SEGMENTATION OF THE COMPLETE SUPERIOR CEREBELLAR PEDUNCLES USING A MULTI-OBJECT GEOMETRIC DEFORMABLE MODEL

Area of Science:

• Neuroimaging
• Neuroanatomy
• Diffusion Tensor Imaging

Background:

• Superior cerebellar peduncles (SCPs) are crucial white matter tracts connecting the cerebellum and thalamus.
• Existing diffusion tensor imaging (DTI) methods struggle to accurately segment SCPs, especially at the decussation of the SCPs (dSCP), due to crossing fibers.
• Inaccurate segmentation limits understanding of cerebellar pathways.

Purpose of the Study:

• To develop a novel method for volumetric segmentation of the complete superior cerebellar peduncles (SCPs).
• To accurately delineate both non-crossing SCP fibers and the decussation of the SCPs (dSCP).
• To improve the analysis of cerebellar white matter tracts using DTI.

Main Methods:

• Modeling non-crossing SCPs and the dSCP as distinct objects.
• Employing a multi-object geometric deformable model.
• Utilizing forces derived from diffusion properties and the primary eigenvector (PEV) to define boundaries.

Main Results:

• Successfully resolved fiber crossing issues at the dSCP.
• Achieved accurate volumetric segmentation of the complete SCPs.
• Demonstrated repeatability of the segmentation method in phantom and real subject data.

Conclusions:

• The proposed method enables accurate and complete segmentation of the superior cerebellar peduncles (SCPs).
• This technique overcomes the limitations of prior methods in handling fiber crossings at the dSCP.
• Offers improved potential for detailed neuroanatomical and clinical studies involving cerebellar pathways.