Incorporation of geometrical uncertainties into 3D treatment planning of prostate cancer

The positional uncertainty of the clinical target volume (CTV) during the multifractionated treatment of prostate cancer is determined by organ motion and patient setup deviations during treatment. The prostate motion has been quantified by performing repeat CT scans during treatment and could be described by a rotation around the left right axis with the apex of the prostate as centre of rotation. Setup deviations have been measured, using portal imaging, and could be reduced by applying appropriate setup verification and correction procedures. Both organ position and setup uncertainties have to be incorporated into the treatment planning by defining a planning target volume (PTV) around the CTV. The purpose of this study was to investigate whether an appropriate description of the margin around the CTV for prostate cancer can be established based on the statistics of the geometrical uncertainties. The uncertainties can be described by random and systematic components. The random components of translational and rotational deviations can be incorporated into the treatment planning by a convolution of the 3D dose distribution. The shrinkage of the 95% isodose surface was fitted with an erosion operator (representing an anisotropic margin) and the approximation, inherent to this fit, was evaluated by inspection of the dose volume histogram (DVH) of the convolved dose distribution and the volume within the 'eroded 95% isodose surface' of the static dose distribution. For some of the prostate cases, an anisotropic margin was not a good approximation, since the distance between the two 95% isodose surfaces was strongly place dependent. To account for the systematic components, the distribution of DVHs of the CTV can be studied as a function of the magnitude of the margin around the CTV and corresponding 3D dose distributions. Based on the resulting data, a margin for systematic deviations can be established. For the prostate cases, even an anisotropic margin was not a good solution and other means to generate a PTV from the CTV are necessary, such as the calculation of iso-coverage probability surfaces and/or iso-nominal probability surfaces. In conclusion, for prostate cancer an anistropic margin around the CTV is not the preferred solution to incorporate geometrical uncertainties into the treatment planning