Segmentation of Brain MRI Images using Multi-Kernel FCM EHO Method

Area of Science:

• Medical Imaging
• Computer Vision
• Artificial Intelligence

Background:

• Image segmentation is challenging due to variations in shape, location, and intensity.
• Brain tumor detection and segmentation are critical in medical diagnostics.
• Accurate segmentation aids in treatment planning and patient outcomes.

Purpose of the Study:

• To segment brain Magnetic Resonance Imaging (MRI) images into tumor and non-tumor regions.
• To develop an automated methodology for precise brain tumor segmentation.
• To enhance the visibility and delineation of brain tumors in medical images.

Main Methods:

• Utilized MRI images from the BraTS2020 database.
• Applied contrast enhancement using thresholding.
• Implemented image denoising with a fourth-order partial differential equation.
• Employed an elephant herding algorithm for centroid optimization.
• Performed image segmentation using multi-kernel fuzzy c-means clustering.

Main Results:

• Performance evaluated using Peak Signal-to-Noise Ratio, Mean Square Error, sensitivity, specificity, and accuracy.
• The proposed method demonstrated superior performance compared to conventional techniques.
• Quantitative metrics indicate enhanced accuracy in tumor segmentation.

Conclusions:

• The developed methodology is a more effective technique for brain tumor segmentation.
• The proposed approach offers improved accuracy and reliability over existing methods.
• This technique shows promise for clinical application in brain tumor analysis.