Brain magnetic resonance image (MRI) segmentation using multimodal optimization.

Area of Science:

• Medical imaging analysis
• Computational neuroscience
• Artificial intelligence in healthcare

Background:

• Accurate brain tumor segmentation from MRI is crucial for early diagnosis and treatment planning.
• Manual segmentation is time-consuming and subjective, necessitating automated solutions.
• Existing multi-level segmentation methods often require manual selection of the number of segments, posing a challenge.

Purpose of the Study:

• To develop an automated method for brain tumor segmentation in MRI.
• To address the challenge of automatically determining the optimal number of segments in image analysis.
• To improve the efficiency and accuracy of brain tumor detection and diagnosis.

Main Methods:

• A modified 3D Histogram-based segmentation approach is proposed.
• The method utilizes a Gaussian filter for smoothing 3D RGB histograms.
• Particle swarm optimization identifies histogram peaks, followed by non-Euclidean distance-based pixel clustering.

Main Results:

• The algorithm was tested on TCIA and brain MRI datasets for tumor detection.
• Performance was compared against Fuzzy C-Means (FCM), FCM_FWCW, and FCM_FW clustering methods.
• The proposed method demonstrated superior performance, achieving the top mean rank across all metrics.

Conclusions:

• The developed algorithm effectively automates brain tumor segmentation in MRI.
• It accurately determines the appropriate number of segments, outperforming existing clustering methods.
• This automated approach holds significant potential for improving early cancer diagnosis and patient outcomes.