High-performance reconstruction of CT medical images by using out-of-core methods in GPU

Area of Science:

• Medical Imaging
• High-Performance Computing
• Computational Science

Background:

• Computed Tomography (CT) is a widely used medical imaging technique.
• Reducing patient radiation exposure is critical, with sparse sampling being a key approach.
• Efficient algebraic methods are required for real-time CT image reconstruction, necessitating high-performance computing.

Purpose of the Study:

• To develop and optimize GPU software for Computed Tomography (CT) image reconstruction.
• To implement QR factorization with out-of-core (OOC) techniques for efficient reconstruction.
• To minimize data transfer times and overlap I/O operations with computations.

Main Methods:

• Developed GPU software utilizing QR factorization with out-of-core (OOC) techniques.
• Optimized data transfer between disk, CPU, and GPU memory.
• Implemented block cache on main page-locked memory for enhanced efficiency.

Main Results:

• The GPU implementation achieved up to 6.5 times faster performance compared to a CPU version.
• The block cache on main page-locked memory proved more efficient than GPU or mirrored caches.
• The reconstructed images demonstrated improved quality compared to other methods.

Conclusions:

• The developed software provides an optimized QR factorization for GPU, enabling high-quality CT image reconstruction.
• Performance is comparable to state-of-the-art clinical methods.
• This approach effectively reduces patient exposure time and radiation dose.