| Summary: | Safe and effective delivery of radiosurgery demands a steep dose fall-off outside the target, in addition to highly conformal treatment and sub-millimetre overall accuracy. This thesis concerns the CyberKnife system - an image-guided radiosurgery system capable of treating both intra- and extracranial targets. Plan-specific QA performed using an ionisation chamber and radiochromic film confirmed that the dose distributions produced by MultiPlan software accurately reflect the treatment delivered, and therefore subsequent dosimetric studies using MultiPlan are valid. The relationship between prescription isodose value and external dose gradient (measured by the Gradient Index) was explored for solitary intracranial spherical targets, and then irregularly-shaped lesions. For smaller targets the steepest dose fall-off was achieved by prescribing to as close to the 50 % isodose as could be achieved. For larger targets the effect of changing the prescription isodose value was less marked but the optimum value was in the range 60 – 70 %. A planning method to optimise dose fall-off whilst maintaining other aspects of plan quality has been proposed. An additional study looked at optimising dose falloff on one aspect of a target situated close to the brainstem. It was demonstrated that using “VOI” hard limits in treatment planning can reduce the brainstem dose substantially without any significant compromise on other important plan parameters. Finally, a dosimetric comparison between CyberKnife and the Gamma Knife system was performed for solitary intracranial targets. Overall, there was no significant difference in conformality and external dose gradient across the lesions studied. However the results suggested that Gamma Knife dosimetry may be superior for small lesions, and CyberKnife for larger ones. Whilst the experimental findings in this thesis relate to intracranial dosimetry, they may also be relevant to extracranial treatment planning using the CyberKnife system: this is a suggested area of future research.
|
|---|
