Image reconstruction in fluorescence molecular tomography with sparsity-initialized maximum-likelihood expectation...

Area of Science:

• Biomedical imaging
• Optical tomography
• Image reconstruction

Background:

• Fluorescence molecular tomography (FMT) is a powerful imaging technique.
• Accurate modeling of Poisson noise is crucial for high-quality FMT reconstruction.
• Existing methods may struggle with noise and convergence speed.

Purpose of the Study:

• To develop an improved reconstruction method for FMT.
• To effectively model Poisson noise in FMT.
• To enhance image quality and reconstruction efficiency.

Main Methods:

• A hybrid approach combining sparse reconstruction (SVD-FISTA) with maximum-likelihood expectation maximization (MLEM).
• MLEM is initialized using the output of the sparse reconstruction method.
• The method incorporates sparsity information and models Poisson noise.

Main Results:

• Significantly improved image quality, both qualitatively and quantitatively.
• Over 20 times faster convergence compared to uniformly initialized MLEM.
• Enhanced robustness to noise and reduced background noise compared to pure sparse reconstruction.

Conclusions:

• The proposed method effectively models Poisson noise in FMT reconstruction.
• This framework offers improved performance for FMT imaging.
• The approach provides a robust and efficient solution for FMT image reconstruction.