Connectivity-based segmentation of the brainstem by probabilistic tractography.

Diffusion magnetic resonance imaging is a non-invasive tool increasingly used for the investigation of brain connectivity in vivo. In this paper we propose a method that allows segmentation of the brainstem to four subregions (frontopontine, motor, sensory and reticular) based on connections to supratentorial structures, thereby eliminating the need for using anatomical landmarks within the brainstem for the identification of these subregions. The feasibility of connectivity-based brainstem segmentation was investigated in a group of healthy subjects (n = 20). Multifiber probabilistic tractography was performed using the FMRIB Software Library, and connections between a pontomesencephalic seed mask and four supratentorial target regions (anterior and posterior limbs of the internal capsule, sensory and medial thalamus) were used to determine connectivity maps of the brainstem. Results were compared with a neuroanatomy atlas and histological sections, confirming good anatomic correspondence. The four subregions detected by the connectivity-based segmentation showed good intersubject reproducibility. The presented method may be a potential tool to investigate brainstem connectivity in diseases that distort normal anatomy, and quantitative analyses of the diffusion-related parameters may provide additional information on the involvement of brainstem pathways in certain disease states (e.g., traumatic brain injury, demyelinating disorders, brainstem tumors). The potential clinical applicability of the method is demonstrated in two cases of severe traumatic brain injury.