Unet++

A nested u-net architecture for medical image segmentation

Zongwei Zhou, Md Mahfuzur Rahman Siddiquee, Nima Tajbakhsh, Jianming Liang

Research output: Chapter in Book/Report/Conference proceedingConference contribution

10 Citations (Scopus)

Abstract

In this paper, we present UNet++, a new, more powerful architecture for medical image segmentation. Our architecture is essentially a deeply-supervised encoder-decoder network where the encoder and decoder sub-networks are connected through a series of nested, dense skip pathways. The re-designed skip pathways aim at reducing the semantic gap between the feature maps of the encoder and decoder sub-networks. We argue that the optimizer would deal with an easier learning task when the feature maps from the decoder and encoder networks are semantically similar. We have evaluated UNet++ in comparison with U-Net and wide U-Net architectures across multiple medical image segmentation tasks: nodule segmentation in the low-dose CT scans of chest, nuclei segmentation in the microscopy images, liver segmentation in abdominal CT scans, and polyp segmentation in colonoscopy videos. Our experiments demonstrate that UNet++ with deep supervision achieves an average IoU gain of 3.9 and 3.4 points over U-Net and wide U-Net, respectively.

Original languageEnglish (US)
Title of host publicationDeep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - 4th International Workshop, DLMIA 2018 and 8th International Workshop, ML-CDS 2018 Held in Conjunction with MICCAI 2018
EditorsLena Maier-Hein, Tanveer Syeda-Mahmood, Zeike Taylor, Zhi Lu, Danail Stoyanov, Anant Madabhushi, João Manuel R.S. Tavares, Jacinto C. Nascimento, Mehdi Moradi, Anne Martel, Joao Paulo Papa, Sailesh Conjeti, Vasileios Belagiannis, Hayit Greenspan, Gustavo Carneiro, Andrew Bradley
PublisherSpringer Verlag
Pages3-11
Number of pages9
ISBN (Print)9783030008888
DOIs
StatePublished - Jan 1 2018
Event4th International Workshop on Deep Learning in Medical Image Analysis, DLMIA 2018 and 8th International Workshop on Multimodal Learning for Clinical Decision Support, ML-CDS 2018 Held in Conjunction with MICCAI 2018 - Granada, Spain
Duration: Sep 20 2018Sep 20 2018

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11045 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other4th International Workshop on Deep Learning in Medical Image Analysis, DLMIA 2018 and 8th International Workshop on Multimodal Learning for Clinical Decision Support, ML-CDS 2018 Held in Conjunction with MICCAI 2018
CountrySpain
CityGranada
Period9/20/189/20/18

Fingerprint

Computerized tomography
Medical Image
Encoder
Image segmentation
Image Segmentation
Segmentation
Liver
Pathway
Microscopic examination
Semantics
Nodule
Microscopy
Nucleus
Dose
Experiments
Series
Architecture
Demonstrate
Experiment

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Zhou, Z., Rahman Siddiquee, M. M., Tajbakhsh, N., & Liang, J. (2018). Unet++: A nested u-net architecture for medical image segmentation. In L. Maier-Hein, T. Syeda-Mahmood, Z. Taylor, Z. Lu, D. Stoyanov, A. Madabhushi, J. M. R. S. Tavares, J. C. Nascimento, M. Moradi, A. Martel, J. P. Papa, S. Conjeti, V. Belagiannis, H. Greenspan, G. Carneiro, ... A. Bradley (Eds.), Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - 4th International Workshop, DLMIA 2018 and 8th International Workshop, ML-CDS 2018 Held in Conjunction with MICCAI 2018 (pp. 3-11). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11045 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-030-00889-5_1

Unet++ : A nested u-net architecture for medical image segmentation. / Zhou, Zongwei; Rahman Siddiquee, Md Mahfuzur; Tajbakhsh, Nima; Liang, Jianming.

Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - 4th International Workshop, DLMIA 2018 and 8th International Workshop, ML-CDS 2018 Held in Conjunction with MICCAI 2018. ed. / Lena Maier-Hein; Tanveer Syeda-Mahmood; Zeike Taylor; Zhi Lu; Danail Stoyanov; Anant Madabhushi; João Manuel R.S. Tavares; Jacinto C. Nascimento; Mehdi Moradi; Anne Martel; Joao Paulo Papa; Sailesh Conjeti; Vasileios Belagiannis; Hayit Greenspan; Gustavo Carneiro; Andrew Bradley. Springer Verlag, 2018. p. 3-11 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11045 LNCS).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Zhou, Z, Rahman Siddiquee, MM, Tajbakhsh, N & Liang, J 2018, Unet++: A nested u-net architecture for medical image segmentation. in L Maier-Hein, T Syeda-Mahmood, Z Taylor, Z Lu, D Stoyanov, A Madabhushi, JMRS Tavares, JC Nascimento, M Moradi, A Martel, JP Papa, S Conjeti, V Belagiannis, H Greenspan, G Carneiro & A Bradley (eds), Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - 4th International Workshop, DLMIA 2018 and 8th International Workshop, ML-CDS 2018 Held in Conjunction with MICCAI 2018. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11045 LNCS, Springer Verlag, pp. 3-11, 4th International Workshop on Deep Learning in Medical Image Analysis, DLMIA 2018 and 8th International Workshop on Multimodal Learning for Clinical Decision Support, ML-CDS 2018 Held in Conjunction with MICCAI 2018, Granada, Spain, 9/20/18. https://doi.org/10.1007/978-3-030-00889-5_1
Zhou Z, Rahman Siddiquee MM, Tajbakhsh N, Liang J. Unet++: A nested u-net architecture for medical image segmentation. In Maier-Hein L, Syeda-Mahmood T, Taylor Z, Lu Z, Stoyanov D, Madabhushi A, Tavares JMRS, Nascimento JC, Moradi M, Martel A, Papa JP, Conjeti S, Belagiannis V, Greenspan H, Carneiro G, Bradley A, editors, Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - 4th International Workshop, DLMIA 2018 and 8th International Workshop, ML-CDS 2018 Held in Conjunction with MICCAI 2018. Springer Verlag. 2018. p. 3-11. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-030-00889-5_1
Zhou, Zongwei ; Rahman Siddiquee, Md Mahfuzur ; Tajbakhsh, Nima ; Liang, Jianming. / Unet++ : A nested u-net architecture for medical image segmentation. Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - 4th International Workshop, DLMIA 2018 and 8th International Workshop, ML-CDS 2018 Held in Conjunction with MICCAI 2018. editor / Lena Maier-Hein ; Tanveer Syeda-Mahmood ; Zeike Taylor ; Zhi Lu ; Danail Stoyanov ; Anant Madabhushi ; João Manuel R.S. Tavares ; Jacinto C. Nascimento ; Mehdi Moradi ; Anne Martel ; Joao Paulo Papa ; Sailesh Conjeti ; Vasileios Belagiannis ; Hayit Greenspan ; Gustavo Carneiro ; Andrew Bradley. Springer Verlag, 2018. pp. 3-11 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{1e3fbce37fb641d8b7646aae4260dce2,
title = "Unet++: A nested u-net architecture for medical image segmentation",
abstract = "In this paper, we present UNet++, a new, more powerful architecture for medical image segmentation. Our architecture is essentially a deeply-supervised encoder-decoder network where the encoder and decoder sub-networks are connected through a series of nested, dense skip pathways. The re-designed skip pathways aim at reducing the semantic gap between the feature maps of the encoder and decoder sub-networks. We argue that the optimizer would deal with an easier learning task when the feature maps from the decoder and encoder networks are semantically similar. We have evaluated UNet++ in comparison with U-Net and wide U-Net architectures across multiple medical image segmentation tasks: nodule segmentation in the low-dose CT scans of chest, nuclei segmentation in the microscopy images, liver segmentation in abdominal CT scans, and polyp segmentation in colonoscopy videos. Our experiments demonstrate that UNet++ with deep supervision achieves an average IoU gain of 3.9 and 3.4 points over U-Net and wide U-Net, respectively.",
author = "Zongwei Zhou and {Rahman Siddiquee}, {Md Mahfuzur} and Nima Tajbakhsh and Jianming Liang",
year = "2018",
month = "1",
day = "1",
doi = "10.1007/978-3-030-00889-5_1",
language = "English (US)",
isbn = "9783030008888",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "3--11",
editor = "Lena Maier-Hein and Tanveer Syeda-Mahmood and Zeike Taylor and Zhi Lu and Danail Stoyanov and Anant Madabhushi and Tavares, {Jo{\~a}o Manuel R.S.} and Nascimento, {Jacinto C.} and Mehdi Moradi and Anne Martel and Papa, {Joao Paulo} and Sailesh Conjeti and Vasileios Belagiannis and Hayit Greenspan and Gustavo Carneiro and Andrew Bradley",
booktitle = "Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - 4th International Workshop, DLMIA 2018 and 8th International Workshop, ML-CDS 2018 Held in Conjunction with MICCAI 2018",

