High-performance image reconstruction in fluorescence tomography on desktop computers and graphics hardware

Area of Science:

• Biomedical optics
• Computational imaging
• Scientific computing

Background:

• Fluorescence optical tomography (FOT) image reconstruction is a complex, 3D nonlinear inverse problem.
• Solving these problems is computationally intensive, often requiring significant processing time.

Purpose of the Study:

• To demonstrate accelerated image reconstruction for FOT using modern hardware.
• To present methods for solving both forward and inverse problems efficiently.

Main Methods:

• Utilized state-of-the-art numerical algorithms and hardware-optimized implementations.
• Employed massively parallel programming on graphics processors (GPUs) for forward and inverse problems.
• Compared optimized CPU and GPU implementations.

Main Results:

• Successfully solved the large-scale inverse problem in seconds on standard desktop PCs.
• Achieved acceleration factors of approximately 15x using GPU hardware compared to CPU.
• Maintained accuracy in reconstructed images despite significant speedup.

Conclusions:

• Massively parallel GPU programming enables rapid FOT image reconstruction.
• Hardware optimization is key to solving complex inverse problems efficiently.
• This approach makes advanced FOT analysis more accessible on standard computing hardware.