Brain tissue segmentation using fuzzy clustering techniques

Area of Science:

• Medical Imaging
• Neuroscience
• Computer Vision

Background:

• Automated medical image segmentation is crucial for accurate image analysis.
• Manual segmentation is time-consuming and less accurate than automated methods.
• Accurate segmentation of MR brain images aids in diagnosing neurological conditions.

Purpose of the Study:

• To automatically segment MR brain images into Grey Matter (GM), White Matter (WM), and Cerebro-Spinal Fluid (CSF).
• To improve tissue classification for diagnosing diseases like tumors, Alzheimer's, and stroke.
• To develop a robust segmentation technique for noisy MR brain images.

Main Methods:

• An unsupervised clustering technique, Fuzzy C-Means (FCM), was adapted.
• A novel algorithm, Reformulated Fuzzy Local information C-Means Clustering (RFLICM), was developed.
• RFLICM incorporates spatial and gray level information, replacing spatial distance with local coefficient of variation.

Main Results:

• Experiments validated the proposed RFLICM technique on brain MR images.
• The RFLICM algorithm achieved 99.86% efficiency in segmenting images with salt and pepper noise.
• Comparative analysis demonstrated RFLICM's superior performance over standard FCM and FLICM.

Conclusions:

• The RFLICM method provides accurate segmentation and classification of tissues in brain MR images.
• The proposed approach is robust and effective, particularly in the presence of noise.
• RFLICM is a suitable technique for clinical applications requiring precise brain tissue segmentation.