Unsupervised domain adaptation method for segmenting cross-sectional CCA images.


Journal

Computer methods and programs in biomedicine
ISSN: 1872-7565
Titre abrégé: Comput Methods Programs Biomed
Pays: Ireland
ID NLM: 8506513

Informations de publication

Date de publication:
Oct 2022
Historique:
received: 16 03 2022
revised: 20 07 2022
accepted: 21 07 2022
pubmed: 31 7 2022
medline: 28 9 2022
entrez: 30 7 2022
Statut: ppublish

Résumé

Automatic vessel segmentation in ultrasound is challenging due to the quality of the ultrasound images, which is affected by attenuation, high level of speckle noise and acoustic shadowing. Recently, deep convolutional neural networks are increasing in popularity due to their great performance on image segmentation problems, including vessel segmentation. Traditionally, large labeled datasets are required to train a network that achieves high performance, and is able to generalize well to different orientations, transducers and ultrasound scanners. However, these large datasets are rare, given that it is challenging and time-consuming to acquire and manually annotate in-vivo data. In this work, we present a model-based, unsupervised domain adaptation method that consists of two stages. In the first stage, the network is trained on simulated ultrasound images, which have an accurate ground truth. In the second stage, the network continues training on in-vivo data in an unsupervised way, therefore not requiring the data to be labelled. Rather than using an adversarial neural network, prior knowledge on the elliptical shape of the segmentation mask is used to detect unexpected outputs. The segmentation performance was quantified using manually segmented images as ground truth. Due to the proposed domain adaptation method, the median Dice similarity coefficient increased from 0 to 0.951, outperforming a domain adversarial neural network (median Dice 0.922) and a state-of-the-art Star-Kalman algorithm that was specifically designed for this dataset (median Dice 0.942). The results show that it is feasible to first train a neural network on simulated data, and then apply model-based domain adaptation to further improve segmentation performance by training on unlabeled in-vivo data. This overcomes the limitation of conventional deep learning approaches to require large amounts of manually labeled in-vivo data. Since the proposed domain adaptation method only requires prior knowledge on the shape of the segmentation mask, performance can be explored in various domains and applications in future research.

Sections du résumé

BACKGROUND AND OBJECTIVES OBJECTIVE
Automatic vessel segmentation in ultrasound is challenging due to the quality of the ultrasound images, which is affected by attenuation, high level of speckle noise and acoustic shadowing. Recently, deep convolutional neural networks are increasing in popularity due to their great performance on image segmentation problems, including vessel segmentation. Traditionally, large labeled datasets are required to train a network that achieves high performance, and is able to generalize well to different orientations, transducers and ultrasound scanners. However, these large datasets are rare, given that it is challenging and time-consuming to acquire and manually annotate in-vivo data.
METHODS METHODS
In this work, we present a model-based, unsupervised domain adaptation method that consists of two stages. In the first stage, the network is trained on simulated ultrasound images, which have an accurate ground truth. In the second stage, the network continues training on in-vivo data in an unsupervised way, therefore not requiring the data to be labelled. Rather than using an adversarial neural network, prior knowledge on the elliptical shape of the segmentation mask is used to detect unexpected outputs.
RESULTS RESULTS
The segmentation performance was quantified using manually segmented images as ground truth. Due to the proposed domain adaptation method, the median Dice similarity coefficient increased from 0 to 0.951, outperforming a domain adversarial neural network (median Dice 0.922) and a state-of-the-art Star-Kalman algorithm that was specifically designed for this dataset (median Dice 0.942).
CONCLUSIONS CONCLUSIONS
The results show that it is feasible to first train a neural network on simulated data, and then apply model-based domain adaptation to further improve segmentation performance by training on unlabeled in-vivo data. This overcomes the limitation of conventional deep learning approaches to require large amounts of manually labeled in-vivo data. Since the proposed domain adaptation method only requires prior knowledge on the shape of the segmentation mask, performance can be explored in various domains and applications in future research.

Identifiants

pubmed: 35907375
pii: S0169-2607(22)00419-9
doi: 10.1016/j.cmpb.2022.107037
pii:
doi:

Types de publication

Journal Article

Langues

eng

Sous-ensembles de citation

IM

Pagination

107037

Informations de copyright

Copyright © 2022 The Author(s). Published by Elsevier B.V. All rights reserved.

Déclaration de conflit d'intérêts

Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Auteurs

Luuk van Knippenberg (L)

Department of Electrical Engineering, Eindhoven University of Technology, the Netherlands. Electronic address: l.a.e.m.v.knippenberg@tue.nl.

Ruud J G van Sloun (RJG)

Department of Electrical Engineering, Eindhoven University of Technology, the Netherlands; Department of Anesthesiology, Catharina Hospital Eindhoven, the Netherlands.

Massimo Mischi (M)

Department of Electrical Engineering, Eindhoven University of Technology, the Netherlands; Department of Anesthesiology, Catharina Hospital Eindhoven, the Netherlands.

Joerik de Ruijter (J)

Department of Biomedical Engineering, Eindhoven University of Technology, the Netherlands.

Richard Lopata (R)

Department of Biomedical Engineering, Eindhoven University of Technology, the Netherlands; Department of Anesthesiology, Catharina Hospital Eindhoven, the Netherlands.

R Arthur Bouwman (RA)

Department of Anesthesiology, Catharina Hospital Eindhoven, the Netherlands.

Articles similaires

Humans United States Aged Cross-Sectional Studies Medicare Part C

Selecting optimal software code descriptors-The case of Java.

Yegor Bugayenko, Zamira Kholmatova, Artem Kruglov et al.
1.00
Software Algorithms Programming Languages
Humans Adult Male Female Video Games

Classifications MeSH