ESTRO 36 Abstract Book

96 / 1082

S83 ESTRO 36 2017
_______________________________________________________________________________________________
the tabletop for the patients’ positioning devices has to
provide a reproducible position. The treatment plan will
be adapted to the patients’ position, anatomical variation
and organ motion of the day. The aim of this study was to
characterize the accuracy of patient set-up with a table
indexing system in the absence of lasers and skin marks.
Material and Methods
This investigation was performed on a conventional MRI.
MR-scans were acquired at 3 different time points from 8
volunteers. The pelvis was chosen as anatomical region of
interest. The tabletop of this MRI is comparable with the
tabletop of the MR-Linac. A head support and a knee
support, indexed on the table, were used for stability and
reproducibility. The first MR-scan was defined as
reference scan and the following two MR-scans were
registered to this reference on bony anatomy. The setup
variability was analyzed in terms of group mean (M),
systematic (Σ) and random errors (σ) for both translations
and rotations. The results were compared to retrospective
set-up data of 79 patients treated for rectum carcinoma
(5x5 Gy), aligned with lasers and skin marks and measured
with CT-scan and Cone Beam-CT position verification.
Because the group of volunteers is relatively small, the
comparison to the rectal cancer patient group is on a
descriptive basis only.
Results
When comparing retrospective set-up data of rectal
cancer patients to the group of volunteers in this
investigation, for translations, the group mean for the
patient group seem to show a better set-up reproducibility
in the LR and CC direction as compared to the volunteers.
This resulted in group means closer to zero with
corresponding smaller Σ errors and σ errors. In the AP
direction, the mean and standard errors did not seem to
show apparent differences. For rotations the results for
both groups were comparable. The results are presented
in table 1.
Conclusion
For volunteers, without the use of laser alignment,
translations seem to be larger in LR and CC direction.
Rotations were comparable for both groups. However, for
daily practice, the impact of this increased uncertainty is
likely small relative to uncertainties of internal organ
motion that can be in the cm range. In daily on-line
corrections, the combination needs to be considered in
positioning pelvic cancer patients without skin marks, on
an MR-Linac.
OC-0165 TPUS vs CBCT: comparison of daily inter-
modality derived setup shifts for prostate
radiotherapy.
E.P.P. Pang1,2, K. Knight2 , M. Baird2, J.M.Q. Loh1, E.T.Y.
Chen 1, G.K. Low1, C.C.C. Yap1, A.H.S. Boo1, J.K.L. Tuan1,3
1 National Cancer Centre Singapore, Division of Radiation
Oncology, Singapore, Singapore
2 Monash University, Faculty of Medicine- Nursing and
Health Sciences Medical Imaging & Radiation Sciences,
Melbourne, Australia
3 Duke-NUS, Graduate Medical School, Singapore,
Singapore
Purpose or Objective
Kilovoltage cone-beam computed tomography (kVCBCT)
has often been regarded as the preferred imaging modality
for the visualisation of soft tissues and verification of
treatment position due to its superior spatial resolution
[1-3]. Transperineal ultrasound (TPUS) is an alternative
imaging tool that can be employed for pre-treatment
verification and in-treatment monitoring as it is non-
invasive and does not involve additional imaging dose [4,
5]. This study aimed to compare the daily inter-modality
derived setup shifts using TPUS versus kVCBCT (gold
standard) for prostate radiotherapy.
Material and Methods
A total of 1927 paired datasets (TPUS versus kVCBCT) from
55 patients were compared in three directions (i.e. x,y,z
shifts representing left/right, anterior/posterior and
superior/inferior directions respectively). The derived
setup shifts were reported to the nearest mm. Data were
analysed using PASW for windows, version 20.0 (SPSS Inc,
Chicago, IL). Observed differences in the derived shifts for
each imaging modality were reported. Statistical tests
were conducted under a two-tailed significance level, at
a minimum 95% confidence interval.
Results
A Shapiro-Wilk test revealed that the data was not
normally distributed (p<0.05). A non-parametric Wilcoxon
Signed Ranks test demonstrated no statistically significant
difference between the derived setup shifts from TPUS
and kVCBCT for all planes; x (p= 0.376), y (p=0.244) and z
(p=0.253). The proportion (%) of datasets where the
difference in the derived shifts between the two imaging
modalities were within 5/4/3mm in the x, y and z
directions are reported in Table 1. Spearman’s rank
correlation coefficients of the derived shifts were
moderate (0.612-0.671) for all three directions (p<0.005),
signifying that the accuracy of TPUS-derived setup shifts
was comparable to kVCBCT.
Table 1: Proportion of datasets where the difference in
the derived shifts between the two imaging modalities
were within 5/4/3mm in the x, y and z directions.
Conclusion
Measured differences were acceptable considering the
planning target volume (PTV) margin expansion was 10mm
in all directions, except posteriorly (6mm). Findings were
in agreement with the recent report by Trivali et al. [6]
who found no significance difference in the x, y and z
coordinates between TPUS and fiducial-based CT
localisation of the prostate gland. With specialised
training and user experience, TPUS is a promising imaging
modality in treatment setup and verification for prostate
radiotherapy without the need for additional exposure to
ionising radiation.