Towards an Automated Segmentation of the Ventro-Intermediate Thalamic Nucleus

Abstract

The ventro-intermediate nucleus (Vim), as the others thalamic subparts, cannot be directly visualized by current standard magnetic resonance imaging (MRI), in daily clinical practice. Hence, for treatment of tremor in functional neurosurgery, where the commonly used target is the Vim, the targeting procedure is done indirectly. We present a novel direct automated segmentation of the Vim using only subject-related MRI information, specifically, diffusion MRI at 3T and susceptibility weighted images (SWI) acquired at 7T. With a state-of-the-art method based on local diffusion MR properties for automated subdivision of the thalamus, we first restrain the region of interest to the group of motor-related nuclei. Then, this thalamic part is further subdivided, in graph parcellation manner, using the intensity-related features provided by SWI together with prior knowledge of the Vim localization inside the motor thalamic segment. Our framework was tested in four healthy elderly subjects, for eight thalami in total, and the results were evaluated by an experienced neurosurgeon, showing the ability to directly detect the Vim area. The qualitative inspection indicated that the proposed method outperforms standard multi-atlas based techniques.