June 2015
Volume 56, Issue 7
Free
ARVO Annual Meeting Abstract  |   June 2015
Evaluation of Fundus Image Illumination Normalization in Longitudinal Diabetic Retinopathy Screening
Author Affiliations & Notes
  • Kedir Adal
    Rotterdam Ophthalmic Inst, Rotterdam, Netherlands
    Imaging Physics, Delft University of Technology, Delft, Netherlands
  • Peter G. Van Etten
    Rotterdam Eye Hospital, Rotterdam, Netherlands
  • Jose P. Martinez
    Rotterdam Eye Hospital, Rotterdam, Netherlands
  • Lucas J. van Vliet
    Imaging Physics, Delft University of Technology, Delft, Netherlands
  • Koenraad Arndt Vermeer
    Rotterdam Ophthalmic Inst, Rotterdam, Netherlands
  • Footnotes
    Commercial Relationships Kedir Adal, None; Peter Van Etten, None; Jose Martinez, None; Lucas van Vliet, None; Koenraad Vermeer, None
  • Footnotes
    Support None
Investigative Ophthalmology & Visual Science June 2015, Vol.56, 5249. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Kedir Adal, Peter G. Van Etten, Jose P. Martinez, Lucas J. van Vliet, Koenraad Arndt Vermeer; Evaluation of Fundus Image Illumination Normalization in Longitudinal Diabetic Retinopathy Screening. Invest. Ophthalmol. Vis. Sci. 2015;56(7 ):5249.

      Download citation file:


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

      ×
  • Supplements
Abstract
 
Purpose
 

To evaluate if fundus images that are normalized for variations in luminance and contrast improve the sensitivity of diabetic retinopathy (DR) related lesion detection and the efficiency of longitudinal screening for DR and diagnosis.

 
Methods
 

Four-field fundus image sets captured from either eye of 30 diabetic patients who had retinal exams in 2012 and 2013 as part of an ongoing diabetic retinopathy screening program in the Rotterdam Eye Hospital were included. The fundus images were normalized for local variations in illumination, yielding uniform luminance and considerably increased contrast, especially near the edges of the fundus images (Adal et. al. WBIR 2014). Image sets of both exams were registered to a common coordinate system after which color and normalized mosaics (fig. 1) were produced for each retinal exam (Adal et. al., ARVO Abstract 444, 2014).<br /> The fundus mosaics of each eye were shown to two expert graders to examine early DR symptoms. DR-related retinal abnormalities were then annotated on the color mosaics and, in a second round, on the normalized fundus mosaics. The ground truth was defined as all abnormalities that were annotated by both graders in either mosaic, provided that at least one of the graders identified the abnormalities in the color fundus mosaic.

 
Results
 

30 color and 30 normalized fundus mosaics were independently annotated for DR-related lesions and annotated by two expert graders. The results are summarized in Table 1. A far higher sensitivity was found for both experts for normalized mosaics than for color mosaics (97.2% vs 62.5%), at the cost of a somewhat higher number of false positives per eye (6.0 vs 0.7). On average, the experts annotated 1.7 and 6.2 abnormalities per eye in the color and normalized mosaics, respectively.

 
Conclusions
 

Normalization of fundus images improved the sensitivity of longitudinal screening for diabetic retinopathy. The normalized fundus image enhances the visibility of retinal features and the detection of lesions such as microaneurysms and hemorrhages.  

 
Fig 1. (a) Example annotations in a color and a normalized fundus mosaic. Some lesions are missed in the color images but are detected in the normalized fundus images. (b) Example in which several abnormalities were found on the normalized mosaic although none of the graders found an abnormality on the color mosaic.
 
Fig 1. (a) Example annotations in a color and a normalized fundus mosaic. Some lesions are missed in the color images but are detected in the normalized fundus images. (b) Example in which several abnormalities were found on the normalized mosaic although none of the graders found an abnormality on the color mosaic.
 

 
×
×

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.

×