Greedy reconstruction algorithm for fluorescence molecular tomography by means of truncated singular value...

Area of Science:

• Biomedical Imaging
• Medical Physics
• Optical Imaging

Background:

• Fluorescence molecular tomography (FMT) enables in vivo 3D visualization of fluorescent targets in small animals.
• Standard L2-norm regularization in FMT leads to smoothing, limiting spatial resolution due to loss of high-frequency features.
• The ill-posed nature of FMT necessitates advanced reconstruction techniques.

Purpose of the Study:

• To enhance the spatial resolution and signal-to-noise ratio (SNR) of FMT reconstruction.
• To address the limitations of traditional L2-norm regularization methods in FMT.
• To develop a compressed sensing (CS) based algorithm for improved FMT image quality.

Main Methods:

• Implemented truncated singular value decomposition (TSVD) to ensure the feasibility of CS for the FMT inverse problem.
• Developed an ameliorated stagewise orthogonal matching pursuit (SOMP) algorithm with adaptive thresholds and a specific halting condition.
• Evaluated the proposed CS-SOMP algorithm against TSVD with L2-norm regularization using numerical simulations and phantom experiments.

Main Results:

• The proposed CS-SOMP algorithm achieved significantly higher spatial resolution compared to the TSVD method.
• The algorithm demonstrated a superior signal-to-noise ratio (SNR) in reconstruction results.
• Numerical simulations and phantom experiments validated the effectiveness of the developed method.

Conclusions:

• Exploiting sparsity via compressed sensing significantly improves FMT reconstruction quality.
• The developed CS-SOMP algorithm offers a promising solution for high-resolution FMT imaging.
• This approach overcomes the resolution limitations associated with traditional L2-norm regularization in FMT.