April 2009
Volume 50, Issue 13
Free
ARVO Annual Meeting Abstract  |   April 2009
A Machine Learning Approach to the Detection of Intermediate Stage of Age-Related Macular Degeneration
Author Affiliations & Notes
  • D. E. Freund
    Milton S Eisenhower Research Center, Johns Hopkins University Applied Physics Laboratory, Laurel, Maryland
  • P. M. Burlina
    Milton S Eisenhower Research Center, Johns Hopkins University Applied Physics Laboratory, Laurel, Maryland
    Computer Science Department, Johns Hopkins University, Baltimore, Maryland
  • N. M. Bressler
    Wilmer Eye Institute, Johns Hopkins University School of Medicine, Baltimore, Maryland
  • Footnotes
    Commercial Relationships  D.E. Freund, None; P.M. Burlina, None; N.M. Bressler, None.
  • Footnotes
    Support  Johns Hopkins University
Investigative Ophthalmology & Visual Science April 2009, Vol.50, 243. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      D. E. Freund, P. M. Burlina, N. M. Bressler; A Machine Learning Approach to the Detection of Intermediate Stage of Age-Related Macular Degeneration. Invest. Ophthalmol. Vis. Sci. 2009;50(13):243.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose: : To develop an automated age-related macular degeneration (AMD) diagnostic system based on fundus image analysis and machine learning algorithms initially for detection and monitoring of the intermediate stage of AMD.

Methods: : The approach differs from past drusen detection methods, and consists of posing the problem as anomaly detection (i.e. a single class classification problem) by characterizing normal fundus tissues, and by exploiting novel non-parametric anomaly detection methods. We use multiscale analysis by constructing a feature vector based on wavelet response at different scales and orientations of a fundus image. Image areas corresponding to blood vessels are automatically removed using edge detection and morphology techniques. Methods are based on Support Vector Data Description (SVDD) to carry out detection of macular anomalies. The SVDD method employs a non-linear kernel which results in a tight non-linear decision boundary. These methods do not assume any specific probability distribution models, are data driven, offer good generalization to unseen data, and result in sparse representations of the data support that minimizes over-fitting.

Results: : We performed experiments with fundus images taken from patients with no AMD and intermediate stage of AMD. Using 7 fundus images (6 with AMD), all were diagnosed correctly. A detection example is shown in Figure 1.

Conclusions: : A method based on SVDD was proposed and tested to detect the intermediate stage of AMD. The method could allow automated detection and monitoring of the intermediate stage of AMD. Figure 1: Example of an original AMD image on the left and anomalous regions found using SVDD on the right

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

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.

×