A comparison of the different algorithms for essential tremor and parkinson s disease tremor differentiation based on hand tremor
Loading...
Date
2018-05
Authors
Boey, Keen Huang
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Essential tremor (ET) and Parkinson’s disease tremor (PD) are the two most
common types of tremor. Misdiagnosis of among these two groups of tremors often
occurs when using clinical observations, due to overlapping of symptoms between ET
and PD at the early stage of disease. To assist specialist in making decisions when
diagnosing the tremor, a tremor monitoring and differential diagnosis system is
implemented using a wireless inertia measurement unit, to collect hand tremor data from
patients and perform classifications of tremor.
Four different types of tremor classification algorithms, namely the Tremor
Stability Index (TSI), Mean Harmonic Peak Power (MHPP), Relative Energy (RE) and
Empirical Mode Decomposition – Singular Value Decomposition (EMD-SVD) analysis
had been tested with 153 postural tremor recordings and 154 rest tremor recordings
collected from ET and PD patients. A tremor detection algorithm based on tremor
intensity, dominant frequency and length of tremor had been used to select recordings
significance tremor of ET and PD for analysis. 43 postural recordings (ET, n= 8 and PD,
n=35) and 43 rest tremor recordings (ET, n=4 and PD, n=39) had been selected. The
distribution of features extracted from each algorithm was tested with Mann Whitney U
test, and the sensitivity, specificity and accuracy for each algorithms in correctly classify
ET patients were analysed using receiver operating curves (ROC).
The results had shown that are distinct differences between the distributions of
MHPP (p=0.001) and RE (p=0.019) among ET and PD. The ROC results had showed that
the MHPP had the highest accuracy in classify ET and PD (85.7%), with sensitivity and
specificity of 88.6% and 87.5% respectively.