The differential diagnosis of tremor is mainly based on clinical criteria.Nevertheless, these criteria are in some cases not sufficient to differentiate between different tremor forms. Long-term EMG has proven to be a valid and reliable method for the quantification of pathological tremors. The aim of the study was to develop a long-term EMG-based automated analysis procedure that separates parkinsonian tremor from essential tremor. Using longterm EMG tremor was recorded in 45 consecutive patients, 26 with Parkinson's disease (PD) and 19 with essential tremor (ET). Eight tremor parameters were generated automatically. By stepwise backward regression a subset of these criteria was extracted to achieve an automated classification of the tremor by a mathematical model. The obtained model was then tested on a new group of 13 patients in early stages of the disease. Significant differences between groups were found for tremor occurrence, tremor asymmetry, mean tremor frequency and standard deviation of phase of antagonistic muscles. Due to data overlap a classification of the two tremor forms was not possible based on a single tremor parameter. Using logistic regression, a linear formula based on the three parameters tremor occurrence, mean tremor frequency and standard deviation of phase was established and predicted the correct diagnosis in 93% of patients. The validation of the model on the new group of patients in early stages of the tremor disease yielded a correct diagnosis in 100% of cases. We conclude that long-term EMG recording allows a rater-independent classification of parkinsonian versus essential tremor.
Read full abstract