Analysis of pathological tremors using the autoregression model.

Abstract

The usefulness of analysis of acceleration data using an autoregression model (AR) for differential diagnosis of Parkinson's disease and other diseases with tremors was investigated. The order of the AR model used in this study was 7, in accordance with Akaike's final prediction error criterion. The subjects included 19 patients with Parkinson's disease; 21 patients with essential tremor, which mainly appears in old people, as well as Parkinson's disease; and 13 healthy old people as a control group. The results of analysis of acceleration data showed that the first prediction coefficient, just as the main tremor frequency, was a useful parameter for differentiating patients in the Parkinson's disease patient group and essential tremor patient group. The seventh prediction coefficient was found to be a useful parameter for distinguishing pathological tremors observed in Parkinson's disease and essential tremor disease from physiological tremors observed in healthy people. Although the usefulness of other prediction coefficients for differential diagnosis of Parkinson's disease and other diseases with tremors has not yet been clarified, the results of this study showed that information obtained from AR model parameters in addition to information on main tremor frequency is useful for the diagnosis of Parkinson's disease.