Program Information
Uncertainties in Convolution-Based Organ Dose Estimation in Tube-Current Modulated CT
W Fu*, K Choudhury , A Kapadia , P Segars , E Samei , Duke University Medical Center, Durham, NC
Presentations
TH-CD-601-10 (Thursday, August 3, 2017) 10:00 AM - 12:00 PM Room: 601
Purpose: Convolution-based methods allow for fast organ dose estimation under tube-current modulated (TCM) CT; however, the organ dose is subject to considerable uncertainty. The purpose of this study was to estimate the uncertainty in organ dose in clinical TCM studies and use the information to improve the estimation accuracy.
Methods: Fifty-eight anatomical computational phantoms (XCAT) were scanned using a simulation of two scanner models (GE VCT, Siemens Definition Flash) with six common CT protocols under TCM conditions. The organ doses were derived using a validated Monte Carlo program (true dose) and estimated using a convolution-based method (estimated dose). The convolution-based method estimates organ dose based on organ dose coefficients (h factors) with the dose field adjusted for TCM. The dose field under TCM was quantified by convolving a dose spread function with the tube current profile. To simulate a clinical scenario with variable and unknown patient anatomy, the phantom anatomy was approximated by averaging five XCAT phantoms with closest body-region height and using the standard deviation to determine the uncertainty in patient/XCAT match. The patient width was taken into account using organ dose coefficients.
Results: On average, the estimated organ doses differed from true doses by less than 10%. For abdomen-pelvis and chest protocols, the estimated organ doses showed a linear correlation with true doses (R²>0.5, for most organs within the scan coverage). Organs within the CT field-of-view (FOV) showed low relative error between the estimated and true dose. Organs outside the CT FOV showed clear correlations between the relative error and true dose (R²>0.5 for most organs outside the CT FOV). The estimated dose vs. dose relationships varied for different protocols.
Conclusion: In this study, we estimated the uncertainty associated with convolution-based estimation for organ doses. The uncertainties could be used to improve organ dose estimations.
Contact Email: