Cone-beam X-ray luminescence computed tomography via Laplacian scale mixture prior-driven variational Bayesian method

Area of Science:

• Medical Imaging
• Biophysics
• Computational Biology

Background:

• Cone-beam X-ray luminescence computed tomography (CB-XLCT) is an emerging optical molecular imaging technique for monitoring dynamic changes in early-stage tumors.
• Reconstruction in CB-XLCT is challenging due to low absorption and high scattering properties of biological tissues.
• Sparse signal recovery is crucial for accurate CB-XLCT imaging.

Purpose of the Study:

• To propose a novel variational Bayesian method for sparse signal modeling and recovery in CB-XLCT.
• To enhance the reconstruction accuracy of CB-XLCT imaging.
• To improve the potential of CB-XLCT for preclinical and clinical applications.

Main Methods:

• Developed a variational Bayesian framework utilizing a Laplacian scale mixture prior for sparse signal modeling.
• Employed an inverse gamma distribution for the scale variable, allowing adaptive sparsity representation.
• Applied Laplace approximation to convert the intractable posterior distribution into a Gaussian distribution.
• Utilized the expectation-maximization (EM) algorithm within the variational Bayesian structure for signal recovery.

Main Results:

• The proposed method demonstrated substantial improvements in CB-XLCT reconstruction accuracy.
• Numerical simulations and implantation experiments confirmed superior performance in source localization.
• The method achieved enhanced morphological restoration of signals.
• The variational Bayesian approach effectively addressed the challenges posed by tissue scattering and low absorption.

Conclusions:

• The developed variational Bayesian method significantly enhances sparse signal recovery for CB-XLCT.
• The technique offers improved accuracy in source localization and morphological restoration.
• This advancement holds considerable potential for the future development of CB-XLCT in preclinical and clinical settings.