Adaptive Grouping Block Sparse Bayesian Learning Method for Accurate and Robust Reconstruction in Bioluminescence Tomography

Objective:Bioluminescence tomography (BLT) is a promising modality that is designed to provide non-invasive quantitative three-dimensional information regarding the tumor distribution in living animals. However, BLT suffers from inferior reconstructions due to its ill-posedness. This study aims to improve the reconstruction performance of BLT. Methods: We propose an adaptive grouping block sparse Bayesian learning (AGBSBL) method, which incorporates the sparsity prior, correlation of neighboring mesh nodes, and anatomical structure prior to balance the sparsity and morphology in BLT. Specifically, an adaptive grouping prior model is proposed to adjust the grouping according to the intensity of the mesh nodes during the optimization process. Results: Numerical simulations and in vivo experiments demonstrate that AGBSBL yields a high position and morphology recovery accuracy, stability, and practicality. Conclusion: The proposed method is a robust and effective reconstruction algorithm for BLT. Moreover, the proposed adaptive grouping strategy can further increase the practicality of BLT in biomedical applications.