Improved Block Sparse Bayesian Learning Method Using K-Nearest Neighbor Strategy for Accurate Tumor Morphology Reconstruction in Bioluminescence Tomography

Objective: Bioluminescence tomography (BLT) is a non-invasive technique designed to enable three-dimensional (3D) visualization and quantification of viable tumor cells in living organisms. However, despite the excellent sensitivity and specificity of bioluminescence imaging (BLI), BLT is limited by the photon scattering effect and ill-posed inverse problem. If the complete structural information of a light source is considered when solving the inverse problem, reconstruction accuracy will be improved. Methods: This paper proposed a block sparse Bayesian learning method based on K-nearest neighbor strategy (KNNBSBL), which incorporated several types of a priori information including sparsity, spatial correlations among neighboring points, and anatomical information to balance over-sparsity and morphology preservation in BLT. Furthermore, we considered the Gaussian weighted distance prior in a light source and proposed a KNN-GBSBL method to further improve the performance of KNN-BSBL. Results: The results of numerical simulations and in vivo glioma-bearing mouse experiments demonstrated that KNNBSBL and KNN-GBSBL achieved superior accuracy for tumor spatial positioning and morphology reconstruction. Conclusion: The proposed method KNN-BSBL incorporated several types of a priori information is an efficient and robust reconstruction method for BLT.