Geodesic image regression with a sparse parameterization of diffeomorphisms

James Fishbaugh, Marcel Prastawa, Guido Gerig, Stanley Durrleman

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    Abstract

    Image regression allows for time-discrete imaging data to be modeled continuously, and is a crucial tool for conducting statistical analysis on longitudinal images. Geodesic models are particularly well suited for statistical analysis, as image evolution is fully characterized by a baseline image and initial momenta. However, existing geodesic image regression models are parameterized by a large number of initial momenta, equal to the number of image voxels. In this paper, we present a sparse geodesic image regression framework which greatly reduces the number of model parameters. We combine a control point formulation of deformations with a L1 penalty to select the most relevant subset of momenta. This way, the number of model parameters reflects the complexity of anatomical changes in time rather than the sampling of the image. We apply our method to both synthetic and real data and show that we can decrease the number of model parameters (from the number of voxels down to hundreds) with only minimal decrease in model accuracy. The reduction in model parameters has the potential to improve the power of ensuing statistical analysis, which faces the challenging problem of high dimensionality.

    Original languageEnglish (US)
    Title of host publicationGeometric Science of Information - First International Conference, GSI 2013, Proceedings
    Pages95-102
    Number of pages8
    Volume8085 LNCS
    DOIs
    StatePublished - 2013
    Event1st International SEE Conference on Geometric Science of Information, GSI 2013 - Paris, France
    Duration: Aug 28 2013Aug 30 2013

    Publication series

    NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
    Volume8085 LNCS
    ISSN (Print)03029743
    ISSN (Electronic)16113349

    Other

    Other1st International SEE Conference on Geometric Science of Information, GSI 2013
    CountryFrance
    CityParis
    Period8/28/138/30/13

    Fingerprint

    Parameterization
    Diffeomorphisms
    Geodesic
    Regression
    Statistical Analysis
    Statistical methods
    Momentum
    Voxel
    Model
    Decrease
    Image Model
    Control Points
    Dimensionality
    Penalty
    Baseline
    Regression Model
    Discrete-time
    Imaging
    Sampling
    Imaging techniques

    ASJC Scopus subject areas

    • Computer Science(all)
    • Theoretical Computer Science

    Cite this

    Fishbaugh, J., Prastawa, M., Gerig, G., & Durrleman, S. (2013). Geodesic image regression with a sparse parameterization of diffeomorphisms. In Geometric Science of Information - First International Conference, GSI 2013, Proceedings (Vol. 8085 LNCS, pp. 95-102). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8085 LNCS). https://doi.org/10.1007/978-3-642-40020-9_9

    Geodesic image regression with a sparse parameterization of diffeomorphisms. / Fishbaugh, James; Prastawa, Marcel; Gerig, Guido; Durrleman, Stanley.

    Geometric Science of Information - First International Conference, GSI 2013, Proceedings. Vol. 8085 LNCS 2013. p. 95-102 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8085 LNCS).

