Modeling the impact of treatment uncertainties in radiotherapy

The aim of this thesis is to model the impact (dosimetric and radiobiological) of prostate radiotherapy treatment uncertainties and techniques for displaying potential dose errors. A Monte Carlo code is written to simulate both the treatment planning and delivery phases. Distributions of systematic organ motion, systematic positioning error, random organ motion, random positioning error, and interpatient cell sensitivity are taken from the literature. These distributions are used to sample the position and size of the prostate and rectum at each fraction of treatment. Dose to organ voxels are assigned from precalculated dose distributions (four-field box using 6 MV photons with margins around the clinical target volume of 0.5, 1.0, and 1.5 cm). In particular, the poor estimation of prostate mean position from one planning CT series is investigated and shown to introduce large dose uncertainty. A convolution kernel is suggested that considers multiple pre-treatment CT scans to better estimate mean prostate position. The radiobiological impact of the treatment uncertainties is evaluated in terms of tumor control probability for prostate, normal tissue complication probability for rectal bleeding and uncomplicated tumor control probability using established models. Local prostate control is highly sensitive to treatment uncertainties with inadequate margin size and to interpatient radiosensitivity. Rectal bleeding is more sensitive to margin size than to deformation. A number of ways to display the dosimetric uncertainty are presented. The display tools proposed allow the planner to consider potential variations in dose visually and may assist with margin size considerations.