Boundary Learning by Optimization with Topological Constraints

By Viren Jain, B. Bollmann, Mark Richardson, D. Berger, Moritz Helmstaedter, Kevin L. Briggman, Winfried Denk, Jared B. Bowden, J. Mendenhall, Wickliffe C Abraham and 7 more. 2010

155 citations recorded by the source corpus.

Recent studies have shown that machine learning can improve the accuracy of detecting object boundaries in images. In the standard approach, a boundary detector is trained by minimizing its pixel-level disagreement with human boundary tracings. This naive metric is problematic because it is overly sensitive to boundary locations. This problem is solved by metrics provided with the Berkeley Segmentation Dataset, but these can be insensitive to topo-logical differences, such as gaps in boundaries. Furthermore, the Berkeley metrics have not been useful as cost functions for supervised learning. Using concepts from digital topology, we propose a new metric called the warping error that tolerates disagreements over boundary location, penalizes topological disagreements, and can be used directly as a cost function for learning boundary detection, in a method that we call Boundary Learning by Optimization with Topological Constraints (BLOTC). We trained boundary detectors on electron microscopic images of neurons, using both BLOTC and standard training. BLOTC produced substantially better performance on a 1.2 million pixel test set, as measured by both the warping error and the Rand index evaluated on segmentations generated from the boundary labelings. We also find our approach yields significantly better segmentation performance than either gPb-OWT-UCM or multiscale normalized cut, as well as Boosted Edge Learning trained directly on our data.

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

How researchers see this paper

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

In Biophysics

In Computational Theory and Mathematics

In Computer Vision and Pattern Recognition

Citations in the corpus

C3D: Generic Features for Video Analysis (2014, background): “3 D ConvNets were proposed for human action recognition [20, 46] and for medical image segmentation [19, 48].”
Parallel Multi-Dimensional LSTM, With Application to Fast Biomedical Volumetric Image Segmentation (2015, background): “• Warping error [23]: topological disagreements (object splits and mergers)”
Gland segmentation in colon histology images: The glas challenge contest (2016, methodology): “vere penalty on mismatch of the objects. One could replace these measures by less conservative ones, for example, adjusted Rand index [Hubert and Arabie, 1985] or a topology preserving warping error [Jain et al., 2010] for segmentation and elastic distance [Younes, 1998, Joshi et al., 2007] for shap”
Crowdsourcing the creation of image segmentation algorithms for connectomics (2015, methodology): “Warping error , based on digital topology, has been proposed as a metric and used as a cost function for machine learning (Jain et al., 2010a).”
Deep Neural Networks Segment Neuronal Membranes in Electron Microscopy Images (2012, background): “Warping error: a segmentation metric designed to account for topological disagreements [19]; it accounts for the number of neuron splits and mergers required to obtain the candidate segmentation from ground truth.”
Volume electron microscopy for neuronal circuit reconstruction. (2012, unknown): “In the medium-to-long term, we anticipate that the recent progress in automating and speeding the acquisition of volume EM data will carry over to automation of the analysis of these large datasets [41,42], eventually permitting complete reconstruction of mammalian circuits across millimeter scales.”
Learning computationally efficient approximations of complex image segmentation metrics (2013, background): “…been proposed that emphasize particular types of error (e.g., over- and under-segmentation [2], topological disagreements such as region splitting and merging [3]), can be robust to noise (e.g., small boundary displacement [4], white noise [5]), or correlate well with human intuition [3], [5],…”
High-precision automated reconstruction of neurons with flood-filling networks (2017, methodology): “This is conceptually different from using a segmentation-performance-dependent cost function 23,24 .”