Personalized Alzheimer's Disease Progression Forecasting through Temporal Attention-based Analysis
Poster No:
159
Submission Type:
Abstract Submission
Authors:
Zhiyuan Song1, Sarah Morgan2, Shahid Zaman1
Institutions:
1Department of Psychiatry, University of Cambridge, Cambridge, United Kingdom, 2School of Biomedical Engineering and Imaging Sciences, King's College London, London, United Kingdom
First Author:
Co-Author(s):
Sarah Morgan
School of Biomedical Engineering and Imaging Sciences, King's College London
London, United Kingdom
School of Biomedical Engineering and Imaging Sciences, King's College London
London, United Kingdom
Introduction:
Recent advances in machine learning methodologies, coupled with the increasing availability of large-scale medical datasets, have enabled innovative approaches for early detection and progression forecasting of Alzheimer's disease (AD) spectrum disorders(Arya et al., 2023; Tanveer et al., 2020). However, existing models often fail to reveal which specific biomarkers drive disease progression at distinct timepoints for individual patients, limiting their clinical interpretability and usefulness for targeted interventions.
Methods:
To address these gaps, here we propose a multi-dimensional attention-based framework that integrates diverse biomarkers to both improve predictive accuracy and pinpoints the most important features at each stage of disease progression for individual patients. Our framework predicts clinical diagnoses at the next time point for each patient, leveraging all previously observed biomarker and clinical data up to that stage in their disease trajectory. To that end, we used The Alzheimer's Disease Prediction of Longitudinal Evolution (TADPOLE) dataset(Marinescu et al., 2021), a subset of the Alzheimer's Disease Neuroimaging Initiative (ADNI)(Petersen et al., 2010). The dataset included N=1,667 subjects (N=602 normal cognition (NC) cases, N=931 mild cognitive impairment (MCI) cases, and N=347 AD patients). Multiple data modalities were available, including demographic information, clinical diagnoses, cognitive assessment scores, neuroimaging biomarkers, and genetic data. We split the data into a discovery set and held-out test set (90:10 split).
We first compared a range of approaches for feature selection, including random forest, mutual information, and attention-based methods, and found that the attention-based approach provided optimal predictive accuracy. To capture the temporal patterns of AD progression, we then implemented a Recurrent Neural Network (RNN) architecture for time-series forecasting(Yu et al., 2019). By integrating feature-level and temporal-level attention mechanisms with the RNN(Niu et al., 2021) we identified the most informative features at different time points for each individual patient's AD progression prediction. This approach revealed likely dominant factors in progression from the prodromal phase to receiving a clinical diagnosis of Alzheimer's, at an individual level.
We first compared a range of approaches for feature selection, including random forest, mutual information, and attention-based methods, and found that the attention-based approach provided optimal predictive accuracy. To capture the temporal patterns of AD progression, we then implemented a Recurrent Neural Network (RNN) architecture for time-series forecasting(Yu et al., 2019). By integrating feature-level and temporal-level attention mechanisms with the RNN(Niu et al., 2021) we identified the most informative features at different time points for each individual patient's AD progression prediction. This approach revealed likely dominant factors in progression from the prodromal phase to receiving a clinical diagnosis of Alzheimer's, at an individual level.
Results:
Our model achieved a receiver operating characteristic (ROC) area under the curve (AUC) of 0.888 in the discovery dataset and AUC=0.803 in the held-out test set.
Our multi-dimensional attention framework then allows us to visualize feature importance across disease progression stages through heatmaps of feature importance across time, providing personalized insights. The model leverages patients' complete historical records to predict the next diagnostic status, with the heatmap visualizing the attention weights assigned to each feature in making these predictions. For example, in Figure 1, the model demonstrates how different biomarkers' predictive importance varies across the disease trajectory for a single subject, highlighting diagnosis, age, ADAS13, and Ventricles as key features. The temporal patterns of feature importance reveal dynamic shifts in biomarker relevance with progression of the disease for individuals, enabling early identification of critical transition points and providing actionable insights for personalized disease monitoring and patient stratification.
Our multi-dimensional attention framework then allows us to visualize feature importance across disease progression stages through heatmaps of feature importance across time, providing personalized insights. The model leverages patients' complete historical records to predict the next diagnostic status, with the heatmap visualizing the attention weights assigned to each feature in making these predictions. For example, in Figure 1, the model demonstrates how different biomarkers' predictive importance varies across the disease trajectory for a single subject, highlighting diagnosis, age, ADAS13, and Ventricles as key features. The temporal patterns of feature importance reveal dynamic shifts in biomarker relevance with progression of the disease for individuals, enabling early identification of critical transition points and providing actionable insights for personalized disease monitoring and patient stratification.
·Figure 1: Example of a personalized temporal-aware feature importance heatmap, for a single subject.
Conclusions:
In conclusion, this study presents a multi-dimensional attention-based framework that effectively integrates temporal patterns with feature importance analysis to forecast AD progression. By providing temporally resolved, patient-specific insights into biomarker relevance, our approach opens new avenues for personalized disease management and optimized patient stratification in clinical settings.
Disorders of the Nervous System:
Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) 1
Lifespan Development:
Aging
Modeling and Analysis Methods:
Classification and Predictive Modeling 2
Methods Development
Multivariate Approaches
Keywords:
Aging
Computational Neuroscience
Computing
Data analysis
Machine Learning
Modeling
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?
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:
For human MRI, what field strength scanner do you use?
Which processing packages did you use for your study?
Provide references using APA citation style.
Marinescu, R. V., Oxtoby, N. P., Young, A. L., Bron, E. E., Toga, A. W., Weiner, M. W., Barkhof, F., Fox, N. C., Eshaghi, A., Toni, T., Salaterski, M., Lunina, V., Ansart, M., Durrleman, S., Lu, P., Iddi, S., Li, D., Thompson, W. K., Donohue, M. C., … Alexander, D. C. (2021). The Alzheimer’s Disease Prediction Of Longitudinal Evolution (TADPOLE) Challenge: Results after 1 Year Follow-up. Machine Learning for Biomedical Imaging, 1(December 2021), 1–60. https://doi.org/10.59275/j.melba.2021-2dcc
Niu, Z., Zhong, G., & Yu, H. (2021). A review on the attention mechanism of deep learning. Neurocomputing, 452, 48–62. https://doi.org/10.1016/j.neucom.2021.03.091
Petersen, R. C., Aisen, P. S., Beckett, L. A., Donohue, M. C., Gamst, A. C., Harvey, D. J., Jack, C. R., Jagust, W. J., Shaw, L. M., Toga, A. W., Trojanowski, J. Q., & Weiner, M. W. (2010). Alzheimer’s Disease Neuroimaging Initiative (ADNI): Clinical characterization. Neurology, 74(3), 201–209. https://doi.org/10.1212/WNL.0b013e3181cb3e25
Tanveer, M., Richhariya, B., Khan, R. U., Rashid, A. H., Khanna, P., Prasad, M., & Lin, C. T. (2020). Machine learning techniques for the diagnosis of alzheimer’s disease: A review. ACM Transactions on Multimedia Computing, Communications and Applications, 16(1s). https://doi.org/10.1145/3344998
Yu, Y., Si, X., Hu, C., & Zhang, J. (2019). A Review of Recurrent Neural Networks: LSTM Cells and Network Architectures. Neural Computation, 31(7), 1235–1270. https://doi.org/10.1162/neco_a_01199