Crowdsourcing the creation of image segmentation algorithms for connectomics
2015 (English)In: Frontiers in Neuroanatomy, ISSN 1662-5129, E-ISSN 1662-5129, Vol. 9, no 142Article in journal (Refereed) PublishedText
To stimulate progress in automating the reconstruction of neural circuits, we organized the first international challenge on 2D segmentation of electron microscopic (EM) images of the brain. Participants submitted boundary maps predicted for a test set of images, and were scored based on their agreement with a consensus of human expert annotations. The winning team had no prior experience with EM images, and employed a convolutional network. This "deep learning" approach has since become accepted as a standard for segmentation of FM images. The challenge has continued to accept submissions, and the best so far has resulted from cooperation between two teams. The challenge has probably saturated, as algorithms cannot progress beyond limits set by ambiguities inherent in 2D scoring and the size of the test dataset. Retrospective evaluation of the challenge scoring system reveals that it was not sufficiently robust to variations in the widths of neurite borders. We propose a solution to this problem, which should be useful for a future 3D segmentation challenge.
Place, publisher, year, edition, pages
FRONTIERS MEDIA SA , 2015. Vol. 9, no 142
connectomics; electron microscopy; image segmentation; machine learning; reconstruction
IdentifiersURN: urn:nbn:se:liu:diva-123808DOI: 10.3389/fnana.2015.00142ISI: 000365846500001PubMedID: 26594156OAI: oai:DiVA.org:liu-123808DiVA: diva2:892912
Funding Agencies|NIH [1R01NS075314-01]; ARO [W911NF-12-1-0594]; DARPA [HR0011-14-2-0004]; Human Frontier Science Program; Mathers Foundation; Gatsby Charitable Foundation; Howard Hughes Medical Institute; [CZ.1.07/2.3.00/20.0094]2016-01-112016-01-112016-05-04