Meta-learning for Medical Image Segmentation Uncertainty Quantification

Sabri Can Cetindag; Mert Yergin; Deniz Alis; Ilkay Oksuz

doi:10.1007/978-3-031-09002-8_51

Meta-learning for Medical Image Segmentation Uncertainty Quantification

Sabri Can Cetindag^*, Mert Yergin, Deniz Alis, Ilkay Oksuz

^*Corresponding author for this work

Department of Computer Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

1 Citation (Scopus)

Abstract

Inter-rater and intra-rater variability is a major challenge in medical image segmentation. Inconsistencies of manual segmentations between different experts can challenge development of deterministic automated medical image analysis tools. QUBIQ 2021 is a challenge to enable the successful development of automated machine learning tools, when there are inconsistencies between the labels of different annotators. In this paper, we propose to use meta-learning for quantifying uncertainty in biomedical image quantification. We first train a segmentation network for each expert separately with extensive data augmentation using the nnUnet framework. Then, a meta learner model based on a conventional U-net architecture is trained using the average of all annotators as ground truth, and output of all models that have been trained for each radiologist as input. We compared our results of meta-learning with ensemble methods for various image segmentation tasks and illustrate improved performance.

Original language	English
Title of host publication	Brainlesion
Subtitle of host publication	Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries - 7th International Workshop, BrainLes 2021, Held in Conjunction with MICCAI 2021, Revised Selected Papers
Editors	Alessandro Crimi, Spyridon Bakas
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	578-584
Number of pages	7
ISBN (Print)	9783031090011
DOIs	https://doi.org/10.1007/978-3-031-09002-8_51
Publication status	Published - 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

Bibliographical note

Publisher Copyright:
© 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Keywords

Medical image segmentation
Meta-learning
Uncertainty quantification

Access to Document

10.1007/978-3-031-09002-8_51

Cite this

Cetindag, S. C., Yergin, M., Alis, D., & Oksuz, I. (2022). Meta-learning for Medical Image Segmentation Uncertainty Quantification. In A. Crimi, & S. Bakas (Eds.), Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries - 7th International Workshop, BrainLes 2021, Held in Conjunction with MICCAI 2021, Revised Selected Papers (pp. 578-584). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12963 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-09002-8_51

Cetindag, Sabri Can ; Yergin, Mert ; Alis, Deniz et al. / Meta-learning for Medical Image Segmentation Uncertainty Quantification. Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries - 7th International Workshop, BrainLes 2021, Held in Conjunction with MICCAI 2021, Revised Selected Papers. editor / Alessandro Crimi ; Spyridon Bakas. Springer Science and Business Media Deutschland GmbH, 2022. pp. 578-584 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{e2764f57d0e7473aac69969cf76bf971,

title = "Meta-learning for Medical Image Segmentation Uncertainty Quantification",

abstract = "Inter-rater and intra-rater variability is a major challenge in medical image segmentation. Inconsistencies of manual segmentations between different experts can challenge development of deterministic automated medical image analysis tools. QUBIQ 2021 is a challenge to enable the successful development of automated machine learning tools, when there are inconsistencies between the labels of different annotators. In this paper, we propose to use meta-learning for quantifying uncertainty in biomedical image quantification. We first train a segmentation network for each expert separately with extensive data augmentation using the nnUnet framework. Then, a meta learner model based on a conventional U-net architecture is trained using the average of all annotators as ground truth, and output of all models that have been trained for each radiologist as input. We compared our results of meta-learning with ensemble methods for various image segmentation tasks and illustrate improved performance.",

keywords = "Medical image segmentation, Meta-learning, Uncertainty quantification",

author = "Cetindag, {Sabri Can} and Mert Yergin and Deniz Alis and Ilkay Oksuz",

note = "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",

doi = "10.1007/978-3-031-09002-8_51",

language = "English",

isbn = "9783031090011",

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

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "578--584",

editor = "Alessandro Crimi and Spyridon Bakas",

booktitle = "Brainlesion",

address = "Germany",

}

Cetindag, SC, Yergin, M, Alis, D & Oksuz, I 2022, Meta-learning for Medical Image Segmentation Uncertainty Quantification. in A Crimi & S Bakas (eds), Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries - 7th International Workshop, BrainLes 2021, Held in Conjunction with MICCAI 2021, Revised Selected Papers. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 12963 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 578-584, 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_51

Meta-learning for Medical Image Segmentation Uncertainty Quantification. / Cetindag, Sabri Can; Yergin, Mert; Alis, Deniz et al.
Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries - 7th International Workshop, BrainLes 2021, Held in Conjunction with MICCAI 2021, Revised Selected Papers. ed. / Alessandro Crimi; Spyridon Bakas. Springer Science and Business Media Deutschland GmbH, 2022. p. 578-584 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12963 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Meta-learning for Medical Image Segmentation Uncertainty Quantification

AU - Cetindag, Sabri Can

AU - Yergin, Mert

AU - Alis, Deniz

AU - Oksuz, Ilkay

PY - 2022

Y1 - 2022

N2 - Inter-rater and intra-rater variability is a major challenge in medical image segmentation. Inconsistencies of manual segmentations between different experts can challenge development of deterministic automated medical image analysis tools. QUBIQ 2021 is a challenge to enable the successful development of automated machine learning tools, when there are inconsistencies between the labels of different annotators. In this paper, we propose to use meta-learning for quantifying uncertainty in biomedical image quantification. We first train a segmentation network for each expert separately with extensive data augmentation using the nnUnet framework. Then, a meta learner model based on a conventional U-net architecture is trained using the average of all annotators as ground truth, and output of all models that have been trained for each radiologist as input. We compared our results of meta-learning with ensemble methods for various image segmentation tasks and illustrate improved performance.

AB - Inter-rater and intra-rater variability is a major challenge in medical image segmentation. Inconsistencies of manual segmentations between different experts can challenge development of deterministic automated medical image analysis tools. QUBIQ 2021 is a challenge to enable the successful development of automated machine learning tools, when there are inconsistencies between the labels of different annotators. In this paper, we propose to use meta-learning for quantifying uncertainty in biomedical image quantification. We first train a segmentation network for each expert separately with extensive data augmentation using the nnUnet framework. Then, a meta learner model based on a conventional U-net architecture is trained using the average of all annotators as ground truth, and output of all models that have been trained for each radiologist as input. We compared our results of meta-learning with ensemble methods for various image segmentation tasks and illustrate improved performance.

KW - Medical image segmentation

KW - Meta-learning

KW - Uncertainty quantification

UR - http://www.scopus.com/inward/record.url?scp=85135159914&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-09002-8_51

DO - 10.1007/978-3-031-09002-8_51

M3 - Conference contribution

AN - SCOPUS:85135159914

SN - 9783031090011

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

SP - 578

EP - 584

BT - Brainlesion

A2 - Crimi, Alessandro

A2 - Bakas, Spyridon

PB - Springer Science and Business Media Deutschland GmbH

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

Y2 - 27 September 2021 through 27 September 2021

ER -

Cetindag SC, Yergin M, Alis D, Oksuz I. Meta-learning for Medical Image Segmentation Uncertainty Quantification. In Crimi A, Bakas S, editors, Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries - 7th International Workshop, BrainLes 2021, Held in Conjunction with MICCAI 2021, Revised Selected Papers. Springer Science and Business Media Deutschland GmbH. 2022. p. 578-584. (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_51