PADDIT: Probabilistic Augmentation of Data using Diffeomorphic Image Transformation

Mauricio Orbes-Arteaga; Lauge Sørensen; Jorge  Cardoso; Marc  Modat; Sebastien Ourselin; Stefan Horst Sommer; Mads Nielsen; Christian Igel; Akshay Sadananda Uppinakudru Pai

doi:10.1117/12.2512520

PADDIT: Probabilistic Augmentation of Data using Diffeomorphic Image Transformation

Mauricio Orbes-Arteaga, Lauge Sørensen, Jorge Cardoso , Marc Modat , Sebastien Ourselin, Stefan Horst Sommer, Mads Nielsen, Christian Igel, Akshay Sadananda Uppinakudru Pai

1 Citationer (Scopus)

Abstract

For proper generalization performance of convolutional neural networks (CNNs) in medical image segmentation, the learnt features should be invariant under particular non-linear shape variations of the input. To induce invariance in CNNs to such transformations, we propose Probabilistic Augmentation of Data using Diffeomorphic Image Transformation (PADDIT) – a systematic framework for generating realistic transformations that can be used to augment data for training CNNs. The main advantage of PADDIT is the ability to produce transformations that capture the morphological variability in the training data. To this end, PADDIT constructs a mean template which represents the main shape tendency of the training data. A Hamiltonian Monte Carlo(HMC) scheme is used to sample transformations which warp the training images to the generated mean template. Augmented images are created by warping the training images using the sampled transformations. We show that CNNs trained with PADDIT outperforms CNNs trained without augmentation and with generic augmentation (0.2 and 0.15 higher dice accuracy respectively) in segmenting white matter hyperintensities from T1 and FLAIR brain MRI scans.

Originalsprog	Engelsk
Titel	Proceedings, SPIE 10949, Medical Imaging 2019: Image Processing
Antal sider	6
Forlag	SPIE - International Society for Optical Engineering
Publikationsdato	2019
Kapitel	109490S
DOI	https://doi.org/10.1117/12.2512520
Status	Udgivet - 2019
Begivenhed	SPIE Medical Imaging - San Diego, USA Varighed: 16 feb. 2019 → 21 feb. 2019

Konference

Konference	SPIE Medical Imaging
Land/Område	USA
By	San Diego
Periode	16/02/2019 → 21/02/2019

Adgang til dokumentet

10.1117/12.2512520

Citationsformater

Orbes-Arteaga, M., Sørensen, L., Cardoso , J., Modat , M., Ourselin, S., Sommer, S. H., Nielsen, M., Igel, C., & Pai, A. S. U. (2019). PADDIT: Probabilistic Augmentation of Data using Diffeomorphic Image Transformation. I Proceedings, SPIE 10949, Medical Imaging 2019: Image Processing SPIE - International Society for Optical Engineering. https://doi.org/10.1117/12.2512520

PADDIT: Probabilistic Augmentation of Data using Diffeomorphic Image Transformation. / Orbes-Arteaga, Mauricio; Sørensen, Lauge; Cardoso , Jorge et al.
Proceedings, SPIE 10949, Medical Imaging 2019: Image Processing . SPIE - International Society for Optical Engineering, 2019.

Publikation: Bidrag til bog/antologi/rapport › Konferencebidrag i proceedings › Forskning › peer review

Orbes-Arteaga, M, Sørensen, L, Cardoso , J, Modat , M, Ourselin, S, Sommer, SH , Nielsen, M , Igel, C & Pai, ASU 2019, PADDIT: Probabilistic Augmentation of Data using Diffeomorphic Image Transformation. i Proceedings, SPIE 10949, Medical Imaging 2019: Image Processing . SPIE - International Society for Optical Engineering, SPIE Medical Imaging, San Diego, USA, 16/02/2019. https://doi.org/10.1117/12.2512520

