Quantitative inversion of 3D multiwavelength photoacoustic images using an adjoint radiance Monte Carlo model (Conference Presentation)

Abstract

Quantitative photoacoustic (PA) tomography aims to recover absolute chromophore concentrations from multiwavelength PA images. Challenges include the accurate prediction of the fluence, the accuracy of the initial pressure distribution reconstructed from measured data, and the large scale of the inverse problem involving high resolution 3D images. In this study, a radiance Monte-Carlo (RMC) light model was used to predict the fluence inside tissue phantoms. Gradients of the scattering coefficient and the chromophore concentrations were calculated using the adjoint formalism. The gradient descent efficiency was significantly improved by using adaptive moment estimation. 3D maps of chromophore concentrations and the scattering coefficient were recovered from measured PA images. The inversion scheme was validated on measured images of a tissue phantom consisting of a scattering liquid and chromophore-filled polymer tubes immersed at different depths. The images were acquired at visible and near-infrared wavelengths using a Fabry-Perot scanner with a planar detection geometry. Amplitude mismatches in the reconstructed initial pressure images due to limited view detection were corrected using an ad hoc correction method. The inversion was stabilized by introducing a calibrated absorber in the imaged volume, or an absolute calibration of the setup. 3D maps of absolute chromophore concentrations, their ratios, and the global scattering coefficient were accurately recovered. The recovery of chromophore concentrations in the image background where SNR is low was identified as a significant new challenge for quantitative PA imaging.