Sparse reconstruction based on dictionary learning and group structure strategy for cone-beam X-ray luminescence...

Area of Science:

• Biomedical Imaging
• Medical Physics
• Computational Biology

Background:

• Cone-beam X-ray luminescence computed tomography (CB-XLCT) is a promising dual-modal imaging technique for early 3D tumor detection in small animals.
• Tissue optical properties (low absorption, high scattering) create ill-conditioned inverse problems, hindering satisfactory CB-XLCT reconstruction.
• Existing methods struggle with accuracy due to the complex nature of light propagation in biological tissues.

Purpose of the Study:

• To develop and evaluate a novel reconstruction strategy for CB-XLCT that addresses the challenges of ill-posed inverse problems.
• To enhance the accuracy and practical applicability of CB-XLCT for preclinical tumor imaging.
• To leverage advanced computational techniques for improved 3D imaging performance.

Main Methods:

• A novel strategy utilizing dictionary learning and group structure (DLGS) was proposed for CB-XLCT reconstruction.
• Group structure was employed to model the clustering of nanophosphors, enhancing inter-element relationships.
• Dictionary learning was implemented to effectively capture sparse features within the imaging data.

Main Results:

• The proposed DLGS method demonstrated superior reconstruction performance in numerical simulations and in vivo experiments.
• Key improvements were observed in location accuracy, target shape fidelity, and robustness against noise.
• Enhanced dual-source resolution and in vivo practicability were validated, showcasing the method's effectiveness.

Conclusions:

• The DLGS strategy significantly improves CB-XLCT reconstruction performance for preclinical tumor detection.
• This method offers a robust and accurate solution for overcoming the inherent challenges in CB-XLCT imaging.
• The findings support the potential of DLGS-enhanced CB-XLCT for early and precise 3D tumor visualization in small animal models.