Premium
Evaluation of different automated fundus photograph analysis algorithms for detecting and grading diabetic retinopathy
Author(s) -
VALVERDE C,
GARCíA M,
HORNERO R,
DIEZ A,
LOPEZ MI
Publication year - 2010
Publication title -
acta ophthalmologica
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.534
H-Index - 87
eISSN - 1755-3768
pISSN - 1755-375X
DOI - 10.1111/j.1755-3768.2010.4117.x
Subject(s) - artificial intelligence , computer science , diabetic retinopathy , support vector machine , multilayer perceptron , pattern recognition (psychology) , radial basis function , feature selection , grading (engineering) , computer aided diagnosis , artificial neural network , logistic regression , medicine , algorithm , machine learning , diabetes mellitus , civil engineering , engineering , endocrinology
Purpose Diabetic Retinopathy (DR) is an important cause of visual impairment in industrialised countries. Automatic detection of DR primary lesions can contribute to the diagnosis and grading of the disease. The aim of this study was to develop an automated algorithm and a computer assisted diagnosis system for the detection of DR and diabetic macular edema (DME). Methods To achieve this goal, we extracted a set of colour and shape features from image regions and performed feature selection using logistic regression. Four neural network (NN) based classifiers were subsequently used to obtain the final segmentation of Hard exudates, Microaneurysms and Haemorraghes; multilayer perceptron (MLP), radial basis function (RBF), support vector machine (SVM) and a combination of these three NNs using a majority voting (MV) schema. The database was composed of 117 images. Results Attending to performance and complexity criteria, the best results for primary lesions were obtained for RBF. Using an image‐based criterion, a mean sensitivity of 100%, mean specificity of 70.4% and mean accuracy of 88.1% were achieved for hard exudates and a mean sensitivity of 100%, and mean accuracy of 83.08% for haemorraghes. Conclusion This study shows that algorithms based in NNs for computer‐aided diagnosis are a reliable alternative to time‐consuming manual screening of diabetic retinopathy.