Multi-scale 3D UNet: Multi-scale 3D UNet for Brain Tumor Segmentation

Parvez Ahmad; Saqib Qamar; Linlin Shen; Syed Qasim Afser Rizvi; Aamir Ali; Girija Chetty

doi:10.1007/978-3-031-09002-8_3

Multi-scale 3D UNet: Multi-scale 3D UNet for Brain Tumor Segmentation

Parvez Ahmad
, Saqib Qamar
, Linlin Shen
, Syed Qasim Afser Rizvi
, Aamir Ali
, Girija Chetty

Information Systems

Research output: A Conference proceeding or a Chapter in Book › Conference contribution › peer-review

11 Citations (Scopus)

Abstract

A deep convolutional neural network (CNN) achieves remarkable performance for medical image analysis. UNet is the primary source in the performance of 3D CNN architectures for medical imaging tasks, including brain tumor segmentation. The skip connection in the UNet architecture concatenates multi-scale features from image data. The multi-scaled features play an essential role in brain tumor segmentation. Researchers presented numerous multi-scale strategies that have been excellent for the segmentation task. This paper proposes a multi-scale strategy that can further improve the final segmentation accuracy. We propose three multi-scale strategies in MS UNet. Firstly, we utilize densely connected blocks in the encoder and decoder for multi-scale features. Next, the proposed residual-inception blocks extract local and global information by merging features of different kernel sizes. Lastly, we utilize the idea of deep supervision for multiple depths at the decoder. We validate the MS UNet on the BraTS 2021 validation dataset. The dice (DSC) scores of the whole tumor (WT), tumor core (TC), and enhancing tumor (ET) are 91.938 %, 86.268 %, and 82.409 %, respectively.

Original language	English
Title of host publication	International MICCAI Brainlesion Workshop
Subtitle of host publication	Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries - 7th International Workshop, BrainLes 2021, Held in Conjunction with MICCAI 2021
Editors	Alessandro Crimi, Spyridon Bakas
Publisher	Springer
Pages	30-41
Number of pages	12
ISBN (Electronic)	9783031090028
ISBN (Print)	9783031090011
DOIs	https://doi.org/10.1007/978-3-031-09002-8_3
Publication status	Published - 15 Jul 2022
Event	7th International Brain Lesion Workshop, BrainLes 2021, held in conjunction with the Medical Image Computing and Computer Assisted Intervention, MICCAI 2021 - Virtual, Online Duration: 27 Sept 2021 → 27 Sept 2021

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	12963 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	7th International Brain Lesion Workshop, BrainLes 2021, held in conjunction with the Medical Image Computing and Computer Assisted Intervention, MICCAI 2021
City	Virtual, Online
Period	27/09/21 → 27/09/21

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 3 Good Health and Well-being

Access to Document

10.1007/978-3-031-09002-8_3

Cite this

Ahmad, P., Qamar, S., Shen, L., Rizvi, S. Q. A., Ali, A., & Chetty, G. (2022). Multi-scale 3D UNet: Multi-scale 3D UNet for Brain Tumor Segmentation. In A. Crimi, & S. Bakas (Eds.), International MICCAI Brainlesion Workshop: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries - 7th International Workshop, BrainLes 2021, Held in Conjunction with MICCAI 2021 (pp. 30-41). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12963 LNCS). Springer. https://doi.org/10.1007/978-3-031-09002-8_3

Ahmad, Parvez ; Qamar, Saqib ; Shen, Linlin et al. / Multi-scale 3D UNet : Multi-scale 3D UNet for Brain Tumor Segmentation. International MICCAI Brainlesion Workshop: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries - 7th International Workshop, BrainLes 2021, Held in Conjunction with MICCAI 2021. editor / Alessandro Crimi ; Spyridon Bakas. Springer, 2022. pp. 30-41 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{dd5a0dd0007b495ab71ddfdfc497903e,

title = "Multi-scale 3D UNet: Multi-scale 3D UNet for Brain Tumor Segmentation",

abstract = "A deep convolutional neural network (CNN) achieves remarkable performance for medical image analysis. UNet is the primary source in the performance of 3D CNN architectures for medical imaging tasks, including brain tumor segmentation. The skip connection in the UNet architecture concatenates multi-scale features from image data. The multi-scaled features play an essential role in brain tumor segmentation. Researchers presented numerous multi-scale strategies that have been excellent for the segmentation task. This paper proposes a multi-scale strategy that can further improve the final segmentation accuracy. We propose three multi-scale strategies in MS UNet. Firstly, we utilize densely connected blocks in the encoder and decoder for multi-scale features. Next, the proposed residual-inception blocks extract local and global information by merging features of different kernel sizes. Lastly, we utilize the idea of deep supervision for multiple depths at the decoder. We validate the MS UNet on the BraTS 2021 validation dataset. The dice (DSC) scores of the whole tumor (WT), tumor core (TC), and enhancing tumor (ET) are 91.938 \%, 86.268 \%, and 82.409 \%, respectively.",

keywords = "Brain tumor segmentation, CNN, Contextual information, Dense connections, Residual inception blocks, UNet",

author = "Parvez Ahmad and Saqib Qamar and Linlin Shen and Rizvi, \{Syed Qasim Afser\} and Aamir Ali and Girija Chetty",

note = "Funding Information: Acknowledgment. This work is supported by the National Natural Science Foundation of China under Grant No. 91959108. Publisher Copyright: {\textcopyright} 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.; 7th International Brain Lesion Workshop, BrainLes 2021, held in conjunction with the Medical Image Computing and Computer Assisted Intervention, MICCAI 2021 ; Conference date: 27-09-2021 Through 27-09-2021",

