159 / 1082
S146
ESTRO 36 2017
_______________________________________________________________________________________________
algorithms. This work investigates the feasibility of a-
priori estimation and correction of OOPM.
Material and Methods
Data from a thoraco-abdominal numeric MRI phantom
developed in-house were used
2
. A 10-phases 4DMRI,
simulating the planning dataset, was registered to the
exhale volume using 3D optical flow
3
, thus measuring in-
plane motion (IPM
3D
P
) and OOPM
P
along the three
orthogonal slices intersecting in the GTV. In addition,
IPM
2D
P
was obtained with 2D slice-to-slice optical flow
3
registration and the difference C = IPM
3D
P
− IPM
2D
P
represented the phase-specific a-priori correction.
A 36-frames volume sequence (duration 5.4s) represented
treatment data: sagittal/coronal/axial slices simulated
cine-MRI sequences, whereas 3D volumes served as
ground-truth. The diaphragm position measured on each
sagittal slice was used to identify the corresponding
breathing phase within the 4DMRI. Each axial and coronal
slice of the sequence was registered to the corresponding
exhale slices of the 4DMRI (IPM
2D
T
) and the phase-specific
correction was applied (IPM
COR
T
= IPM
2D
T
+ C). The average
end-point distances (EPD) against ground-truth IPM
(obtained through 3D registration) were measured with
and without correction. OOPM was estimated for each
frame as OOPM
P
measured in the corresponding 4DMRI
phase. Finally, the planning GTV was propagated from the
4DMRI exhale phase to each treatment frame using: (1)
IPM
2D
T
with OOPM = 0 and (2) IPM
COR
T
combined with
OOPM
P
. Dice indexes against ground-truth GTVs were
calculated for both scenarios. The sagittal slice, showing
OOPM < 1 mm, was excluded from the analysis.
Results
GTV motion amplitude was (4.0, 1.7, 0.2) mm (SI, AP, LR)
in the 4DMRI and (5.1, 1.2, 0.6) mm in treatment data.
Fig.1 reports EPDs and Dice indexes as a function of the
ground-truth OOPM. On average, the a-priori
correction/estimation approach resulted in EPD reduction
and in Dice index increase with respect to the scenario
without IPM correction and OOPM estimation (Tab.1).
Conclusion
A-priori information from 4DMRI provides a breathing
phase-specific approximation of OOPM and can be used to
correct OOPM in slice-to-slice registrations. Such
procedure significantly improved GTV position estimation
when relevant OOPM is observed, i.e. on the axial slice.
The corrected IPM represents a more accurate
approximation of the motion field that would be measured
if full 3D volumes were acquired and registered in real-
time to the planning data. Future work should focus on
robustness to inter-fraction variations in patients’ data.
[1]Mutic
et al
2014
Semin Radiat Oncol
[2]Paganelli
et al
2015
MICCAI
[3]Zachiu
et al
2015
PMB
PV-0283 Gated liver SBRT based on internal
electromagnetic motion monitoring
E. Worm
1
, M. Høyer
2,3
, R. Hansen
1
, L.P. Larsen
4
, B.
Weber
1
, C. Grau
1,3
, P. Poulsen
1,3
1
Aarhus University Hospital, Department of Oncology,
Aarhus, Denmark
2
Aarhus University Hospital, The Danish Centre for
Particle Therapy, Aarhus, Denmark
3
Aarhus University, Institute of Clinical Medicine,
Aarhus, Denmark
4
Aarhus University Hospital, Department of Radiology,
Aarhus, Denmark
Purpose or Objective
To present our results with the new technique of
respiratory gated liver SBRT based on internal
electromagnetic motion monitoring. The study presents
the geometric and dosimetric improvements in treatment
accuracy of the gating compared to standard CBCT-guided
non-gated treatment.
Material and Methods
Thirteen patients with primary liver cancer or metastases
had three electromagnetic transponders (Calypso)
implanted near the target and received three-fraction
gated liver SBRT at a TrueBeam Linac. The PTV was
created by a 5mm axial and 7mm (n=10) or 10mm (n=3)
cranio-caudal (CC) expansion of the CTV as defined on an
exhale breath-hold CT. A mean homogenous dose between
45 and 61.8Gy was prescribed to the CTV using 7-field
IMRT or 3D conformal planning. The PTV was covered with
67% of the prescribed dose. Treatment was delivered in
free-breathing but gated to the exhale breathing phase
according to the continuously monitored (25Hz)
transponder centroid position. Gate ON windows were set
to +/- 3mm LR/AP and +/-4 mm CC around the exhale
position of the transponders. The couch was adjusted
remotely if baseline drifts above ~1mm of the exhale
transponder position occurred. Post-treatment, log files of