PDF, English Download (13MB) | Terms of use

Abstract

Magnetic resonance imaging for image guidance of proton therapy could greatly reduce dosimetric uncertainties during treatment, as uncertainties due to set-up or anatomical changes can be detected and corrected for. For dose calculation, however, MR images lack information on the stopping power of the tissue. The use of synthetic computed tomography (CT) images derived from MR images was thus investigated using the example of prostate cancer in this thesis. First, the impact of bulk density (BD) overwrites in the planning CT for small (< 5) sets of tissue classes was investigated. For the modified CT image with BD values for five tissues, sufficient dosimetric agreement (mean gamma pass rates (GPRs) above 98.5 % for the PTV) and mean absolute range difference below 1.8 mm were observed. The same set of tissue classes was subsequently identified in the rigidly registered MR image and overwritten with the same BD values as before. The mean GPR decreased to 95.2 %, and the absolute range difference increased to 4.0 mm. The observed differences are unacceptable and emphasize the necessity of MR guidance for proton therapy of prostate cancer. Using a deformably registered MR image to overcome the anatomical differences between the CT and the MR, the dosimetric agreement was found clinically acceptable for most cases (mean GPR = 97.6 %), and the mean absolute range difference decreased to 2.7 mm. The workflow implemented in this thesis allows fast and robust generation of synthetic CT images for proton therapy. The presented conversion technique is compatible with different MR scanners and can be extended to different anatomical sites with little extra effort. While the full potential of MR images has not yet been tapped, synthetic CT images are a useful tool for MR-guided proton therapy.