FMT-ReconNet: Fluorescence Molecular Tomography Reconstruction using Prior Knowledge and Deformation Neural Network.

Area of Science:

• Biomedical Imaging
• Medical Physics
• Computational Biology

Background:

• Fluorescence molecular tomography (FMT) enables 3D reconstruction of internal fluorescent sources.
• Current FMT methods face limitations in spatial resolution due to simplified models and ill-posed inverse problems.

Purpose of the Study:

• To introduce FMT-ReconNet, a novel deep neural network designed to enhance spatial resolution and accuracy in FMT.
• To leverage deep learning for improved 3D reconstruction of fluorescent sources within biological tissues.

Main Methods:

• FMT-ReconNet integrates a spatial transformer network (STN) for source transformation and a V-Net for reconstruction.
• The STN generates prior knowledge by transforming a template source based on target FMT surface data.
• The V-Net utilizes this prior knowledge combined with target surface data for accurate source prediction and reconstruction.

Main Results:

• FMT-ReconNet significantly improves the spatial resolution of FMT imaging.
• The network accurately reconstructs the 3D distribution of internal fluorescent sources.
• This deep learning approach overcomes limitations of traditional FMT reconstruction methods.

Conclusions:

• FMT-ReconNet represents a substantial advancement in precise FMT imaging reconstruction.
• The proposed deep neural network offers a powerful solution for overcoming spatial resolution challenges in FMT.
• This technique holds promise for more accurate in vivo molecular imaging and diagnostics.