Dense cellular segmentation for EM using 2D–3D neural network ensembles
Abstract Biologists who use electron microscopy (EM) images to build nanoscale 3D models of whole cells and their organelles have historically been limited to small numbers of cells and cellular features due to constraints in imaging and analysis. This has been a major factor limiting insight into t...
Main Authors: | , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Nature Publishing Group
2021-01-01
|
Series: | Scientific Reports |
Online Access: | https://doi.org/10.1038/s41598-021-81590-0 |
id |
doaj-41216f2c37bc4441ab20307a221baaf2 |
---|---|
record_format |
Article |
spelling |
doaj-41216f2c37bc4441ab20307a221baaf22021-01-31T16:19:53ZengNature Publishing GroupScientific Reports2045-23222021-01-0111111110.1038/s41598-021-81590-0Dense cellular segmentation for EM using 2D–3D neural network ensemblesMatthew D. Guay0Zeyad A. S. Emam1Adam B. Anderson2Maria A. Aronova3Irina D. Pokrovskaya4Brian Storrie5Richard D. Leapman6National Institute of Biomedical Imaging and Bioengineering, NIHNational Institute of Biomedical Imaging and Bioengineering, NIHNational Institute of Biomedical Imaging and Bioengineering, NIHNational Institute of Biomedical Imaging and Bioengineering, NIHUniversity of Arkansas for Medical SciencesUniversity of Arkansas for Medical SciencesNational Institute of Biomedical Imaging and Bioengineering, NIHAbstract Biologists who use electron microscopy (EM) images to build nanoscale 3D models of whole cells and their organelles have historically been limited to small numbers of cells and cellular features due to constraints in imaging and analysis. This has been a major factor limiting insight into the complex variability of cellular environments. Modern EM can produce gigavoxel image volumes containing large numbers of cells, but accurate manual segmentation of image features is slow and limits the creation of cell models. Segmentation algorithms based on convolutional neural networks can process large volumes quickly, but achieving EM task accuracy goals often challenges current techniques. Here, we define dense cellular segmentation as a multiclass semantic segmentation task for modeling cells and large numbers of their organelles, and give an example in human blood platelets. We present an algorithm using novel hybrid 2D–3D segmentation networks to produce dense cellular segmentations with accuracy levels that outperform baseline methods and approach those of human annotators. To our knowledge, this work represents the first published approach to automating the creation of cell models with this level of structural detail.https://doi.org/10.1038/s41598-021-81590-0 |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Matthew D. Guay Zeyad A. S. Emam Adam B. Anderson Maria A. Aronova Irina D. Pokrovskaya Brian Storrie Richard D. Leapman |
spellingShingle |
Matthew D. Guay Zeyad A. S. Emam Adam B. Anderson Maria A. Aronova Irina D. Pokrovskaya Brian Storrie Richard D. Leapman Dense cellular segmentation for EM using 2D–3D neural network ensembles Scientific Reports |
author_facet |
Matthew D. Guay Zeyad A. S. Emam Adam B. Anderson Maria A. Aronova Irina D. Pokrovskaya Brian Storrie Richard D. Leapman |
author_sort |
Matthew D. Guay |
title |
Dense cellular segmentation for EM using 2D–3D neural network ensembles |
title_short |
Dense cellular segmentation for EM using 2D–3D neural network ensembles |
title_full |
Dense cellular segmentation for EM using 2D–3D neural network ensembles |
title_fullStr |
Dense cellular segmentation for EM using 2D–3D neural network ensembles |
title_full_unstemmed |
Dense cellular segmentation for EM using 2D–3D neural network ensembles |
title_sort |
dense cellular segmentation for em using 2d–3d neural network ensembles |
publisher |
Nature Publishing Group |
series |
Scientific Reports |
issn |
2045-2322 |
publishDate |
2021-01-01 |
description |
Abstract Biologists who use electron microscopy (EM) images to build nanoscale 3D models of whole cells and their organelles have historically been limited to small numbers of cells and cellular features due to constraints in imaging and analysis. This has been a major factor limiting insight into the complex variability of cellular environments. Modern EM can produce gigavoxel image volumes containing large numbers of cells, but accurate manual segmentation of image features is slow and limits the creation of cell models. Segmentation algorithms based on convolutional neural networks can process large volumes quickly, but achieving EM task accuracy goals often challenges current techniques. Here, we define dense cellular segmentation as a multiclass semantic segmentation task for modeling cells and large numbers of their organelles, and give an example in human blood platelets. We present an algorithm using novel hybrid 2D–3D segmentation networks to produce dense cellular segmentations with accuracy levels that outperform baseline methods and approach those of human annotators. To our knowledge, this work represents the first published approach to automating the creation of cell models with this level of structural detail. |
url |
https://doi.org/10.1038/s41598-021-81590-0 |
work_keys_str_mv |
AT matthewdguay densecellularsegmentationforemusing2d3dneuralnetworkensembles AT zeyadasemam densecellularsegmentationforemusing2d3dneuralnetworkensembles AT adambanderson densecellularsegmentationforemusing2d3dneuralnetworkensembles AT mariaaaronova densecellularsegmentationforemusing2d3dneuralnetworkensembles AT irinadpokrovskaya densecellularsegmentationforemusing2d3dneuralnetworkensembles AT brianstorrie densecellularsegmentationforemusing2d3dneuralnetworkensembles AT richarddleapman densecellularsegmentationforemusing2d3dneuralnetworkensembles |
_version_ |
1724316500245872640 |