Three-dimensional photoacoustic tomography based on graphics-processing-unit-accelerated finite element method.

Area of Science:

• Biomedical Imaging
• Computational Physics
• Medical Physics

Background:

• Photoacoustic tomography (PAT) is a powerful imaging modality.
• Frequency-domain finite element method (FEM) offers accuracy and flexibility for PAT reconstruction.
• High computational cost of 3D FEM limits its practical application.

Purpose of the Study:

• To enhance the computational efficiency of the FEM-based PAT reconstruction algorithm.
• To implement a parallel computing strategy using graphics processing units (GPUs).

Main Methods:

• Implementation of a parallel computational strategy within the FEM framework.
• Utilizing the Compute Unified Device Architecture (CUDA) for GPU parallelization.
• Validation through a series of simulation experiments.

Main Results:

• The parallelized FEM algorithm maintains the accuracy of the original method.
• Computational cost is reduced by a factor of 38.9 using a GTX 580 graphics card.
• Demonstrated significant acceleration of the 3D FEM reconstruction process.

Conclusions:

• Parallel GPU implementation effectively accelerates FEM-based photoacoustic tomography.
• The developed method offers a computationally efficient solution for complex 3D reconstructions.
• This advancement can facilitate wider clinical adoption of PAT.