Robust transformed l₁ metric for fluorescence molecular tomography

Area of Science:

• Biomedical imaging
• Molecular imaging
• Tomography

Background:

• Fluorescence molecular tomography (FMT) is a key non-invasive 3D imaging technique for visualizing fluorescent probes in vivo.
• Traditional FMT reconstruction algorithms often struggle with noise and errors, limiting robustness.
• Existing methods using squared l2-norm can amplify noise, compromising accuracy.

Purpose of the Study:

• To introduce a novel robust transformed l1 (TL1) metric for FMT.
• To develop a robust FMT model that minimizes noise influence.
• To enhance the accuracy and reliability of FMT reconstructions.

Main Methods:

• Proposed a transformed l1 (TL1) metric interpolating l0 and l1 norms.
• Developed a robust FMT model utilizing the TL1 metric for noise reduction.
• Implemented a continuous optimization method, converting the problem to difference in convex programming (DCATL1).

Main Results:

• The DCATL1 algorithm demonstrated superior robustness compared to state-of-the-art methods.
• Achieved improved source localization and morphology recovery in simulations and in vivo experiments.
• Verified performance through numerical simulations and bead-implanted mouse studies.

Conclusions:

• The DCATL1 algorithm effectively visualizes fluorescent probe distribution in biological tissues.
• Demonstrated feasibility and effectiveness for preclinical applications in small animals.
• The proposed method enhances the robustness and accuracy of FMT imaging.