Roto-translation covariant convolutional networks for medical image analysis

Onderzoeksoutput: Hoofdstuk in Boek/Rapport/CongresprocedureConferentiebijdrageAcademicpeer review

4 Citaties (Scopus)

Uittreksel

We propose a framework for rotation and translation covariant deep learning using SE(2) group convolutions. The group product of the special Euclidean motion group SE(2) describes how a concatenation of two roto-translations results in a net roto-translation. We encode this geometric structure into convolutional neural networks (CNNs) via SE(2) group convolutional layers, which fit into the standard 2D CNN framework, and which allow to generically deal with rotated input samples without the need for data augmentation. We introduce three layers: a lifting layer which lifts a 2D (vector valued) image to an SE(2)-image, i.e., 3D (vector valued) data whose domain is SE(2); a group convolution layer from and to an SE(2)-image; and a projection layer from an SE(2)-image to a 2D image. The lifting and group convolution layers are SE(2) covariant (the output roto-translates with the input). The final projection layer, a maximum intensity projection over rotations, makes the full CNN rotation invariant. We show with three different problems in histopathology, retinal imaging, and electron microscopy that with the proposed group CNNs, state-of-the-art performance can be achieved, without the need for data augmentation by rotation and with increased performance compared to standard CNNs that do rely on augmentation.

TaalEngels
TitelMedical Image Computing and Computer Assisted Intervention – MICCAI 2018 - 21st International Conference, 2018, Proceedings
RedacteurenJulia A. Schnabel, Christos Davatzikos, Carlos Alberola-López, Gabor Fichtinger, Alejandro F. Frangi
Plaats van productieCham
UitgeverijSpringer
Pagina's440-448
Aantal pagina's9
ISBN van elektronische versie978-3-030-00928-1
ISBN van geprinte versie978-3-030-00927-4
DOI's
StatusGepubliceerd - 1 jan 2018
Evenement21st International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2018 - Granada, Spanje
Duur: 16 sep 201820 sep 2018

Publicatie series

NaamLecture Notes in Computer Science
Volume11070
ISSN van geprinte versie0302-9743
ISSN van elektronische versie1611-3349

Congres

Congres21st International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2018
LandSpanje
StadGranada
Periode16/09/1820/09/18

Vingerafdruk

Medical Image Analysis
Image analysis
Neural networks
Convolution
Neural Networks
Data Augmentation
Projection
Electron microscopy
Rotation Invariant
Concatenation
Electron Microscopy
Augmentation
Geometric Structure
3D Image
Imaging techniques
Euclidean
Imaging
Motion
Output

Trefwoorden

    Citeer dit

    Bekkers, E. J., Lafarge, M. W., Veta, M., Eppenhof, K. A. J., Pluim, J. P. W., & Duits, R. (2018). Roto-translation covariant convolutional networks for medical image analysis. In J. A. Schnabel, C. Davatzikos, C. Alberola-López, G. Fichtinger, & A. F. Frangi (editors), Medical Image Computing and Computer Assisted Intervention – MICCAI 2018 - 21st International Conference, 2018, Proceedings (blz. 440-448). (Lecture Notes in Computer Science; Vol. 11070). Cham: Springer. DOI: 10.1007/978-3-030-00928-1_50
    Bekkers, Erik J. ; Lafarge, Maxime W. ; Veta, Mitko ; Eppenhof, Koen A.J. ; Pluim, Josien P.W. ; Duits, Remco. / Roto-translation covariant convolutional networks for medical image analysis. Medical Image Computing and Computer Assisted Intervention – MICCAI 2018 - 21st International Conference, 2018, Proceedings. redacteur / Julia A. Schnabel ; Christos Davatzikos ; Carlos Alberola-López ; Gabor Fichtinger ; Alejandro F. Frangi. Cham : Springer, 2018. blz. 440-448 (Lecture Notes in Computer Science).
    @inproceedings{a1a1ff2ec964483db9039246064ccf23,
    title = "Roto-translation covariant convolutional networks for medical image analysis",
    abstract = "We propose a framework for rotation and translation covariant deep learning using SE(2) group convolutions. The group product of the special Euclidean motion group SE(2) describes how a concatenation of two roto-translations results in a net roto-translation. We encode this geometric structure into convolutional neural networks (CNNs) via SE(2) group convolutional layers, which fit into the standard 2D CNN framework, and which allow to generically deal with rotated input samples without the need for data augmentation. We introduce three layers: a lifting layer which lifts a 2D (vector valued) image to an SE(2)-image, i.e., 3D (vector valued) data whose domain is SE(2); a group convolution layer from and to an SE(2)-image; and a projection layer from an SE(2)-image to a 2D image. The lifting and group convolution layers are SE(2) covariant (the output roto-translates with the input). The final projection layer, a maximum intensity projection over rotations, makes the full CNN rotation invariant. We show with three different problems in histopathology, retinal imaging, and electron microscopy that with the proposed group CNNs, state-of-the-art performance can be achieved, without the need for data augmentation by rotation and with increased performance compared to standard CNNs that do rely on augmentation.",
    keywords = "Cell boundary segmentation, Group convolutional network, Mitosis detection, Roto-translation group, Vessel segmentation",
    author = "Bekkers, {Erik J.} and Lafarge, {Maxime W.} and Mitko Veta and Eppenhof, {Koen A.J.} and Pluim, {Josien P.W.} and Remco Duits",
    year = "2018",
    month = "1",
    day = "1",
    doi = "10.1007/978-3-030-00928-1_50",
    language = "English",
    isbn = "978-3-030-00927-4",
    series = "Lecture Notes in Computer Science",
    publisher = "Springer",
    pages = "440--448",
    editor = "Schnabel, {Julia A.} and Christos Davatzikos and Carlos Alberola-L{\'o}pez and Gabor Fichtinger and Frangi, {Alejandro F.}",
    booktitle = "Medical Image Computing and Computer Assisted Intervention – MICCAI 2018 - 21st International Conference, 2018, Proceedings",
    address = "Germany",

