Physical-space refraction-corrected transmission ultrasound computed tomography made computationally practical

Area of Science:

• Medical Imaging
• Computational Physics
• Ultrasound Technology

Background:

• Acoustic refraction significantly impacts Transmission Ultrasound Computed Tomography (CT) image quality.
• Accurate modeling of refractive effects is essential for high-fidelity reconstructions.
• Current methods, while effective, are computationally intensive.

Purpose of the Study:

• To accelerate Eikonal solvers for refractive Transmission Ultrasound CT.
• To evaluate computational performance across different hardware platforms.
• To enable interactive clinical practice without compromising image quality.

Main Methods:

• Implemented and accelerated three Eikonal solvers: Fast Marching Method (FMM), Fast Sweeping Method (FSM), and Fast Iterative Method (FIM).
• Utilized diverse computational platforms: Graphics Processing Units (GPUs), multi-core CPUs, and cluster CPUs.
• Integrated solvers within an iterative reconstruction framework for refractive Transmission Ultrasound CT.

Main Results:

• Achieved significant acceleration of Eikonal solvers on GPUs and multi-core/cluster CPUs.
• Demonstrated the suitability of various architectures for acoustic wave-front tracking.
• Validated that acceleration does not lead to a loss in reconstruction quality.

Conclusions:

• Optimized Eikonal solvers enable faster acoustic wave-front tracking in Ultrasound CT.
• The developed framework meets interactive demands for clinical applications.
• This work enhances the practical utility of refractive Transmission Ultrasound CT.