To recover absolute blood oxygenation from high resolution PAT images, the fluence needs to be corrected. However, qPAT is a large scale inverse problem. Finite difference gradient methods or gradient-free parameter searches are unsuitable for inverting 3D images. Adjoint methods by contrast provide the parameter search gradient efficiently. An adjoint radiative transfer equation formulation using a Monte-Carlo fluence approximation is used and validated in silico and on tissue phantom data. Convergence and limitations are discussed.

This presentation is part of Minisymposium “MS61 - Imaging with Light and Sound (3 parts)”
organized by: Felix Lucka (CWI & UCL) , Tanja Tarvainen (University of Eastern Finland) .

Authors:

Jan Laufer (Martin-Luther-Univ Halle-Wittenberg )

Bernhard Kaplan (Zuse Institute Berlin)

Keywords:

image segmentation, inverse problems, nonlinear optimization, statistical inverse estimation methods