Discrimination of Stabilograms by a Neural Network in Patients with Equilibrium Disturbances

This study attempted to discriminate stabilograms using a neural network that can learn with experience in patients with vertigo and equilibrium disturbances.Stabilograms obtained from 60 patients with disorders of the labyrinth, the cerebellum and the basal ganglia were used in this study. Area, length per second, length per unit area, displacements in X- and Y-axes and Romberg ratio of the stabilograms were used as input signals. The learning of the network was carried out with a specially designed program.1 Stabilograms of patients with disorders of the labyrinth and the central nervous system were discriminated with an square error margin of 3.36E-02.2 Stabilograms of patients with bleeding in the cerebellar vermis, bleeding or infarction of the cerebellar hemisphere and spinocerebellar degeneration were discreminated with an square error margin of 4.98E-03.3 Stabilograms of patients with positive phenomena, negative phenomena a and b in Parkinson disease were discriminated with an square error margin of 4.98E-03.Neural networks obtained from learning using stabilograms were useful for discrimination of lesion sites and pathophysiologies of various diseases.