Quantification of shape variation of prostate and seminal vesicles during external beam radiotherapy

The prostate is known to translate and rotate under influence of rectal filling changes and many studies have addressed the magnitude of these motions. However, prostate shape variations also have been reported. For image-guided radiotherapy, it is essential to know the relative magnitude of translations, rotations, and shape variation so that the most appropriate correction strategy can be chosen. However, no quantitative analysis of shape variation has been performed. It is, therefore, the purpose of this article to develop a method to determine shape variation of complex organs and apply it to determine shape variation during external beam radiotherapy of a GTV (gross tumor volume) consisting of prostate and seminal vesicles. For this study, the data of 19 patients with prostate cancer were used. Each patient received a planning computed tomography (CT) scan and 8–12 (11 on average) repeat CT scans that were made during the course of conformal radiotherapy. One observer delineated the GTV in all scans, and volume variations were measured. After matching the GTVs for each patient for translation and rotation, a coverage probability matrix was constructed and the 50% isosurface was taken to determine the average GTV surface. Perpendicular distances between the average GTV and the individual GTVs were calculated for each point of the average GTV, and their variation was expressed in terms of local standard deviation (SD). The local SDs of the shape variation of all 19 patients were mapped onto a reference case by matching and morphing of the individual average GTVs. Repeated delineation of the GTV was done for 6 patients to determine intraobserver variation. Finally, the measured shape variation was corrected for intraobserver variation to estimate the “real” shape variation. No significant variations in GTV volume were observed. The measured shape variation (including delineation variation) was largest at the tip of the vesicles (SD = 2.0 mm), smallest at the left and right side of the prostate (SD = 1.0 mm), and average elsewhere (SD = 1.5 mm). At the left, right, and cranial sides of the prostate, the intraobserver variation was of the same order of magnitude as the measured shape variation; elsewhere it was smaller. However, the accuracy of the estimated SD for intraobserver variation was about half of the accuracy of the estimated SD for the measured shape variation, giving an overall uncertainty of maximum 0.6 mm SD in the estimate of the “real” shap