Diagnosis of Parkinson’s Disease Using Empirical Mode Decomposition of Ground Reaction Force Signal

Abstract:
Parkinson’s disease is a neurodegenerative disorder characterized by a decrease in dopamine secretion and alterations in the basal ganglia, which are likely contributors to motor dysfunction. The condition is associated with symptoms such as limb tremors, imbalance during ambulation, muscle rigidity, and various motor difficulties. As the disease progresses, these symptoms typically worsen over time. Therefore, early diagnosis is crucial, as timely intervention can significantly manage symptoms and enhance the patient’s quality of life. Traditional diagnostic methods, including brain imaging, speech analysis, and handwriting assessment, require advanced equipment and complex processing. In contrast, we utilize ground reaction force (GRF) signal data collected from instrumented footwear for the diagnosis of Parkinson’s disease, leveraging the accessibility and cost-effectiveness of wearable technology. This approach enables real-time monitoring of walking patterns and provides insights for improving mobility and rehabilitation in individuals with motor impairments. We employed the empirical mode decomposition (EMD) technique to extract the intrinsic functions of the GRF signal. Subsequently, the signal’s domain information is obtained through the Hilbert transform. By employing a range of nonlinear bivariate and univariate features, including sample entropy and three statistical features, we quantify the differences in walking signals between healthy individuals and those with Parkinson’s disease. Finally, a sparse coding method is applied for classification and disease detection. The results indicate that the proposed algorithm, utilizing a minimal number of foot sensors, achieves a disease detection accuracy of 91.88%.