Adaptable fuzzy C-Means for improved classification as a preprocessing procedure of brain parcellation

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Neuroscience

Background:

• Brain parcellation subdivides brain regions for analysis.
• Fuzzy clustering, particularly Fuzzy C-Means (FCM), is used for preprocessing in brain segmentation.
• Conventional FCM struggles with magnetic resonance imaging (MRI) artifacts like partial volume effect and inhomogeneity, leading to gray matter misclassification.

Purpose of the Study:

• To improve the accuracy of brain parcellation, especially for gray matter structures like the basal ganglia and thalamus.
• To overcome the limitations of conventional Fuzzy C-Means (FCM) algorithms in MRI analysis.
• To develop a more user-friendly and efficient parcellation tool.

Main Methods:

• Implemented successive fuzzy clustering for improved classification accuracy.
• Developed a modified graphic user interface (GUI) for the parcellation module.
• Utilized MRI data, addressing challenges like partial volume effect and intensity inhomogeneity.

Main Results:

• Successfully improved the classification of gray matter, reducing misclassification into white matter.
• The modified GUI enhanced user convenience and streamlined the parcellation process.
• The successive fuzzy clustering approach demonstrated greater robustness in handling MRI data characteristics.

Conclusions:

• The proposed successive fuzzy clustering method offers a more accurate and reliable approach to brain parcellation from MRI data.
• The enhanced GUI facilitates easier and more efficient use of the parcellation tool for researchers.
• This work contributes to more precise neuroimaging analysis by improving segmentation of critical brain structures.