Harmonised Segmentation of Neonatal Brain MRI: A Domain Adaptation Approach

TY - CHAP

T1 - Harmonised Segmentation of Neonatal Brain MRI

T2 - 1st International Workshop on Advances in Simplifying Medical UltraSound, ASMUS 2020, and the 5th International Workshop on Perinatal, Preterm and Paediatric Image Analysis, PIPPI 2020, held in conjunction with the 23rd International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2020

AU - Grigorescu, Irina

AU - Cordero-Grande, Lucilio

AU - Batalle, Dafnis

AU - Edwards, A. David

AU - Hajnal, Joseph V.

AU - Modat, Marc

AU - Deprez, Maria

PY - 2020

Y1 - 2020

N2 - Medical image deep learning segmentation has shown great potential in becoming an ubiquitous part of the clinical analysis pipeline. However, these 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. In this work, we aim to predict tissue segmentation maps on an unseen dataset, which has both different acquisition parameters and population bias when compared to our training data. We achieve this by investigating two unsupervised domain adaptation (UDA) techniques with the objective of finding the best solution for our problem. We compare the two methods with a baseline fully-supervised segmentation network in terms of cortical thickness measures.

AB - Medical image deep learning segmentation has shown great potential in becoming an ubiquitous part of the clinical analysis pipeline. However, these 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. In this work, we aim to predict tissue segmentation maps on an unseen dataset, which has both different acquisition parameters and population bias when compared to our training data. We achieve this by investigating two unsupervised domain adaptation (UDA) techniques with the objective of finding the best solution for our problem. We compare the two methods with a baseline fully-supervised segmentation network in terms of cortical thickness measures.

KW - Domain adaptation

KW - Segmentation

UR - http://www.scopus.com/inward/record.url?scp=85092746942&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-60334-2_25

DO - 10.1007/978-3-030-60334-2_25

M3 - Conference paper

AN - SCOPUS:85092746942

SN - 9783030603335

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

SP - 253

EP - 263

BT - Medical Ultrasound, and Preterm, Perinatal and Paediatric Image Analysis - 1st International Workshop, ASMUS 2020, and 5th International Workshop, PIPPI 2020, Held in Conjunction with MICCAI 2020, Proceedings

A2 - Hu, Yipeng

A2 - Licandro, Roxane

A2 - Noble, J. Alison

A2 - Hutter, Jana

A2 - Melbourne, Andrew

A2 - Aylward, Stephen

A2 - Abaci Turk, Esra

A2 - Torrents Barrena, Jordina

A2 - Torrents Barrena, Jordina

PB - Springer Science and Business Media Deutschland GmbH

Y2 - 4 October 2020 through 8 October 2020

ER -