Purchase this article with an account.
SA Barman, A Toniappa, AR Fielder, MJ Moseley; Application of an Adaptive Local Contrast Enhancement Technique to Retinal Images of Premature Infants Taken With the Retcam 120TM Digital Fundus Camera . Invest. Ophthalmol. Vis. Sci. 2002;43(13):1254.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Purpose:To apply and assess the usefulness of an adaptive local contrast enhancement technique when applied to fundus images of premature infants with respect to the clinical assessment of the condition retinopathy of prematurity (ROP). Methods:We applied an adaptive local contrast enhancement algorithm based on the local statistics of a fundus image of a premature infant. The technique has been shown previously to be successful when applied to adult retinal images. We applied the algorithm with a patch size of between 10 and 150 pixels and minimum variance values between 8 and 64. The effect in terms of clinical usefulness of the local contrast enhancement technique was compared against the standard technique of histogram equalisation. Results:A data set of 30 images, showing a full range of image quality and image type was used to test the algorithm. The data set ranged from images showing no ROP to those exhibiting advanced ROP. The parameters to control the enhancement features of the algorithm, were judged on their effect on clinical usefulness in relation to diagnosis of the condition. A patch size of 60 pixels and a minimum variance of 32 were judged by clinicians expert in the field to be optimal parameters on all 30 images. The local enhancement technique was judged by clinicians to be more effective than standard histogram equalisation in increasing clinical usefulness in all 30 images, particularly around the ROP ridge area at the periphery of the retina. Conclusion:The application of adaptive local contrast enhancement is a reliable technique to improve the clinical usefulness of ROP images. The algorithm performs well, using identical patch size and variance parameters, across the wide range and quality of images typical of preterm infant fundus images.
This PDF is available to Subscribers Only