Boundary Learning by Optimization with Topological Constraints

Jain, Viren; Bollmann, Benjamin; Richardson, Mark A.; Berger, Daniel R.; Helmstaedter, Moritz; Briggman, Kevin L.; Denk, Winfried; Bowden, Jared B.; Mendenhall, John M.; Abraham, Wickliffe C.; Harris, Kristen M.; Kasthuri, Narayanan; Hayworth, Kenneth J.; Schalek, Richard; Tapia, Juan Carlos; Lichtman, Jeff W.; Seung, H. Sebastian

doi:10.1109/cvpr.2010.5539950

Cited by 100 publications

(93 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Splits and Mergers Warping error is a segmentation metric that penalizes topological disagreements between the two labelings [6]. The warping error is the squared Euclidean distance between Y and the "best warping" L ofŶ onto Y such that the warping L is from the class Λ that preserve topological structure:…”

Section: Methodsmentioning

confidence: 99%

Anisotropic ssTEM Image Segmentation Using Dense Correspondence across Sections

Laptev

Vezhnevets

Dwivedi

et al. 2012

Medical Image Computing and Computer-Assisted Intervention – MICCAI 2012

View full text Add to dashboard Cite

Abstract. Connectomics based on high resolution ssTEM imagery requires reconstruction of the neuron geometry from histological slides. We present an approach for the automatic membrane segmentation in anisotropic stacks of electron microscopy brain tissue sections. The ambiguities in neuronal segmentation of a section are resolved by using the context from the neighboring sections. We find the global dense correspondence between the sections by SIFT Flow algorithm, evaluate the features of the corresponding pixels and use them to perform the segmentation. Our method is 3.6 and 6.4% more accurate in two different accuracy metrics than the algorithm with no context from other sections.

show abstract

Section: Methodsmentioning

confidence: 99%

Anisotropic ssTEM Image Segmentation Using Dense Correspondence across Sections

Laptev

Vezhnevets

Dwivedi

et al. 2012

Medical Image Computing and Computer-Assisted Intervention – MICCAI 2012

View full text Add to dashboard Cite

show abstract

“…Convolutional neural networks have shown good performance for neurite boundary detection [32,17,10], and may well perform better than the boundary classification method we used for our segmentation evaluation. Our choice was motivated by the fact that the state-of-the-art 3-D convolutional neural network approach for this problem is currently far from a settled matter, and we believe our method to be similar in performance; furthermore, it would have been highly impractical to spend the several weeks to months 5 of GPU time required to the train the network for each variant and data split that we tested.…”

Section: Comparison Of Segmentation Accuracymentioning

confidence: 99%

“…On SBEM and FIBSEM volumes without these artifacts, the ability to view cross sections along arbitrary axes has aided humans tasked with manually tracing neurites and detecting synapses [16,15], and automated algorithms for reconstructing neurite morphology (via segmentation) have depended on 3-D features. [18,33,17,19,4,3] To eliminate these artifacts and enable truly 3-D analysis of such image volumes, we propose a coarse-to-fine optimization-based procedure EMISAC (EM Image Stack Artifact Correction). We note that a single per-section brightness and contrast adjustment (i.e.…”

Section: Introductionmentioning

confidence: 99%

Optimization-Based Artifact Correction for Electron Microscopy Image Stacks

Azadi

Maitin-Shepard

Abbeel

2014

Computer Vision – ECCV 2014

View full text Add to dashboard Cite

Abstract. Investigations of biological ultrastructure, such as comprehensive mapping of connections within a nervous system, increasingly rely on large, high-resolution electron microscopy (EM) image volumes. However, discontinuities between the registered section images from which these volumes are assembled, due to variations in imaging conditions and section thickness, among other artifacts, impede truly 3-D analysis of these volumes. We propose an optimization procedure, called EMISAC (EM Image Stack Artifact Correction), to correct these discontinuities. EMISAC optimizes the parameters of spatially varying linear transformations of the data in order to minimize the squared norm of the gradient along the section axis, subject to detail-preserving regularization. Assessment on a mouse cortex dataset demonstrates the effectiveness of our approach. Relative to the original data, EMISAC produces a large improvement both in NIQE score, a measure of statistical similarity between orthogonal cross-sections and the original image sections, as well as in accuracy of neurite segmentation, a critical task for this type of data. Compared to a recent independently-developed gradient-domain algorithm, EMISAC achieves significantly better NIQE image quality scores, and equivalent segmentation accuracy; future segmentation algorithms may be able to take advantage of the higher image quality. In addition, on several time-lapse videos, EMISAC significantly reduces lighting artifacts, resulting in greatly improved video quality. A software release is available at http://rll.berkeley.edu/2014_ECCV_ EMISAC.

show abstract

“…This makes them focus on measuring a segmentation's boundary accuracy at the pixel level but not take into account its topological correctness. However, in EM segmentation evaluation, measuring the degree of the topological correctness of a segmentation is also important because obtaining accurate analysis of the neural circuits relies on topologically correct reconstructions [5].…”

Section: Evaluation Metric: Warping Errormentioning

confidence: 99%

Ground Truth Estimation by Maximizing Topological Agreements in Electron Microscopy Data

Yang

Choe

2011

Advances in Visual Computing

View full text Add to dashboard Cite

Abstract. Manual editing can correct segmentation errors produced by automated segmentation algorithms, but it also introduces a practical challenge: the combination of multiple users' annotations of an image to obtain an estimation of the true, unknown labeling. Current estimation methods are not suited for electron microscopy (EM) images because they typically do not take into account topological correctness of a segmentation that can be critical in EM analysis. This paper presents a ground truth estimation method for EM images. Taking a collection of alternative segmentations, the algorithm seeks an estimated segmentation that is topologically equivalent and geometrically similar to the true, unknown segmentation. To this end, utilizing warping error as the evaluation metric, which measures topological disagreements between segmentations, the algorithm iteratively modifies the topology of an estimated segmentation to minimize the topological disagreements between this estimated segmentation and the given segmentations. Our experimental results obtained using EM images with densely packed cells demonstrate that the proposed method is superior to majority voting and STAPLE commonly used for combining multiple segmentation results.

show abstract

Boundary Learning by Optimization with Topological Constraints

Cited by 100 publications

References 29 publications

Anisotropic ssTEM Image Segmentation Using Dense Correspondence across Sections

Anisotropic ssTEM Image Segmentation Using Dense Correspondence across Sections

Optimization-Based Artifact Correction for Electron Microscopy Image Stacks

Ground Truth Estimation by Maximizing Topological Agreements in Electron Microscopy Data

Contact Info

Product

Resources

About