A hierarchical Bayesian brain parcellation framework for fusion of functional imaging datasets

One important barrier in the development of complex models of human brain organization is the lack of a large and comprehensive task-based neuro-imaging dataset. Therefore, current atlases of functional brain organization are mainly based on single and homogeneous resting-state datasets. Here, we propose a hierarchical Bayesian framework that can learn a probabilistically defined brain parcellation across numerous task-based and resting-state datasets, exploiting their combined strengths. The framework is partitioned into a spatial arrangement model that defines the probability of a specific individual brain parcellation, and a set of dataset-specific emission models that defines the probability of the observed data given the individual brain organization. We show that the framework optimally combines information from different datasets to achieve a new population-based atlas of the human cerebellum. Furthermore, we demonstrate that, using only 10 min of individual data, the framework is able to generate individual brain parcellations that outperform group atlases. ### Competing Interest Statement The authors have declared no competing interest.