A multiscale and multiblock fuzzy C-means classification method for brain MR images.

Area of Science:

• Medical Imaging
• Computer Vision
• Biomedical Engineering

Background:

• Magnetic resonance (MR) image classification is crucial for clinical and research applications.
• Challenges in MR image classification include noise, intensity inhomogeneity, and partial volume effects.
• Conventional "hard" classification and standard Fuzzy C-means (FCM) methods have limitations in accuracy and robustness.

Purpose of the Study:

• To develop an accurate and robust MR brain image classification method.
• To address the limitations of conventional FCM in handling noise and intensity inhomogeneity.
• To introduce a modified multiscale and multiblock FCM (MsbFCM) for improved MR image classification.

Main Methods:

• An automatic multiscale and multiblock fuzzy C-means (MsbFCM) classification method with MR intensity correction was developed.
• A bilateral filter was used to create a multiscale image series.
• The image was separated into blocks, and a multiscale FCM was applied from coarse to fine levels, with coarse scale results guiding finer scales.

Main Results:

• The MsbFCM method demonstrated superior performance compared to conventional FCM, MFCM, and MsFCM.
• Validation studies on synthesized, simulated, and real MR images showed consistent improvements.
• The MsbFCM method achieved an overlap ratio of 91% or higher, proving its accuracy and robustness.

Conclusions:

• The proposed MsbFCM method offers accurate and robust MR brain classification.
• The method's ability to handle intensity variations without assuming Gaussian distribution allows for broader applications in tissue classification and quantification.
• This automatic classification method serves as a valuable tool for neuroimaging and other fields.