Development of an algorithm to improve the accuracy of dose delivery in Gamma Knife radiosurgery
Cernica, George Dumitru
MetadataShow full item record
Gamma Knife stereotactic radiosurgery has demonstrated decades of successful treatments. Despite its high spatial accuracy, the Gamma Knife's planning software, GammaPlan, uses a simple exponential as the TPR curve for all four collimator sizes, and a skull scaling device to acquire ruler measurements to interpolate a threedimensional spline to model the patient's skull. The consequences of these approximations have not been previously investigated. The true TPR curves of the four collimators were measured by blocking 200 of the 201 sources with steel plugs. Additional attenuation was provided through the use of a 16 cm tungsten sphere, designed to enable beamlet measurements along one axis. TPR, PDD, and beamlet profiles were obtained using both an ion chamber and GafChromic EBT film for all collimators. Additionally, an in-house planning algorithm able to calculate the contour of the skull directly from an image set and implement the measured beamlet data in shot time calculations was developed. Clinical and theoretical Gamma Knife cases were imported into our algorithm. The TPR curves showed small deviations from a simple exponential curve, with average discrepancies under 1%, but with a maximum discrepancy of 2% found for the 18 mm collimator beamlet at shallow depths. The consequences on the PDD of the of the beamlets were slight, with a maximum of 1.6% found with the 18 mm collimator beamlet. Beamlet profiles of the 4 mm, 8 mm, and 14 mm showed some underestimates of the off-axis ratio near the shoulders (up to 10%). The toes of the profiles were underestimated for all collimators, with differences up to 7%. Shot times were affected by up to 1.6% due to TPR differences, but clinical cases showed deviations by no more than 0.5%. The beamlet profiles affected the dose calculations more significantly, with shot time calculations differing by as much as 0.8%. The skull scaling affected the shot time calculations the most significantly, with differences of up to 5.6% for single shot placements, and 11% for multiple-shot plans. Despite the sometimes large differences in shot times, prescription isodose lines did not shift significantly, with prescription isodose lines not shifting by more than 0.5 mm to 1.0 mm.