Recurrence Network Approach for Severity Classification of Parkinson's Disease Using Multimodal Data

Deepa Raj K, Jyothish Lal G, Gopalakrishnan EA, Orozco Arroyave JR (2026)

Publication Type: Journal article

Publication year: 2026

Journal

IEEE Access Institute of Electrical and Electronics Engineers (IEEE)

Book Volume: 14

Pages Range: 70613-70634

DOI: 10.1109/ACCESS.2026.3691536

Abstract

Parkinson's disease (PD) a multi-factorial progressive neurodegenerative disorder characterized by the loss of dopamine neurons in the substantia nigra, leading to motor and non-motor impairments. Although there is a significant advancement in the clinical assessment, its underlying mechanisms remain only partially understood, while demanding the integration of computational, signal processing, and systems-level approaches for deeper insights. Recent years have witnessed a paradigm shift in PDs research, pioneering clinical evaluation to data-driven frameworks that unleash the efficacy of network science, and multimodal sensing technologies. This study proposes to model neurodynamics in PD from a complex systems theory in the context of Recurrence Network (RN) - a transformation of a time-series data to a complex network. We classified the severity stages in PD patients using RN features extracted from the network corresponding to speech and gait signals. As the motor impairments observed in these biomarkers manifests intermittently rather than uniformly, we followed a frame-based approach in the classification framework. Precisely, we evaluated the frame-based analysis under three threshold conditions: 40%, 50%, and 60% correctly classified frames. Experimental results show that (under the strict threshold condition of 60%) the proposed multimodal approach achieves approximately a 37% and 18% improvement in accuracy compared to that of the speech-only and gait-only modalities, respectively. Besides, we investigated the qualitative and quantitative characterization of the RN topology and measures, respectively, across different PD severity stages.

Involved external institutions

Amrita University / Amrita Vishwa Vidyapeetham IN India (IN) Universidad de Antioquía (UDEA) CO Colombia (CO)

How to cite

APA:

Deepa Raj, K., Jyothish Lal, G., Gopalakrishnan, E.A., & Orozco Arroyave, J.R. (2026). Recurrence Network Approach for Severity Classification of Parkinson's Disease Using Multimodal Data. IEEE Access, 14, 70613-70634. https://doi.org/10.1109/ACCESS.2026.3691536

MLA:

Deepa Raj, K., et al. "Recurrence Network Approach for Severity Classification of Parkinson's Disease Using Multimodal Data." IEEE Access 14 (2026): 70613-70634.

BibTeX: Download