Similarity Measure-Based Possibilistic FCM With Label Information for Brain MRI Segmentation

Area of Science:

• Medical Imaging
• Computer Vision
• Biomedical Engineering

Background:

• Magnetic Resonance Imaging (MRI) is crucial in clinical diagnostics.
• Accurate brain image segmentation is vital for detecting abnormalities.
• Intensity inhomogeneity and noise in MRI data complicate precise segmentation.

Purpose of the Study:

• To propose an improved possibilistic fuzzy c-means (FCM) method for enhanced MRI brain image segmentation.
• To address limitations of existing FCM methods, including cluster-size sensitivity and handling complex data distributions.
• To improve the robustness and accuracy of brain image segmentation in the presence of noise and intensity variations.

Main Methods:

• Developed an improved possibilistic fuzzy c-means (FCM) algorithm incorporating a novel similarity measure.
• The new similarity measure is designed to handle non-spherical data distributions effectively.
• Integrated local label information to preserve image details and suppress noise during segmentation.

Main Results:

• The proposed method demonstrates superior clustering performance for data with non-spherical distributions.
• Effectively alleviates the "cluster-size sensitivity" issue common in FCM-based techniques.
• Exhibits strong resistance to noise and preserves fine image details, validated on synthetic and clinical MRI datasets.

Conclusions:

• The improved possibilistic FCM method offers enhanced accuracy and robustness for MRI brain image segmentation.
• Successfully mitigates common challenges like cluster-size sensitivity and noise.
• Shows significant potential for clinical applications requiring precise brain image analysis.