Abstract:

ObjectiveTo systematically evaluate the differences in CT values between virtual non-contrast （VNC） images and true non-contrast （TNC） images generated from dual-energy CT （DECT），and to validate the feasibility of VNC images replacing TNC images in dose calculations for photon and proton radiotherapy plans.MethodsA retrospective analysis was conducted on the imaging data of 40 patients with solid tumors （20 cranial，10 thoracic and 10 abdominal cases） who underwent DECT scans at Cancer Hospital of Shandong First Medical University from February 2022 to May 2023. VNC and TNC images were registered slice-by-slice. The differences in CT values of anatomical structures were compared，and Pearson correlation analysis was used to evaluate the correlation of CT values of different anatomical structures in VNC and TNC images. For structures with significant differences，linear regression models （TNC=β×VNC+α） were established using the least squares method. In the Varian Eclipse 15.5 treatment planning system，photon and proton radiotherapy plans based on TNC images and VNC images，as well as the proton radiotherapy plan based on the VNC images corrected by the regression models，were respectively designed. Dose differences of radiotherapy plans designed based on the two images were evaluated. To evaluate dose variations in regions adjacent to the clinical target volume （CTV），two 2-mm-thick annular reference structures were generated on the axial slice containing the largest cross-section of the CTV，extending cranially and caudally from the CTV. These structures were designated as Ring_p and Ring_d，respectively.ResultsThe differences in CT values between VNC and TNC images were mainly concentrated in the bony structure. The CT values difference between TNC and VNC images was （409.07±53.38） HU for the skull in 20 cranial tumor patients （t=13.88，P<0.001），and （118.66±20.90） HU for the vertebral bone in 10 thoracic and 10 abdominal tumor patients （t=10.43，P<0.001）. The CT values of the skull and spine showed high correlation between TNC and VNC images （r=0.98，P<0.001；r=0.99，P<0.001）. The regression models established respectively were： TNC=1.859×VNC+33.896 （skull），and TNC=1.827×VNC+5.491 （spine）. For photon radiotherapy plans based on TNC and VNC images，the D_meanof the CTV were （60.00±0.00） and （60.00±0.00） Gy respectively，with D_meanof Ring_p were （61.17±1.69） and （61.01±1.67） Gy，and Ring_d were （55.26±2.06） and （55.20±1.94） Gy，respectively. The relative dose differences in D_meanbetween the two image types were 0 （t<0.01，P>0.999），0.33% （t=0.30，P=0.766），and 0.19% （t=0.07，P=0.947），all with no statistically significant differences. For proton radiotherapy plans based on TNC and VNC images，the D_meanof the CTV were （61.73±0.32） and （61.67±0.26） Gy（RBE），respectively，with D_meanof Ring_p were （61.19±0.44） and （60.53±1.22） Gy（RBE），and Ring_d were （60.97±0.67） and （59.80±4.26） Gy（RBE），respectively. The relative dose differences in D_meanbetween the two image types were 0.24% （t=0.63，P=0.530），1.80% （t=1.45，P=0.156），and 3.56% （t=2.26，P=0.030），with a statistically significant difference in the Ring_d region. In the proton radiotherapy plan designed based on the corrected VNC images，the D_meanof the CTV was （61.75±0.32） Gy（RBE），Ring_p was （61.43±0.71） Gy（RBE），and Ring_d was （59.96±2.80） Gy（RBE）. The relative dose differences in D_meanbetween TNC images and corrected VNC images were 0.16% （t=0.19，P=0.850），0.76% （t=1.32，P=0.196），and 2.22% （t=1.93，P=0.061），respectively，with no statistically significant differences.ConclusionsThe differences in CT values between VNC and TNC images in DECT mainly exist in bony structures，particularly in the skull and vertebrae. For patients with cranial tumors，VNC images can be directly used in photon radiotherapy planning. In contrast，for proton therapy，after being corrected by the regression model，VNC images can effectively replace TNC images for the dose calculations of radiotherapy plan.

Key words:Radiotherapy,Proton therapy,Radiotherapy dosage,Dual-energy CT,Virtual non-contrast scan,True non-contrast scan