@inproceedings{dec667bf64d54544bec8b04dd6d68afc,

title = "PADDIT: Probabilistic Augmentation of Data using Diffeomorphic Image Transformation",

abstract = "For proper generalization performance of convolutional neural networks (CNNs) in medical image segmentation, the learnt features should be invariant under particular non-linear shape variations of the input. To induce invariance in CNNs to such transformations, we propose Probabilistic Augmentation of Data using Diffeomorphic Image Transformation (PADDIT) – a systematic framework for generating realistic transformations that can be used to augment data for training CNNs. The main advantage of PADDIT is the ability to produce transformations that capture the morphological variability in the training data. To this end, PADDIT constructs a mean template which represents the main shape tendency of the training data. A Hamiltonian Monte Carlo(HMC) scheme is used to sample transformations which warp the training images to the generated mean template. Augmented images are created by warping the training images using the sampled transformations. We show that CNNs trained with PADDIT outperforms CNNs trained without augmentation and with generic augmentation (0.2 and 0.15 higher dice accuracy respectively) in segmenting white matter hyperintensities from T1 and FLAIR brain MRI scans. ",

author = "Mauricio Orbes-Arteaga and Lauge S{\o}rensen and Jorge Cardoso and Marc Modat and Sebastien Ourselin and Sommer, {Stefan Horst} and Mads Nielsen and Christian Igel and Pai, {Akshay Sadananda Uppinakudru}",

year = "2019",

doi = "10.1117/12.2512520",

language = "English",

booktitle = "Proceedings, SPIE 10949, Medical Imaging 2019: Image Processing",

publisher = "SPIE - International Society for Optical Engineering",

note = "SPIE Medical Imaging ; Conference date: 16-02-2019 Through 21-02-2019",

}

TY - GEN

T1 - PADDIT

T2 - SPIE Medical Imaging

AU - Orbes-Arteaga, Mauricio

AU - Sørensen, Lauge

AU - Cardoso , Jorge

AU - Modat , Marc

AU - Ourselin, Sebastien

AU - Sommer, Stefan Horst

AU - Nielsen, Mads

AU - Igel, Christian

AU - Pai, Akshay Sadananda Uppinakudru

PY - 2019

Y1 - 2019

N2 - For proper generalization performance of convolutional neural networks (CNNs) in medical image segmentation, the learnt features should be invariant under particular non-linear shape variations of the input. To induce invariance in CNNs to such transformations, we propose Probabilistic Augmentation of Data using Diffeomorphic Image Transformation (PADDIT) – a systematic framework for generating realistic transformations that can be used to augment data for training CNNs. The main advantage of PADDIT is the ability to produce transformations that capture the morphological variability in the training data. To this end, PADDIT constructs a mean template which represents the main shape tendency of the training data. A Hamiltonian Monte Carlo(HMC) scheme is used to sample transformations which warp the training images to the generated mean template. Augmented images are created by warping the training images using the sampled transformations. We show that CNNs trained with PADDIT outperforms CNNs trained without augmentation and with generic augmentation (0.2 and 0.15 higher dice accuracy respectively) in segmenting white matter hyperintensities from T1 and FLAIR brain MRI scans.

AB - For proper generalization performance of convolutional neural networks (CNNs) in medical image segmentation, the learnt features should be invariant under particular non-linear shape variations of the input. To induce invariance in CNNs to such transformations, we propose Probabilistic Augmentation of Data using Diffeomorphic Image Transformation (PADDIT) – a systematic framework for generating realistic transformations that can be used to augment data for training CNNs. The main advantage of PADDIT is the ability to produce transformations that capture the morphological variability in the training data. To this end, PADDIT constructs a mean template which represents the main shape tendency of the training data. A Hamiltonian Monte Carlo(HMC) scheme is used to sample transformations which warp the training images to the generated mean template. Augmented images are created by warping the training images using the sampled transformations. We show that CNNs trained with PADDIT outperforms CNNs trained without augmentation and with generic augmentation (0.2 and 0.15 higher dice accuracy respectively) in segmenting white matter hyperintensities from T1 and FLAIR brain MRI scans.

U2 - 10.1117/12.2512520

DO - 10.1117/12.2512520

M3 - Article in proceedings

BT - Proceedings, SPIE 10949, Medical Imaging 2019: Image Processing

PB - SPIE - International Society for Optical Engineering

Y2 - 16 February 2019 through 21 February 2019

ER -