Gisoo Fathi; Peyvand Ghaderyan
Volume 15, Issue 2 , August 2021, , Pages 161-174
Abstract
Parkinson’s Disease (PD) is one of the most common neurodegenerative diseases that cause abnormal gait patterns by affecting central nervous system. Since this disease is incurable, the reliable diagnosis can lead to slowing disease progression, reducing the risk of physical injuries and improving ...
Read More
Parkinson’s Disease (PD) is one of the most common neurodegenerative diseases that cause abnormal gait patterns by affecting central nervous system. Since this disease is incurable, the reliable diagnosis can lead to slowing disease progression, reducing the risk of physical injuries and improving the quality of patient's life. In this regard, the development of fast, cost-effective and reliable detection systems is essential. This study has therefore proposed a detection method using vertical ground reaction force signals, which provide a non-invasive and useful index of the motor control function. It is based on generalized singular value decomposition, K-Nearest Neighbor (KNN) and Probabilistic Neural Network (PNN). The performance of the algorithm has been evaluated by gait signal of 93 individuals with PD and 73 healthy controls. The results have demonstrated that the proposed new symmetric feature is able to achieve 96.19% and 95.67% accuracy rates, 97.22% and 93.35% sensitivity rates, 95.02% and 97.33% specificity rates using the KNN and PNN classifiers, respectively. Furthermore, average accuracy rates of 98.23% and 98.51%, sensitivity rates of 93.5% and 100%, specificity rates of 100% and 96.53% have been obtained for stage classification using these two classifiers. The obtained high average accuracy rates have confirmed the promising capability of the proposed non-invasive and cost-effective method in PD detection and stage classification, which makes it suitable for clinical applications.