Title: Improving post-implant evaluation of permanent seed implant brachytherapy for prostate cancer
Other Titles: Optimalisatie van post-implant evaluatie van permanente zaadimplant brachytherapie voor prostaatkanker
Authors: De Brabandere, Marisol; M9309400
Issue Date: 17-Jun-2013
Abstract: Seed implant brachytherapy has been widely accepted as a treatment option for localised T1-T2 prostate tumours, a technique in which radioactive sources (seeds) are implanted permanently in the prostate gland. Apart from some patient-related factors such as PSA, Gleason score and clinical stage, implant quality has been identified as a crucial factor affecting tumour control and complication rates. Implant quality can be measured by performing a dosimetric post-implant evaluation (postplanning), in which the delivered dose (which never equals the intended prescription dose) to the prostate and organs at risk is assessed. For this, post-implant images are acquired on which the implanted seeds and the OAR are identified, and on which the dose is calculated using a dedicated TPS. Accurate dosimetry provides feedback to the brachytherapy team, fostering technical changes to improve implant quality and providing information to correlate dose to treatment outcome.Currently, the relationship between dose (with D90 being the most important parameter) and tumour response is not clear. To a large extent, this may be attributed to the uncertainties related to post-implant dosimetry. The calculated dose depends on the accuracy of seed localisation and prostate delineation.CT is the most widely used technique for postplanning, with adequate seed imaging properties, but with poor soft tissue contrast for prostate delineation. MRI provides superior prostate visualisation (especially T2-weighted MRI), but the seeds are not well depicted. To solve this conundrum, fusion techniques have been proposed, combining e.g. CT with MRI-T2, or matching an MRI sequence optimised for seed imaging with MRI-T2. However, each of these postplan techniques comes with specific dosimetric uncertainties, which are currently unknown. In this thesis we aimed at improving the quality of post-implant evaluation in order to obtain more consistent and reliable dosimetric postplan data for evaluation of implant quality and for determination of dose-response relationships.In the first two chapters of this project, we evaluated seed imaging quality and reconstruction accuracy using prostate simulating phantoms. In chapter 2, a solid PMMA phantom implanted with a set of inactive seeds was used to test CT based seed reconstruction accuracy in function of seed type and various scanning parameters. It was found that seed type, CT tube current, FOV and CT scanning mode (axial versus spiral) had no influence on seed reconstruction accuracy. Slice thickness was the only parameter affecting seed localisation precision: a larger slice thickness (> 4 mm) was associated with increased seed reconstruction uncertainty. In the current study there was no improvement in reconstruction accuracy by decreasing slice thickness below 3 mm. These results were confirmed in a mailed multi-centric study involving six European centres. The results were neither dependent on the TPS used nor on the scanning parameters, regardless of the CT scanner and practitioner performing the seed localisationprocedure. In chapter 3, we designed and constructed a gel-based phantom to investigate and compare seed reconstruction accuracy based on MRI and CT images. The gel core was prepared in such a way that the imaging properties were comparable to those of prostate tissue (and hence the seeds) on both CT and MRI. With this phantom, a suitable T1-weighted GE MRI scan sequence was developed for seed visualisation. Consequently, phantom images with 3, 4 and 5 mm slice thickness were acquired on CT and MRI. For CT, seed reconstruction deviations increased with slice thickness, with a mean reconstruction error of 0.8 mm for 3 mm slice spacing to 2.1 mm for 5 mm spacing. This trend was in agreement with our findings for the solid CT phantom. Surprisingly, on MRI the largest reconstruction uncertainties were observed for 3 mm slice thickness (mean of 2.1 mm), while 4 mm and 5 mm spacing yielded better results. This could be attributed to the typical seed (void) display, which was longer than the actual seed dimension and not always related to signal (void) intensity. This complicated MRI seed detection in longitudinal direction, especially on 3 mmsets where seeds were