Cargando…

Evaluating Very Deep Convolutional Neural Networks for Nucleus Segmentation from Brightfield Cell Microscopy Images

Advances in microscopy have increased output data volumes, and powerful image analysis methods are required to match. In particular, finding and characterizing nuclei from microscopy images, a core cytometry task, remains difficult to automate. While deep learning models have given encouraging resul...

Descripción completa

Detalles Bibliográficos
Autores principales: Ali, Mohammed A. S., Misko, Oleg, Salumaa, Sten-Oliver, Papkov, Mikhail, Palo, Kaupo, Fishman, Dmytro, Parts, Leopold
Formato: Online Artículo Texto
Lenguaje:English
Publicado: SAGE Publications 2021
Materias:
Acceso en línea:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8458686/
https://www.ncbi.nlm.nih.gov/pubmed/34167359
http://dx.doi.org/10.1177/24725552211023214
_version_ 1784571348840873984
author Ali, Mohammed A. S.
Misko, Oleg
Salumaa, Sten-Oliver
Papkov, Mikhail
Palo, Kaupo
Fishman, Dmytro
Parts, Leopold
author_facet Ali, Mohammed A. S.
Misko, Oleg
Salumaa, Sten-Oliver
Papkov, Mikhail
Palo, Kaupo
Fishman, Dmytro
Parts, Leopold
author_sort Ali, Mohammed A. S.
collection PubMed
description Advances in microscopy have increased output data volumes, and powerful image analysis methods are required to match. In particular, finding and characterizing nuclei from microscopy images, a core cytometry task, remains difficult to automate. While deep learning models have given encouraging results on this problem, the most powerful approaches have not yet been tested for attacking it. Here, we review and evaluate state-of-the-art very deep convolutional neural network architectures and training strategies for segmenting nuclei from brightfield cell images. We tested U-Net as a baseline model; considered U-Net++, Tiramisu, and DeepLabv3+ as latest instances of advanced families of segmentation models; and propose PPU-Net, a novel light-weight alternative. The deeper architectures outperformed standard U-Net and results from previous studies on the challenging brightfield images, with balanced pixel-wise accuracies of up to 86%. PPU-Net achieved this performance with 20-fold fewer parameters than the comparably accurate methods. All models perform better on larger nuclei and in sparser images. We further confirmed that in the absence of plentiful training data, augmentation and pretraining on other data improve performance. In particular, using only 16 images with data augmentation is enough to achieve a pixel-wise F1 score that is within 5% of the one achieved with a full data set for all models. The remaining segmentation errors are mainly due to missed nuclei in dense regions, overlapping cells, and imaging artifacts, indicating the major outstanding challenges.
format Online
Article
Text
id pubmed-8458686
institution National Center for Biotechnology Information
language English
publishDate 2021
publisher SAGE Publications
record_format MEDLINE/PubMed
spelling pubmed-84586862021-09-24 Evaluating Very Deep Convolutional Neural Networks for Nucleus Segmentation from Brightfield Cell Microscopy Images Ali, Mohammed A. S. Misko, Oleg Salumaa, Sten-Oliver Papkov, Mikhail Palo, Kaupo Fishman, Dmytro Parts, Leopold SLAS Discov Article Advances in microscopy have increased output data volumes, and powerful image analysis methods are required to match. In particular, finding and characterizing nuclei from microscopy images, a core cytometry task, remains difficult to automate. While deep learning models have given encouraging results on this problem, the most powerful approaches have not yet been tested for attacking it. Here, we review and evaluate state-of-the-art very deep convolutional neural network architectures and training strategies for segmenting nuclei from brightfield cell images. We tested U-Net as a baseline model; considered U-Net++, Tiramisu, and DeepLabv3+ as latest instances of advanced families of segmentation models; and propose PPU-Net, a novel light-weight alternative. The deeper architectures outperformed standard U-Net and results from previous studies on the challenging brightfield images, with balanced pixel-wise accuracies of up to 86%. PPU-Net achieved this performance with 20-fold fewer parameters than the comparably accurate methods. All models perform better on larger nuclei and in sparser images. We further confirmed that in the absence of plentiful training data, augmentation and pretraining on other data improve performance. In particular, using only 16 images with data augmentation is enough to achieve a pixel-wise F1 score that is within 5% of the one achieved with a full data set for all models. The remaining segmentation errors are mainly due to missed nuclei in dense regions, overlapping cells, and imaging artifacts, indicating the major outstanding challenges. SAGE Publications 2021-06-24 2021-10 /pmc/articles/PMC8458686/ /pubmed/34167359 http://dx.doi.org/10.1177/24725552211023214 Text en © The Author(s) 2021 https://creativecommons.org/licenses/by/4.0/This article is distributed under the terms of the Creative Commons Attribution 4.0 License (https://creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without further permission provided the original work is attributed as specified on the SAGE and Open Access page (https://us.sagepub.com/en-us/nam/open-access-at-sage).
spellingShingle Article
Ali, Mohammed A. S.
Misko, Oleg
Salumaa, Sten-Oliver
Papkov, Mikhail
Palo, Kaupo
Fishman, Dmytro
Parts, Leopold
Evaluating Very Deep Convolutional Neural Networks for Nucleus Segmentation from Brightfield Cell Microscopy Images
title Evaluating Very Deep Convolutional Neural Networks for Nucleus Segmentation from Brightfield Cell Microscopy Images
title_full Evaluating Very Deep Convolutional Neural Networks for Nucleus Segmentation from Brightfield Cell Microscopy Images
title_fullStr Evaluating Very Deep Convolutional Neural Networks for Nucleus Segmentation from Brightfield Cell Microscopy Images
title_full_unstemmed Evaluating Very Deep Convolutional Neural Networks for Nucleus Segmentation from Brightfield Cell Microscopy Images
title_short Evaluating Very Deep Convolutional Neural Networks for Nucleus Segmentation from Brightfield Cell Microscopy Images
title_sort evaluating very deep convolutional neural networks for nucleus segmentation from brightfield cell microscopy images
topic Article
url https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8458686/
https://www.ncbi.nlm.nih.gov/pubmed/34167359
http://dx.doi.org/10.1177/24725552211023214
work_keys_str_mv AT alimohammedas evaluatingverydeepconvolutionalneuralnetworksfornucleussegmentationfrombrightfieldcellmicroscopyimages
AT miskooleg evaluatingverydeepconvolutionalneuralnetworksfornucleussegmentationfrombrightfieldcellmicroscopyimages
AT salumaastenoliver evaluatingverydeepconvolutionalneuralnetworksfornucleussegmentationfrombrightfieldcellmicroscopyimages
AT papkovmikhail evaluatingverydeepconvolutionalneuralnetworksfornucleussegmentationfrombrightfieldcellmicroscopyimages
AT palokaupo evaluatingverydeepconvolutionalneuralnetworksfornucleussegmentationfrombrightfieldcellmicroscopyimages
AT fishmandmytro evaluatingverydeepconvolutionalneuralnetworksfornucleussegmentationfrombrightfieldcellmicroscopyimages
AT partsleopold evaluatingverydeepconvolutionalneuralnetworksfornucleussegmentationfrombrightfieldcellmicroscopyimages