May 2005
Volume 46, Issue 13
Free
ARVO Annual Meeting Abstract  |   May 2005
Elaboration of a Diabetic Retinopathy Images Data Base for the Test of New Computer Algorithms of Content Base Image Retrieval
Author Affiliations & Notes
  • B. Cochener
    Ophtalmologie, CHU Morvan, Brest, France
  • W. Daccache
    Ophtalmologie, CHU Morvan, Brest, France
  • C. Cochard
    Ophtalmologie, CHU Morvan, Brest, France
  • G. Cazuguel
    Inserm U650, Latim, Brest, France
  • M. Lamard
    Inserm U650, Latim, Brest, France
  • Footnotes
    Commercial Relationships  B. Cochener, None; W. Daccache, None; C. Cochard, None; G. Cazuguel, None; M. Lamard, None.
  • Footnotes
    Support  None.
Investigative Ophthalmology & Visual Science May 2005, Vol.46, 351. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      B. Cochener, W. Daccache, C. Cochard, G. Cazuguel, M. Lamard; Elaboration of a Diabetic Retinopathy Images Data Base for the Test of New Computer Algorithms of Content Base Image Retrieval . Invest. Ophthalmol. Vis. Sci. 2005;46(13):351.

      Download citation file:


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

      ×
  • Supplements
Abstract

Abstract: : Purpose:Create a data base of photographic and angiographic images of diabetic retinopathy manually classified with the aim of validating new algorithms of content base image retrieval. Methods:The photographic and angiographic images of the diabetic patients recruited in the ophthalmology department of Brest teaching hospital since January 2003 are studied and classified by expert doctors according to the criteria of classification of «International Clinical Diabetic Retinopathy Disease Severity Scale» thus allowing the creation of a computerized data base. The techniques of images indexation by digital contents use the contribution of JPEG2000 wavelet transform–based compression standard. Multiple coefficients of the image decomposition in the transformed space generate a unique digital signature for every image. Measure of distances between all the signatures allows finding the similar images. The manual semantic indexation allows quantifying the quality of the automatic algorithm. Results:The standard of compression JPEG2000 offers numerous possibilities of calculation for signature and measure. The signature using the "bit planes" with a dedicated measure is the most adapted to our searches notably in the detection of neovessels. The results of this study show that the reliability of retrieval of retinal images varies with the used algorithm and with the lesion type. The number of images of the data base remains however limited to cover all the nuances and the combinations between lesions. Conclusions:The data base containing this day 1800 indexed images allows the tests of numerous algorithms. The first results of these computer methods are promising and let consider a new way in the automatic indexation of retinal images, first stage in the elaboration of helping tools for the diagnosis of severity of diabetic retinopathy.

Keywords: diabetic retinopathy • 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.

×