Compressive sensing for ultrasound RF echoes using a-Stable Distributions.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Data Science

Background:

• Biomedical ultrasonic signal acquisition often faces challenges with data sparsity and noise.
• Traditional compressive sensing methods may not optimally handle the non-Gaussian nature of some biomedical signals.
• Stable distributions offer a flexible model for characterizing complex signal properties.

Purpose of the Study:

• To develop a novel compressive sensing framework for biomedical ultrasonic signals.
• To improve signal reconstruction quality by modeling data with stable distributions.
• To introduce an efficient algorithm for ℓ(p) norm minimization tailored for alpha-stable data.

Main Methods:

• Development of a compressive sensing framework utilizing stable distribution modeling.
• Proposal of an iteratively reweighted least squares (IRLS) algorithm adapted for ℓ(p) norm minimization.
• Parameter 'p' in the IRLS algorithm is determined by the characteristic exponent of alpha-stable distributed data.
• Introduction of the S ± S-IRLS algorithm.

Main Results:

• The S ± S-IRLS algorithm demonstrates superior performance in reconstructing biomedical ultrasonic signals.
• Visual inspection and Peak Signal-to-Noise Ratio (PSNR) metrics confirm the algorithm's effectiveness.
• The proposed method outperforms established ℓ(1) minimization algorithms like basis pursuit and orthogonal matching pursuit.

Conclusions:

• The novel S ± S-IRLS framework provides enhanced compressive sensing for biomedical ultrasound.
• Modeling data with stable distributions and adapting IRLS significantly improves signal reconstruction.
• This approach offers a valuable advancement for biomedical signal processing applications.