Conditional Neural ODE for Longitudinal Parkinson’s Disease Progression Forecasting
Poster No:
251
Submission Type:
Late-Breaking Abstract Submission
Authors:
Xiaoda Wang1, Yuji Zhao2, Kaiqiao Han3, Xiao Luo3, Sanne van Rooij1, Jennifer Stevens1, Lifang He4, Liang Zhan5, Yizhou Sun3, Wei Wang3, Carl Yang1
Institutions:
1Emory University, Atlanta, GA, 2Illinois Institute of Technology, Chicago, IL, 3University of California, Los Angeles, Los Angeles, CA, 4Lehigh University, Bethlehem, PA, 5University of Pittsburgh, Pittsburgh, PA
First Author:
Co-Author(s):
Introduction:
Parkinson's disease (PD) is a progressive neurodegenerative disorder characterized by complex and heterogeneous changes in brain morphometry over time (Armstrong, 2020). Modeling these longitudinal trajectories is essential for understanding disease pathology, guiding treatment development, and enabling personalized "digital twin" simulations to forecast the evolution of PD under various hypothetical interventions (Gratwicke, 2015). However, existing methods usually adopt recurrent neural networks and transformer architectures, which rely on discrete, regularly sampled data and struggle to handle the irregular and sparse magnetic resonance imaging (MRI) in PD cohorts. Moreover, these methods have difficulty in capturing individual heterogeneity including variations in disease onset, progression rate, and symptom severity, which is a hallmark of PD (Lian, 2024). To address these challenges, we propose CNODE (Conditional Neural ODE), a novel framework for continuous, individualized PD progression forecasting as shown in Fig. 1. The core of CNODE is to model morphological brain changes as continuous temporal processes using a neural ODE model. In addition, we jointly learn patient-specific initial time and progression speed to align individual trajectories into a shared progression trajectory. We validate CNODE on the Parkinson's Progression Markers Initiative (PPMI) dataset (Marek, 2011). Experimental results show that our method outperforms state-of-the-art baselines in forecasting PD progression, which can pave the way for deeper insights into PD dynamics and improved clinical decision support.
Methods:
CNODE first extracts vertex-wise medial thickness from MRI scans, along with subcortical structure volumes and metadata (Fischl, 2012). Then CNODE constructs a shared progression trajectory, aligning patient data through two neural networks that predict for individual start time and progression speed. To ensure that visits with similar shape features are mapped closer in time, we integrate contrastive learning (Oord, 2018). This is achieved by computing feature-based similarity scores and optimizing a contrastive loss to refine the shared trajectory. Next, we employ a Neural ODE (Chen, 2018) to model the continuous evolution of shape features. The model consists of an encoder that maps observed features to a latent space, an ODE solver that predicts latent state dynamics over time, and a decoder that reconstructs shape features from the latent representations.
Results:
We evaluate our model using the PPMI dataset, leveraging T1-weighted MRI scans from 161 PD patients, with 50 individuals having three visits and another 111 having two visits. The average visit interval is 1.11 years, with a maximum of 2.27 years and a minimum of 0.61 years. CNODE demonstrating superior performance in MSE, RMSE, and R^2 compared with RNN-based methods (Sherstinsky, 2020), Transformer (Vaswani, 2017) and LLMTime (Gruver, 2023). Ablation studies confirm the effectiveness of contrastive learning and progression speed alignment. To further demonstrate the interpretability of CNODE, we visualize the predicted disease progression trajectories for individual patients in Fig. 2. It highlights CNODE's ability to generate biologically plausible progression patterns that closely align with real patient data. This interpretability is particularly valuable for identifying high-risk individuals and enabling timely interventions.
Conclusions:
In this work, we introduced CNODE that models continuous Parkinson's disease progression. By harnessing subcortical shape features and aligning patients within a shared progression trajectory, CNODE captures individualized disease dynamics while addressing common challenges such as missing timepoints and heterogeneous progression rates. Experimental results demonstrate the superiority of our proposed CNODE over state-of-the-art methods, highlighting its potential for personalized disease monitoring and therapeutic decision support.
Disorders of the Nervous System:
Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) 1
Modeling and Analysis Methods:
Classification and Predictive Modeling 2
Methods Development
Keywords:
Computational Neuroscience
Design and Analysis
Machine Learning
MRI
Other - Neural ODE; Parkinson’s disease progression; Contrastive learning; Time series forecasting
1|2Indicates the priority used for review
Abstract Information
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.
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:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Are you Internal Review Board (IRB) certified? Please note: Failure to have IRB, if applicable will lead to automatic rejection of abstract.
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.
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.
Please indicate which methods were used in your research:
Provide references using APA citation style.
[2] Lian, J., Luo, X., Shan, C., Han, D., Zhang, C., Vardhanabhuti, V., ... & Qiu, L. (2024). Personalized progression modelling and prediction in Parkinson’s disease with a novel multi-modal graph approach. npj Parkinson's Disease, 10(1), 229.
[3] Gratwicke, J., Jahanshahi, M., & Foltynie, T. (2015). Parkinson’s disease dementia: a neural networks perspective. Brain, 138(6), 1454-1476.
[4] Marek, K., Jennings, D., Lasch, S., Siderowf, A., Tanner, C., Simuni, T., ... & Parkinson Progression Marker Initiative. (2011). The Parkinson progression marker initiative (PPMI). Progress in neurobiology, 95(4), 629-635.
[5] Fischl, B. (2012). FreeSurfer. Neuroimage, 62(2), 774-781.
[6] Oord, A. V. D., Li, Y., & Vinyals, O. (2018). Representation learning with contrastive predictive coding. arXiv preprint arXiv:1807.03748.
[7] Chen, R. T., Rubanova, Y., Bettencourt, J., & Duvenaud, D. K. (2018). Neural ordinary differential equations. Advances in neural information processing systems, 31.
[8] Gruver, N., Finzi, M., Qiu, S., & Wilson, A. G. (2023). Large language models are zero-shot time series forecasters. Advances in Neural Information Processing Systems, 36, 19622-19635.
[9] Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., ... & Polosukhin, I. (2017). Attention is all you need. Advances in neural information processing systems, 30.
[10] Sherstinsky, A. (2020). Fundamentals of recurrent neural network (RNN) and long short-term memory (LSTM) network. Physica D: Nonlinear Phenomena, 404, 132306.