# Computationally efficient x‐ray simulation framework using parameterized material attenuation models in anatomically detailed imaging

## TL;DR

This paper introduces a fast and accurate x-ray simulation framework using parameterized material models, validated in breast and whole-body imaging.

## Contribution

A new x-ray simulation framework using parameterized material attenuation models to reduce computational cost while maintaining accuracy.

## Key findings

- The maximum attenuation coefficient error was 0.007%, well below biological variability.
- Projection and reconstruction errors remained within ±0.006% across all cases.
- Simulation times decreased significantly with acceleration factors scaling linearly with phantom complexity.

## Abstract

Virtual clinical trials provide an efficient alternative to clinical imaging trials for evaluating imaging technologies. In x‐ray simulations, however, modeling material‐specific attenuation becomes computationally intensive as anatomical complexity and material heterogeneity in digital phantoms increase. Parameterization models offer a potential solution by representing material properties with a compact set of coefficients.

To develop and validate an x‐ray simulation framework that models material attenuation using parameterization models, reducing computational cost while maintaining accuracy.

Material attenuation was modeled with a five‐coefficient parameterization derived from physical cross‐section data. Unlike conventional ray‐tracing, which projects each material separately, the proposed method projects only the five parameter maps, making computational cost independent of phantom complexity. This framework was evaluated in two scenarios: breast imaging with 10 compressed breast phantoms with varying fibro‐glandular content, and whole‐body imaging with head and abdomen phantoms. Accuracy was assessed by computing percent errors in attenuation coefficients, sinograms, and reconstructed images relative to the conventional approach. For whole‐body imaging only, additional analyses included the impact of resolution loss and noise, the comparison with errors introduced by different projector models to place results in the context of standard simulation variability, and computational time measurements.

Across all materials and both applications, the maximum attenuation coefficient error was 0.007% (breast skin tissue), far below reported biological variability. Projection and reconstruction errors remained within ± 0.006% for all cases. In whole‐body imaging, these errors were well below those from projector model differences (± 0.5%), and image modification routines further concentrated the error distribution around zero. Simulation times decreased significantly, with acceleration factors scaling linearly with the number of materials within the phantoms.

The proposed framework achieves accurate and efficient simulation of material attenuation in x‐ray imaging, especially in anatomically complex scenarios. Validated in both breast and whole‐body imaging, it offers a robust and efficient alternative to conventional methods, supporting the development of advanced virtual clinical trials and spectral imaging research.