Distinguishing between physiological and essential tremor using discriminant and cluster analyses of parameters derived from the frequency spectrum

Abstract

Three parameters, root mean square (rms) acceleration, frequency and shape, were calculated from the frequency spectrum in 20 patients with essential tremor. Tremor was measured in two postures (with the arms outstretched – `postural' tremor and with the arm supported and hand relaxed – `rest' tremor) in the dominant and non-dominant hands using a piezo-electric accelerometer. These data were compared with data obtained from accelerometric measurements of physiological tremor in 81 adults. Step-wise discriminant analysis was applied to the data to identify which variables best discriminated between essential and physiological tremor. K-means cluster analysis was used to identify any sub-groups within the essential tremor data. The analysis identified four variables, which together, best distinguished between physiological and essential tremor, these were postural amplitude and frequency, and rest shape (non-dominant hand) and postural shape (dominant hand). Using these variables 94.74% of cases were correctly classified into the physiological and essential tremor groups. Cluster analysis on the amplitude values for postural tremor (but not for the other parameters or postures measured) suggested that two sub-groups were present in this sample of essential tremor patients. Discriminant analysis and K-means cluster analysis on parameters derived from the frequency spectrum of tremor measured in different postures can be used to distinguish pathological tremors from physiological tremor and provide a way of identifying possible sub-groups in essential tremor.