Analysis of Parkinson's Disease Gait using Computational Intelligence

Abstract

Millions of individuals throughout the world are living with Parkinson’s disease (PD), a neurodegenerative condition whose symptoms are difficult to differentiate from those of other disorders. Freezing of gait (FOG) is one of the signs of Parkinson’s disease that have been utilized as the main diagnostic factor. Bradykinesia, tremors, depression, hallucinations, cognitive impairment, and falls are all common symptoms of Parkinson’s disease (PD). This research uses a dataset that captures data on individuals with PD who suffer from freezing of gait. This dataset includes data for medication in both the “On” and “Off” stages (denoting whether patients have taken their medicines or not). The dataset is comprised of four separate experiments, which are referred to as Voluntary Stop, Timed Up and Go (TUG), Simple Motor Task, and Dual Motor and Cognitive Task. Each of these tests has been carried out over a total of three separate attempts (trials) to verify that they are both reliable and accurate. The dataset was used for four significant challenges. The first challenge is to differentiate between people with Parkinson’s disease and healthy volunteers, and the second task is to evaluate effectiveness of medicines on the patients. The third task is to detect episodes of FOG in each individual, and the last task is to predict the FOG episode at the time of occurrence. For the last task, the author proposed. a new framework to make real-time predictions for detecting FOG, in which the results demonstrated the effectiveness of the approach. It is worth mentioning that techniques from many classifiers have been combined in order to reduce the likelihood of being biased toward a single approach. Multilayer Perceptron, K-Nearest Neighbors, random Forest, and Decision Tree Classifier all produced the best results when applied to the first three tasks with an accuracy of more than 90% amongst the classifiers that were investigated.