Harmonized segmentation of neonatal brain MRI

Irina Grigorescu; Lucy Vanes; Alena Uus; Dafnis Batalle; Lucilio Cordero-Grande; Chiara Nosarti; David Edwards; Joseph Hajnal; Marc Modat; Maria Deprez

doi:10.3389/fnins.2021.662005

Harmonized segmentation of neonatal brain MRI

Irina Grigorescu, Lucy Vanes, Alena Uus, Dafnis Batalle, Lucilio Cordero-Grande, Chiara Nosarti, David Edwards, Joseph Hajnal, Marc Modat, Maria Deprez

Biomedical Image Technologies, ETSI Telecomunicacion, Universidad Politecnica de Madrid and CIBER-BBN, Madrid, 28040, Spain.

Research output: Contribution to journal › Article › peer-review

10 Citations (Scopus)

75 Downloads (Pure)

Abstract

Deep learning based medical image segmentation has shown great potential in becoming a key part of the clinical analysis pipeline. However, many of these models rely on the assumption that the train and test data come from the same distribution. This means that such methods cannot guarantee high quality predictions when the source and target domains are dissimilar due to different acquisition protocols, or biases in patient cohorts. Recently, unsupervised domain adaptation techniques have shown great potential in alleviating this problem by minimizing the shift between the source and target distributions, without requiring the use of labeled data in the target domain. In this work, we aim to predict tissue segmentation maps on T2-weighted magnetic resonance imaging data of an unseen preterm-born neonatal population, which has both different acquisition parameters and population bias when compared to our training data. We achieve this by investigating two unsupervised domain adaptation techniques with the objective of finding the best solution for our problem. We compare the two methods with a baseline fully-supervised segmentation network and report our results in terms of Dice scores obtained on our source test dataset. Moreover, we analyse tissue volumes and cortical thickness measures of the harmonized data on a subset of the population matched for gestational age at birth and postmenstrual age at scan. Finally, we demonstrate the applicability of the harmonized cortical gray matter maps with an analysis comparing term and preterm-born neonates and a proof-of-principle investigation of the association between cortical thickness and a language outcome measure.

Original language	English
Article number	662005
Journal	Frontiers in Neuroscience
Volume	15
Early online date	25 May 2021
DOIs	https://doi.org/10.3389/fnins.2021.662005
Publication status	Published - 25 May 2021

Access to Document

10.3389/fnins.2021.662005Licence: CC BY

Harmonized segmentation of neonatal_GRIGORESCU_Accepted21April2021_GREEN AAM (CC BY)Final published version, 6.13 MBLicence: CC BY

Cite this

@article{5a0e60adf8ba4605bf5af4d258a439a9,

title = "Harmonized segmentation of neonatal brain MRI",

abstract = "Deep learning based medical image segmentation has shown great potential in becoming a key part of the clinical analysis pipeline. However, many of these models rely on the assumption that the train and test data come from the same distribution. This means that such methods cannot guarantee high quality predictions when the source and target domains are dissimilar due to different acquisition protocols, or biases in patient cohorts. Recently, unsupervised domain adaptation techniques have shown great potential in alleviating this problem by minimizing the shift between the source and target distributions, without requiring the use of labeled data in the target domain. In this work, we aim to predict tissue segmentation maps on T2-weighted magnetic resonance imaging data of an unseen preterm-born neonatal population, which has both different acquisition parameters and population bias when compared to our training data. We achieve this by investigating two unsupervised domain adaptation techniques with the objective of finding the best solution for our problem. We compare the two methods with a baseline fully-supervised segmentation network and report our results in terms of Dice scores obtained on our source test dataset. Moreover, we analyse tissue volumes and cortical thickness measures of the harmonized data on a subset of the population matched for gestational age at birth and postmenstrual age at scan. Finally, we demonstrate the applicability of the harmonized cortical gray matter maps with an analysis comparing term and preterm-born neonates and a proof-of-principle investigation of the association between cortical thickness and a language outcome measure.",

author = "Irina Grigorescu and Lucy Vanes and Alena Uus and Dafnis Batalle and Lucilio Cordero-Grande and Chiara Nosarti and David Edwards and Joseph Hajnal and Marc Modat and Maria Deprez",

note = "Funding Information: We thank everyone who was involved in acquisition and analysis of the datasets. We thank all participants and their families. This paper is an extension of our previous work (Grigorescu et al., 2020). Funding. This work was supported by the Academy of Medical Sciences Springboard Award [SBF004\1040], Medical Research Council (Grant nos. [MR/K006355/1] and [MR/S026460/1]), European Research Council under the European Union's Seventh Framework Programme [FP7/20072013]/ERC grant agreement no. 319456 dHCP project, the EPSRC Research Council as part of the EPSRC DTP (grant Ref: [EP/R513064/1]), the Wellcome/EPSRC Centre for Medical Engineering at King's College London [WT 203148/Z/16/Z], the NIHR Clinical Research Facility (CRF) at Guy's and St Thomas', and by the National Institute for Health Research Biomedical Research Centre based at Guy's and St Thomas' NHS Foundation Trust and King's College London. The EPrime study was funded by the National Institute for Health Research (NIHR) under its Programme Grants for Applied Research Programme (grant reference no. [RP-PG-0707-10154]). Funding Information: This work was supported by the Academy of Medical Sciences Springboard Award [SBF004\1040], Medical Research Publisher Copyright: {\textcopyright} Copyright {\textcopyright} 2021 Grigorescu, Vanes, Uus, Batalle, Cordero-Grande, Nosarti, Edwards, Hajnal, Modat and Deprez. Copyright: Copyright 2021 Elsevier B.V., All rights reserved.",

