Implementation and Validation of the Multi-scale Gradient Smoothing Strategy for Breast Full Waveform Inversion.

Area of Science:

• Medical Imaging
• Computational Physics
• Biomedical Engineering

Background:

• Full waveform inversion (FWI) is crucial for high-resolution sound speed imaging in ultrasound computed tomography.
• FWI is prone to cycle skipping, leading to inaccurate reconstructions and local minima.
• Mitigating cycle skipping is essential for reliable quantitative sound speed measurements.

Purpose of the Study:

• To propose a modified FWI algorithm to address cycle skipping.
• To improve the convergence and accuracy of sound speed image reconstruction.
• To enhance the robustness of FWI in ultrasound computed tomography.

Main Methods:

• Incorporation of low-wavenumber extraction in the image domain using Gaussian gradient filtering (GGF).
• Development of a multi-scale GGF strategy with progressively reduced low-pass filter standard deviation.
• Application of gradient smoothing for stable updates in the inversion process.

Main Results:

• The GGF + FWI method significantly reduced root mean square error and improved structural similarity compared to conventional FWI.
• In vivo breast experiments demonstrated the robustness of the proposed approach.
• Multi-scale GGF achieved the highest contrast-to-noise ratio in comparative analysis.

Conclusions:

• The proposed GGF-based FWI approach improves optimization convergence properties.
• The method effectively mitigates artefacts associated with cycle skipping.
• This technique enhances the reliability of quantitative sound speed imaging.