year = "2022",

month = jul,

day = "15",

doi = "10.1007/978-3-031-09002-8\_3",

language = "English",

isbn = "9783031090011",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer",

pages = "30--41",

editor = "Alessandro Crimi and Spyridon Bakas",

booktitle = "International MICCAI Brainlesion Workshop",

address = "Netherlands",

}

Ahmad, P, Qamar, S, Shen, L, Rizvi, SQA, Ali, A & Chetty, G 2022, Multi-scale 3D UNet: Multi-scale 3D UNet for Brain Tumor Segmentation. in A Crimi & S Bakas (eds), International MICCAI Brainlesion Workshop: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries - 7th International Workshop, BrainLes 2021, Held in Conjunction with MICCAI 2021. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 12963 LNCS, Springer, pp. 30-41, 7th International Brain Lesion Workshop, BrainLes 2021, held in conjunction with the Medical Image Computing and Computer Assisted Intervention, MICCAI 2021, Virtual, Online, 27/09/21. https://doi.org/10.1007/978-3-031-09002-8_3

Multi-scale 3D UNet: Multi-scale 3D UNet for Brain Tumor Segmentation. / Ahmad, Parvez; Qamar, Saqib; Shen, Linlin et al.
International MICCAI Brainlesion Workshop: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries - 7th International Workshop, BrainLes 2021, Held in Conjunction with MICCAI 2021. ed. / Alessandro Crimi; Spyridon Bakas. Springer, 2022. p. 30-41 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12963 LNCS).

Research output: A Conference proceeding or a Chapter in Book › Conference contribution › peer-review

TY - GEN

T1 - Multi-scale 3D UNet

T2 - 7th International Brain Lesion Workshop, BrainLes 2021, held in conjunction with the Medical Image Computing and Computer Assisted Intervention, MICCAI 2021

AU - Ahmad, Parvez

AU - Qamar, Saqib

AU - Shen, Linlin

AU - Rizvi, Syed Qasim Afser

AU - Ali, Aamir

AU - Chetty, Girija

N1 - Funding Information: Acknowledgment. This work is supported by the National Natural Science Foundation of China under Grant No. 91959108. Publisher Copyright: © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

PY - 2022/7/15

Y1 - 2022/7/15

N2 - A deep convolutional neural network (CNN) achieves remarkable performance for medical image analysis. UNet is the primary source in the performance of 3D CNN architectures for medical imaging tasks, including brain tumor segmentation. The skip connection in the UNet architecture concatenates multi-scale features from image data. The multi-scaled features play an essential role in brain tumor segmentation. Researchers presented numerous multi-scale strategies that have been excellent for the segmentation task. This paper proposes a multi-scale strategy that can further improve the final segmentation accuracy. We propose three multi-scale strategies in MS UNet. Firstly, we utilize densely connected blocks in the encoder and decoder for multi-scale features. Next, the proposed residual-inception blocks extract local and global information by merging features of different kernel sizes. Lastly, we utilize the idea of deep supervision for multiple depths at the decoder. We validate the MS UNet on the BraTS 2021 validation dataset. The dice (DSC) scores of the whole tumor (WT), tumor core (TC), and enhancing tumor (ET) are 91.938 %, 86.268 %, and 82.409 %, respectively.

AB - A deep convolutional neural network (CNN) achieves remarkable performance for medical image analysis. UNet is the primary source in the performance of 3D CNN architectures for medical imaging tasks, including brain tumor segmentation. The skip connection in the UNet architecture concatenates multi-scale features from image data. The multi-scaled features play an essential role in brain tumor segmentation. Researchers presented numerous multi-scale strategies that have been excellent for the segmentation task. This paper proposes a multi-scale strategy that can further improve the final segmentation accuracy. We propose three multi-scale strategies in MS UNet. Firstly, we utilize densely connected blocks in the encoder and decoder for multi-scale features. Next, the proposed residual-inception blocks extract local and global information by merging features of different kernel sizes. Lastly, we utilize the idea of deep supervision for multiple depths at the decoder. We validate the MS UNet on the BraTS 2021 validation dataset. The dice (DSC) scores of the whole tumor (WT), tumor core (TC), and enhancing tumor (ET) are 91.938 %, 86.268 %, and 82.409 %, respectively.

KW - Brain tumor segmentation

KW - CNN

KW - Contextual information

KW - Dense connections

KW - Residual inception blocks

KW - UNet

UR - https://www.scopus.com/pages/publications/85135188304

UR - https://sano.science/news/miccai-brainles-workshop-co-organized-by-sanos-computer-vision-team-leader-dr-alessandro-crimi/

U2 - 10.1007/978-3-031-09002-8_3

DO - 10.1007/978-3-031-09002-8_3

M3 - Conference contribution

AN - SCOPUS:85135188304

SN - 9783031090011

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

SP - 30

EP - 41

BT - International MICCAI Brainlesion Workshop

A2 - Crimi, Alessandro

A2 - Bakas, Spyridon

PB - Springer

Y2 - 27 September 2021 through 27 September 2021

ER -

Ahmad P, Qamar S, Shen L, Rizvi SQA, Ali A, Chetty G. Multi-scale 3D UNet: Multi-scale 3D UNet for Brain Tumor Segmentation. In Crimi A, Bakas S, editors, International MICCAI Brainlesion Workshop: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries - 7th International Workshop, BrainLes 2021, Held in Conjunction with MICCAI 2021. Springer. 2022. p. 30-41. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-09002-8_3