ESTRO 36 Abstract Book

936 / 1082

S920 ESTRO 36 2017
_______________________________________________________________________________________________
of volumes according to the Dice similarity coefficient
(DSC) and the distance between contours as average
Hausdorff distance (AHD) were calculated to estimate the
inter-observer variability. The size of volumes obtained
with visible and hidden surgical clips was also evaluated.
Results
A summary of the calculated metrics is reported in Fig.1,
with the boxplot representing the distribution of the
metrics obtained from all the possible pairwise operator
comparisons. A relative median increase of 48.7% in DSC
(absolute difference 0.21) and a relative median decrease
of 50.7% and 57.1% in CMD and AHD (absolute difference -
0.68 mm and -2.21 mm), respectively, were observed
when surgical clips are visible. The differences between
“w/o clips” and “with clips” contours were always
statistically significant (p<0.001), with the surgical clips
showing an important contribute in diminishing the inter-
operator variability. Greater volumes were obtained when
surgical clips are visible (relative median increased
volume +46.4% with visible surgical clips, p-value<<0.001).
However, a larger variability in segmented volumes
between operators is observed when surgical clips are not
visible, as shown in Table 1.
Conclusion
Discrepancies in contouring target structures can
undermine the precision of ERT. The correct identification
of tumour bed is important to avoid target miss in the
subsequent ERT, to spare this area from the overdosage if
the patient underwent IORT and to allow for reliable
multi-centric studies. The use of surgical clips supports
the radiation oncologist during the contouring process. In
this study, we demonstrated that the inter-operator
variability decreases with the guidance of surgical clips. A
dosimetric analysis will be performed as further
development in order to estimate possible underdosage of
the target or overdosage of the surrounding normal tissue.
[The study was partially supported by AIRC and Accuray
Inc.]
EP-1705 MR imaging of internal mammary lymph
nodes and organs at risk in supine breast radiotherapy
T. Van Heijst1, E.M. Aalbers 1, E. Alberts1, H.J.G.D. Van
den Bongard1, J.J.W. Lagendijk1, B. Van Asselen1, M.E.P.
Philippens 1
1 UMC Utrecht, Radiotherapy, Utrecht, The Netherlands
Purpose or Objective
Standard radiotherapy (RT) for breast cancer patients is
performed with CT guidance. Addition of MRI in breast RT
planning is being investigated. We have developed MRI
sequences to visualize individual axillary lymph nodes
(LNs) [van Heijst et al. 2016, BJR]. The next step is to
optimize visualization of internal mammary LNs (IMLNs)
and organs at risk (OARs). The purpose of this study is to
image the IMLNs, the heart including the coronary arteries
(left anterior descending artery (LAD), right coronary
artery (RCA), and circumflex artery (CX)) – which are
difficult to image due to influence of respiratory and
cardiac motion – as well as the brachial plexus, body
contour, and chest wall, using MRI in supine treatment
position.
Material and Methods
Ten healthy volunteers were scanned, arms abducted, on
a 1.5 T scanner (Ingenia, Philips). An anterior receive coil
was placed on an adjustable bridge, from the abdomen
until the mandible without touching the patient; a
posterior receive coil was located in the scanner table. No
contrast agent was administered. Three MRI sequences
were optimized and evaluated qualitatively for structure
identification in RT planning:
- Coronal 3-dimensional (3D) T2-weighted (T2w) fast spin
echo with short-tau inversion recovery fat suppression
(T2-STIR), covering the supraclavicular and axillary lymph
node regions and the apex of the heart as caudal edge
- Transverse 3D balanced spoiled gradient (bSPGR)
covering the heart using:
spectral inversion recovery for fat suppression (SPIR);
electrocardiography cardiac triggering on the diastolic
heart phase; respiratory gating on a 1-D MRI navigator
placed on the long-liver interface.
- Transverse 3D T1w SPGR, with mDixon water-fat
separation (T1-SPGR) covering the body contour.
Results
The heart and the RCA, LAD, CX arteries were clearly
identified on bSPGR, with high contrast to the surrounding
tissues, without apparent motion artifacts [fig 1A]. Small
(4 mm diameter) IMLNs located close to the heart were
also identified on bSPGR. The chest wall was visualized in
bSPGR [fig 1C] and T1-SPGR; the latter distinctly showed
the body contour [fig 2B]. The brachial plexus was clearly
visualized with T2-STIR, showing intermediate signal
intensity (SI), while main axillary arteries showed high SI
[fig 2A]. The plexus could be followed from the spinal cord
to the axillary nerves into the arm, on both sides.