Tensor decomposition of hyperspectral images to study autofluorescence in age-related macular degeneration

Neel Dey, Sungmin Hong, Thomas Ach, Yiannis Koutalos, Christine A. Curcio, R. Theodore Smith, Guido Gerig

    Research output: Contribution to journalArticle

    Abstract

    Autofluorescence is the emission of light by naturally occurring tissue components on the absorption of incident light. Autofluorescence within the eye is associated with several disorders, such as Age-related Macular Degeneration (AMD) which is a leading cause of central vision loss. Its pathogenesis is incompletely understood, but endogenous fluorophores in retinal tissue might play a role. Hyperspectral fluorescence microscopy of ex-vivo retinal tissue can be used to determine the fluorescence emission spectra of these fluorophores. Comparisons of spectra in healthy and diseased tissues can provide important insights into the pathogenesis of AMD. However, the spectrum from each pixel of the hyperspectral image is a superposition of spectra from multiple overlapping tissue components. As spectra cannot be negative, there is a need for a non-negative blind source separation model to isolate individual spectra. We propose a tensor formulation by leveraging multiple excitation wavelengths to excite the tissue sample. Arranging images from different excitation wavelengths as a tensor, a non-negative tensor decomposition can be performed to recover a provably unique low-rank model with factors representing emission and excitation spectra of these materials and corresponding abundance maps of autofluorescent substances in the tissue sample. We iteratively impute missing values common in fluorescence measurements using Expectation-Maximization and use L2 regularization to reduce ill-posedness. Further, we present a framework for performing group hypothesis testing on hyperspectral images, finding significant differences in spectra between AMD and control groups in the peripheral macula. In the absence of ground truth, i.e. molecular identification of fluorophores, we provide a rigorous validation of chosen methods on both synthetic and real images where fluorescence spectra are known. These methodologies can be applied to the study of other pathologies presenting autofluorescence that can be captured by hyperspectral imaging.

    Original languageEnglish (US)
    Pages (from-to)96-109
    Number of pages14
    JournalMedical Image Analysis
    Volume56
    DOIs
    StatePublished - Aug 1 2019

    Fingerprint

    Macular Degeneration
    Tensors
    Tissue
    Decomposition
    Fluorophores
    Fluorescence
    Light
    Wavelength
    Blind source separation
    Fluorescence microscopy
    Pathology
    Fluorescence Microscopy
    Pixels
    Control Groups
    Testing

    Keywords

    • Age-related macular degeneration
    • Functional data analysis
    • Hyperspectral fluorescence microscopy imaging
    • Non-negative tensor decompositions
    • Unsupervised machine learning

    ASJC Scopus subject areas

    • Radiological and Ultrasound Technology
    • Radiology Nuclear Medicine and imaging
    • Computer Vision and Pattern Recognition
    • Health Informatics
    • Computer Graphics and Computer-Aided Design

    Cite this

    Tensor decomposition of hyperspectral images to study autofluorescence in age-related macular degeneration. / Dey, Neel; Hong, Sungmin; Ach, Thomas; Koutalos, Yiannis; Curcio, Christine A.; Smith, R. Theodore; Gerig, Guido.

    In: Medical Image Analysis, Vol. 56, 01.08.2019, p. 96-109.

    Research output: Contribution to journalArticle

    Dey, Neel ; Hong, Sungmin ; Ach, Thomas ; Koutalos, Yiannis ; Curcio, Christine A. ; Smith, R. Theodore ; Gerig, Guido. / Tensor decomposition of hyperspectral images to study autofluorescence in age-related macular degeneration. In: Medical Image Analysis. 2019 ; Vol. 56. pp. 96-109.
    @article{d9e7328281474b10914aa9a29f89c75e,
    title = "Tensor decomposition of hyperspectral images to study autofluorescence in age-related macular degeneration",
    abstract = "Autofluorescence is the emission of light by naturally occurring tissue components on the absorption of incident light. Autofluorescence within the eye is associated with several disorders, such as Age-related Macular Degeneration (AMD) which is a leading cause of central vision loss. Its pathogenesis is incompletely understood, but endogenous fluorophores in retinal tissue might play a role. Hyperspectral fluorescence microscopy of ex-vivo retinal tissue can be used to determine the fluorescence emission spectra of these fluorophores. Comparisons of spectra in healthy and diseased tissues can provide important insights into the pathogenesis of AMD. However, the spectrum from each pixel of the hyperspectral image is a superposition of spectra from multiple overlapping tissue components. As spectra cannot be negative, there is a need for a non-negative blind source separation model to isolate individual spectra. We propose a tensor formulation by leveraging multiple excitation wavelengths to excite the tissue sample. Arranging images from different excitation wavelengths as a tensor, a non-negative tensor decomposition can be performed to recover a provably unique low-rank model with factors representing emission and excitation spectra of these materials and corresponding abundance maps of autofluorescent substances in the tissue sample. We iteratively impute missing values common in fluorescence measurements using Expectation-Maximization and use L2 regularization to reduce ill-posedness. Further, we present a framework for performing group hypothesis testing on hyperspectral images, finding significant differences in spectra between AMD and control groups in the peripheral macula. In the absence of ground truth, i.e. molecular identification of fluorophores, we provide a rigorous validation of chosen methods on both synthetic and real images where fluorescence spectra are known. These methodologies can be applied to the study of other pathologies presenting autofluorescence that can be captured by hyperspectral imaging.",
    keywords = "Age-related macular degeneration, Functional data analysis, Hyperspectral fluorescence microscopy imaging, Non-negative tensor decompositions, Unsupervised machine learning",
    author = "Neel Dey and Sungmin Hong and Thomas Ach and Yiannis Koutalos and Curcio, {Christine A.} and Smith, {R. Theodore} and Guido Gerig",
    year = "2019",
    month = "8",
    day = "1",
    doi = "10.1016/j.media.2019.05.009",
    language = "English (US)",
    volume = "56",
    pages = "96--109",
    journal = "Medical Image Analysis",
    issn = "1361-8415",
    publisher = "Elsevier",

