Nonlinear Feature Extraction Methods Based on Dual-Tree Complex Wavelet Transform Subimages of Brain Magnetic...

Area of Science:

• Medical Imaging
• Machine Learning
• Signal Processing

Background:

• Magnetic Resonance Imaging (MRI) is crucial for detecting brain diseases.
• Current MRI classification methods have room for improvement in accuracy and efficiency.
• Advanced feature extraction and classification techniques are needed for precise multi-class brain disease diagnosis.

Purpose of the Study:

• To develop a novel non-linear feature extraction method for brain MRI.
• To propose a computationally efficient Hybrid Radial Basis Function (RBF) network for multi-class brain disease classification.
• To achieve high classification accuracy and reduce computational complexity compared to existing methods.

Main Methods:

• Non-linear feature extraction from MRI sub-images using three levels of 2D Dual Tree Complex Wavelet Transform (2D DT-CWT).
• Feature reduction using Spectral Regression Discriminant Analysis (SRDA).
• Classification using a Hybrid RBF network integrating k-means and Recursive Least Squares (RLS) algorithms.

Main Results:

• The Hybrid RBF network achieved 100% classification accuracy for both two-class and multi-class (8 and 10 classes) brain diseases.
• The proposed method demonstrated superior performance compared to Support Vector Machines (SVM) and K-Nearest Neighbors (KNN).
• The method offers a significant reduction in computational complexity while maintaining high precision.

Conclusions:

• The developed non-linear feature extraction and Hybrid RBF network provide a computationally reasonable and highly accurate approach for brain MRI disease classification.
• This method enhances the reliability and efficiency of diagnosing multiple brain diseases from MRI scans.
• The study presents a promising low-complexity, high-accuracy solution for clinical applications in neuroimaging.