Projected algebraic reconstruction technique-network for high-fidelity diffuse fluorescence tomography...

Area of Science:

• Biomedical optics
• Medical imaging
• Computational imaging

Background:

• Diffuse fluorescence tomography (DFT) is crucial for in vivo imaging.
• Traditional methods like Algebraic Reconstruction Technique (ART) face limitations in image quality and robustness.
• Integrating model-based approaches with neural networks offers a promising avenue for improvement.

Purpose of the Study:

• To develop an advanced algorithm for high-fidelity image reconstruction in diffuse fluorescence tomography.
• To enhance the noise robustness and quantitative accuracy of DFT imaging.
• To validate the proposed method using numerical simulations, phantom experiments, and in vivo studies.

Main Methods:

• Proposed a model-driven projected algebraic reconstruction technique (PART)-network (PART-Net).
• Incorporated nonnegative prior information into the ART iteration process.
• Combined PART with a residual convolutional neural network for high-fidelity reconstruction.

Main Results:

• PART-Net demonstrated significant improvements in noise robustness and reconstruction accuracy (1-2 times higher) compared to traditional ART.
• The algorithm showed superior spatial resolution and quantification, particularly for small targets (r=2mm).
• Phantom and in vivo experiments confirmed the effectiveness and strong generalization capability of PART-Net.

Conclusions:

• PART-Net effectively enhances image quality in diffuse fluorescence tomography.
• The algorithm offers superior performance in noise robustness, accuracy, and spatial resolution.
• PART-Net shows great potential for practical applications in biomedical imaging.