Adaptively Regularized Bases-Expansion Subspace Optimization Methods for Electrical Impedance Tomography.

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Computational Science

Background:

• Electrical Impedance Tomography (EIT) faces challenges with regularization parameter selection and low spatial resolution.
• Existing EIT methods struggle with accurate reconstruction of complex geometries.

Purpose of the Study:

• To introduce novel adaptive regularization techniques for difference EIT.
• To address limitations in parameter selection and spatial resolution in EIT.

Main Methods:

• Developed two adaptively regularized bases-expansion subspace optimization methods (AR-BE-SOMs).
• Incorporated adaptive L1-norm total variation and adaptive weighted L2-norm multiplicative regularization.
• Utilized conjugate gradient method for optimization, alternating updates between induced contrast current and conductivity domains.

Main Results:

• AR-BE-SOMs demonstrated superior edge preservation and anti-noise capabilities compared to standard BE-SOM.
• Achieved lower relative errors and higher structure similarity indexes in numerical and experimental tests.
• Successfully reconstructed challenging geometries, including sharp corners like "heart and lung" phantoms.

Conclusions:

• The proposed AR-BE-SOMs effectively overcome regularization parameter selection difficulties in EIT.
• Adaptive regularization factors are obtained dynamically during the optimization process.
• AR-BE-SOMs offer a promising approach for high-quality EIT applications, including clinical lung health monitoring.