On the diagnosis of idiopathic Parkinson’s disease using continuous wavelet transform complex plot

Abstract

Gait analysis provides valuable motor deficit quantitative information about Parkinson’s disease patients. Detection of gait abnormalities is key to preserving healthy mobility. The goal of this paper is to propose a novel gait analysis and continuous wavelet transform-based approach to diagnose idiopathic Parkinson’s disease. First, we eliminate the noise resulting from orientation changes of test subjects by filtering the continuous wavelet transform output below 0.8 Hz. Next, we analyze the complex plot output above 0.8 Hz, which takes an ellipse, and calculate the area using $$95\%$$ confidence level. We found out that this ellipse area, along with the mean continuous wavelet transform output value, and the peak of the temporal signal are excellent features for classification. Experiments using Artificial Neural Networks on the Physionet database produced an accuracy of $$97.6\%$$ . Furthermore, we have shown an association between the Parkinson’s disease severity stage and the ellipse complex plot area with a 97.8% overall accuracy. Based on the results, we could effectively recognize the gait patterns and distinguish apart Parkinson’s disease patients with varying severity from healthy individuals.