    }

    TY - JOUR

    T1 - Tensor decomposition of hyperspectral images to study autofluorescence in age-related macular degeneration

    AU - Dey, Neel

    AU - Hong, Sungmin

    AU - Ach, Thomas

    AU - Koutalos, Yiannis

    AU - Curcio, Christine A.

    AU - Smith, R. Theodore

    AU - Gerig, Guido

    PY - 2019/8/1

    Y1 - 2019/8/1

    N2 - Autofluorescence is the emission of light by naturally occurring tissue components on the absorption of incident light. Autofluorescence within the eye is associated with several disorders, such as Age-related Macular Degeneration (AMD) which is a leading cause of central vision loss. Its pathogenesis is incompletely understood, but endogenous fluorophores in retinal tissue might play a role. Hyperspectral fluorescence microscopy of ex-vivo retinal tissue can be used to determine the fluorescence emission spectra of these fluorophores. Comparisons of spectra in healthy and diseased tissues can provide important insights into the pathogenesis of AMD. However, the spectrum from each pixel of the hyperspectral image is a superposition of spectra from multiple overlapping tissue components. As spectra cannot be negative, there is a need for a non-negative blind source separation model to isolate individual spectra. We propose a tensor formulation by leveraging multiple excitation wavelengths to excite the tissue sample. Arranging images from different excitation wavelengths as a tensor, a non-negative tensor decomposition can be performed to recover a provably unique low-rank model with factors representing emission and excitation spectra of these materials and corresponding abundance maps of autofluorescent substances in the tissue sample. We iteratively impute missing values common in fluorescence measurements using Expectation-Maximization and use L2 regularization to reduce ill-posedness. Further, we present a framework for performing group hypothesis testing on hyperspectral images, finding significant differences in spectra between AMD and control groups in the peripheral macula. In the absence of ground truth, i.e. molecular identification of fluorophores, we provide a rigorous validation of chosen methods on both synthetic and real images where fluorescence spectra are known. These methodologies can be applied to the study of other pathologies presenting autofluorescence that can be captured by hyperspectral imaging.

    AB - Autofluorescence is the emission of light by naturally occurring tissue components on the absorption of incident light. Autofluorescence within the eye is associated with several disorders, such as Age-related Macular Degeneration (AMD) which is a leading cause of central vision loss. Its pathogenesis is incompletely understood, but endogenous fluorophores in retinal tissue might play a role. Hyperspectral fluorescence microscopy of ex-vivo retinal tissue can be used to determine the fluorescence emission spectra of these fluorophores. Comparisons of spectra in healthy and diseased tissues can provide important insights into the pathogenesis of AMD. However, the spectrum from each pixel of the hyperspectral image is a superposition of spectra from multiple overlapping tissue components. As spectra cannot be negative, there is a need for a non-negative blind source separation model to isolate individual spectra. We propose a tensor formulation by leveraging multiple excitation wavelengths to excite the tissue sample. Arranging images from different excitation wavelengths as a tensor, a non-negative tensor decomposition can be performed to recover a provably unique low-rank model with factors representing emission and excitation spectra of these materials and corresponding abundance maps of autofluorescent substances in the tissue sample. We iteratively impute missing values common in fluorescence measurements using Expectation-Maximization and use L2 regularization to reduce ill-posedness. Further, we present a framework for performing group hypothesis testing on hyperspectral images, finding significant differences in spectra between AMD and control groups in the peripheral macula. In the absence of ground truth, i.e. molecular identification of fluorophores, we provide a rigorous validation of chosen methods on both synthetic and real images where fluorescence spectra are known. These methodologies can be applied to the study of other pathologies presenting autofluorescence that can be captured by hyperspectral imaging.

    KW - Age-related macular degeneration

    KW - Functional data analysis

    KW - Hyperspectral fluorescence microscopy imaging

    KW - Non-negative tensor decompositions

    KW - Unsupervised machine learning

    UR - http://www.scopus.com/inward/record.url?scp=85067183128&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=85067183128&partnerID=8YFLogxK

    U2 - 10.1016/j.media.2019.05.009

    DO - 10.1016/j.media.2019.05.009

    M3 - Article

    C2 - 31203169

    AN - SCOPUS:85067183128

    VL - 56

    SP - 96

    EP - 109

    JO - Medical Image Analysis

    JF - Medical Image Analysis

    SN - 1361-8415

    ER -