FNST-DT: A federated neuro-symbolic framework for Parkinson’s disease progression
DOI:
https://doi.org/10.18488/76.v13i4.5353Keywords:
Digital twin, Explainable artificial intelligence, Federated learning, Multimodal data analysis, Neuro-symbolic AI, Parkinson’s disease, Privacy-preserving healthcare, Transformer networks.Abstract
Parkinson’s disease (PD) progression modeling remains a challenging research problem due to variations in patient conditions, temporal changes in clinical characteristics, and the need for secure handling of healthcare data. Existing centralized deep learning approaches often face limitations related to data privacy, limited transparency, and difficulties in collaborative analysis across healthcare institutions. This study introduces FNST-DT, a Federated Neuro-Symbolic Framework with Digital Twin integration for privacy-preserving Parkinson’s disease progression modeling. The framework integrates federated learning for decentralized model training, transformer-based temporal learning for extracting progression patterns from longitudinal clinical data, and neuro-symbolic reasoning for improving the interpretability of prediction outcomes. In addition, the digital twin module provides a patient-specific representation of disease progression to support personalized monitoring and clinical decision-making. The proposed framework was evaluated using the Parkinson’s Telemonitoring dataset under a simulated federated learning environment. The results show that FNST-DT achieved a prediction accuracy of 94%, which is higher than MLP-LSVM (82%), PFMOO (85%), and PQFL (88%). The framework also obtained 95% privacy preservation and 93% computational efficiency, indicating its ability to balance predictive performance, security, and computational requirements. The findings suggest that FNST-DT can support future healthcare applications by enabling secure multi-institutional collaboration, continuous patient monitoring, and more transparent clinical decision support for Parkinson’s disease management.
