From Cross-Sectional Snapshots to Longitudinal Brain Aging Trajectories: AI for Precision Neurodegeneration Modeling

Brisbane Convention & Exhibition Centre 

Aging and disease heterogeneity remain major obstacles to achieving precision diagnostics. Artificial intelligence (AI) advances have introduced promising approaches to tackle this complexity, particularly by identifying imaging-derived biomarkers capable of predicting disease progression and mortality. Our laboratory has developed a range of machine learning frameworks designed to characterize this heterogeneity and derive subtypes by modeling transformations from a reference cohort to target/patient populations. Traditional clustering approaches identify subtypes directly within patient populations—often confounded by disease-unrelated variability—but also impose rigid group boundaries failing to capture overlapping or continuous variation, and lack the capacity to model longitudinal progression. Our approaches include weakly supervised models such as Smile-GAN (PMID: 34862382), which identifies imaging-derived disease subtypes, and Surreal-GAN (PMID: 39147830), which models continuous disease heterogeneity through cross-sectional representation learning using generative adversarial networks (GANs). In this talk, we will summarize these methods and will further introduce CCL-NMF, a novel framework integrating both cross-sectional deviations and personalized aging trajectories from longitudinal data to uncover dominant patterns of brain aging. We will also highlight the applications of these models in diverse contexts, including normal aging and disorders such as Alzheimer's disease, along with the public NiChart platform (neuroimagingchart.com), which enables easy deployment of these models in new datasets. This talk will exemplify how advanced AI and imaging techniques can model disease heterogeneity to advance precision medicine.

The results presented in this talk are based on the work of Dr. Ioanna Skampardoni (presenter) and on the work of Drs. Zhijian Yang and Junhao Wen (co-organizer), under the supervision of Dr. Christos Davatzikos.