Full time-resolved diffuse fluorescence tomography accelerated with parallelized Fourier-series truncated diffusion...

Area of Science:

• Biomedical Optics
• Medical Imaging
• Computational Science

Background:

• Diffuse fluorescence tomography (DFT) provides rich data for fluorophore distribution in tissues.
• Current time-domain DFT methods are computationally intensive, limiting 3D imaging applications.
• Efficient image reconstruction is crucial for advancing DFT in biological and medical research.

Purpose of the Study:

• To develop a computationally efficient scheme for diffuse fluorescence tomography (DFT) image reconstruction.
• To reduce the lengthy reconstruction time associated with explicit time-domain DFT methods.
• To enable more practical applications of DFT, especially for 3D or large-volume imaging.

Main Methods:

• Proposed a computationally efficient DFT image reconstruction scheme using Fourier-series expansion of time-dependent photon density.
• Solved independent frequency-domain diffusion equations at multiple sampling frequencies.
• Employed a hybrid parallelization strategy combining multicore CPU (coarse-grain) and multithread GPU (fine-grain) processing.
• Developed and validated both time- and frequency-domain inversion procedures.

Main Results:

• The proposed parallelized Fourier-series truncated diffusion approximation significantly reduces computational time.
• Reconstruction accuracy is comparable to the explicit time-domain scheme.
• Validated effectiveness and accuracy through simulative and phantom experiments.
• Demonstrated feasibility for practical DFT image reconstruction.

Conclusions:

• The developed method offers a computationally efficient alternative for DFT image reconstruction.
• Significant reduction in computational time achieved without compromising image quality.
• This approach enhances the applicability of DFT for in vivo and 3D imaging of fluorophore distribution.