Higher-order correlation based real-time beamforming in photoacoustic imaging

TL;DR

This paper introduces a higher-order correlation based beamformer for photoacoustic imaging that enhances image quality and can be implemented in real-time, outperforming traditional methods in resolution, contrast, and SNR.

Contribution

The paper proposes a novel higher-order correlation beamformer that improves photoacoustic image quality while maintaining real-time computational efficiency.

Findings

51% improvement in FWHM over DAS

81% improvement in SNR over DAS

Comparable computational time to DAS and DMAS

Abstract

Linear-array based photoacoustic images are reconstructed using the conventional delay-and-sum (DAS) beamforming method. Although the DAS beamformer is well suited for PA image formation, reconstructed images are often afflicted by noises, sidelobes, and other intense artifacts due to inaccurate assumptions of PA signal correlation. The work aims to develop an inversion method that reduces the occurrence of sidelobes and artifacts and improves image quality performance. We present a novel beamformer based on higher-order signal correlation, where more number of delayed PA signals are combined and summed up than the conventional delay-multiply-and-sum (DMAS). The proposed technique provides efficient improvements in resolution, contrast, and SNR compared to the traditional beamformers. Computational complexity in this method is shrunk to the same order of DAS $O(N)$. Therefore, this beamformer can be implemented in real-time PA image reconstruction. A GPU based study was performed on computation time. Proposed method almost executes in the same time frame of DAS and real-time DMAS. A validation study of the algorithm was accomplished both numerically and experimentally. Higher-order DMAS beamformers demonstrate superior reconstruction in all cases. The quantitative evaluation of the ex-vivo phantom shows that the proposed method leads to 51% and 6% improvement in FWHM, 81% and 39% improvement in SNR compared to DAS and DMAS, respectively. Conclusively, the proposed algorithm is very much potential and promising in real-time photoacoustic imaging and its applications.