A Systematic Review of Artificial Intelligence Techniques for Parkinson's Disease Prediction and Diagnosis

Abstract

This is a systematic review of the use of artificial intelligence (AI) in the prediction and diagnosis of Parkinson's disease (PD) 2010−2025. We conducted a review of 75 studies, out of an original set of 1247 articles, in accordance with the Preferred Reporting Items of Systematic Reviews and Meta-Analyses guidelines, being interested in the use of machine learning and deep learning schemes in various data modalities. The highest reported classifier was the support vector machine (19.9%), then ensemble methods (18.1%), and then the Convolutional Neural Networks (CNNs) (16.9%). The best median accuracy (95.8%) was with CNNs on neuroimaging data. Voice research showed a high potential in the screening of PD in its early stages, with 89.7% sensitivity in prodromal stages. The accuracy of multimodal data integration was 4.8% higher than that of single-modality methods. Small sizes (median of 209 subjects) of datasets, a marginal outside validation (only 9.6% of the studies), and a paucity of prospective clinical examinations (3.6%) complicate promising outcomes. The future directions are to deal with multimodal integration, longitudinal disease monitoring, model interpretability, heterogeneous patient groups, and prospective clinical validation. The AI methods have a high promise of revolutionizing the management of PDs by means of earlier diagnosis, treatment according to individual preferences, and enhanced monitoring of the disease.

Received:  26 March 2025 | Revised: 9 January 2026 | Accepted: 10 March 2026

Conflicts of Interest

The authors state that they have no conflicts of interest concerning it. The authors do not have any financial, personal, or professional connections with the research, and no agency funded the research, thus having no impact on the study design, analysis, interpretation, or publication of the present manuscript.

Data Availability Statement

The present research is a systematic review, and it is not associated with the creation of new data. Each data mentioned in this paper is based on the already published and publicly available datasets. UPDRS Dataset: https://archive.ics.uci.edu/ml/datasets/parkinsons; Parkinson Speech Dataset: https://archive.ics.uci.edu/ml/datasets/Parkinsons; PPMI Database: https://www.ppmi-info.org/access-data-specimens/download-data; Gait in Parkinson’s Disease: https://physionet.org/content/gaitpdb/1.0.0/; mPower Study: https://www.synapse.org/#!Synapse:syn4993293; Oxford Parkinson’s Disease Detection Dataset: https://archive.ics.uci.edu/datasets?search=parkinson; DREAM Challenge: https://www.synapse.org/.

Author Contribution Statement

Chandrakantha Tholalemane Sathyanarayana: Conceptualization, Methodology, Investigation, Data Curating, Formal analysis, Writing – original draft, Visualizing. Basavaraj Ningappa Jagadale: Conceptualization, Methodology, Supervision, Validation, Writing – review & editing, Project administration, Resources. Madhuri Gurram Ramesh: Formal analysis, Investigation, Data curation, Writing – review & editing, Validation.