Harmonics in frequency spectra: a contribution to differential diagnosis of tremor

Introduction: For quantification and differential diagnosis of tremor, electromyography (EMG) and accelerometry can be used supplementary to clinical assessment. Accelerometry essentially provides information regarding tremor amplitude and frequency. As there is a significant overlap in the frequencies of different tremor species, this parameter, however, is of limited value for specific diagnosis. EMG provides information about the patterns of agonistic and antagonistic muscle activity (alternating or co-contraction), but the specificity of this procedure is insufficient for individual diagnosis.

For these reasons, new approaches focus on more sophisticated analyses. Time series analysis reported here are particularly promising.

Objective: In the present work we analysed whether harmonics in the tremor spectra are of differential diagnostic value.

Methods: Accelerometry signals in parkinsonian tremor (PT, n=32) and essential tremor (ET, n=15) were compared. Accelerometers were employed to measure resting tremor and postural hand tremor for 32 second epochs while the patient was sitting in a chair with armrest, and the results for the tremor-dominant side assessed by means of spectral analysis (Hamming window).

Results: PT exhibited a higher number of harmonics in fast Fourier transformation (FFT) spectra than did ET. Spectral analysis improved the selectivity of this parameter for tremor type (p<0.05). This difference in number of harmonic peaks between PT and ET is due to a striking deviation of acceleration signals from the sinusoidal in PT that is not evident in the displacement time series.

The number of harmonics proved to be a more selective parameter for tremor type than frequency or amplitude. Multivariate analysis (discriminant analysis) with frequency and harmonic number as factors allowed an almost complete separation of ET and PT.

Conclusion: These results are in accordance with the hypothesis of a more complex central oscillator in at least a subgroup of PT patients. Our findings are consistent with those of Deuschl et al. (1995), who, employing a different approach, also identified features of PT time series not consistent with a simple sinusoidal pattern. Nevertheless, more and rarer tremor types should be included in further studies to allow a practical contribution of complex parameter analysis to problems of clinical differential diagnosis.

Keywords: Harmonics, tremor spectra, tremor, accelerometry, differential diagnosis