Beyond the Abstract - Inter-observer variability in contouring the penile bulb on CT images for prostate cancer treatment planning, by Lucia Perna, PhD

BERKELEY, CA ( - Erectile dysfunction (ED) is known to be an adverse side effect after radiotherapy for prostate cancer.

The growing numbers of young patients interested in conserving their potency is leading clinicians and researchers to devote more attention to this issue, as preservation of erectile functionality can have a significant impact on the quality of life of quite a large number of patients likely to be long-term survivors after curative radiotherapy for prostate cancer. From the literature there is some evidence of a vascular ethiopathogenesis of radiation-induced ED, suggesting that irradiation of the penile bulb (PB), the crura and the corpora cavernosa could cause post radiotherapy ED.

Although it has been suggested that post radiotherapy ED may be related to the unnecessary irradiation of erectile structures, dose constraints have not yet been clearly assessed. Possible clinical causes of the differences reported in a number of studies have been associated with the difficulty in assessing ED, the use of anti-impotence drugs, and of hormonal therapy.

Moreover, due to the particular position of the erectile structures, mainly the penile bulb, technical/dosimetry uncertainties could play a role. Firstly, the position of PB next to the caudal limit of the irradiation field may introduce additional uncertainty due to the day-by-day set-up position of the beams. Still more important could be the uncertainty in delineating PB and other erectile structures, especially with computed tomography (CT), owing to the recognized limits of this imaging technique - mainly the low contrast in the pelvic area. Due to these limitations, Magnetic Resonance Imaging (MRI) has been proposed as the most appropriate imaging modality for accurate localization of the erectile structures.

In April 2010, a prospective multi-Institute study (Disfunzione Urinaria Erettile, DUE-01) was activated after approval from the ethics committee. The purpose of the DUE-01 study is to prospectively assess the predictive parameters of genito-urinary toxicity and ED, including the possible correlation between ED and PB dose-volume parameters. In this type of study, involving many Institutes aiming at evaluating the possible correlation between normal tissue complication and dose distribution, it is mandatory to investigate the impact of contouring uncertainties on dose-volume parameters. Accordingly, the first step of the DUE-01 study was the activation of a dummy run exercise for the contouring of PB with the aim of: 1) assessing the impact of contouring uncertainty on PB dose-volume parameters potentially predictive of ED; 2) suggesting possible methods/strategies to minimize their impact; 3) giving individual recommendations to reduce inter-observer variations in case of operators “significantly far from the average.”

CT images of ten prostate patients were randomly selected for the dummy run exercise, and fifteen physicians involved in the treatment of prostate cancer in the different Institutes enrolled in the DUE-01 study were asked to draw the PB. Before starting the dummy run, both patients and observers were anonymised. In order to standardize the PB definition, all physicians were instructed to adhere strictly to some guidelines previously defined by the steering committee of the study.

In order to evaluate the impact of contouring inter-variability on dose-volume histogram (DVH) parameters, for each patient, a prostate treatment plan was simulated using an 18 MV X-Ray four-field box technique and prescribing a dose of 76 Gy to the original target volume.

PB inter-observer variations were analyzed in terms of volume differences and cranial/caudal limit variations. For DVH analysis, the values of PB mean dose and the volume of PB receiving more than 50 Gy and 70 Gy (V50 and V70, respectively) were collected for each patient and each observer - both as absolute (cc) as well as relative (%) values. The rationale for the selection of these DVH parameters was that mean dose and V(50) as surrogate of a threshold dose for ED, whereas V(70) was representative of the overlap between the penile bulb and the target volume.

Seven observers systematically overestimated or underestimated the PB volume with significant deviations (p<0.05) from the average volumes ranging between -48% and +34%. These high deviations could be explained by a lower quality of CT images for some patients as shown from a patient-based analysis. An analysis of the cranial and caudal borders, although it detected some random variations, did not show significant systematic differences. Therefore, presumably, most deviations were in the lateral and/or anterior-posterior directions.

Inter-observer contouring variability strongly impacts on DVH parameters. When considering %DVH parameters, inter-patient differences were greater than inter-observer differences: 14.5 Gy versus 6.8 Gy for mean PB dose; 23.0% versus 11.0% for V50; 16.8% versus 9.3% for V70. On the contrary, when considering absolute (cc) DVH parameters, the impact of inter-observer variability was comparable with inter-patient variability: 1.38 cc versus 1.41 cc for V50 and 1.00cc versus 1.03cc for V70.

The contouring uncertainty in dose-volume modelling studies has been generally neglected, or at least under-reported/under-estimated, despite its potential impact. An important general point concerns the need for clear and simple guidelines for organ delineation; a successful application of such guidelines was demonstrated in the case of the rectum in rectal toxicity dose-volume relationship studies where a robust anatomically based definition of the cranial and caudal borders of the rectum guaranteed sufficient reliability of the DVHs collected in a large multi-centric trial.

The potential impact of PB contouring uncertainty in the context of prostate cancer radiotherapy presents a number of special features: in particular, the proximity of PB to the caudal limit of the PTV suggests that PB dose-volume parameters are highly sensitive to this uncertainty. Another specific point concerns the fact that slight deviations in contouring among different observers led to large relative changes in volume and DVH parameters, due to the relatively small volume (generally between 5 and 20 cc) of this structure. As a consequence of volume variability, the impact of contouring uncertainty on dose-volume parameters of PB was found to be great as well. An important result was that, without any intervention to reduce it, inter-observer variability of absolute (cc) DVH parameters is as large as inter-patient variability. This result shows that the dose-volume relationship for PB would be completely hidden only due to contouring uncertainty. On the other hand, inter-patient variability was found to be two-fold greater than inter-observer variability when considering mean PB dose and % DVH parameters.

As an example of the impact of inter-observer variability, with regard to the constraint V50 <50%, our results (1 SD for inter-observer variability on V50: 11%) suggest that with a V50 value of around 35% there is still a probability of about 10% that V50 is higher than the constraint; inversely, if V50 is around 65%, there is a probability of about 10% that V50 is below our constraint. Although our result suggests a significant impact of contouring variability, in the presence of a large cohort of patients, as in the DUE-01 study, in which more than 500 patients are expected to be enrolled, the existence of a dose-volume relationship could be detected. Further investigation on the expected impact of these uncertainties on the predictive power of our study is warranted. In any case, it is clear that % DVH should be used to search for correlation, while absolute DVH should be ignored.

This study suggests that the reliability of the quantification of dose-volume effects of penile bulb defined on CT images may be significantly reduced in multi-institutional studies. Attempts to reduce the impact of contouring variability include both a re-contouring after an MRI tutorial and specific advice to those observers for whom the largest systematic deviations from the average values of PB mean dose and % DVH were detected. Other possible solutions may be “a posteriori” contouring by a single observer and/or improvement of the agreement among observers after critical review and repetition of the dummy run procedure.

Written by:
Lucia Perna, PhD as part of Beyond the Abstract on This initiative offers a method of publishing for the professional urology community. Authors are given an opportunity to expand on the circumstances, limitations etc... of their research by referencing the published abstract.

Medical Phys. Dept.
San Raffalele Hospital
Via Olgettina, 60 – 20132 Milano

Inter-observer variability in contouring the penile bulb on CT images for prostate cancer treatment planning - Abstract Prostate Cancer Section

Read other Beyond The Abstract submissions

More Information about Beyond the Abstract