Second-order adjoint sensitivities for fluorescence optical tomography based on the SPN approximation.

Abstract

Use of second-order sensitivity information has been shown in the literature to yield faster convergence, better noise tolerance, and localization besides enhanced post-reconstruction analysis capabilities. In this paper, we derive adjoint-based second-order derivatives for SPN-approximation-modeled fluorescence optical tomography. We modify the regularizing Levenberg-Marquardt method to use second-order sensitivity information through a predictor-corrector framework. Reconstruction studies presented for the fluorophore absorption coefficient in low as well as high scattering tissue-mimicking phantoms in both ideal and differential fluorophore-uptake settings show consistently superior noise tolerance and contrast recovery with the second-order scheme as compared to its first-order counterpart.