Dose tracking in radiation oncology using daily CBCT: effects of physical parameters on dose calculation accuracy.

The availability of cone beam computerized tomography (CBCT) images at the time of treatment has opened possibilities for recalculations and tracking of the delivered dose, becoming an effective tool for adaptive radiotherapy. A significant component in the accuracy of dose recalculation on CBCT images is the calibration of the Hounsfield units (HU) to density. The aim of this thesis, developed at the Policlinic of Modena, is to assert a methodology for the characterization of the HU-to-density calibration curve, and to evaluate the impact of the calibration phantom inserts composition and phantom volume on dose calculation accuracy. The HU-to-density calibration curves from four different phantoms were measured and compared. The HU output of a kV CBCT scan depends on the physical parameters of the phantom density inserts, with particular reference to the atomic number (Z), due to the photoelectric effect which is the main radiation-matter interaction at kV energies. The phantom radial diameter also influences the HU values on the image. The effect of scattering and beam hardening was examined as a function of phantom diameter, founding a high deviation on the HU value of the same density insert when varying the radial diameter of the phantom, especially for high densities. When we are dealing with CBCTs also the acquisition parameters influence the resulting image, that’s why we will show that a protocol-specific calibration curve is needed. The resulting calibration curves were used to compare the calculated doses against planned ones. The percent difference between recalculated and planned dose was obtained for chosen clinically important dose levels and a box plot analysis was conducted. Results show that the best calibration curve for dose recalculation on CBCT images has been obtained when a human-tissue-equivalent inserts are used and when the radial diameter of the phantom is close to the dimensions of the real patient.