PAT image reconstruction using augmented sparsity regularization with practicable tuning of regularization weight

TL;DR

This paper presents a semi-automated method for tuning regularization weights in photo-acoustic tomography (PAT) image reconstruction, improving image quality without extensive parameter adjustments across different noise levels and image details.

Contribution

It introduces a relative smoothness constraint parameter that maps to the regularization weight, which remains stable across noise variations and image structures, along with an integrated reconstruction algorithm.

Findings

The method achieves high-quality PAT reconstructions with a nominal parameter value.

The regularization weight mapping is robust to noise and image content variations.

Experimental results validate the effectiveness of the proposed tuning approach.

Abstract

Among all tissue imaging modalities, photo-acoustic tomography (PAT) has been getting increasing attention in the recent past due to the fact that it has high contrast, high penetrability, and has capability of retrieving high resolution. Model-based regularized reconstruction methods used in PAT plays a crucial role in the applicability of PAT because such methods can yield high quality images from measurements obtained using fewer number of transducers. A crucial factor that determines the quality of reconstruction in such methods is the choice of regularization weight. Unfortunately, an appropriately tuned value of regularization weight varies significantly with variation in the noise level, as well as, with the variation in the high resolution contents of the image, in a way that has not been well understood. There has been attempts to determine optimum regularization weight from the measured data in the context of using elementary and general purpose regularizations. In this paper, we develop a practicable method for semi-automated tuning of the regularization weight in the context of using a modern type of regularization that is specifically designed for PAT image reconstruction. As a first step, we introduce a relative smoothness constraint with a parameter; this parameter computationally maps into the actual regularization weight, but, its tuning does not vary significantly with variation in the noise level, and with the variation in the high resolution contents of the image. Next, we construct an algorithm that integrates the task of determining this mapping along with obtaining the reconstruction. Finally we demonstrate experimentally that we can run this algorithm with a nominal value of the relative smoothness parameter -- a value independent of the noise level and the structure of the underlying image -- to obtain good quality reconstructions.