1998 Volume 57 Issue 4 Pages 389-395
Several stabilometric studies of the characteristics of labyrinthine equilibrium disturbance have been reported, but it is difficult to directly discriminate the degrees of labyrinthine damage and the stages of labyrinthine equilibrium disturbances (stages of disturbances, coordination and compensation) using stabilometric findings. The present study was designed to discriminate the degrees and stages of disturbance by neural network (NN) evaluation after learnings using measurement values of stabilograms from patients with labyrinthine disturbances. On stabilograms, envelope area, length/time, length/area, deviation on X-and Y-axes, and Romberg's coefficient were measured. Learning and evaluation by NN were performed using a program developed by Anima corporation. As the method of learning, we compiled a data file for learning and designated the composition of net, then practiced. As the method of evaluation, we compiled a data file for evaluation and evaluated the data by a weight file which was acquired in the process of learning. Learning and evaluation by NN were carried out using the following 2 supervised signals : (1) percent of canal paresis (CP%) obtained from caloric test. (2) stage of labyrinthine disturbance diagnosed by clinical findings.
Results : (1) For discrimination of the degree of labyrinthine damage stabiometric results were noncontributory. (2) Stabilograms of patients with various stages of disturbances, coordination and compensation were discriminated with a square error of 0.06. The NN was useful for discriminating the stages of labyrinthine disturbances.
