Monte Carlo kilovoltage X-ray tube simulation: A statistical analysis and compact simulation method.

Area of Science:

• Medical Physics
• Computational Imaging

Background:

• Monte Carlo (MC) simulations enhance X-ray imaging quality.
• Accurate X-ray source modeling is crucial for MC simulations in computed tomography (CBCT).
• GPU implementation of MC simulations faces efficiency and memory challenges.

Purpose of the Study:

• Develop a compact, memory-efficient X-ray source model for GPU-accelerated MC simulations.
• Analyze photon distribution statistics from MC X-ray tube simulations.
• Improve the feasibility of complex MC modeling for CBCT.

Main Methods:

• Performed MC simulations of an X-ray tube using BEAMnrc.
• Categorized photons into primary, scatter, and off-focal radiation (OFR).
• Applied statistical linear transformations (PCA, ZCA, ICA) and geometrical transformations for compact data representation.

Main Results:

• Identified strong correlations between photon position and direction across all categories.
• Successfully achieved compact representation of primary (error < 2%) and scatter (error < 6%) photon data.
• Reduced component correlations using the novel compact representation method.

Conclusions:

• Statistical linear transforms offer a memory-efficient approach for simulating X-ray sources in GPU MC systems.
• The proposed method reduces memory requirements by 3.8 times for all photon types.
• Memory usage for primary and scatter data reduced from gigabytes to kilobytes.