Novel Stopping Criteria for Optimization-Based Microwave Breast Imaging Algorithms

Area of Science:

• Medical Imaging
• Computational Electromagnetics
• Applied Mathematics

Background:

• Microwave breast imaging utilizes the Contrast Source Inversion (CSI) algorithm, often implemented with a discontinuous Galerkin formulation (DGM-CSI).
• Frequency-cycling reconstruction techniques enhance image quality but require appropriate stopping points for frequency shifts and algorithm termination.
• Previous DGM-CSI methods used fixed iteration counts, leading to arbitrary and potentially suboptimal reconstruction.
• Tissue-dependent geometrical mapping has improved initial guesses in frequency hopping, but lacked automated control over the inversion process.

Purpose of the Study:

• To introduce automated stopping criteria for the DGM-CSI algorithm in microwave breast imaging.
• To enhance the efficiency and objectivity of frequency-cycling reconstruction.
• To improve the determination of optimal times for frequency shifts and global algorithm termination.

Main Methods:

• Modification of the discontinuous Galerkin formulation of the Contrast Source Inversion algorithm (DGM-CSI).
• Implementation of automated stopping criteria based on statistical analysis of data error.
• Utilizing the two-sample Kolmogorov-Smirnov (K-S) test to analyze data error distribution patterns across past iterations.
• Applying these criteria to determine optimal frequency shifts and algorithm termination points.

Main Results:

• The developed stopping criteria intelligently identify suitable moments to shift imaging frequencies during reconstruction.
• The criteria provide an objective method for globally terminating the DGM-CSI algorithm.
• The automated stopping criteria improve the efficiency of DGM-CSI reconstructions.
• Image quality achieved with the new criteria is comparable to reconstructions using a fixed, often overestimated, number of iterations.

Conclusions:

• Automated stopping criteria based on the Kolmogorov-Smirnov test offer an intelligent and efficient approach to DGM-CSI in microwave breast imaging.
• This method enhances the practical application of frequency-cycling reconstruction by removing arbitrary iteration limits.
• The findings suggest improved objectivity and performance in dielectric property imaging of breast tissue models.