A spatio-temporal fusion-based approach for multi-dimensional classification of Parkinson’s disease progression using multi-modal dataset
Context: The progressive neurodegenerative disorder Parkinson’s disease (PD) features diverse symptom presentation that progresses at different speeds and demands effective disease classification with precise patient management. The traditional methods demonstrate insufficient ability to detect the...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-06-01
|
| Series: | Results in Engineering |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2590123025013878 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Context: The progressive neurodegenerative disorder Parkinson’s disease (PD) features diverse symptom presentation that progresses at different speeds and demands effective disease classification with precise patient management. The traditional methods demonstrate insufficient ability to detect the complex spatial-temporal connections between various clinical data types which highlights the need for advanced computational approaches. Objective: The proposed study pioneered a state-of-the-art Spatio-temporal fusion and prediction (STFP) algorithm model approach that utilizes Graph neural network (GNN) and Temporal convolutional network (TCN) for multi-task classification of PD covering three significant dimensions: disease stage, progression speed, symptom subtype classification. Methods: By employing the fusing approach for structural medical resonance imaging (MRI) and dopamine transporter scan (DaTSCAN) along with clinical biomarker data to examine spatial and temporal relationships a fused feature representation was developed. PPMI dataset was employed for training and evaluation of the multi-task classification model with Cohen’s Kappa, cumulative link model (CLM), the area under the precision-recall curve (AUC-PR), and Brier score prominent evaluation metrics for analyzing the robustness of the model. Key Findings: A Cohen’s Kappa of 0.89, a CLM score of 0.91, and an AUC-PR of 0.94 have been recorded for early-stage classification, slow progression, and motor symptom classification respectively demonstrating the outperformance of the proposed model. The superiority of the model has been established with p-values (<0.05) for ANOVA and Wilcoxon tests. Conclusion: The STFP-based model advances the modeling of PD progression by improving classification accuracy and diagnostic clarity. Subsequent research will investigate more data modalities and their practical application in healthcare. |
|---|---|
| ISSN: | 2590-1230 |