In this paper a contribution for supporting diabetic symptoms detection in retinal images is proposed by synthesizing a Cellular Neurofuzzy Network able to provide informations on vague pale regions of fundus images with suspect diabetic damages. After highlighting pale regions in input images by an Intensity Difference Map Evaluation, their contrast is enhanced by means of a CNN-based Fuzzy Subnet. After an adaptive thresholding evaluation, contrast-enhanced images with bimodal histograms are globally segmented by a CNN-based subsystem, providing binary output images, in which suspect diabetic areas are easily isolated. Performances are evaluated by means of the Correct Recognition Rate, which provides percentage measures of exactness in the detection of suspect damaged areas. Results are discussed and compared with other researchers' ones.
A Cellular Neurofuzzy Network for Supporting Detection of Diabetic Symptoms in Retinal Images / Carnimeo, Leonarda; Giaquinto, A. - In: 2007 International Symposium On Signals, Circuits and Systems, ISSCS 2007; Iasi; Romania; 12-13 July 2007[s.l], 2007. - ISBN 978-1-4244-0968-6. - pp. 257-260 (( convegno 2007 IEEE Int. Symposium on Signals, Circuits and Systems tenutosi a Iasi nel July 12-13 2007 [10.1109/ISSCS.2007.4292700].
A Cellular Neurofuzzy Network for Supporting Detection of Diabetic Symptoms in Retinal Images
CARNIMEO, Leonarda;
2007-01-01
Abstract
In this paper a contribution for supporting diabetic symptoms detection in retinal images is proposed by synthesizing a Cellular Neurofuzzy Network able to provide informations on vague pale regions of fundus images with suspect diabetic damages. After highlighting pale regions in input images by an Intensity Difference Map Evaluation, their contrast is enhanced by means of a CNN-based Fuzzy Subnet. After an adaptive thresholding evaluation, contrast-enhanced images with bimodal histograms are globally segmented by a CNN-based subsystem, providing binary output images, in which suspect diabetic areas are easily isolated. Performances are evaluated by means of the Correct Recognition Rate, which provides percentage measures of exactness in the detection of suspect damaged areas. Results are discussed and compared with other researchers' ones.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
