Machine learning and deep learning approach to Parkinson’s disease detection: present state-of-the-art and a bibliometric review

Parkinson’s disease (PD) is fatal, severe and irreversible neurological disorder. With the aging of the world population, the prevalence of PD is on the rise, making it the second most common neurodegenerative condition after Alzheimer's. Early recognition can be tricky because the disease's phenotype is essential for anticipating and halting its course. If PD is detected early and accurately, medical therapies may be initiated that may enhance the patient's quality of life and minimize its course. Medical image analysis, diagnostics, and medical management has been greatly improved with Machine Learning (ML) and Deep Learning (DL) based models. Our research analyzed present state-of-art and also the bibliometric review regarding ML and DL techniques used for Parkinson’s Disease Detection (PDD) in order to measure the effectiveness of these technologies. This study analyses the various algorithms, modalities used, their limitations and possible future direction of ML/DL in PDD and at the same time, presents the quantitative analysis of publications in the mentioned research area. We have searched the Scopus database using major and auxiliary keywords pertaining to PD detection using ML and DL research articles published between 2013 and 2023. Based on the keyword input, 1674 articles were returned for bibliometric analysis. In order to visualize and analyze bibliometric networks, software tools such as VOSviewer and Biblioshiny are being used. Additionally, the article offers perceptions on networks of collaboration and bibliographic coupling. Our study will help researchers to easily identify techniques being used, relationships between publications and authors, uncovering new research opportunities and knowledge in ML/DL based PDD.