Multivariate statistical analysis in stabilometry in human upright standing (second report) —Pattern recognition of a stabilogram—

Abstract

A pattern of a figure which is displayed by an X-Y recorder or by a polygraph in stabilometry may present several qualitative facts about the standing ability of a test-subject. The purpose of this study was to make a quantitative evaluation of the body stability in both normal and ataxic patients who were standing erect. The technique utilised was pattern-recognition, composed of prinpal component analysis and discriminant analysis. Stabilometrical examinations were carried out in 20 normal subjects, 39 with peripheral vestibular disturbance and 11 with disturbance of the central origin. Each examinee was ordered to keep an upright posture for 60 seconds both with the eyes open and with the eyes closed. The following parameters were used : the area of the ellipse for rejection by a statokinesigram, the maximum width and the total length of body excursion, the root mean square, the velocity, the acceleration and the average frequency in the horizontal plane. These were the same parameters that were used in the first report. Each component predicting a specific character in stabilometry was identified by this study of principal component analysis. The relationship between the components in order and its prediction were as follows : the first component predicted the size of the body movement, the second the force, the third the density, the fourth the difference between body movement with the eyes open and with the eyes closed and the fifth the direction of body sway. Furthermore, the principal components were successfully utilized to allow the classification of the test subjects into three groups : those who were normal, those with peripheral vestibular disturbance and those with disturbance of the central origin. Discriminant functions were also used to classify the above mentioned three groups and another five groups, namely : those who were normal, those with vestibular neuronitis, those with positional vertigo of the benign paroxysmal type, those with Meniere's disease and those with cerebellar ataxia. In classification of the three groups, the significant parameters used for discrimination were as follows (P<0.05) : the root mean square with the eyes open, the distance and the maximum width of body excursion in the Y direction with the eyes open, the area of the ellipse for rejection both with the eyes open and with the eyes closed, the maximum width and the average frequency in the X direction with the eyes closed. On the contrary, the significant parameters used to classify the other five groups were as follows (P<0.05) : the length of excursion, the acceleration and the average frequency in the Y direction with the eyes open, the maximum width in the X direction with the eyes open, the root mean square in both the X and Y direction and the root mean square with the eyes closed. These discriminant functions may be available in differential diagnosis since they had a specificity and sensitivity of sufficiently high grade to classify both the three groups and the five groups, mentioned above. In conclusion, stabilograms were successfully classified into either three or five groups and they were discriminated quantitatively against each other by using discriminant functions. © 1989, The Oto-Rhino-Laryngological Society of Japan, Inc. All rights reserved.