Poster No:
1129
Submission Type:
Abstract Submission
Authors:
Hyewon Shin1, Yong Jeong1
Institutions:
1KAIST, Daejeon, Yuseong-gu
First Author:
Co-Author:
Introduction:
Parkinson's disease (PD) is a progressive neurodegenerative disorder characterized by motor impairments due to disruptions in structural brain connectivity. Accurate prediction of motor symptom progression, measured by the Unified Parkinson's Disease Rating Scale Part III (UPDRS-III), is critical for personalized treatment and timely interventions. However, conventional models often overlook individual brain features and inter-subject variability. Additionally, PD's progressive nature necessitates temporal modeling.
To address these challenges, we propose a hierarchical graph convolutional network (GCN) framework combining subject-level and population-level GCNs with a long short-term memory (LSTM) network. The subject-level GCN models region-to-region brain connectivity, while the population-level GCN captures inter-subject relationships, leveraging shared structural patterns. The LSTM tracks longitudinal changes, enabling multi-year motor symptom forecasting. GNNExplainer enhances interpretability by identifying biologically relevant regions linked to motor decline.
Methods:
We constructed structural brain connectivity networks using diffusion tensor imaging (DTI)-derived fractional anisotropy (FA) maps, covering 130 regions of interest (ROIs) from AAL116 and ATAG atlases. Each subject's brain was modeled as a graph with nodes representing ROIs and edges weighted by FA-based connectivity.
The proposed framework includes three components. First, a subject-level GCN (s-GCN) processes individual brain graphs, capturing localized structural disruptions. Next, an LSTM models longitudinal changes from multi-year s-GCN embeddings. Finally, a population-level GCN (p-GCN) aggregates inter-subject embeddings to model shared disease progression patterns using semi-supervised learning.
The model was trained on 1,917 PD patients from the PPMI database. Performance was assessed using mean squared error (MSE) and the coefficient of determination (R²). Comparative experiments against baseline models (e.g., SVM, KNN, RF) and state-of-the-art GCN models (ST-GCN, BrainNetCNN, EigenPoolingGCN) were conducted. GNNExplainer was used to enhance interpretability by linking salient ROIs to motor symptoms.
Results:
The proposed framework demonstrated superior performance in predicting motor symptom progression in PD. For current UPDRS-III scores, it achieved MSE = 3.24 ± 0.57 and R² = 0.91 ± 0.02, outperforming SVM, RF, and KNN models. Longitudinal prediction over three years maintained accuracy with MSE = 4.78 ± 0.81 and R² = 0.87 ± 0.03, surpassing ST-GCN, BrainNetCNN, and EigenPoolingGCN.
An ablation study highlighted each component's importance. Removing the s-GCN increased prediction error by 21%, while excluding the p-GCN reduced R² by 15%. Omitting the LSTM resulted in a 26% performance drop. The full integration of all components yielded the best results. GNNExplainer revealed key motor-related ROIs, including the putamen, caudate nucleus, thalamus, and pallidum in the cortico-basal ganglia-thalamo-cortical loop, consistent with established PD pathology. The cerebellum's lobules 4 and 5 also emerged, reflecting compensatory motor coordination roles. Lower white matter integrity in these regions correlated with increased motor impairment, offering biologically relevant insights.
Conclusions:
This study presents a novel hierarchical GCN framework integrating subject-level connectivity, temporal dynamics, and inter-subject relationships for predicting motor symptom progression in PD. The model achieved superior predictive accuracy with biologically interpretable results. By combining deep learning with DTI biomarkers, it provides a robust, scalable, and interpretable tool for clinical applications. Future work includes extending to multi-modal data and other neurological disorders.
Disorders of the Nervous System:
Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) 2
Modeling and Analysis Methods:
Classification and Predictive Modeling 1
Connectivity (eg. functional, effective, structural)
Diffusion MRI Modeling and Analysis
Neuroanatomy, Physiology, Metabolism and Neurotransmission:
White Matter Anatomy, Fiber Pathways and Connectivity
Keywords:
Aging
Machine Learning
Tractography
White Matter
Other - Diffusion Tensor Imaging, Parkinson's Disease, Graph Convolutional Neural Networks
1|2Indicates the priority used for review
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
I accept
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information.
Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
I do not want to participate in the reproducibility challenge.
Please indicate below if your study was a "resting state" or "task-activation” study.
Resting state
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Patients
Was this research conducted in the United States?
No
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel?
NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Not applicable
Were any animal research approved by the relevant IACUC or other animal research panel?
NOTE: Any animal studies without IACUC approval will be automatically rejected.
Not applicable
Please indicate which methods were used in your research:
Diffusion MRI
For human MRI, what field strength scanner do you use?
3.0T
If Other, please list
-
DTI
Which processing packages did you use for your study?
SPM
FSL
Free Surfer
Provide references using APA citation style.
[1] Pyatigorskaya, N., Gallea, C., Garcia-Lorenzo, D., Vidailhet, M., & Lehericy, S. (2014). A review of the use of magnetic resonance imaging in Parkinson's disease. Therapeutic advances in neurological disorders, 7(4), 206–220. https://doi.org/10.1177/1756285613511507
[2] Dorsey, E. R., Sherer, T., Okun, M. S., & Bloem, B. R. (2018). The Emerging Evidence of the Parkinson Pandemic. Journal of Parkinson's disease, 8(s1), S3–S8. https://doi.org/10.3233/JPD-181474
[3] Tysnes, O. B., & Storstein, A. (2017). Epidemiology of Parkinson's disease. Journal of neural transmission (Vienna, Austria : 1996), 124(8), 901–905. https://doi.org/10.1007/s00702-017-1686-y
[4] Schwab, A. D., Thurston, M. J., Machhi, J., Olson, K. E., Namminga, K. L., Gendelman, H. E., & Mosley, R. L. (2020). Immunotherapy for Parkinson's disease. Neurobiology of disease, 137, 104760. https://doi.org/10.1016/j.nbd.2020.104760
[5] Koeglsperger, T., Rumpf, S. L., Schließer, P., Struebing, F. L., Brendel, M., Levin, J., Trenkwalder, C., Höglinger, G. U., & Herms, J. (2023). Neuropathology of incidental Lewy body & prodromal Parkinson's disease. Molecular neurodegeneration, 18(1), 32. https://doi.org/10.1186/s13024-023-00622-7
[6] Gómez-Benito, M., Granado, N., García-Sanz, P., Michel, A., Dumoulin, M., & Moratalla, R. (2020). Modeling Parkinson's Disease With the Alpha-Synuclein Protein. Frontiers in pharmacology, 11, 356. https://doi.org/10.3389/fphar.2020.00356
[7] McGregor, M. M., & Nelson, A. B. (2019). Circuit Mechanisms of Parkinson's Disease. Neuron, 101(6), 1042–1056. https://doi.org/10.1016/j.neuron.2019.03.004
[8] Heng, N., Malek, N., Lawton, M. A., Nodehi, A., Pitz, V., Grosset, K. A., Ben-Shlomo, Y., & Grosset, D. G. (2023). Striatal Dopamine Loss in Early Parkinson's Disease: Systematic Review and Novel Analysis of Dopamine Transporter Imaging. Movement disorders clinical practice, 10(4), 539–546. https://doi.org/10.1002/mdc3.13687
[9] Kumaresan, M., & Khan, S. (2021). Spectrum of Non-Motor Symptoms in Parkinson's Disease. Cureus, 13(2), e13275. https://doi.org/10.7759/cureus.13275
[10] Murakami, H., Shiraishi, T., Umehara, T., Omoto, S., & Iguchi, Y. (2023). Recent Advances in Drug Therapy for Parkinson's Disease. Internal medicine (Tokyo, Japan), 62(1), 33–42. https://doi.org/10.2169/internalmedicine.8940-21
No