Iterative spatial fuzzy clustering for 3D brain magnetic resonance image supervoxel segmentation

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Neuroscience

Background:

• Supervoxel segmentation methods are used for brain MRI analysis.
• Existing methods struggle with complex brain structures and partial volume effects, leading to unsatisfactory performance.

Purpose of the Study:

• To present a novel iterative spatial fuzzy clustering (ISFC) algorithm for 3D supervoxel generation in brain MRI.
• To improve the accuracy and robustness of brain MRI segmentation.

Main Methods:

• The ISFC algorithm utilizes prior knowledge and common human brain topology.
• Seed templates are derived from population-based MRI and projected onto individual brains.
• An iterative fuzzy clustering approach allocates voxels to seeds to generate supervoxels, mitigating partial volume effects.

Main Results:

• The algorithm's performance was evaluated on two public brain MRI datasets.
• Results were compared against three other state-of-the-art methods, demonstrating effectiveness.

Conclusions:

• The proposed ISFC algorithm offers improved 3D supervoxel generation for brain MRI.
• Potential applications include tissue segmentation, tumor detection, functional parcellation, and registration.