Teaching learning based optimization-functional link artificial neural network filter for mixed noise reduction from...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Signal Processing

Background:

• Clinical magnetic resonance imaging (MRI) images are susceptible to corruption from mixed noise types (Rician, Gaussian, impulse).
• Existing filtering algorithms are often noise-specific, linear, and lack adaptivity, limiting their effectiveness.

Purpose of the Study:

• To develop a nonlinear adaptive filter capable of suppressing mixed noise in MRI images.
• To create a filter that dynamically adapts to specific noise conditions for improved performance.

Main Methods:

• Proposed a novel nonlinear adaptive filter based on a functional link artificial neural network (FLANN).
• Employed the teaching learning based optimization (TLBO) technique, a derivative-free meta-heuristic, for training FLANN weights.
• Implemented and evaluated the proposed FLANN-TLBO filter.

Main Results:

• The proposed FLANN-TLBO filter demonstrated superior performance compared to five other adaptive filters.
• Quantitative metrics and nonparametric statistical tests confirmed the filter's effectiveness in noise suppression.
• Analysis of convergence curves and computational time indicated high efficiency.

Conclusions:

• The developed FLANN-TLBO filter significantly outperforms existing adaptive filters for mixed noise removal in MRI.
• The proposed filter offers potential for hybridization with other evolutionary techniques for enhanced artifact removal in various medical images.