year = "2021",

month = may,

day = "25",

doi = "10.3389/fnins.2021.662005",

language = "English",

volume = "15",

journal = "Frontiers in Neuroscience",

issn = "1662-453X",

publisher = "Frontiers Media S.A.",

}

TY - JOUR

T1 - Harmonized segmentation of neonatal brain MRI

AU - Grigorescu, Irina

AU - Vanes, Lucy

AU - Uus, Alena

AU - Batalle, Dafnis

AU - Cordero-Grande, Lucilio

AU - Nosarti, Chiara

AU - Edwards, David

AU - Hajnal, Joseph

AU - Modat, Marc

AU - Deprez, Maria

N1 - Funding Information: We thank everyone who was involved in acquisition and analysis of the datasets. We thank all participants and their families. This paper is an extension of our previous work (Grigorescu et al., 2020). Funding. This work was supported by the Academy of Medical Sciences Springboard Award [SBF004\1040], Medical Research Council (Grant nos. [MR/K006355/1] and [MR/S026460/1]), European Research Council under the European Union's Seventh Framework Programme [FP7/20072013]/ERC grant agreement no. 319456 dHCP project, the EPSRC Research Council as part of the EPSRC DTP (grant Ref: [EP/R513064/1]), the Wellcome/EPSRC Centre for Medical Engineering at King's College London [WT 203148/Z/16/Z], the NIHR Clinical Research Facility (CRF) at Guy's and St Thomas', and by the National Institute for Health Research Biomedical Research Centre based at Guy's and St Thomas' NHS Foundation Trust and King's College London. The EPrime study was funded by the National Institute for Health Research (NIHR) under its Programme Grants for Applied Research Programme (grant reference no. [RP-PG-0707-10154]). Funding Information: This work was supported by the Academy of Medical Sciences Springboard Award [SBF004\1040], Medical Research Publisher Copyright: © Copyright © 2021 Grigorescu, Vanes, Uus, Batalle, Cordero-Grande, Nosarti, Edwards, Hajnal, Modat and Deprez. Copyright: Copyright 2021 Elsevier B.V., All rights reserved.

PY - 2021/5/25

Y1 - 2021/5/25

N2 - Deep learning based medical image segmentation has shown great potential in becoming a key part of the clinical analysis pipeline. However, many of these models rely on the assumption that the train and test data come from the same distribution. This means that such methods cannot guarantee high quality predictions when the source and target domains are dissimilar due to different acquisition protocols, or biases in patient cohorts. Recently, unsupervised domain adaptation techniques have shown great potential in alleviating this problem by minimizing the shift between the source and target distributions, without requiring the use of labeled data in the target domain. In this work, we aim to predict tissue segmentation maps on T2-weighted magnetic resonance imaging data of an unseen preterm-born neonatal population, which has both different acquisition parameters and population bias when compared to our training data. We achieve this by investigating two unsupervised domain adaptation techniques with the objective of finding the best solution for our problem. We compare the two methods with a baseline fully-supervised segmentation network and report our results in terms of Dice scores obtained on our source test dataset. Moreover, we analyse tissue volumes and cortical thickness measures of the harmonized data on a subset of the population matched for gestational age at birth and postmenstrual age at scan. Finally, we demonstrate the applicability of the harmonized cortical gray matter maps with an analysis comparing term and preterm-born neonates and a proof-of-principle investigation of the association between cortical thickness and a language outcome measure.

AB - Deep learning based medical image segmentation has shown great potential in becoming a key part of the clinical analysis pipeline. However, many of these models rely on the assumption that the train and test data come from the same distribution. This means that such methods cannot guarantee high quality predictions when the source and target domains are dissimilar due to different acquisition protocols, or biases in patient cohorts. Recently, unsupervised domain adaptation techniques have shown great potential in alleviating this problem by minimizing the shift between the source and target distributions, without requiring the use of labeled data in the target domain. In this work, we aim to predict tissue segmentation maps on T2-weighted magnetic resonance imaging data of an unseen preterm-born neonatal population, which has both different acquisition parameters and population bias when compared to our training data. We achieve this by investigating two unsupervised domain adaptation techniques with the objective of finding the best solution for our problem. We compare the two methods with a baseline fully-supervised segmentation network and report our results in terms of Dice scores obtained on our source test dataset. Moreover, we analyse tissue volumes and cortical thickness measures of the harmonized data on a subset of the population matched for gestational age at birth and postmenstrual age at scan. Finally, we demonstrate the applicability of the harmonized cortical gray matter maps with an analysis comparing term and preterm-born neonates and a proof-of-principle investigation of the association between cortical thickness and a language outcome measure.

UR - http://www.scopus.com/inward/record.url?scp=85107776575&partnerID=8YFLogxK

U2 - 10.3389/fnins.2021.662005

DO - 10.3389/fnins.2021.662005

M3 - Article

SN - 1662-453X

VL - 15

JO - Frontiers in Neuroscience

JF - Frontiers in Neuroscience

M1 - 662005

ER -

Harmonized segmentation of neonatal brain MRI

Abstract

Access to Document

Other files and links

Fingerprint

Cite this