Three-dimensional optoacoustic reconstruction using fast sparse representation

Abstract

Optoacoustic tomography based on insufficient spatial sampling of ultrasound waves leads to loss of contrast and artifacts on the reconstructed images. Compared to reconstructions based on L2-norm regularization, sparsity-based reconstructions may improve contrast and reduce image artifacts but at a high computational cost, which has so far limited their use to 2D optoacoustic tomography. Here we propose a fast, sparsity-based reconstruction algorithm for 3D optoacoustic tomography, based on gradient descent with Barzilai-Borwein line search (L1-GDBB). Using simulations and experiments, we show that the L1-GDBB offers fourfold faster reconstruction than the previously reported L1-norm regularized reconstruction based on gradient descent with backtracking line search. Moreover, the new algorithm provides higher-quality images with fewer artifacts than the L2-norm regularized reconstruction and the back-projection reconstruction.

Photoacoustic imaging depth comparison at 532-, 800-, and 1064-nm wavelengths: Monte Carlo simulation and experimental validation

“…Reconstruction algorithms are applied on the acquired signals to form images of the internal structure of the object. [18][19][20] On the other hand, PAM generally employs confocal alignment of optical illumination and US detection, and by raster scanning of the sample, three-dimensional volumetric information of the object is obtained. Such alignment provides PAM the advantage of better imaging resolution, but at a cost of relatively poor imaging depth.…”

A Synthetic Total Impulse Response Characterization Method for Correction of Hand-Held Optoacoustic Images

“…Nevertheless, numerical simulations may contain assumptions and simplifications which do not reflect the actual system modelled and result in loss of image quality. Different aspects of the optoacoustic image reconstruction problem have been investigated to improve image quality; e.g., regularization schemes [12], [13], image/signal processing methods [14], [15] and modelling the underlying physics [16]- [18]. Ideally a suitable combination of all these methods is expected to achieve the best image quality.…”

Maximum Entropy Based Non-Negative Optoacoustic Tomographic Image Reconstruction

“…For example, 2 -or 1 -norm minimization of the total variation of an image minimizes the edges of the reconstructed image. Using this notion, negative artifacts can then be eliminated by applying an explicit non-negativity constraint along with 2 -norm minimization [13], [15]. Another image metric that has been considered for eliminating negative values is the entropy of an image [16], [17].…”