A Generalizable Lifespan Brain Age Framework: Advancing Deep learning Algorithm for Mental Health
Poster No:
1553
Submission Type:
Abstract Submission
Authors:
Niousha Dehestani1, Zijiao Chen1, Yichi Zhang1, Zijian Dong2, Juan Helen Zhou1
Institutions:
1National University of Singapore, Singapore, Singapore, 2National University of Singapore, Singapore, N/A
First Author:
Co-Author(s):
Introduction:
Brain aging serves as a critical marker of neurological health and is associated with cognitive decline and neurodegenerative diseases such as Alzheimer's, schizophrenia, and Depression. The brain age gap-the difference between predicted brain age and chronological age-has emerged as a biomarker for aging-related studies. While applying brain age metrics to youth has gained momentum for its potential links to clinical outcomes, inconsistencies remain. For instance, some studies, such as Drobinin et al. (2022), report significant associations between the brain age gap and psychopathology in youth using youth-specific models. In contrast, lifespan-trained models often fail to replicate such findings (Han et al., 2024). These discrepancies highlight the need for models that accurately capture the distinct developmental trajectories of youth brains. Existing brain age algorithms are predominantly biased toward elderly brain development due to the overrepresentation of aging cohorts in training data (Leonardsen et al., 2022). To address these challenges, we proposed hierarchical vision transformer captures richer anatomical representations across the lifespan. By employing selective machine unlearning to remove outdated data and continual learning to preserve critical parameters, our model dynamically refines its focus on age-specific cohorts.
Methods:
We developed a hierarchical Vision Transformer (ViT)-based model pre-trained on extensive, heterogeneous imaging datasets (Dosovitskiy et al., 2020; Wu et al., 2023) (Human Connectome Project: Development, Aging, Young Adulthood; ABCD; UK Biobank; Amsterdam Open MRI Collection). The architecture processes 3D inputs as 16×16×16 patches and utilizes 12 transformer blocks, a hidden dimension of 768, and 12 attention heads. Lightweight adapter modules enable efficient fine-tuning for domain specificity. A regression head estimates chronological age from extracted anatomical features.To ensure flexibility and compliance with evolving data management standards, we introduced a Fisher-based machine unlearning technique(Kirkpatrick et al.,2017 ; Xu et al., 2024). This approach enables selective data removal-focusing the model on particular age groups (e.g., youth) without full retraining. Recognizing that neuroimaging research evolves as new cohorts become available, we integrated Elastic Weight Consolidation (EWC, ) to preserve critical parameters, mitigating catastrophic forgetting. This allows seamless incorporation of additional datasets (e.g., SG70, ADHD-200, SLIM). Finally, we investigated the utility of brain age gap in association with clinical outcomes from the ABCD dataset. Mixed-effects models accounted for age, sex (fixed effects), and family (random effects), providing robust statistical control.
Results:
Our ViT-based framework achieved high accuracy in brain age prediction, with a correlation coefficient of r = 0.91 and a mean absolute error (MAE) of 1.20 years. The integration of EWC significantly improved the model's adaptability to new datasets without catastrophic forgetting. Additionally, the Fisher-based machine unlearning technique enabled the selective removal of data, allowing specialization in youth-specific modeling. Using the ABCD dataset, we demonstrated that a positive brain PAD-indicative of accelerated brain aging-was significantly associated after FDR correction with higher symptoms of internalizing problems (cross-sectional: T = 4.23, P < 0.01; longitudinal: T = 6.01, P < 0.01) and externalizing problems (cross-sectional: T = 3.19, P < 0.01; longitudinal: T = 5.48, P < 0.01). These findings emphasize the utility of our framework for early identification of youth at risk for psychopathology.
Conclusions:
This study introduces a scalable, brain age modeling framework that addresses critical gaps in current methodologies. By leveraging this innovative learning strategies, the model provides a robust biomarker for mental health risks, supporting early interventions
Disorders of the Nervous System:
Psychiatric (eg. Depression, Anxiety, Schizophrenia) 2
Modeling and Analysis Methods:
Methods Development 1
Keywords:
Aging
Computational Neuroscience
Computing
Development
Machine Learning
STRUCTURAL MRI
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.
2. Longitudinal Brain Age in Young People With First-Episode Mania Treated With Lithium or Quetiapine Han, Laura et al. Biological Psychiatry, Volume 95, Issue 10, S18
3. Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., & Zisserman, A. (2020). An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929.
4. Wu, J., Ji, W., Liu, Y., Fang, H., Wang, Z.-Y., Xu, Y., & Arbel, T. (2023). Medical SAM Adapter: Adapting Segment Anything Model for Medical Image Segmentation. arXiv preprint arXiv:2304.12620.
5. Kirkpatrick, J., Pascanu, R., Rabinowitz, N., Veness, J., Desjardins, G., Rusu, A.A., Milan, K., Quan, J., Ramalho, T., Grabska-Barwinska, A. and Hassabis, D. (2017). Overcoming catastrophic forgetting in neural networks. Proceedings of the national academy of sciences. 114(13), 3521-3526.
6. Xu, J., Wu, Z., Wang, C., & Jia, X. (2024). Machine unlearning: Solutions and challenges. IEEE Transactions on Emerging Topics in Computational Intelligence.