ESTRO 36 Abstract Book

567 / 1096

S552 ESTRO 36
_______________________________________________________________________________________________
Conclusion
Inter- and intrafraction variations were analysed fo r three
different cranial fixation systems. In trafraction
translations were small for all systems, while interfraction
rotations could be significant. The addition of an
individual head support does not seem to decrease the
interfraction rotations, and for intrafraction variations the
results seem to even indicate a slight improvement when
using a standard head support with a shape that provides
good fixation for the head. Individual supports might have
added value for patients with a deviating anatomy.
Poster: RTT track: Imaging acquisition and registration,
OAR and target definition
PO-1001 Evaluation of target volume delineation of
the regional lymph nodes in breast cancer patients
M. Mast1, E. Gagesteijn1, T. Stam1, N. Knotter2, E.
Kouwenhoven1, A. Petoukhova1, E. Coerkamp3, J. Van der
Steeg1, J. Van Egmond1, H. Struikmans1
1 Haaglanden Medical Centre Location Antoniushove,
Radiation therapy, Den Haag, The Netherlands
2 Leiden University Medical Centre, Clinical oncology,
Leiden, The Netherlands
3 Haaglanden Medical Centre Location Westeinde,
Radiology, Den Haag, The Netherlands
Purpose or Objective
New ESTRO guidelines have been developed for the
delineation of the Clinical Target Volumes (CTVs) of the
regional lymph nodes of the breast. Until now we used the
methods based on the article of ‘Dijkema et al.’. In
response to these new insights, we decided to develop a
tool to implement this new ESTRO guideline. The main
question, which will be answered, is: “What are the
differences between delineating the regional lymph nodes
of breast cancer according to the method ‘Dijkema et al.’,
‘the ESTRO guidelines’ and ‘the Tool combined with
ESTRO guidelines (‘Tool’)’?”
Material and Methods
In ten patients CTVs of the regional lymph nodes of the
breast were delineated (in Pinnacle [1]) by three
dedicated radiation oncologists, according to the two
different guidelines and the ‘Tool’. The ‘Tool’ is a method
where the subclavian and the axillary vessels are
delineated by a radiation therapist and is expanded in all
directions with 5 mm. This volume is than adjusted by the
radiation oncologist on the basis of the prescribed
anatomical boundaries of ‘the ESTRO guidelines’. After
that, all CTVs were exported to MATLAB to calculate the
Conformity Index generalized (CIgen ) [2]. In MATLAB the
differences in the various directions on the axial coupes of
the treatment planning-Computed Tomography scans
were analysed. Also the volumes of the CTVs were
calculated in Pinnacle. Finally, the required delineating
times per patient, per guideline and per radiation
oncologist were compared and analyses were carried out
using SPSS [3].
[1]: Pinnacle Treatment Planning System, version 9.10
(Philips Healthcare)
[2]: E. Kouwenhoven, 2009, Phys Med Biol.
[3]: IBM SPSS Statistics for Windows, IBM Corp., Armonk,
NY, USA
Results
The MATLAB analyses showed that the ‘Tool’ had the
highest CIgen (0.64 and σ = 0.05) relative to the other
two methods (p<0.04) (table 1). Furthermore, the
delineating time was shortest (13.6 min and σ = 2.4) by
using the ‘Tool’. The use of the ESTRO guideline without
the ‘Tool’ resulted in the smallest average CTV volume
(150.6 cm3 and σ = 41.0). Furthermore, we saw a clear
decrease of the standard deviations in most delineating
directions when using the ‘Tool’, except in the ventral
direction.
Table 1. The differences of the CIgen between the three
methods.
CIgenDijkema et al. CIgenESTRO CIgenTool
Average 0,58 0,57 0,64
Standard deviation
(σ) 0,05 0,06 0,05
CIgen = Conformity Index generalized
Conclusion
Using the ‘Tool’ we found a significantly higher CIgen and
a smaller CTV volume (compared with the method
'Dijkema et al.'). The advice is to use the ‘Tool’ as
delineating method for delineating the CTV of the regional
lymph nodes of breast cancer patients due to the increase
of the CIgen combined with the shortest delineation time
and the smallest standard deviation per delineating
direction. We also recommend performing second reading
to improve the concordance between radiation
oncologists. Finally, further research is required because
the CIgen did not reach a level higher than 0.8.
PO-1002 Comparison of Best Commercial Model and
Atlas based segmentation with CT and MR in brain
cancer.
J.Y. Woo 1, T.Y. KIM2, J.Y. SEOK3, T.M. KIM 1, Y.W. CHO 1,
S.Y. KIM1, J.G. BAEK1, J.H. KIM1, J.H. CHO1
1Yonsei Cancer Center, Radiation Oncology, Seoul, Korea
Republic of
2National Cancer Center, Radiation Oncology, Gyeonggi-
do, Korea Republic of
3Seoul National University Bundang Hospital, Radiation
Oncology, Gyeonggi-do, Korea Republic of
Purpose or Objective
It is important to accurately delineate critical organ such
as optic chiasm, pituitary gland and brainstem when
radiation therapy is delivered in brain cancer. MR images
were usually used to delineate critical organ accurately in
most brain cases. But manually delineated contours by
different users sometimes have different shape and region
in the same planning CT. Even if different users delineate
contours, we would expect to get more accurate and
regular critical organ if using auto contouring method.
Recently there are many commercial auto contouring
softwares including model based segmentation (MBS) and
atlas based segmentation (ABS) softwares even supporting
MR images. This study aims to compare auto contouring
methods to delineate critical organ accurately and to have
certain shape and region.
Material and Methods
It is multi-center study. We selected 10 patients. We used
three MBS software solutions and ABS software solution
(MIM_software ver. 6.5.5.) to generate the automatic
contouring on the planning CT. All MBS software just made
contours without any preparation, and we chose the best
result among 3 MBS solutions for comparison. But ABS
software should have subjects (who are already registered
for ABS to work on auto contouring and also they are not
the patients involved in this study). We made two groups
of atlas, 60 subjects of CT based and 20 subjects of MR
based. We used two matching techniques for MR based
ABS, Majority-vote and STAPLE. We analyzed auto
contouring with 4 classified groups - best MBS (BM), CT
based ABS for 60 subjects (CA), and MR based ABS using
Majority-vote (MR_MV) and MR based ABS using STAPLE
(MR_ST). We gained brain stem, optic chiasm, and
pituitary gland contours. Average Dice Similarity
Coefficients (DSC) was calculated for each structure to
compare against 'gold” standard contours which are
manually defined of 4 groups respectively. Values closer
to 1 indicate higher accuracy.