Determination of geometric uncertainties and their inclusion into margins for three dimensional conformal radiotherapy of the prostate

Abstract

With the advent of three dimensional conformal radiotherapy (3D CRT) techniques, and more recently intensity modulated radiotherapy (IMRT), there has been a general trend to reduce the dose to normal tissues treated. This can be achieved by reducing the margin applied to the clinical target volume (CTV) to form the planning target volume (PTV). As the size of this margin has approached the magnitude of the geometric uncertainties involved in 3D CRT, determining these uncertainties for individual centres and treatment techniques has become increasingly necessary. A determination of set-up, organ motion and target delineation uncertainty was undertaken. Random and systematic set-up uncertainty was determined by an electronic portal imaging (EPI) study. EPI images for 62 patients, with an average of 10 images per patient, were analysed by matching bony anatomy on EPI images to the same anatomy on reference digitally reconstructed radiographs (DRRs). The random set-up uncertainty (1 standard deviation), after the application of a patient correction protocol, was 2.4mm lefthight (LR), 2.06mm anteriorþosterior (AP), 2.0mm superior/inferior (SI) and the systematic component was l.57mm LR' 1.37mm AP 1.02mm SI. Inter-fraction organ motion for the prostate and the rectum were estimated by the movement of the centroid of each structure outlined on sequential CT scans. Several methods for calculating centroid co-ordinates from structure outlines were investigated. It was resolved that the centroid as determined by the planning system was insuff,rciently accurate for use in detecting the size of movements expected. An Excel macro was developed and used to calculate the centroids that were subsequently used to determine both organ motion and intra-observer tatget delineation. Organ motion uncertainty (1 standard deviation), of the prostate was 1.2mm LR, 3.1mm AP, 4.6mm SI and 1.4mm LR, 3.5mm AP, 8.3mm SI for the rectum. Prostate motion was found to correlate with the fractional change in rectal volume. To determine intra-observer variability the prostate and rectum were outlined 3 times, for 6 patients on 12 CT scans by a single observer. The position of the centroid for each outline was used to quantitate the intra-observer target delineation uncertainty (root mean square of the individual standard deviations); which was 0.37mm LR, 1.10mm AP, 1.45mm SI for the prostate and 0.43mm LR, 1.30mm AP, 2.66mm SI for the rectum. These uncertainties were combined to calculate a CTV-PTV margin. The CTVPTV margin was calculated using two methods; one based on physical constraints (McKenzie et al, 2000) and one based on biological constraints (van Herk et al, 2000). The CTV-PTV margin calculated by using the McKenzie et al (2000) method yielded a margin of 2.7mm LR, 7.6mm AP,l2.6mm SI. The van Herk et al (2000) margin calculation resulted in a CTV-PTV margin of 3.7mm LR, 8.1mm AP, 12.3mm SI. The values for the uncertainties determined in the study were consistent with previously published studies. This study showed the ability of an offline set-up correction protocol to reduce systematic set-up errors, and highlighted the larger uncertainties of organ motion and target delineation. These uncertainties have been shown to be non-trivial and must be considered when a decision is made to reduce margins. A CTV-PTV margin was calculated which will account for these uncertainties.