Blind Deconvolution of Ultrasound Images Using l1 -Norm-Constrained Block-Based Damped Variable Step-Size...

Area of Science:

• Medical Imaging
• Signal Processing
• Biomedical Engineering

Background:

• Improving medical ultrasound image resolution is crucial for accurate diagnosis.
• Ultrasound image quality is degraded by convolution effects from the system's point spread function (PSF).
• Estimating tissue reflectivity functions (TRFs) and PSF directly from radio-frequency (RF) data is challenging due to noise and nonstationarity.

Purpose of the Study:

• To develop a blind method for estimating TRFs and PSF from noisy, nonstationary ultrasound RF data without prior knowledge.
• To improve the resolution and accuracy of medical ultrasound image reconstruction.
• To enhance the performance of ultrasound imaging systems.

Main Methods:

• A modified block-based cross-relation equation was developed to handle nonstationarity and incomplete data acquisition.
• An l1-norm regularized cost function was formulated for blind TRF estimation using a new l1-norm block-based multichannel least-mean square (l1-bMCLMS) algorithm.
• A damped variable step-size was incorporated for noise compensation and faster convergence.
• The PSF was estimated using the regularized multiple-input/output inverse theorem.

Main Results:

• The proposed l1-bMCLMS algorithm achieved significant improvements in TRF estimation, with resolution gain (RG) of 0.12–5.2 and normalized projection misalignment (NPM) of 3.34–22.82 dB.
• The PSF estimation showed substantial improvement, with a shifted normalized mean square error (snMSE) improvement of the order 10^2–10^4.
• The method demonstrated efficacy on simulation, experimental phantom, and in vivo RF data without requiring basis functions for TRFs or PSF.

Conclusions:

• The developed blind estimation technique effectively improves ultrasound image resolution by accurately estimating TRFs and PSF.
• The method offers a robust solution for enhancing medical ultrasound imaging without requiring prior information about the system's PSF.
• The proposed approach provides significant quantitative improvements in image quality metrics, making it a valuable tool for medical ultrasound applications.