Half Thresholding Pursuit Algorithm for Fluorescence Molecular Tomography.

Area of Science:

• Biomedical Imaging
• Optical Imaging
• Computational Imaging

Background:

• Fluorescence Molecular Tomography (FMT) is a key optical imaging technique for small animal research.
• The ill-posed nature of FMT inverse problems necessitates advanced reconstruction algorithms.
• L1/2-norm regularization offers sparsity enhancement but requires efficient algorithms for nonconvex models.

Purpose of the Study:

• To develop and evaluate an efficient algorithm for solving nonconvex regularized models in FMT.
• To improve the accuracy and speed of FMT image reconstruction.
• To address the challenges posed by the ill-posed inverse problem in FMT.

Main Methods:

• Proposed a Half Thresholding Pursuit Algorithm (HTPA) for solving the L1/2-norm regularized FMT model.
• Integrated a pursuit strategy to accelerate iterative reconstruction.
• Implemented a parameter optimization scheme for robust performance.

Main Results:

• HTPA demonstrated superior location accuracy in reconstructed images compared to existing methods.
• Achieved higher reconstructed fluorescent yield with HTPA.
• The proposed algorithm offered a reduced time cost for image reconstruction.
• Validated performance using both simulated and experimental FMT data.

Conclusions:

• The HTPA, coupled with parameter optimization, provides an efficient and robust reconstruction approach for FMT.
• This method advances the capabilities of optical imaging in small animal studies.
• The algorithm effectively handles the ill-posed inverse problem inherent in FMT.