Segmentation of brain MRI using SOM-FCM-based method and 3D statistical descriptors

Area of Science:

• Medical Imaging
• Neuroimaging
• Computer-Aided Diagnosis

Background:

• Medical imaging systems offer high resolution and contrast, necessitating advanced image processing for computer-aided diagnosis (CAD).
• Image segmentation is crucial for identifying neuroanatomical structures within brain scans.
• Existing methods may struggle with challenges like the partial volume effect (PVE).

Purpose of the Study:

• To propose a novel, unsupervised image segmentation technique for brain tissues.
• To leverage 3D statistical features for enhanced segmentation accuracy.
• To develop a method that does not rely on prior information.

Main Methods:

• Utilizing unsupervised vector quantization and fuzzy clustering techniques.
• Extracting 3D statistical features directly from volumetric brain images.
• Implementing a fuzzy segmentation approach to delineate neuroanatomical tissues.

Main Results:

• The proposed fuzzy segmentation method effectively addresses the partial volume effect.
• The technique was validated using real brain images from the Internet Brain Image Repository (IBSR).
• The unsupervised approach demonstrates potential for accurate neuroanatomical segmentation without prior knowledge.

Conclusions:

• The developed segmentation technique offers a robust solution for brain image analysis.
• This method enhances the utility of medical imaging for computer-aided diagnosis systems.
• The unsupervised and PVE-addressing nature of the technique makes it a valuable tool in neuroimaging research.