Bioluminescence tomography (BLT) is an important optical molecular imaging modality aimed at visualizing physiological and pathological processes at cellular and molecular levels. While the forward process of light propagation is described by the diffusion approximation to radiative transfer equation, BLT is the inverse problem to reconstruct the 3D localization and quantification of internal bioluminescent sources distribution. Due to the inherent ill-posedness of the BLT problem, regularization is generally indispensable to obtain more favorable reconstruction. In particular, total variation (TV) regularization is known to be effective for piecewise-constant source distribution which can permit sharp discontinuities and preserve edges. However, total variation regularization generally suffers from the unsatisfactory staircasing effect. In this work, we introduce the Bregman iterative regularization to alleviate this degeneration and enhance the numerical reconstruction of BLT. Based on the existing Landweber method (LM), we put forward the Bregman-LM-TV algorithm for BLT. Numerical experiments are carried out and preliminary simulation results are reported to evaluate the proposed algorithms. It is found that Bregman-LM-TV can significantly outperform the individual Landweber method for BLT when the source distribution is piecewise-constant.

Date Published: 17 September 2014
PDF: 7 pages
Proc. SPIE 9230, Twelfth International Conference on Photonics and Imaging in Biology and Medicine (PIBM 2014), 92300I (17 September 2014); doi: 10.1117/12.2068841

Published in SPIE Proceedings Vol. 9230:
Twelfth International Conference on Photonics and Imaging in Biology and Medicine (PIBM 2014)
Qingming Luo; Lihong V. Wang; Valery V. Tuchin, Editor(s)