Improving mesoscopic fluorescence molecular tomography via preconditioning and regularization.

Area of Science:

• Biomedical imaging
• Optical tomography
• Molecular imaging

Background:

• Mesoscopic fluorescence molecular tomography (MFMT) offers 3-D molecular probe distribution imaging in biological tissues.
• The technique suffers from an ill-conditioned inverse problem, limiting reconstruction performance.
• High-resolution imaging exacerbates sensitivity matrix correlation, hindering advanced algorithms.

Purpose of the Study:

• To develop computational methods to improve MFMT reconstruction quality.
• To address the ill-conditioned nature of the MFMT inverse problem.
• To enhance the convergence rate of the inverse solver.

Main Methods:

• Applying a compressed sensing (CS) based preconditioner to reduce sensitivity matrix column coherence.
• Employing a regularization method based on the weight iterative improvement method (WIIM) for improved convergence.
• Validating strategies through numerical simulations and phantom experiments.

Main Results:

• Reduced coherence in the sensitivity matrix using CS preconditioners.
• Faster convergence of the inverse solver with the WIIM regularization method.
• Significant improvement in MFMT reconstruction quality in both simulated and experimental cases.

Conclusions:

• The proposed computational methods effectively enhance MFMT reconstruction.
• CS preconditioning and WIIM regularization mitigate ill-conditioning and improve solver convergence.
• These strategies offer a pathway to higher quality molecular imaging in biological tissues.