Click here to


Are you sure ?

Yes, do it No, cancel

A Real-Time Monte-Carlo Simulation Technique for Dose and Scatter Estimation in Virtual Clinical Trials for CT Imaging

S Sharma1*, E Abadi2 , P Segars3 , A Kapadia4 , E Samei5 , (1) Duke University, Durham, NC, (2) Duke University, Durham, NC, (3) Department of Radiology, Duke University Medical Center, Durham, North Carolina, (4) Duke University Medical Center, Durham, NC, (5) Duke University Medical Center, Durham, NC


(Thursday, 8/2/2018) 10:00 AM - 12:00 PM

Room: Room 202

Purpose: The growing requirement for task-based optimization in CT has created the need for virtual clinical trials (VCTs) for optimizing image quality and dose. Such VCTs require patient and scanner-specific dose estimations for an exhaustive population of virtual patient models. In addition, VCTs also require the estimation of scatter signal for enhancing the realism of simulated images. The established technique of using Monte-Carlo (MC) simulations for such estimations is computationally expensive, hindering possibility of prospective studies. In this study, we present a comprehensive GPU-based MC tool for rapid quasi real-time estimation of patient and scanner-specific dose and scatter signal for image-based VCTs.

Methods: An original GPU-based code (MC-GPU, FDA) was modified to incorporate a bowtie filter, tube current modulation, curved detector, and an anti-scatter grid based on a commercial CT system. The dose estimates were validated using AAPM’s TG-195 datasets and previously published dose studies. The scatter and dose results were also validated against physical measurements obtained on a commercial CT scanner (Definition Flash, Siemens) using a lead-augmented 32 cm CTDI phantom and an anthropomorphic phantom with embedded dosimeters.

Results: The dose estimates from our tool were consistent with values reported in AAPM’s TG-195 within 3% and the published dose study within 5% for organs within the primary field. The dose results were also found to be within 10% of physical measurements. The scatter-adjusted intensity profile agreed with that obtained on a physical scanner with a coefficient-of-variation of 0.048. The simulations were performed within 120s for a typical chest CT simulation using a single Nvidia Titan-Xp GPU.

Conclusion: The validated GPU-based MC method allows accurate and rapid estimation of dose and scatter in CT imaging, and can be used as an integral component of VCTs by providing patient-specific dose and scatter estimates for a virtual population of XCAT patient models.

Funding Support, Disclosures, and Conflict of Interest: Ehsan Samei: Unrelated to this study, active reseach grants with Siemens and GE and advisory board member of medInt Holdings, LLC. Nothing else to disclose.


Not Applicable / None Entered.


Not Applicable / None Entered.

Contact Email