Contrastive Self-Supervised Learning for Parkinson’s Disease Classification and A Comparative Evaluation with Supervised Deep Models

DOI

10.2991/978-94-6239-616-6_79How to use a DOI?

Keywords

Parkinson’s Disease Classification; Self-Supervised Learning (SSL); Supervised Deep Learning (SDL); Contrastive Learning; Biomedical Data Analysis

Abstract

Accurately identifying Parkinson’s disease and distinguishing between its variations remains a significant challenge for traditional machine learning (ML) models due to the limited availability of labelled data, feature redundancy, and poor generalization across complex biomedical patterns. Conventional methods often depend on hand-engineered features or linear classifiers, which struggle to capture the non-linear and latent relationships present in Parkinson’s subtypes. To address these limitations, this study proposes a Self-Supervised Learning (SSL) framework based on contrastive learning, which enables robust representation learning from unlabelled data. The approach is evaluated on a dataset comprising 195 biomedical voice recordings, capturing acoustic markers relevant to Parkinson’s disease. Once trained, the SSL model’s embeddings are evaluated using a small portion of labelled data through a lightweight supervised classifier. For comparison, a Supervised Deep Learning (SDL) model is also developed, integrating numerical biomarkers with contextual features extracted via Sentence-BERT embeddings. Experimental results demonstrate that the SSL model, despite being trained without labels, achieves a high classification accuracy of 95.3% when evaluated with labels, while the SDL model reaches 96.1% accuracy. These findings highlight the potential of SSL as a practical solution in label-scarce clinical settings, offering a scalable and generalizable alternative to fully supervised models for the classification of Parkinson’s disease.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Ratnam Dodda
AU  - Sureshbabu Alladi
AU  - Y. B. Sai Prasad
AU  - J. R. Vishweshwara Sai
AU  - Gayas Khan Mohammad
AU  - Nityam Kethan Upadhyay
PY  - 2026
DA  - 2026/03/31
TI  - Contrastive Self-Supervised Learning for Parkinson’s Disease Classification and A Comparative Evaluation with Supervised Deep Models
BT  - Proceedings of the International Conference on Artificial Intelligence and Secure Data Analytics (ICAISDA 2025)
PB  - Atlantis Press
SP  - 1091
EP  - 1104
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6239-616-6_79
DO  - 10.2991/978-94-6239-616-6_79
ID  - Dodda2026
ER  -