Parallel Solver for Diffuse Optical Tomography on Realistic Head Models With Scattering and Clear Regions.

Area of Science:

• Biomedical optics
• Medical imaging
• Computational modeling

Background:

• Diffuse optical tomography (DOT) is crucial for non-invasive functional brain imaging.
• Accurate light propagation modeling, considering tissue properties, is vital for DOT precision.
• Current models face challenges in speed and accuracy for complex biological tissues.

Purpose of the Study:

• To develop a novel numerical solver for DOT.
• To improve the accuracy and speed of optical property mapping in the human head.
• To integrate anatomical information from MRI into the DOT model.

Main Methods:

• Developed a numerical solver based on the radiosity-diffusion model.
• Integrated structural MRI data for anatomical accuracy.
• Implemented the solver on parallel heterogeneous platforms (GPUs and CPUs).

Main Results:

• Achieved a 7x speed-up compared to a Monte Carlo engine for a 2 million voxel domain.
• Demonstrated significant improvements in accuracy for optical property reconstruction.
• Enabled computation of full human head light distribution in 116 seconds.

Conclusions:

• The radiosity-diffusion solver enhances DOT's speed and accuracy.
• Integration with MRI provides precise anatomical context for light propagation.
• The parallel computing approach enables efficient, high-resolution brain imaging.