Multi-center anatomical segmentation with heterogeneous labels via landmark-based models

TL;DR

This paper addresses the challenge of multi-center anatomical segmentation with heterogeneous labels by proposing a landmark-based model, HybridGNet, which learns more domain-invariant features and outperforms pixel-level models in this context.

Contribution

The paper introduces HybridGNet, a landmark-based segmentation model that effectively handles heterogeneous labels in multi-center datasets, overcoming domain memorization issues of traditional pixel-level models.

Findings

HybridGNet learns more domain-invariant features.

Pixel-level models fail due to domain memorization and conflicting labels.

Empirical evidence shows HybridGNet's superior performance in chest X-ray segmentation.

Abstract

Learning anatomical segmentation from heterogeneous labels in multi-center datasets is a common situation encountered in clinical scenarios, where certain anatomical structures are only annotated in images coming from particular medical centers, but not in the full database. Here we first show how state-of-the-art pixel-level segmentation models fail in naively learning this task due to domain memorization issues and conflicting labels. We then propose to adopt HybridGNet, a landmark-based segmentation model which learns the available anatomical structures using graph-based representations. By analyzing the latent space learned by both models, we show that HybridGNet naturally learns more domain-invariant feature representations, and provide empirical evidence in the context of chest X-ray multiclass segmentation. We hope these insights will shed light on the training of deep learning models with heterogeneous labels from public and multi-center datasets.