Sparsity constrained contrast source inversion.

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Computational Physics

Background:

• Ultrasound imaging is crucial for breast lesion detection and characterization.
• Contrast Source Inversion (CSI) is an advanced method for solving the nonlinear inverse problem in ultrasound imaging.
• Traditional CSI can diverge in noisy environments, and Total Variation (TV) regularization alone is insufficient for very noisy data.

Purpose of the Study:

• To develop and evaluate novel regularization techniques for CSI to enhance breast lesion detection accuracy.
• To improve the robustness and convergence of CSI in the presence of measurement noise.
• To investigate the efficacy of compressed sensing and combined sparsity-TV regularization for ultrasound imaging.

Main Methods:

• Incorporating compressed sensing principles to enforce sparsity in a transformation domain for CSI regularization.
• Developing a method that combines contrast source sparsity with minimal Total Variation (TV) in the contrast function.
• Estimating contrast source and function by minimizing mean squared error and adding a sparsity penalty term.

Main Results:

• Sparsity-constrained CSI improved normalized mean squared error by up to 36% for data with 1% noise compared to traditional CSI.
• Reconstruction quality improved by up to 70% for data with 5% noise using the proposed methods.
• The novel methods demonstrated consistent convergence to the correct speed-of-sound profile with increasing iterations, even in noisy conditions.

Conclusions:

• Compressed sensing-based regularization significantly enhances the accuracy and robustness of CSI for breast ultrasound imaging.
• The proposed methods offer a substantial improvement over traditional CSI and TV regularization, particularly in noisy conditions.
• These advancements hold promise for more reliable and accurate breast lesion detection and characterization using ultrasound.