving segmentation performance in low-annotation regimes on datasets such as ACDC and CHD. The approach demonstrates state-of-the-art results and provides anatomically informed pre-training representations that transfer effectively to downstream segmentation, with code available at the authors' repository.
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.