Disconnect to Connect: A Data Augmentation Method for Improving Topology Accuracy in Image Segmentation

2025-03-07

AI Summary

In medical image segmentation, thin tubular structures (e.g., vessels) often suffer topological disconnections due to pixel-wise misclassifications, while obtaining topology-accurate ground-truth annotations is costly and unreliable. To address this, we propose CoLeTra—a controllable, disconnection-aware data augmentation method that requires no topology-specific labels. CoLeTra synthesizes training images exhibiting *visual* discontinuities yet preserving *semantic* connectivity, thereby implicitly embedding structural continuity priors—while leaving original pixel-level ground truth unchanged. Our approach is grounded in morphological operations and graph-theoretic connectivity constraints, ensuring compatibility with mainstream architectures and loss functions. We further introduce a topology-sensitive evaluation benchmark. Experiments across multiple benchmarks demonstrate consistent improvements in topological connectivity (e.g., reduced break counts), Dice score, and Hausdorff distance. CoLeTra features intuitive hyperparameters and strong robustness. Code and dataset are publicly available.

Technology Category

• Computer Vision: Segmentation
• Constraint Satisfaction and Optimization: Mixed Discrete/Continuous Optimization
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Security and Privacy: Data transparency and provenance

Abstract

Accurate segmentation of thin, tubular structures (e.g., blood vessels) is challenging for deep neural networks. These networks classify individual pixels, and even minor misclassifications can break the thin connections within these structures. Existing methods for improving topology accuracy, such as topology loss functions, rely on very precise, topologically-accurate training labels, which are difficult to obtain. This is because annotating images, especially 3D images, is extremely laborious and time-consuming. Low image resolution and contrast further complicates the annotation by causing tubular structures to appear disconnected. We present CoLeTra, a data augmentation strategy that integrates to the models the prior knowledge that structures that appear broken are actually connected. This is achieved by creating images with the appearance of disconnected structures while maintaining the original labels. Our extensive experiments, involving different architectures, loss functions, and datasets, demonstrate that CoLeTra leads to segmentations topologically more accurate while often improving the Dice coefficient and Hausdorff distance. CoLeTra's hyper-parameters are intuitive to tune, and our sensitivity analysis shows that CoLeTra is robust to changes in these hyper-parameters. We also release a dataset specifically suited for image segmentation methods with a focus on topology accuracy. CoLetra's code can be found at https://github.com/jmlipman/CoLeTra.

Problem

• Improves topology accuracy in image segmentation of thin structures.
• Addresses challenges in obtaining precise, topologically-accurate training labels.
• Enhances segmentation accuracy using a novel data augmentation strategy.

Innovation

• Data augmentation improves topology accuracy
• CoLeTra integrates prior knowledge of connectivity
• Robust to hyper-parameter changes, intuitive tuning