Machine Learning Applications in the Study of Parkinson’s Disease: A Systematic Review

Abstract

Background: Parkinson’s disease is a common neurodegenerative disorder that has been studied from multiple perspectives using several data modalities. Given the size and complexity of these data, machine learning emerged as a useful approach to analyze them for different purposes. These methods have been successfully applied in a broad range of applications, including the diagnosis of Parkinson’s disease or the assessment of its severity. In recent years, the number of published articles that used machine learning methodologies to analyze data derived from Parkinson’s disease patients have grown substantially. Objective: Our goal was to perform a comprehensive systematic review of the studies that applied machine learning to Parkinson’s disease data Methods: We extracted published articles in PubMed, SCOPUS and Web of Science until March 15, 2022. After selection, we included 255 articles in this review. Results: We classified the articles by data type and we summarized their characteristics, such as outcomes of interest, main algorithms, sample size, sources of data and model performance. Conclusion: This review summarizes the main advances in the use of Machine Learning methodologies for the study of Parkinson’s disease, as well as the increasing interest of the research community in this area.