Use of neural networks for analysis of visual evoked potentials developing a type of "pattern"--preliminary report
Sobolewski, P.; Swiercz, M.; Grusza, M.; Stankiewicz, A.
Klinika Oczna 99(3): 161-164
1997
ISSN/ISBN: 0023-2157 PMID: 9456560 Document Number: 472090
The authors present the "pattern" visual evoked potentials (VEP) analysis with use of the artificial neural networks (ANN). The study involved 11 patients with compressive chiasmal optic neuropathy, 20 patients with optic neuritis, 12 patients with anterior ischaemic optic neuropathy, 20 patients with optic nerve atrophy from neuritis, 8 patients with demyelinative neuropathy, 5 patients with oedema optic nerve, 20 healthy persons. The tests of visual evoked potentials were performed with the use of computer system UTAS-E1000. Classification of potentials was made by correlation of outputs of ANN with results of confirmed neuro-ophthalmology conditions. ANN of different architecture were classified correctly in 80-100% of VEP record samples. The obtained correctness of classification confirms usefulness of VEP analysis as the objective diagnostic method in some neuro-ophthalmological diseases and indicates application of ANN in multifactor analysis.