An MR brain images classifier system via particle swarm optimization and kernel support vector machine.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Neuroscience

Background:

• Automated detection of abnormal brain conditions is crucial for accurate clinical diagnosis.
• Numerous methods have been developed, but improved accuracy and efficiency remain key objectives.

Purpose of the Study:

• To propose a novel hybrid system for classifying MR brain images as normal or abnormal.
• To enhance classification accuracy through optimized feature extraction and machine learning techniques.

Main Methods:

• Feature extraction using digital wavelet transform.
• Dimensionality reduction via principal component analysis (PCA).
• Kernel support vector machine (KSVM) with RBF kernel, optimized by particle swarm optimization (PSO) for parameters C and σ.
• Fivefold cross-validation for model evaluation.

Main Results:

• The proposed hybrid system achieved a classification accuracy of 97.78%.
• This accuracy surpasses that of backpropagation neural networks (BP-NN) at 86.22% and radial basis function neural networks (RBF-NN) at 91.33%.
• Particle swarm optimization (PSO) demonstrated superior effectiveness in selecting optimal KSVM parameters compared to random selection.

Conclusions:

• The developed hybrid system offers a highly accurate and effective approach for automated abnormal brain MR image classification.
• The integration of wavelet transform, PCA, and PSO-optimized KSVM provides a robust framework for clinical diagnostic support.
• This method shows significant potential for improving the early and reliable detection of various brain pathologies.