Dogleg trust-region application in electrical impedance tomography

Area of Science:

• Medical Imaging
• Computational Electromagnetics
• Applied Mathematics

Background:

• Electrical Impedance Tomography (EIT) is a non-invasive imaging technique.
• Accurate image reconstruction is crucial for detecting abnormalities in EIT.
• Traditional methods like Levenberg-Marquardt can be computationally intensive.

Purpose of the Study:

• To implement and evaluate a trust-region algorithm for EIT image reconstruction.
• To assess the performance of the dogleg trust-region method against the Levenberg-Marquardt algorithm.
• To determine the efficiency and accuracy of the dogleg method for abnormality detection in EIT.

Main Methods:

• A trust-region framework was developed for EIT image reconstruction.
• The dogleg trust-region algorithm was applied, approximating Levenberg-Marquardt steps with a quadratic model.
• Performance was evaluated by comparing reconstructed images from both algorithms.

Main Results:

• The dogleg trust-region algorithm demonstrated a significant reduction in execution time, achieving less than 50% of the Levenberg-Marquardt algorithm's time.
• No quantifiable loss in the quality of reconstructed images was observed with the dogleg method.
• The dogleg method proved effective in detecting abnormalities within the tested cases.

Conclusions:

• The dogleg trust-region algorithm offers a computationally efficient alternative for EIT image reconstruction.
• This method provides a viable approach for faster abnormality detection in EIT without compromising image quality.
• The findings suggest practical advantages for implementing the dogleg method in clinical or research settings for EIT.