Evaluation of on-board imager cone beam CT hounsfield units for treatment planning using rigid image registration

To evaluate the on-board imager cone beam CT (OBI-CBCT) Hounsfield units (HUs) for treatment planning.

The HU-electron density (eD) calibration for CBCT, the CATphan504 phantom was used, and the CBCT HU (HUCBCT) consistency was studied by analyzing the CBCT images of Rando phantom and compared with planning CT. The latter study was also performed on CBCT images of 10 H&N patients. For comparison, the structures contoured and treatment plans generated on CT were transferred on to the CBCT after registration. The treatment plans were compared using gamma (g) index analysis and the plan comparison dose volume histograms (DVHPlanComp).

Although the HU-eD calibration curves of both the planning CT and CBCT were found to be linear, differences in mean HU values were found in the region of interest (ROI) corresponding to Acrylic, Derlin, and Teflon, viz., 144 ± 11 HU, 193 ± 5 HU, and 257 ± 7 HU respectively. For all the cases, the consistency and reproducibility of HUCBCT values for low density medium agreed the HU CT except at regions of high density. Overall g-evaluation showed more than 94% pixels pass rate and DVH results showed small difference in the DVHPlanComp, Rando, and large differences in DVHPlanComp, patient for structures contoured at peripheral regions (PV) of CBCT images.

We conclude that the pixel-to-pixel HU corrections for entire range of eD are not necessary for OBI-CBCT images. Application of local correction in the high-density and penumbral regions would facilitate the use of CBCT images for routine treatment planning.