Augmentation based unsupervised domain adaptation

Publikation: Working paper › Preprint › Forskning

Standard

Augmentation based unsupervised domain adaptation. / Orbes-Arteaga, Mauricio; Varsavsky, Thomas; Sørensen, Lauge ; Nielsen, Mads; Pai, Akshay Sadananda Uppinakudru; Ourselin, Sebastien; Modat , Marc ; Cardoso, M. Jorge.

arXiv.org, 2022.

Publikation: Working paper › Preprint › Forskning

Harvard

Orbes-Arteaga, M, Varsavsky, T, Sørensen, L, Nielsen, M, Pai, ASU, Ourselin, S, Modat , M & Cardoso, MJ 2022 'Augmentation based unsupervised domain adaptation' arXiv.org. <https://arxiv.org/abs/2202.11486>

APA

Orbes-Arteaga, M., Varsavsky, T., Sørensen, L., Nielsen, M., Pai, A. S. U., Ourselin, S., Modat , M., & Cardoso, M. J. (2022). Augmentation based unsupervised domain adaptation. arXiv.org. https://arxiv.org/abs/2202.11486

Vancouver

Orbes-Arteaga M, Varsavsky T, Sørensen L, Nielsen M, Pai ASU, Ourselin S o.a. Augmentation based unsupervised domain adaptation. arXiv.org. 2022.

Author

Orbes-Arteaga, Mauricio ; Varsavsky, Thomas ; Sørensen, Lauge ; Nielsen, Mads ; Pai, Akshay Sadananda Uppinakudru ; Ourselin, Sebastien ; Modat , Marc ; Cardoso, M. Jorge. / Augmentation based unsupervised domain adaptation. arXiv.org, 2022.

Bibtex

@techreport{64c4ec21cad6490990d5b0bb4cb0db74,
title = "Augmentation based unsupervised domain adaptation",
abstract = "The insertion of deep learning in medical image analysis had lead to the development of state-of-the art strategies in several applications such a disease classification, as well as abnormality detection and segmentation. However, even the most advanced methods require a huge and diverse amount of data to generalize. Because in realistic clinical scenarios, data acquisition and annotation is expensive, deep learning models trained on small and unrepresentative data tend to outperform when deployed in data that differs from the one used for training (e.g data from different scanners). In this work, we proposed a domain adaptation methodology to alleviate this problem in segmentation models. Our approach takes advantage of the properties of adversarial domain adaptation and consistency training to achieve more robust adaptation. Using two datasets with white matter hyperintensities (WMH) annotations, we demonstrated that the proposed method improves model generalization even in corner cases where individual strategies tend to fail.",
author = "Mauricio Orbes-Arteaga and Thomas Varsavsky and Lauge S{\o}rensen and Mads Nielsen and Pai, {Akshay Sadananda Uppinakudru} and Sebastien Ourselin and Marc Modat and Cardoso, {M. Jorge}",
year = "2022",
language = "English",
publisher = "arXiv.org",
type = "WorkingPaper",
institution = "arXiv.org",

}

RIS

TY - UNPB

T1 - Augmentation based unsupervised domain adaptation

AU - Orbes-Arteaga, Mauricio

AU - Varsavsky, Thomas

AU - Sørensen, Lauge

AU - Nielsen, Mads

AU - Pai, Akshay Sadananda Uppinakudru

AU - Ourselin, Sebastien

AU - Modat , Marc

AU - Cardoso, M. Jorge

PY - 2022

Y1 - 2022

N2 - The insertion of deep learning in medical image analysis had lead to the development of state-of-the art strategies in several applications such a disease classification, as well as abnormality detection and segmentation. However, even the most advanced methods require a huge and diverse amount of data to generalize. Because in realistic clinical scenarios, data acquisition and annotation is expensive, deep learning models trained on small and unrepresentative data tend to outperform when deployed in data that differs from the one used for training (e.g data from different scanners). In this work, we proposed a domain adaptation methodology to alleviate this problem in segmentation models. Our approach takes advantage of the properties of adversarial domain adaptation and consistency training to achieve more robust adaptation. Using two datasets with white matter hyperintensities (WMH) annotations, we demonstrated that the proposed method improves model generalization even in corner cases where individual strategies tend to fail.

AB - The insertion of deep learning in medical image analysis had lead to the development of state-of-the art strategies in several applications such a disease classification, as well as abnormality detection and segmentation. However, even the most advanced methods require a huge and diverse amount of data to generalize. Because in realistic clinical scenarios, data acquisition and annotation is expensive, deep learning models trained on small and unrepresentative data tend to outperform when deployed in data that differs from the one used for training (e.g data from different scanners). In this work, we proposed a domain adaptation methodology to alleviate this problem in segmentation models. Our approach takes advantage of the properties of adversarial domain adaptation and consistency training to achieve more robust adaptation. Using two datasets with white matter hyperintensities (WMH) annotations, we demonstrated that the proposed method improves model generalization even in corner cases where individual strategies tend to fail.

M3 - Preprint

BT - Augmentation based unsupervised domain adaptation

PB - arXiv.org

ER -

ID: 339908099