Proceedings Article | 3 March 2017
KEYWORDS: Acoustics, Sensors, Calibration, Absorption, 3D image processing, Reconstruction algorithms, Scattering, Image segmentation, 3D modeling, Chromophores, Photoacoustic tomography, Model-based design, Image resolution, Nonlinear optics, Distortion, In vivo imaging, Diffusion, Optical properties
Photoacoustic tomography can, in principle, provide quantitatively accurate, high-resolution, images of chromophore distributions in 3D in vivo. However, achieving this goal requires not only dealing with the optical fluence-related spatial and spectral distortion but also having access to high quality, calibrated, measurements and using image reconstruction algorithms free from inaccurate assumptions. Furthermore, accurate knowledge of experimental parameters, such as the positions of the ultrasound detectors and the illumination pattern, is necessary for the reconstruction step. A meticulous and rigorous experimental phantom study was conducted to show that highly-resolved 3D estimation of chromophore distributions can be achieved: a crucial step towards in vivo implementation. The phantom consisted of four 580 μm diameter tubes with different ratios of copper sulphate and nickel sulphate as hemoglobin analogues, submersed in a background medium of intralipid and india ink. The optical absorption, scattering, photostability, and Grüneisen parameter were characterised for all components independently. A V-shaped imaging scanner enabled 3D imaging with the high resolution, high sensitivity, and wide bandwidth characteristic of Fabry-Pérot ultrasound sensors, but without the limited-view disadvantage of single-plane scanners. The optical beam profile and position were determined experimentally. Nine wavelengths between 750 and 1110 nm were used. The images of the chromophore concentrations were obtained using a model-based, two-step, procedure, that did not require image segmentation. First, the acoustic reconstruction was solved with an iterative time-reversal algorithm to obtain images of the initial acoustic pressure at each of the nine wavelengths for an 18×17×13 mm3 volume with 50μm voxels. Then, 3D high resolution estimates of the chromophore concentrations were obtained by using a diffusion model of light transport in an iterative nonlinear optimisation scheme. Among the lessons to be drawn from this study, one is fundamental: in order to obtain accurate estimates of chromophores (or their ratios) it is not only necessary to model the light fluence accurately, but it is just as crucial to obtain accurate estimates of the initial acoustic pressure distributions, and to account for variations in the thermoelastic efficiency (Grüneisen parameter).
