INTERPRETATION OF AUTOMATED PERIMETRY FOR GLAUCOMA BY NEURAL-NETWORK

Purpose. Neural networks were trained to interpret the visual fields from an automated perimeter. The authors evaluated the reliability of the trained neural networks to discriminate between normal eyes and eyes with glaucoma. Methods. Inclusion criteria for glaucomatous and normal eyes were the intraocular pressure and the appearance of the optic nerve; previous visual fields were not used. The authors compared the backpropagation learning method used by automated neural networks to those used by two specialists in glaucoma to classify the central 24 degrees automated perimetric visual fields from 60 normal and 60 glaucomatous eyes. Results. The glaucoma experts and a trained two-layered network were each correct at approximately 67%. The average sensitivity of this test was 59% for the two glaucoma specialists and 65% for the two-layered network. The corresponding specificities were 74% and 71% for the specialists and the two-layered network, respectively. The experts and the network were in agreement about 74% of the time, which indicated no significant disagreement between the methods of testing. Feature analysis with a one-layered network determined the most important visual field positions. Conclusions. The authors conclude that a neural network can be taught to be as proficient as a trained reader in interpreting visual fields for glaucoma.