May 2008
Volume 49, Issue 13
Free
ARVO Annual Meeting Abstract  |   May 2008
Application of Independent Component Analysis to Classify Non-Exudative Amd Phenotypes Quantitatively From Vision-Based Derived Features in Retinal Images
Author Affiliations & Notes
  • B. Davis
    VisionQuest Biomedical, Albuquerque, New Mexico
  • S. R. Russell
    Ophthalmology,
    University of Iowa, Iowa City, Iowa
  • M. D. Abramoff
    Ophthalmology, University od Iowa, Iowa City, Iowa
  • T. Scheetz
    University of Iowa, Iowa City, Iowa
  • S. Murillo
    VisionQuest Biomedical, Albuquerque, New Mexico
  • P. Soliz
    VisionQuest Biomedical, Albuquerque, New Mexico
  • Footnotes
    Commercial Relationships  B. Davis, VisionQuest Biomedical, I; VisionQuest Biomedical, E; S.R. Russell, None; M.D. Abramoff, None; T. Scheetz, None; S. Murillo, VisionQuest Biomedical, E; P. Soliz, VisionQuest Biomedical, I; VisionQuest Biomedical, E.
  • Footnotes
    Support  University of Iowa, Dept. of Ophthalmology unrestricted funds., VisionQuest Biomedical corporate support
Investigative Ophthalmology & Visual Science May 2008, Vol.49, 887. doi:https://doi.org/
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      B. Davis, S. R. Russell, M. D. Abramoff, T. Scheetz, S. Murillo, P. Soliz; Application of Independent Component Analysis to Classify Non-Exudative Amd Phenotypes Quantitatively From Vision-Based Derived Features in Retinal Images. Invest. Ophthalmol. Vis. Sci. 2008;49(13):887. doi: https://doi.org/.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Purpose: : To apply computer-based feature detection to digital images for phenotyping of patients with non-exudative (NE) age-related macular degeneration (AMD).

Methods: : The dataset consisted of 100 digital images from 100 sequential patients with NE AMD. Images were scanned from 35mm slides at 2400x2400 pixels and centered on the fovea. The digital images were visually sorted into 12 fundus phenotypes, based upon macular pattern. A independent component analysis (ICA) algorithm was trained using the ‘leave-one-out’ method, from the two categories, temporal drusen phenotype (TDP, >85% drusen temporal to the fovea), and small, distributed drusen phenotype (SDDP, drusen < 63 microns, no radial inhomogeneity). To test whether ICA could mimic human visual phenotyping, five images from the SDDP and six images from the TDP were evaluated in masked fashion. Independent components (ICs or features) were collected from two regions of the image (512x512 pixels centered on fovea, 512x512 pixels 200 pixels temporal to fovea) for each of the images in the two phenotype classes.

Results: : Two feature scales (6x6 and 12x12 pixels) were found to maximize separation of the two phenotypes based on 32 ICs for each scale. A regression analysis using partial least squares of the 64 features showed perfect classification, in other words 100% accuracy in replicating the human-based classification of phenotypes. Using only either set of features alone produced an 85% accuracy.

Conclusions: : These results suggest that visually recognized fundus patterns (fundus phenotypes) of NE AMD can be detected and defined using computer vision-based methods. ICA demonstrates high accuracy in sorting fundus patterns in digital fundus images and can objectively categorize fundus phenotypes when the phenotypes have been defined based on subjectively chosen representative exemplars. Additional study will be required to determine whether these or other subjectively or objectively defined fundus phenotypes are associated with visual prognosis or genetically meaningful relationships in AMD.

Keywords: age-related macular degeneration • imaging/image analysis: non-clinical • genetics 
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×