Regularized reconstruction based on joint smoothly clipped absolute deviation regularization and graph manifold...

Area of Science:

• Biomedical Imaging
• Optical Imaging
• Preclinical Research

Background:

• Fluorescence Molecular Tomography (FMT) offers sensitive, low-cost 3D biomarker imaging for preclinical cancer studies.
• Current FMT reconstruction methods struggle with accuracy in morphology and location due to light scattering and ill-posed inverse problems.

Purpose of the Study:

• To develop an improved regularized reconstruction method for FMT.
• To enhance the accuracy of fluorescence distribution reconstruction in terms of morphology and location.

Main Methods:

• Introduced a novel regularized reconstruction method: joint smoothly clipped absolute deviation regularization and graph manifold learning (SCAD-GML).
• Combined sparsity of fluorescent sources with latent manifold structure for accurate and sparse reconstruction.
• Employed a non-convex gradient descent iterative method for efficient objective function solution.
• Validated the SCAD-GML method using numerical simulations and in vivo experiments.

Main Results:

• The SCAD-GML method demonstrated superior performance compared to existing methods.
• Achieved enhanced accuracy in both location and shape recovery of fluorescence biomarker distribution.
• Numerical simulations and in vivo experiments confirmed the method's effectiveness.

Conclusions:

• The SCAD-GML method significantly improves FMT reconstruction accuracy.
• This advancement holds potential for enhancing in vivo biological research applications of FMT.
• The method addresses key limitations of current FMT techniques, paving the way for more precise preclinical diagnostics.