}

TY - GEN

T1 - Unet++

T2 - A nested u-net architecture for medical image segmentation

AU - Zhou, Zongwei

AU - Rahman Siddiquee, Md Mahfuzur

AU - Tajbakhsh, Nima

AU - Liang, Jianming

PY - 2018/1/1

Y1 - 2018/1/1

N2 - In this paper, we present UNet++, a new, more powerful architecture for medical image segmentation. Our architecture is essentially a deeply-supervised encoder-decoder network where the encoder and decoder sub-networks are connected through a series of nested, dense skip pathways. The re-designed skip pathways aim at reducing the semantic gap between the feature maps of the encoder and decoder sub-networks. We argue that the optimizer would deal with an easier learning task when the feature maps from the decoder and encoder networks are semantically similar. We have evaluated UNet++ in comparison with U-Net and wide U-Net architectures across multiple medical image segmentation tasks: nodule segmentation in the low-dose CT scans of chest, nuclei segmentation in the microscopy images, liver segmentation in abdominal CT scans, and polyp segmentation in colonoscopy videos. Our experiments demonstrate that UNet++ with deep supervision achieves an average IoU gain of 3.9 and 3.4 points over U-Net and wide U-Net, respectively.

AB - In this paper, we present UNet++, a new, more powerful architecture for medical image segmentation. Our architecture is essentially a deeply-supervised encoder-decoder network where the encoder and decoder sub-networks are connected through a series of nested, dense skip pathways. The re-designed skip pathways aim at reducing the semantic gap between the feature maps of the encoder and decoder sub-networks. We argue that the optimizer would deal with an easier learning task when the feature maps from the decoder and encoder networks are semantically similar. We have evaluated UNet++ in comparison with U-Net and wide U-Net architectures across multiple medical image segmentation tasks: nodule segmentation in the low-dose CT scans of chest, nuclei segmentation in the microscopy images, liver segmentation in abdominal CT scans, and polyp segmentation in colonoscopy videos. Our experiments demonstrate that UNet++ with deep supervision achieves an average IoU gain of 3.9 and 3.4 points over U-Net and wide U-Net, respectively.

UR - http://www.scopus.com/inward/record.url?scp=85054551189&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85054551189&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-00889-5_1

DO - 10.1007/978-3-030-00889-5_1

M3 - Conference contribution

SN - 9783030008888

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 3

EP - 11

BT - Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - 4th International Workshop, DLMIA 2018 and 8th International Workshop, ML-CDS 2018 Held in Conjunction with MICCAI 2018

A2 - Maier-Hein, Lena

A2 - Syeda-Mahmood, Tanveer

A2 - Taylor, Zeike

A2 - Lu, Zhi

A2 - Stoyanov, Danail

A2 - Madabhushi, Anant

A2 - Tavares, João Manuel R.S.

A2 - Nascimento, Jacinto C.

A2 - Moradi, Mehdi

A2 - Martel, Anne

A2 - Papa, Joao Paulo

A2 - Conjeti, Sailesh

A2 - Belagiannis, Vasileios

A2 - Greenspan, Hayit

A2 - Carneiro, Gustavo

A2 - Bradley, Andrew

PB - Springer Verlag

ER -