Semiempirical, parameterized spectrum estimation for x-ray computed tomography.

Area of Science:

• Medical Physics
• Radiological Imaging
• Computational Science

Background:

• Accurate X-ray spectra are crucial for reliable X-ray and CT simulations.
• Existing models may lack precision, necessitating improved spectrum generation methods.

Purpose of the Study:

• To develop a tool for generating accurate, well-validated X-ray spectra.
• To provide spectra for standalone use or integration into an open-access X-ray/CT simulation tool.
• To cover tube voltages from 80-140 kVp and anode takeoff angles from 5°-9°.

Main Methods:

• Developed a novel spectrum parameterization method using spline knots and characteristic line values.
• Initialized spectra using physics-based tools (XSPECT, SpekPy) and refined them with empirical measurements.
• Systematically designed experiments with attenuators and reduced scatter contamination (<1%) on a 64-row CT scanner.
• Utilized two optimization schemes (modulation, gradient descent) to refine spectra against measurements.

Main Results:

• Achieved excellent agreement between parameterized and initial spectra (R² values up to 0.997).
• Empirical measurements covered an X-ray transmission range of up to 3.5 orders of magnitude.
• Optimized spectra improved agreement with measurements by twofold to tenfold compared to unoptimized spectra.

Conclusions:

• The developed, well-validated X-ray spectra are suitable for the Cancer Imaging Toolkit (XCIST) and standalone applications.
• Spectra can be interpolated for arbitrary kVp (80-140 kVp) and takeoff angles (5°-9°).
• A standalone Matlab function for spectral interpolation is publicly available.