Topology-Preserving Deep Image Segmentation

By Xiaoling Hu, Fuxin Li, Dimitris Samaras, Chao Chen. Neural Information Processing Systems, 2019

398 citations recorded by the source corpus.

Segmentation algorithms are prone to make topological errors on fine-scale structures, e.g., broken connections. We propose a novel method that learns to segment with correct topology. In particular, we design a continuous-valued loss function that enforces a segmentation to have the same topology as the ground truth, i.e., having the same Betti number. The proposed topology-preserving loss function is differentiable and we incorporate it into end-to-end training of a deep neural network. Our method achieves much better performance on the Betti number error, which directly accounts for the topological correctness. It also performs superiorly on other topology-relevant metrics, e.g., the Adjusted Rand Index and the Variation of Information. We illustrate the effectiveness of the proposed method on a broad spectrum of natural and biomedical datasets.

Subfields: Computational Theory and Mathematics, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/9c52fd008a37230ecf64af3aefa532fd7af685a6

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computational Theory and Mathematics

In Computer Vision and Pattern Recognition

Citations in the corpus

A Topological Loss Function for Deep-Learning based Image Segmentation using Persistent Homology (2019, methodology): “k to take this approach was our preliminary work in [4], although this idea has subsequently been developed for the speciﬁc case of one-dimensional topological features (i.e. connected components) in [20]. Here, we extend our preliminary work [4] by introducing an explicit topological loss function ”
Dynamic Snake Convolution based on Topological Geometric Constraints for Tubular Structure Segmentation (2023, methodology): “The existing methods [30, 5, 10] use modified Wasserstein distance to compute the best match between the points generated by the output and the points generated by the groundtruth, and the outlier points without optimal pairing are matched to the diagonal and do not participate in the loss calculati”
clDice - a Novel Topology-Preserving Loss Function for Tubular Structure Segmentation (2020, methodology): “proposed the first continuous-valued loss function based on the Betti number and persistent homology [17].”