Parkinson’s disease (PD) is a progressive condition of the brain and central nervous system that causes a gradual deterioration in body movement. Major symptoms of PD are stiffness in the legs, arms, and torso; tremor, slow movement, balance problems, depression, etc. PD can be diagnosed based on the above-mentioned common symptoms. Yet, diagnostic imaging techniques that are noninvasive, such as Positron Emission Tomography (PET) can help the doctors to detect PD. Every year, an estimated 60,000 new cases of PD are diagnosed. Numerous machine learning (ML) and deep learning (DL) approaches have been suggested in the literature for the automated detection of PD. The accurate diagnosis of PD poses a challenge due to the absence of a standardized and objective approach. The integration of ML algorithms into medical diagnostics has enhanced the precision of disease predictions, making the diagnosis more effective.

This paper presents a systematic review of automatic PD detection used machine ML and DL, from papers published between 2010 and 2024. Through the extensive and careful search procedure, 67 papers were selected from a total of 262. We retrieved these articles from various academic databases such as Google Scholar, IEEE Xplore and Scopus ensuring detailed mention of relevant literature.

Through this review, we can understand how ML techniques can be a partner to make the automatic detection of PD faster and efficient. We also addressed the different public database sources of PD such as UCI Parkinson’s Dataset, Parkinson’s Drawing Dataset etc. so that the researchers in this field can easily use these datasets. Furthermore, we identified some benefits, limitations and gaps, which should be addressed in the future.

In conclusion, the field of automatic detection of PD has witnessed remarkable advancements through the integration of machine learning, sensor technologies and various techniques. The reviewed literature highlights the efficacy of various classification algorithms such as Support Vector Machines (SVM), Random Forests (RF), etc., with them consistently demonstrating superior accuracy in distinguishing individuals with PD.