Photoacoustic Image Formation Based on Sparse Regularization of Minimum Variance Beamformer

TL;DR

This paper introduces a modified sparse regularization approach for the minimum variance beamformer in photoacoustic imaging, significantly enhancing image resolution and sidelobe suppression over traditional methods.

Contribution

A novel MS-MV algorithm incorporating L1-norm sparsity constraint is proposed, improving sidelobe suppression and SNR in photoacoustic image formation.

Findings

SNR improved by approximately 19.48 dB in simulations

Experimental SNR increased by about 2.64 dB with MS-MV

Efficient convex optimization enables practical implementation

Abstract

Delay-and-Sum (DAS) is the most common algorithm used in photoacoustic (PA) image formation. However, this algorithm results in a reconstructed image with a wide mainlobe and high level of sidelobes. Minimum variance (MV), as an adaptive beamformer, overcomes these limitations and improves the image resolution and contrast. In this paper, a novel algorithm, named modified-sparse-MV (MS-MV) is proposed in which a L1-norm constraint is added to the MV minimization problem after some modifications, in order to suppress the sidelobes more efficiently, compared to MV. The added constraint can be interpreted as the sparsity of the output of the MV beamformed signals. Since the final minimization problem is convex, it can be solved efficiently using a simple iterative algorithm. The numerical results show that the proposed method, MS-MV beamformer, improves the signal-to-noise (SNR) about 19.48 dB, in average, compared to MV. Also, the experimental results, using a wire-target phantom, show that MS-MV leads to SNR improvement of about 2.64 dB in comparison with the MV.