Total Variation Constrained Graph Manifold Learning Strategy for Cerenkov Luminescence Tomography.

Area of Science:

• Biomedical Imaging
• Medical Physics
• Radiochemistry

Background:

• Cerenkov luminescence tomography (CLT) offers non-invasive visualization of tumor cells.
• Current CLT methods face limitations in spatial resolution and shape recovery due to photon scattering and ill-posed inverse problems.

Purpose of the Study:

• To develop an advanced strategy for improved spatial resolution, dual-source resolution, and tumor morphology in CLT.
• To address the inherent challenges in CLT reconstruction for preclinical applications.

Main Methods:

• Proposed a total variation constrained graph manifold learning (TV-GML) strategy.
• Integrated isotropic total variation and dynamic graph Laplacian constraints for edge preservation and smooth reconstruction.
• Employed a tetrahedral mesh-Cartesian grid pair method and adaptive Barzilai-Borwein optimization for efficient convergence.

Main Results:

• TV-GML demonstrated superior performance in simulation and in vivo experiments.
• Achieved enhanced accuracy in location, dual-source resolution, and shape recovery.
• Showcased robustness and practical applicability for in vivo imaging.

Conclusions:

• The novel TV-GML method significantly improves CLT reconstruction performance.
• This technique is beneficial for quantitative analysis and morphological observation in preclinical CLT applications.
• Facilitates advancements in solving inverse problems within biomedical imaging.