Segmentation of MR Brain Images Through Hidden Markov Random Field and Hybrid Metaheuristic Algorithm

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Neuroscience

Background:

• Accurate segmentation of Magnetic Resonance (MR) images is crucial for clinical analysis.
• Existing segmentation methods often struggle with noise and intensity non-uniformity artifacts (INU).
• A robust and artifact-resistant segmentation technique is needed for reliable MR image analysis.

Purpose of the Study:

• To develop a novel, precise, and artifact-resistant image segmentation method for MR images.
• To improve the performance of Maximum a posteriori (MAP) estimation in Hidden Markov Random Field (HMRF) models.
• To enhance the exploration and exploitation capabilities in the search for optimal segmentation solutions.

Main Methods:

• A new Hidden Markov Random Field (HMRF) model with adaptive parameters was developed.
• A hybrid metaheuristic algorithm, combining Cuckoo Search (CS) and Particle Swarm Optimization (PSO), was introduced.
• The hybrid algorithm was applied to the MAP estimation of the HMRF model, utilizing parallel processing and a solution selection mechanism.

Main Results:

• The proposed method demonstrated satisfactory performance on simulated and real MR brain images, even with noise and intensity inhomogeneity.
• Experimental results indicated that the hybrid metaheuristic approach improved the quality of solutions found by the MAP estimation.
• The new segmentation method outperformed its considered competitors in artifact-prone MR images.

Conclusions:

• The combined HMRF model and hybrid CS-PSO algorithm offers a robust solution for MR image segmentation.
• The method effectively addresses challenges posed by noise and intensity non-uniformity in MR imaging.
• This approach represents a significant advancement in artifact-resistant MR image analysis and segmentation.