Uncovering heterogeneous Tau network diffusion based on new generative modeling

Music City Center 

The progression of brain Tau pathology has been broadly believed to follow a stereotypical pattern. However, the latest advances in Tau-PET imaging suggests the existence of heterogeneous Tau emergence and progression, which could bias interventions if based solely on the established Tau-targeting routine. Meanwhile, most Tau-PET imaging studies are either cross-sectional or with limited follow-ups, bringing further challenges to uncover personalized Tau progression patterns. In this work, we address these hurdles by proposing an innovative generative network-based diffusion models under cross-sectional observations and heterogeneity. Unlike existing models that rely on repeated measurements to characterize propagation networks, our method can uncover spread network even with sparse observations. Based on extensive simulations and data analyses, we demonstrate the superiority of our method.

Network modeling

Generative modeling

Imaging