A hybrid time-wavelet domain approach for early-stage diagnosis of Parkinson’s disease using multichannel EEG signals

Work under submission, more details will be shared once published.

A hybrid time-wavelet domain approach for early-stage detection of PD using inter-channel self-similarity and multi-resolution features extracted in the time domain and wavelet domain respectively has been proposed in this paper. In order to extract inter-channel self-similarity, a set of correlation coefficients is extracted in the time domain. Renyi entropy and Kraskov entropy features are extracted in the wavelet domain. Optimal biorthogonal wavelet filter bank (OBWFB) has been used to carry out. The wavelet transform allows the utilization of the predominant quadrature bandpass frequency components, resulting in the enhancement of the entropy features. In this paper, we have examined each feature as well as the features in conjunction. The classification is carried out using the support vector machine of polynomial degree 3 (CSVM). 10-fold cross-validation is exercised to reduce the overfitting phenomenon. The proposed method provides excellent results with accuracy, specificity and sensitivity as 99.56%, 99.24% and 99.87% respectively.

Additional Materials: to be updated soon…