U.S. flag An official website of the United States government

U.S. Food and Drug Administration

GPU-accelerated Monte Carlo X-ray Simulation Software for Evaluating Medical Imaging Devices with Photon Counting Detectors (MCGPUv1.3_PCD)

Catalog of Regulatory Science Tools to Help Assess New Medical Devices 

This regulatory science tool is a computer model intended to support evaluation of photon counting detectors (PCDs) during the product development phase by generating in-silico X-ray projections of computational anatomical models detected by PCDs.

Technical Description

MCGPUv1.3_PCD is a software tool which may be used to simulate radiographic images and computed tomography (CT) scans using photon counting detectors (PCD). The tool is extended from an existing RST, MCGPU v1.3 [1], which generates in-silico X-ray projections with a traditional energy-integrating detector and includes example codes to apply a realistic detector response that accounts for spectral distortion during the detection process. By extending the software to also simulate a PCD, this tool helps regulatory scientists develop evaluation methods for PCDs to access image quality and material decomposition accuracy.

Inputs and outputs: 

  • User-Defined Inputs: User inputs are largely the same as MCGPU v1.3 [1] but with a new addition related to energy setting of the simulated PCD, including detector size, energy range, and the number of energy bins.
  • Output Data: The tool outputs an array containing the number of photons detected per detector pixel, per energy bin, and per projection.  

MCGPUv1.3_PCD simulates an ideal Photon Counting Detector (PCD) with perfect detection efficiency and precise photon registration at each pixel and energy bin. To account for realistic detector responses and potential imperfections, the tool offers instructions and code examples for processing the output with external software [2]. This allows users to simulate realistic detector response, including potential imperfections. This enables a more comprehensive simulation of how a PCD would perform in real-world scenarios.

Intended Purpose 

MCGPUv1.3_PCD is intended to support quantitative evaluation of Photon Counting Detectors (PCDs) by simulating their performance in realistic scenarios and to help assess image quality and spectral material decomposition. This assists medical device innovators in developing robust testing methods to assess the safety and efficacy of new PCD technologies and benefits companies in optimizing PCD designs during product development.

Testing

The physics of photon interactions in a digital phantom are identical to MCGPU v1.3 and have been tested previously [1].  Internal validation and verification for simulating photon detection by a PCD involves three studies comparing reconstructed parameters (from the output) to ground truth parameters (based on user inputs). 

Study 1: Energy-dependent attenuation coefficients for 3 materials

Three digital phantoms were created: one with water and two with iodine at different concentrations (2 mg and 10 mg). Each phantom was irradiated by a pencil beam from a 120 kVp X-ray source with a 1 mm aluminum filter. Using this tool, the linear attenuation coefficients with respect to photon energy were calculated for each material based on the photon counts recorded by the simulated PCD. The output results are comparable to those theoretically calculated by the NIST XCOM software. The details and codes used in this study are provided in the MCGPUv1.3_PCD repository on the DIDSR GitHub (see supporting documents).

Study 2: Reconstructed images from fan beam CT geometry

A fan beam CT simulation was performed using a digital cylindrical water phantom with two iodine rod inserts of different concentrations (2 mg and 10 mg). A total of 180 projections at different energy levels were used for filtered back projection. The reconstructed image was compared against the cross-sectional image of the input digital phantom, and the results were as expected. The details and codes used in this study are provided in the MCGPUv1.3_PCD repository on the DIDSR GitHub (see supporting documents).

Study 3: Material identifications from Coherent scattering

In this study, we aimed at validating the adequacy of energy registration for scattered photons. A set of caffeine slabs (digital phantoms) with various thicknesses were created and irradiated by a set of monochromatic X-ray beams ranging from 5 keV to 120 keV. For each simulation setup, the recorded photon counts from the simulated PCD were used to create a histogram of momentum transfers, called the X-ray scattering profile. The resulting simulated setup was compared to the true scattering profile, and the results were identical. The details and codes used in this study are discussed in [5, 6].

Limitations

MCGPUv1.3_PCD provides an extended version of MC-GPU that simulates ideal PCDs and includes example codes to apply a realistic detector response that accounts for spectral distortion during the detection process. Instructions are included in the InstructionManual.md. However, this mechanism relies on the assumption of uniform and homogeneous irradiation of a pixel, assuming that all parts of the pixel surface receive the same amount and spectrum of X-rays. In other words, the photons detected in a pixel are assumed to be uniformly distributed within that pixel. This assumption is not valid due to the partial volume effect, which becomes an issue with larger pixel sizes. This limitation should be considered by users when simulating larger pixel sizes (above 500 µm).

Supporting Documentation

References:

[1] Andreu Badal and Aldo Badano, "Accelerating Monte Carlo simulations of photon transport in a voxelized geometry using a massively parallel Graphics Processing Unit", Medical Physics 36, pp. 4878–4880 (2009) https://cdrh-rst.fda.gov/mcgpu-gpu-accelerated-monte-carlo-x-ray-imaging-simulator

[2] Taguchi K. Photon Counting Detector Simulator: Photon Counting Toolkit (PcTK). (In) Spectral, Photon Counting Computed Tomography: Technology and Applications (Devices, Circuits, and Systems). Edited by Taguchi K, Blevis I, Iniewski K. (CRC Press, Taylor & Francis Books, Inc.). July 14, 2020 (ISBN-13: 978-1138598126, ISBN-10: 1138598127) https://pctk.jhu.edu/

[3] Ghammraoui B, Taguchi K, Glick SJ (2023) Inclusion of a GaAs detector model in the Photon Counting Toolkit software for the study of breast imaging systems. PLoS ONE 18(6): e0270387. https://doi.org/10.1371/journal.pone.0270387

[4] Ghammraoui, Bahaa, Andreu Badal, and Lucretiu M. Popescu. "Maximum-likelihood estimation of scatter components algorithm for x-ray coherent scatter computed tomography of the breast." Physics in Medicine & Biology, vol. 61, no. 8, 2016, p. 3164. IOP Publishing, https://dx.doi.org/10.1088/0031-9155/61/8/3164

[5] E. Dahal, Y. L. E. Thompson, S. Amer, O. Sandvold, P. Noël, A. Badano (2023). Effect of spectral binning in X-ray scattering method for non-invasively characterizing amyloids [Poster Presentation]. Clinical Trials on Alzheimer’s Disease 2023, Boston, U.S.

[6] E. Dahal, K. Suresh, S. Amer, Y. L. E. Thompson, B. Ghammraoui, A. Badano. Material characterization in up to 16-cm thick objects using high energy elastic X-ray scattering at small angles. (Manuscript under development)

Contact

Tool Reference