(Sunday, 7/12/2020) [Eastern Time (GMT-4)]
Room: AAPM ePoster Library
Purpose: Dosimetric accuracy in radiation therapy heavily relies on CT images for material characterization of electron density, physical density, or relative stopping power. The widely accepted stoichiometric theoretical Hounsfield unit (HU) calibration procedure relies on extracting coefficients for photon interactions from scans of objects with known chemical compositions. In this work, we studied this accuracy of this calibration over virtual monoenergetic CT (monoCT) images for various human tissues and extended that to artificial metal implants.
Methods: Dual-energy CT scans of two phantoms were taken on a Siemens Twinbeam CT scanner. These were then analyzed in Syngo.via post-processing software. MonoCT images were reconstructed from 40-190 keV. Regions of interest were then placed on each insert, and the reconstructed HU was compared to theoretical HUs calculated from the NIST XCOM database. Based on these results, a material-specific calibration curve from HU to physical density for proton treatment planning was developed for evaluation in both phantom and patient studies.
Results: The 190 keV virtual CT images provided the most accurate HU estimation for all tissues, while tissue-specific analysis showed the 50 keV CT image was most accurate for soft tissues. For the material-specific calibration, we relied on the 190 keV CT image for objects below 0.8 g/cm³ and above 1.14 g/cm³ and the 50 keV image for the remaining tissues. The average reconstructed HU error for the proposed method compared to NIST was -0.45% over all tissues, and -1.15%, -0.56%, and -0.04% for lung, soft tissue, and bone.
Conclusion: The proposed method offered accurate CT reconstruction using a combination of 50 and 190 keV monoCT images. Additionally, the proposed method can accurately estimate the density of metal objects, such as titanium, since the HU for titanium at high energies is below 3071 HU, the maximum CT value for images in DICOM format.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
