L1-L2 norm regularization via forward-backward splitting for fluorescence molecular tomography

Area of Science:

• Biomedical Imaging
• Medical Physics
• Computational Biology

Background:

• Fluorescent molecular tomography (FMT) offers sensitive, noninvasive 3D imaging of fluorescent probes.
• Reconstruction challenges in FMT include light scattering and ill-posed inverse problems, hindering accurate source localization and morphology.
• Existing methods struggle with precise localization and morphology of fluorescence sources.

Purpose of the Study:

• To enhance FMT reconstruction accuracy and robustness for early tumor detection.
• To develop a novel L1-L2 norm regularization method for improved solution sparsity.
• To address the challenges of light scattering and inverse problem ill-posedness in FMT.

Main Methods:

• Proposed a novel L1-L2 norm regularization technique.
• Utilized forward-backward splitting method to solve the nonconvex L1-L2 norm minimization problem.
• Employed analytic solutions for L1-L2 norm proximal operators.

Main Results:

• The proposed algorithm demonstrated superior reconstruction performance in numerical simulations and in vivo experiments.
• Achieved enhanced accuracy in spatial location and dual-source resolution.
• Showcased improved in vivo practicability for glioma mouse models.

Conclusions:

• The novel L1-L2 norm regularization method significantly improves FMT reconstruction.
• This advancement is expected to promote preclinical and clinical applications of FMT in early tumor detection.
• The algorithm offers enhanced accuracy, robustness, and practicability for FMT imaging.