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    Fishbaugh, J, Prastawa, M, Gerig, G & Durrleman, S 2013, Geodesic image regression with a sparse parameterization of diffeomorphisms. in Geometric Science of Information - First International Conference, GSI 2013, Proceedings. vol. 8085 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 8085 LNCS, pp. 95-102, 1st International SEE Conference on Geometric Science of Information, GSI 2013, Paris, France, 8/28/13. https://doi.org/10.1007/978-3-642-40020-9_9
    Fishbaugh J, Prastawa M, Gerig G, Durrleman S. Geodesic image regression with a sparse parameterization of diffeomorphisms. In Geometric Science of Information - First International Conference, GSI 2013, Proceedings. Vol. 8085 LNCS. 2013. p. 95-102. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-642-40020-9_9
    Fishbaugh, James ; Prastawa, Marcel ; Gerig, Guido ; Durrleman, Stanley. / Geodesic image regression with a sparse parameterization of diffeomorphisms. Geometric Science of Information - First International Conference, GSI 2013, Proceedings. Vol. 8085 LNCS 2013. pp. 95-102 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
    @inproceedings{bd960baaf8f542c88171850b1df79e69,
    title = "Geodesic image regression with a sparse parameterization of diffeomorphisms",
    abstract = "Image regression allows for time-discrete imaging data to be modeled continuously, and is a crucial tool for conducting statistical analysis on longitudinal images. Geodesic models are particularly well suited for statistical analysis, as image evolution is fully characterized by a baseline image and initial momenta. However, existing geodesic image regression models are parameterized by a large number of initial momenta, equal to the number of image voxels. In this paper, we present a sparse geodesic image regression framework which greatly reduces the number of model parameters. We combine a control point formulation of deformations with a L1 penalty to select the most relevant subset of momenta. This way, the number of model parameters reflects the complexity of anatomical changes in time rather than the sampling of the image. We apply our method to both synthetic and real data and show that we can decrease the number of model parameters (from the number of voxels down to hundreds) with only minimal decrease in model accuracy. The reduction in model parameters has the potential to improve the power of ensuing statistical analysis, which faces the challenging problem of high dimensionality.",
    author = "James Fishbaugh and Marcel Prastawa and Guido Gerig and Stanley Durrleman",
    year = "2013",
    doi = "10.1007/978-3-642-40020-9_9",
    language = "English (US)",
    isbn = "9783642400193",
    volume = "8085 LNCS",
    series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
    pages = "95--102",
    booktitle = "Geometric Science of Information - First International Conference, GSI 2013, Proceedings",

    }

    TY - GEN

    T1 - Geodesic image regression with a sparse parameterization of diffeomorphisms

    AU - Fishbaugh, James

    AU - Prastawa, Marcel

    AU - Gerig, Guido

    AU - Durrleman, Stanley

    PY - 2013

    Y1 - 2013

    N2 - Image regression allows for time-discrete imaging data to be modeled continuously, and is a crucial tool for conducting statistical analysis on longitudinal images. Geodesic models are particularly well suited for statistical analysis, as image evolution is fully characterized by a baseline image and initial momenta. However, existing geodesic image regression models are parameterized by a large number of initial momenta, equal to the number of image voxels. In this paper, we present a sparse geodesic image regression framework which greatly reduces the number of model parameters. We combine a control point formulation of deformations with a L1 penalty to select the most relevant subset of momenta. This way, the number of model parameters reflects the complexity of anatomical changes in time rather than the sampling of the image. We apply our method to both synthetic and real data and show that we can decrease the number of model parameters (from the number of voxels down to hundreds) with only minimal decrease in model accuracy. The reduction in model parameters has the potential to improve the power of ensuing statistical analysis, which faces the challenging problem of high dimensionality.

    AB - Image regression allows for time-discrete imaging data to be modeled continuously, and is a crucial tool for conducting statistical analysis on longitudinal images. Geodesic models are particularly well suited for statistical analysis, as image evolution is fully characterized by a baseline image and initial momenta. However, existing geodesic image regression models are parameterized by a large number of initial momenta, equal to the number of image voxels. In this paper, we present a sparse geodesic image regression framework which greatly reduces the number of model parameters. We combine a control point formulation of deformations with a L1 penalty to select the most relevant subset of momenta. This way, the number of model parameters reflects the complexity of anatomical changes in time rather than the sampling of the image. We apply our method to both synthetic and real data and show that we can decrease the number of model parameters (from the number of voxels down to hundreds) with only minimal decrease in model accuracy. The reduction in model parameters has the potential to improve the power of ensuing statistical analysis, which faces the challenging problem of high dimensionality.

    UR - http://www.scopus.com/inward/record.url?scp=84884939408&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=84884939408&partnerID=8YFLogxK

    U2 - 10.1007/978-3-642-40020-9_9

    DO - 10.1007/978-3-642-40020-9_9

    M3 - Conference contribution

    AN - SCOPUS:84884939408

    SN - 9783642400193

    VL - 8085 LNCS

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

    SP - 95

    EP - 102

    BT - Geometric Science of Information - First International Conference, GSI 2013, Proceedings

    ER -