Click here to

Session: Imaging General ePoster Viewing [Return to Session]

CT Simulation Realism: What Are the Impacts of Simulation Parameters?

I Montero1,2,4*, S Sotoudeh-Paima1,2,3,5, E Abadi1,2,3,4,5, E Samei1,2,3,4,5, (1) Duke University Health System, Durham, NC, (2) Center for Virtual Imaging Trials, Durham, NC,(3) Carl E. Ravin Advanced Imaging Laboratories, Durham, NC, (4) Medical Physics Graduate Program, Duke University School of Medicine, Durham, NC, (5) Department of Electrical and Computer Engineering, Duke University, Durham, NC

Presentations

PO-GePV-I-15 (Sunday, 7/10/2022)   [Eastern Time (GMT-4)]

ePoster Forums

Purpose: Virtual imaging trials enable medical imaging experiments that are not feasible using patient images. Virtual trials are reliable to the extent that they can replicate clinical experiments. This study aimed to assess the effects of various CT simulation parameters on the closeness of virtual images to experimental images.

Methods: A physical phantom was scanned using two clinical scanners (Siemens NAEOTOM Alpha and Force) at multiple dose levels. A computational version of the phantom was virtually imaged using a CT simulator (DukeSim), mimicking the same scanner models and imaging parameters. The simulations were done with varied phantom voxel size (0.5 – 0.1 mm), source and detector sub-sampling (1-5 on each side). All real and simulated projections images were reconstructed using a vendor-specific software (Siemens ReconCT). We compared the real and simulated images in terms of image contrast, modulation transfer function (MTF), noise magnitude, and noise power spectrum.

Results: The simulated images closely matched real data, especially when optimum simulation parameters were chosen. The error in the MTF measurements was highly sensitive to the phantom voxel size and source and detector sub-sampling, as expected. The results demonstrated that small voxel size and high sub-samplings reduced the MTF f50. Voxel size and sub-sampling did not affect the accuracy of simulations in terms of image contrast, noise magnitude, or noise texture.

Conclusion: To be clinically relevant, CT simulations should be optimized to provide results close to real acquisitions. We studied the effects of two simulation parameters (phantom voxel size and source and detector subsampling) on the accuracy of scanner-specific simulations. With such studies, we can identify and further improve the parameters that highly influence CT simulations.

Funding Support, Disclosures, and Conflict of Interest: R01HL155293-01A1 P41EB028744

Keywords

Simulation, Image Analysis, CT

Taxonomy

Not Applicable / None Entered.

Contact Email

Share: