Model resolution matrix based deconvolution improves over non-quadratic penalization in frequency-domain photoacoustic tomography

• Published in: The Journal of the Acoustical Society of America 2022-09, Vol.152 (3), p.1345-1356
• Main Authors: Nakshatri, Hemanth S., Prakash, Jaya
• Format: Article
• Language: English
• ISSN: 0001-4966; 1520-8524
• DOI: 10.1121/10.0013829

Frequency domain photoacoustic tomography is becoming more attractive due to low-cost and compact light-sources being used; however, frequency-domain implementation suffers from lower signal to noise compared to time-domain implementation. In this work, we have developed a non-quadratic based penalization framework for frequency-domain photoacoustic imaging, and further proposed a two-step model-resolution matrix based deconvolution approach to improve the reconstruction image quality. The model-resolution matrix was developed in the context of different penalty functions like l2-norm, l1-norm, Cauchy, and Geman-McClure. These model-resolution matrices were then used to perform the deconvolution operation using split augmented Lagrangian shrinkage thresholding algorithm in both full-view and limited-view configurations. The results indicated that the two-step approach outperformed the different penalty function (prior constraint) based reconstruction, with an improvement of about 20% in terms of peak signal to noise ratio and 30% in terms of structural similarity index measure. The improved image quality provided using these algorithms will have a direct impact on realizing practical frequency-domain implementation in both limited-view and full-view configurations.