    }

    Bekkers, EJ, Lafarge, MW, Veta, M, Eppenhof, KAJ, Pluim, JPW & Duits, R 2018, Roto-translation covariant convolutional networks for medical image analysis. in JA Schnabel, C Davatzikos, C Alberola-López, G Fichtinger & AF Frangi (redactie), Medical Image Computing and Computer Assisted Intervention – MICCAI 2018 - 21st International Conference, 2018, Proceedings. Lecture Notes in Computer Science, vol. 11070, Springer, Cham, blz. 440-448, Granada, Spanje, 16/09/18. DOI: 10.1007/978-3-030-00928-1_50

    Roto-translation covariant convolutional networks for medical image analysis. / Bekkers, Erik J.; Lafarge, Maxime W.; Veta, Mitko; Eppenhof, Koen A.J.; Pluim, Josien P.W.; Duits, Remco.

    Medical Image Computing and Computer Assisted Intervention – MICCAI 2018 - 21st International Conference, 2018, Proceedings. redactie / Julia A. Schnabel; Christos Davatzikos; Carlos Alberola-López; Gabor Fichtinger; Alejandro F. Frangi. Cham : Springer, 2018. blz. 440-448 (Lecture Notes in Computer Science; Vol. 11070).

    Onderzoeksoutput: Hoofdstuk in Boek/Rapport/CongresprocedureConferentiebijdrageAcademicpeer review

    TY - GEN

    T1 - Roto-translation covariant convolutional networks for medical image analysis

    AU - Bekkers,Erik J.

    AU - Lafarge,Maxime W.

    AU - Veta,Mitko

    AU - Eppenhof,Koen A.J.

    AU - Pluim,Josien P.W.

    AU - Duits,Remco

    PY - 2018/1/1

    Y1 - 2018/1/1

    N2 - We propose a framework for rotation and translation covariant deep learning using SE(2) group convolutions. The group product of the special Euclidean motion group SE(2) describes how a concatenation of two roto-translations results in a net roto-translation. We encode this geometric structure into convolutional neural networks (CNNs) via SE(2) group convolutional layers, which fit into the standard 2D CNN framework, and which allow to generically deal with rotated input samples without the need for data augmentation. We introduce three layers: a lifting layer which lifts a 2D (vector valued) image to an SE(2)-image, i.e., 3D (vector valued) data whose domain is SE(2); a group convolution layer from and to an SE(2)-image; and a projection layer from an SE(2)-image to a 2D image. The lifting and group convolution layers are SE(2) covariant (the output roto-translates with the input). The final projection layer, a maximum intensity projection over rotations, makes the full CNN rotation invariant. We show with three different problems in histopathology, retinal imaging, and electron microscopy that with the proposed group CNNs, state-of-the-art performance can be achieved, without the need for data augmentation by rotation and with increased performance compared to standard CNNs that do rely on augmentation.

    AB - We propose a framework for rotation and translation covariant deep learning using SE(2) group convolutions. The group product of the special Euclidean motion group SE(2) describes how a concatenation of two roto-translations results in a net roto-translation. We encode this geometric structure into convolutional neural networks (CNNs) via SE(2) group convolutional layers, which fit into the standard 2D CNN framework, and which allow to generically deal with rotated input samples without the need for data augmentation. We introduce three layers: a lifting layer which lifts a 2D (vector valued) image to an SE(2)-image, i.e., 3D (vector valued) data whose domain is SE(2); a group convolution layer from and to an SE(2)-image; and a projection layer from an SE(2)-image to a 2D image. The lifting and group convolution layers are SE(2) covariant (the output roto-translates with the input). The final projection layer, a maximum intensity projection over rotations, makes the full CNN rotation invariant. We show with three different problems in histopathology, retinal imaging, and electron microscopy that with the proposed group CNNs, state-of-the-art performance can be achieved, without the need for data augmentation by rotation and with increased performance compared to standard CNNs that do rely on augmentation.

    KW - Cell boundary segmentation

    KW - Group convolutional network

    KW - Mitosis detection

    KW - Roto-translation group

    KW - Vessel segmentation

    UR - http://www.scopus.com/inward/record.url?scp=85054049941&partnerID=8YFLogxK

    U2 - 10.1007/978-3-030-00928-1_50

    DO - 10.1007/978-3-030-00928-1_50

    M3 - Conference contribution

    SN - 978-3-030-00927-4

    T3 - Lecture Notes in Computer Science

    SP - 440

    EP - 448

    BT - Medical Image Computing and Computer Assisted Intervention – MICCAI 2018 - 21st International Conference, 2018, Proceedings

    PB - Springer

    CY - Cham

    ER -

    Bekkers EJ, Lafarge MW, Veta M, Eppenhof KAJ, Pluim JPW, Duits R. Roto-translation covariant convolutional networks for medical image analysis. In Schnabel JA, Davatzikos C, Alberola-López C, Fichtinger G, Frangi AF, redacteurs, Medical Image Computing and Computer Assisted Intervention – MICCAI 2018 - 21st International Conference, 2018, Proceedings. Cham: Springer. 2018. blz. 440-448. (Lecture Notes in Computer Science). Beschikbaar vanaf, DOI: 10.1007/978-3-030-00928-1_50