Brainstem Functional Parcellation Based on Spatial Connectivity Features Using Functional Magnetic Resonance Imaging

The brainstem controls almost all normal functions in the life, such as breathing, memory, movement, and is closely related to many neurological diseases. Despite the importance of the brainstem, the delineation of its functional sub-regions remains largely unexplored. In this study, we aim to explore functional parcellation of the brainstem using functional magnetic resonance imaging (fMRI), and propose a novel framework by combining spatial functional connectivity features of the brainstem and NCut spectral clustering. Firstly, functional connectivity between the brainstem and other cortical and sub-cortical brain regions is estimated using fMRI data. Secondly, the estimated spatial functional connectivity features are used to detect functional sub-regions of the brainstem using NCut spectral clustering. Finally, the Dice coefficient was used to evaluate the reproducibility of brainstem functional parcellation. The results show that the Dice coefficient obtained by the proposed method was 0.74, which is higher than that of the parcellation using temporal features of the brainstem (Dice coefficient of 0.32). In addition, NCut spectral clustering outperformed other clustering methods regarding the reproducibility of brainstem functional parcellation. The proposed method explores the potentials of spatial functional connectivity features for brainstem functional parcellation. It may serve as a promising tool for studying the functions and dysfunctions of the brainstem.