depicted on multiple slices. Due to this larger longitudinal uncertainty, the reconstruction deviations were in general larger for MRI than CT. Although phantoms are valuable tools for accurate assessment of seed reconstruction accuracy, they do not provide anatomical information necessary to calculate the effect on the dose-volume parameters. Therefore, in the second part of this project, postplan uncertainties were further investigated using patient images. We set up three interobserver variability studies to quantify post-implant uncertainties introduced by seed reconstruction, contouring and image fusion. In chapter 4, we assessed and compared the seed reconstruction uncertainty on post-implant CT and MRI-T1 images of three patients in terms of interobserver variability. The seeds were localised in the TPS by seven experienced physicists and compared to a reference seed set. A Matlab program was written to (a) assign the observer’s seeds to the reference seeds, and (b) to calculate interobserver variability, i.e. the mean deviation between the observer and reference positions of the assigned seed pairs. For CT, on average 98% of the reconstructed seeds could be assigned to a reference seed, with a mean interobserver deviation of 1.1 mm (1 SDref). This corresponded well with the results of the phantom studies. For MRI-T1, the mean seeds assignment rate was 93%, with a mean interobserver variability of 3.0 mm (1 SDref). This deviation was slightly larger than the value measured with the gel phantom study. This was plausible given the additional presence of anatomical impurities in prostate tissue. The effect of seed reconstruction variability on prostate dosimetry was calculated for three postplan techniques: (a) CT only, (b) CT fused with MRI-T2 (denoted as CT+T2), and (c) MRI-T1 fused with MRI-T2 (denoted as T1+T2), using fixed CT and MRI-T2 prostate contours. For CT and CT+T2, the D90 interobserver variability (1 SDref) was 1.5% and 1.3%, respectively. For T1+T2, the D90 variability was 6.6%, which was notably larger than the uncertainties in the other postplan techniques. In chapter 5, we quantified the impact of the interobserver variability in post-implant contouring and image fusion and compared it to seed reconstruction for the postplan techniques described above. For each technique, reference plans were defined, consisting of a reference seed geometry, a reference prostate contour, and a reference fusion. For the interobserver study on contouring, eight physicians delineated the CTV-P(rostate) according to the ESTRO/EAU/EORTC recommendations on prostate brachytherapy. The resulting prostate structures showed a large variability in volume. For CT, this could be attributed to the poor soft tissue contrast, enhanced by the obscuring presence of seed metal artefacts. The unexpected large contouring uncertainty on MRI showed however that there was also no uniformity in the interpretation and application of the CTV-P definition. The contouring variability in CT and T2 resulted in a considerable interobserver variability in D90, with SD values of 23% for CT and 17–19% for the other techniques using T2 for contouring. For the interobserver variability study on image fusion six physicists performed a manual image registration for the multi-modality techniques T1+T2 and CT+T2. The impact of fusion variability on D90 was relatively small for T1+T2 (6%, 1 SDref), and larger for CT+T2 (16%, 1 SDref). CT+T2 fusion uncertainty could be reduced to 7% (1 SDref) by using an intermediate registration step using T1int images, making use of the better depicted seeds for registration. In general, we found that dosimetric parameters for prostate post-implant evaluation showed a large technique-dependent interobserver variability, with contouring and image fusion being the ‘weak links’ in the procedure. In order to obtain more reliable dosimetric postplan evaluation data, future prostate dosimetry research should focus mainly on these two aspects. This can be approached by establishing widespread training. Image fusion uncertainties could additionally be tackled with improved fusion procedures or software.
Publication status: published
KU Leuven publication type: TH
Appears in Collections:Laboratory of Experimental Radiotherapy

Files in This Item:
File Status SizeFormat
thesisFINAL.pdf Published 7849KbAdobe PDFView/Open Request a copy

These files are only available to some KU Leuven Association staff members


All items in Lirias are protected by copyright, with all rights reserved.