Reconstruction algorithm for photoacoustic tomography based on the L-alternating direction method of multipliers model

基于L交替方向乘子法的光声层析成像重建算法

阅读:1

Abstract

SIGNIFICANCE: Photoacoustic tomography (PAT) is an emerging biomedical imaging technology that offers high contrast and high resolution, showing great potential for applications in medical imaging. However, existing regularization methods often lead to instability and artifacts in the reconstruction due to imbalanced regularization parameter settings. To address these issues, we propose a reconstruction algorithm based on the L-alternating direction method of multipliers (ADMM) for PAT, which significantly improves image reconstruction quality and has high clinical application potential. AIM: We introduce a nonconvex L1-L2 norm into the variational model and employ the ADMM to decompose the optimization problem into efficiently solvable subproblems. A preconditioned conjugate gradient (PCG) method is further integrated to accelerate the solution of linear systems, thereby improving both reconstruction accuracy and computational efficiency. APPROACH: We propose an L-ADMM framework with adaptive weighted L1-L2 regularization for PAT reconstruction. The method employs ADMM to split the optimization into tractable subproblems and uses PCG to efficiently solve linear systems. It achieves stable, high-quality reconstruction under sparse sampling by enhancing sparsity while preserving structural details. RESULTS: Experiments on vascular and breast models demonstrate that, even with only 64 transducers under sparse sampling, the proposed L-ADMM method achieves peak signal-to-noise ratio values of 37.24 and 36.26 dB and structural similarity index measure values of 0.9766 and 0.9665, respectively. Compared with L2, L1 + L2, L1-L2, TV regularization, and U-Net methods, the proposed algorithm substantially improves image quality, highlighting its feasibility for cost-effective clinical PAT. CONCLUSIONS: The proposed L-ADMM-based reconstruction algorithm, by integrating adaptive regularization with efficient optimization, significantly improves PAT image quality under sparse sampling conditions, offering a feasible solution with strong potential for clinical translation.

特别声明

1、本页面内容包含部分的内容是基于公开信息的合理引用;引用内容仅为补充信息,不代表本站立场。

2、若认为本页面引用内容涉及侵权,请及时与本站联系,我们将第一时间处理。

3、其他媒体/个人如需使用本页面原创内容,需注明“来源:[生知库]”并获得授权;使用引用内容的,需自行联系原作者获得许可。

4、投稿及合作请联系:info@biocloudy.com。