clDice - a Novel Topology-Preserving Loss Function for Tubular Structure Segmentation (2020-03-16T00:00:00.000000Z)

TL;DR

This work theoretically proves that clDice guarantees topology preservation up to homotopy equivalence for binary 2D and 3D segmentation, and pro-poses a computationally efficient, differentiable loss function (soft-clDice) for training arbitrary neural segmentation networks.

Abstract

Accurate segmentation of tubular, network-like structures, such as vessels, neurons, or roads, is relevant to many fields of research. For such structures, the topology is their most important characteristic; particularly preserving connectedness: in the case of vascular networks, missing a connected vessel entirely alters the blood-flow dynamics. We introduce a novel similarity measure termed centerlineDice (short clDice), which is calculated on the inter-section of the segmentation masks and their (morphological) skeleta. We theoretically prove that clDice guarantees topology preservation up to homotopy equivalence for binary 2D and 3D segmentation. Extending this, we pro-pose a computationally efficient, differentiable loss function (soft-clDice) for training arbitrary neural segmentation networks. We benchmark the soft-clDice loss on five public datasets, including vessels, roads and neurons (2D and 3D). Training on soft-clDice leads to segmentation with more accurate connectivity information, higher graph similarity, and better volumetric scores.

Authors

J. Paetzold

3 papers

Suprosanna Shit

3 papers

Bjoern H Menze

10 papers

TL;DR

Abstract

Authors

References59 items

Promoting Connectivity of Network-Like Structures by Enforcing Region Separation

TopoAL: An Adversarial Learning Approach for Topology-Aware Road Segmentation

A distance-based loss for smooth and continuous skin layer segmentation in optoacoustic images

Cellular and Molecular Probing of Intact Human Organs

Learning tree-structured representation for 3D coronary artery segmentation

Image-Conditioned Graph Generation for Road Network Extraction

A Deep Learning Design for Improving Topology Coherence in Blood Vessel Segmentation

A Topological Loss Function for Deep-Learning Based Image Segmentation Using Persistent Homology

Shape-Aware Complementary-Task Learning for Multi-Organ Segmentation

Unsupervised Microvascular Image Segmentation Using an Active Contours Mimicking Neural Network

Large Scale Image Segmentation with Structured Loss Based Deep Learning for Connectome Reconstruction

Topology-Preserving Deep Image Segmentation

Joint Segmentation and Path Classification of Curvilinear Structures

Automated analysis of whole brain vasculature using machine learning

Transfer learning from synthetic data reduces need for labels to segment brain vasculature and neural pathways in 3D

Retinal vessel segmentation of color fundus images using multiscale convolutional neural network with an improved cross-entropy loss function

A Multi-task Network to Detect Junctions in Retinal Vasculature

DeepVesselNet: Vessel Segmentation, Centerline Prediction, and Bifurcation Detection in 3-D Angiographic Volumes

Beyond the Pixel-Wise Loss for Topology-Aware Delineation

Whole-Brain Vasculature Reconstruction at the Single Capillary Level

Vascular Segmentation in TOF MRA Images of the Brain Using a Deep Convolutional Neural Network

3D Convolutional Neural Networks with Graph Refinement for Airway Segmentation Using Incomplete Data Labels

Generalised Dice overlap as a deep learning loss function for highly unbalanced segmentations

Efficient Algorithms for Moral Lineage Tracing

Topology Aware Fully Convolutional Networks for Histology Gland Segmentation

3D U-Net: Learning Dense Volumetric Segmentation from Sparse Annotation

V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation

Deep vessel tracking: A generalized probabilistic approach via deep learning

Metrics for evaluating 3D medical image segmentation: analysis, selection, and tool

U-Net: Convolutional Networks for Biomedical Image Segmentation

Generalized Connectivity Constraints for Spatio-temporal 3D Reconstruction

Tree Shape Priors with Connectivity Constraints Using Convex Relaxation on General Graphs

A Higher-Order CRF Model for Road Network Extraction

Tissue metabolism driven arterial tree generation

Morphological and Pathological Evolution of the Brain Microcirculation in Aging and Alzheimer’s Disease

Probabilistic image segmentation with closedness constraints

Cerebrovascular dysfunction and microcirculation rarefaction precede white matter lesions in a mouse genetic model of cerebral ischemic small vessel disease.

Global connectivity potentials for random field models

Topology cuts: A novel min-cut/max-flow algorithm for topology preserving segmentation in N-D images

Active Contours Under Topology Control—Genus Preserving Level Sets

Graph cut based image segmentation with connectivity priors

A review of vessel extraction techniques and algorithms

Ridge-based vessel segmentation in color images of the retina

A Topology Preserving Level Set Method for Geometric Deformable Models

A 3-subiteration 3D thinning algorithm for extracting medial surfaces

Topological Persistence and Simplification

On Topology Preservation in 2-D and 3-D Thinning

Skeletonization using an extended Euclidean distance transform

A skeletonization algorithm by maxima tracking on Euclidean distance transform

Building Skeleton Models via 3-D Medial Surface/Axis Thinning Algorithms

On topology preservation in 3D thinning

Digital topology: Introduction and survey

Combinatorial Topology, Volume 3.

Towards Reliable Evaluation of Algorithms for Road Network Reconstruction from Aerial Images

Joint 3-D vessel segmentation and centerline extraction using oblique Hough forests with steerable filters

Machine Learning for Aerial Image Labeling

Effective Computational Geometry for Curves and Surfaces Chapter 7 Computational Topology : An Introduction

Digital Topology

Ieee Transactions on Pattern Analysis and Machine Intelligence 1 Reconstructing Curvilinear Networks Using Path Classifiers and Integer Programming

Field of Study

Journal Information

Name

Page

Venue Information

Name

Type

URL

Alternate Names