Multi-center anatomical segmentation with heterogeneous labels via landmark-based models
This addresses the challenge of training deep learning models with inconsistent annotations across medical centers, which is incremental as it adapts an existing method to a specific bottleneck.
The paper tackled the problem of learning anatomical segmentation from multi-center datasets with heterogeneous labels, where state-of-the-art pixel-level models fail due to domain memorization and conflicting labels, and showed that HybridGNet, a landmark-based model, learns more domain-invariant features, achieving empirical evidence in chest X-ray multiclass segmentation.
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.