Fast simulation of x-ray projections of spline-based surfaces using an append buffer

Area of Science:

• Medical Imaging
• Computational Science
• Computer Graphics

Background:

• Realistic x-ray image simulation is crucial for scientific validation but computationally intensive.
• Standard graphics acceleration algorithms are often unsuitable for x-ray transmission-based imaging.
• Efficient simulation methods are needed to accelerate the development of advanced CT reconstruction algorithms.

Purpose of the Study:

• To develop a custom graphics pipeline for simulating x-ray transmission projections using moving spline surface models.
• To leverage graphics hardware (GPU) via OpenCL for significant simulation speed-up.
• To validate the accuracy of the simulated projections for tomographic reconstruction.

Main Methods:

• Implementation of a custom OpenCL graphics pipeline for tessellation, projection, and rendering.
• Utilized moving spline surface models for scene representation.
• Introduced a specialized append buffer for efficient ray-intersection storage and processing.
• Applied an absorption model to generate projection pixel values.

Main Results:

• Achieved high-speed projection generation (640x480 in 254 ms) with moving spline structures.
• Reconstructions using simulated projections showed errors below 1 HU with a sharp kernel.
• Demonstrated that traditional GPU acceleration schemes yield higher errors (up to 9 HU) for this task.
• The new method provides accurate projections suitable for validating novel CT reconstruction algorithms.

Conclusions:

• The developed OpenCL graphics pipeline offers a fast and accurate solution for x-ray image simulation.
• This approach is particularly beneficial for complex simulations, including motion-compensated CT reconstruction.
• The method overcomes limitations of standard GPU acceleration, enabling more efficient research and development in medical imaging.