Automatic generation of probability maps of retinal lesions

Author

Rudas, Jorge ; Toscano, Ricardo ; Sanchez, G.T.

Author_Institution

Ingeniero de Sist., Univ. del Magdalena (Colombia, Santa Marta, Colombia

fYear

2012

fDate

12-14 Sept. 2012

Firstpage

7

Lastpage

12

Abstract

The aim of this paper is the automatic generation of probability lesion maps from a color retinography analysis. A probability map determines the degree of belonging of each pixel in a given set. This process is based on three sequences of procedures: A set of attributes (contrasts enhancement zones), a suppression of artifacts obstacles and a setting of value of each pixel belong to a bright lesion, using probabilistic mapping by supervised classification. The validation of the results was carried out by comparing the subsequent binary map with prior probabilistic diagnosis. The results were analyzed by mean square error (MSE) between resulting diagnosis and diagnostic specialist. MSE was reached of 1.22 on a set of 40 images DIARECTDB1 repository.

Keywords

eye; image colour analysis; mean square error methods; medical image processing; probability; DIARECTDB1 repository; artifacts obstacles; automatic generation; binary map; color retinography analysis; contrasts enhancement zones; mean square error; probabilistic diagnosis; probabilistic mapping; probability lesion maps; probability maps; retinal lesions; supervised classification; Diabetes; Image color analysis; Image segmentation; Lesions; Retina; Silicon compounds; Visualization; Retinography; bright lesions; diabetic retinopathy; exudates;

fLanguage

English

Publisher

ieee

Conference_Titel

Image, Signal Processing, and Artificial Vision (STSIVA), 2012 XVII Symposium of

Conference_Location

Antioquia

Print_ISBN

978-1-4673-2759-6

Type

conf

DOI

10.1109/STSIVA.2012.6340548

Filename

6340548