895 / 1096
S879
ESTRO 36
_______________________________________________________________________________________________
during SBRT. SeedTracker, in conjunction with the Elekta
XVI system, reads the monoscopic images acquired during
treatment and calculates the position of the prostate by
auto segmenting the radiopaque markers implanted in it.
The accurate performance of the SeedTracker was
validated using static and dynamic studies utilising the
phantoms implanted with gold seeds. The system also has
variable angle stereo image reconstruction functionality
for the rapid determination of 3D offsets and position
correction of patients in the situations where intrafraction
motion is observed during treatment.SeedTracker was
utilized for real time monitoring of prostate position for
patients undergoing stereotactic boost treatment within
the PROMETHEUS trial (UTN: U1111-1167-2997) in Sydney
South West Local Health District, Australia. The necessary
ethical and legal approvals were obtained from the local
health district Human Research Ethics Committee
Research Governance Office and Therapeutic Goods
Administration, Department of Health, Australian
Government before its clinical implementation.The
dosimetric accuracy achieved by the utilization of
SeedTracker was studied by incorporating the observed
position offsets in the planned dose.
Results
The performance evaluation study of SeedTracker showed
that the system demonstrated a minimum True Positive
Rate of 88% in studied static and dynamic scenarios with
mean (σ) difference of 0.2(0.5) mm in calculated position
accuracy. At the time of writing this abstract SeedTracker
had been utilized for the real time position monitoring of
twenty six patients’ SBRT treatment (consisting of twenty
two treatment sessions). Eleven occurrences of position
deviations outside the acceptable tolerance limits (3mm)
were observed that led to treatment interruption and
position correction of the patient. The retrospective dose
reconstruction study showed that the V98 to prostate
would have decreased by a maximum 20% compared to the
planned V98 if real time position monitoring had not been
performed and position corrections were not undertaken.
The stereo image based position correction available in
SeedTracker was shown to be minimum 2 mins faster than
the conventional orthogonal image based approach.
Conclusion
The SeedTracker system has been shown to enable the
accurate real time position monitoring and position
corrections during prostate SBRT. The occurance of real
position deviations during dose delivery was identified by
the SeedTracker leading to improved accuracy of dose
delivery to the prostate.
EP-1624 Respiratory gating of an Elekta linac using a
Microsoft Kinect v2 system
D. Edmunds
1
, K. Tang
2
, R. Symonds-Tayler
3
, E. Donovan
1
1
The Royal Marsden NHS Foundation Trust, Physics,
Sutton, United Kingdom
2
University of Surrey, Physics, Guildford, United
Kingdom
3
Institute of Cancer Research, Physics, Sutton, United
Kingdom
Purpose or Objective
To investigate whether it is possible to gate radiation
delivery from an Elekta linac, using a commercial off-the-
shelf (COTS) depth sensor, based on data acquired from
patients in a clinical study. The goal of this work is to
achieve real time breath-hold monitoring and gating for
voluntary breath-hold (VBH) treatments for breast cancer
patients.
Material and Methods
Six participants from the UK HeartSpare trial who had
received left breast radiotherapy while performing VBH
were recruited for this study. The patients were set up on
an Elekta Synergy in a radiotherapy treatment room
exactly as in their original treatment. They then
performed a sequence of 3 breath holds for a period of
approximately 20 seconds each, during a simulated whole
breast treatment with both lateral and medial beams, plus
a VMAT delivery. A Microsoft Kinect Version 2 (Kinect v2)
commodity depth sensor was used to record breathing
traces during this time.
These breathing traces were then used as input to a
programmable motion platform carrying a solid water
phantom placed on the treatment couch, which was
monitored with a Kinect v2. In-house C++ software (see
Fig. 1) was used to set a gating threshold, and when the
phantom moved outside of this threshold, radiation
delivery was paused via signals sent through a fibre optic
connection to the linac’s gating interface. Radiation dose
was verified using a calibrated ionisation chamber and
electrometer, with the chamber positioned inside the
phantom. A dose measurement was performed for a 200
MU radiation delivery, both with and without gating in
place.
Figure 1: Screenshot of in-house C++ Kinect v2 software,
showing a depth image from the camera. The solid water
phantom and linac head can be seen in the centre of the
image. A region of interest (ROI) is drawn as a white
rectangle, and the mean distance of pixels in the ROI is
calculated at 60 Hz. Gating threshold can be set with
software controls (top left), and a control signal is
transmitted to the linac gating interface via fibre-optic
cables.
Results
Kinect v2 was able to acquire all breath holds from each
patient successfully. Extracted traces from each patient
depth file were sent to the motion platform. In the gating
experiments (see Fig. 2), a Kinect v2 was able to track the
phantom motion with a root mean square error of between
0.6 mm and 1.3 mm. The latency of our in-house gating
software was found to range between 30 and 100 ms. In
all cases, the gated radiation delivery dose agreed with
the baseline dose measurement without gating to better
than
0.4%.