932 / 1082
S916
ESTRO 36 2017
_______________________________________________________________________________________________
features in the two CT modalities were compared based
on the intraclass correlation coefficient (ICC1). The
prognostic value of both modalities was assessed using
univariable Cox regression. Additionally, in a previous
analysis we built a local tumor control prognostic model
based on contrast enhanced CT images of 93 patients. It
comprised three radiomic features (HHH large zone high
grey-level emphasis, LLL sum entropy and LLH difference
variance). The performance of this model was
comparatively tested in the two new CT datasets.
Results
118 out of 693 radiomic features showed a good
agreement between native CT and contrast enhanced CT
imaging (ICC>0.8). None of the intensity features was
stable and only 7 of the texture features in the non
transformed map. In the univariable Cox regression 96 and
107 radiomic features in the native and contrast-enhanced
CT images had a significant influence on the local control,
respectively. 62 parameters were prognostic in both
modalities, but only half of them showed a good
agreement between native and contrast enhanced CT
images (ICC>0.8). Two out of the three features of our
previously developed model were stable in respect to the
administration of contrast agent in the CT images
(ICC>0.8). However, our model was only predictive for the
parameter set derived from the contrast enhanced CT
images (CI=0.69, p=0.013). Based on this, patients could
be divided into 2 risk groups (p=0.02, Figure 1).
Figure 1: The patients could be divided into two risk
groups (p=0.02) based on the radiomics of contrast
enhanced CT images and the model derived from our
previous training cohort (contrast enhanced CTs). The
splitting was not significant for radiomics of native CT
images.
Conclusion
Radiomic models can be built on a mixed set of native and
contrast-enhanced CT images but with a reduced number
of suitable radiomic features. A model built exclusively
with contrast enhanced CT images cannot be validated in
a set of native CT Images.
EP-1698 Impact of motion and segmentation on PET
texture features: evaluation with heterogeneous
phantoms.
M. Carles
1,2
, I. Torres-Espallardo
2
, D. Baltas
1
, U. Nestle
1
,
L. Martí-Bonmatí
2
1
University Medical Center, Department of Radiation
Oncology, Freiburg, Germany
2
Hospital Universitario y Politécnico La Fe, Medical
Imaging, Valencia, Spain
Purpose or Objective
A major source of error in quantitative PET/CT of lung
cancer tumors is respiratory motion. Regarding variability
of PET textures features (TF), the impact of respiratory
motion has not been properly studied with experimental
phantoms. The primary aim of this work was to evaluate
the current use of PET texture analysis for heterogeneity
characterization in lesions affected by respiratory motion.
Material and Methods
Twenty-eight heterogeneous lesions were simulated by a
mixture of alginate and 18F- fluoro-2-deoxy-D-glucose
(FDG). Sixteen respiratory patterns were applied. Firstly,
the TF response for different heterogeneous phantoms and
its robustness with respect to the segmentation method
were calculated. Secondly, the variability for TF derived
from PET image with (gated, G-) and without (ungated, U-
) motion compensation was analyzed. Finally, TF
complementarity was assessed.
Results
In the comparison of TF derived from the ideal contour
(VOI_Ideal) with respect to TF derived from 40%-threshold
(VOI_40%) and adaptive-threshold (VOI_COA) PET
contours, 7/8 TF showed strong linear correlation LC
(p<0.001, r>0.75), despite a significant volume
underestimation. Independence on lesion movement (LC
in 100% of the combined pairs of movements, p<0.05) was
obtained for 1/8 TF with U-image (width of the volume-
activity histogram, WH) and 4/8 TF with G-image (WH and
energy ENG, local-homogeneity LH and entropy ENT,
derived from the co-ocurrence matrix). Apart from WH
(r>0.9, p<0.001), no one of these TF has shown LC with
Cmax. Complementarity was observed for the TF pairs:
ENG-LH, CONT-ENT and LH-ENT.
Conclusion
In conclusion, effect of respiratory motion should be taken
into account when heterogeneity of lung cancer is
quantified on PET/CT images. Despite inaccurate volume
delineation, TF derived from 40% and COA contours could
be reliable for their prognostic use. The TF that exhibited
simultaneously added value and independence of lesion
movement were ENG and ENT computed from G-image.
Their use is therefore recommended for heterogeneity
quantification of lesions affected by respiratory motion.
EP-1699 The simulation study of position and biology
of target with PET in high energy X-Ray irradiation
Q. Zhang
1
1
Topgrade Medical - Yiren Hospiatl, Radio- therapy
center, BEIJING, China
Purpose or Objective
To study the possibility of in situ
verification of dose distributions and position in radiation
therapy with PET imaging based on the activity
distribution of
11
C and
15
O produced via photonuclear
reactions in patient irradiated by 45MV X ray from the
LA45 accelerator.
Material and Methods
The method is based on the photonuclear reactions in the
most elemental composition
12
C,
16
O in body tissues
irradiated with high-energy photons with energies up to 45
MeV, resulting primarily in
11
C and
15
O, which are positron
emitting nuclei. The induced positron activity
distributions were obtained with a PET scanner in the
same room of a LA45 accelerator (Top Grade Medical,
Beijing, China). The activity distributions of
11
C and
15
O
were used to verify the dose distributions and position in
tarfet as delivered by the LA45 accelerator.
The experiments were performed with a brain phantom.
Radiation beams were delivered to the phantom according
to realistic radiation therapy treatment plans. The
phantom was immediately transfer to PET anfd was
scanned on the PET after irradiation. The scans were
performed at 20 minutes and 2-5 minutes after irradiation
for
11
C and
15
O, respectively. The interval between the two
scans was 20 minutes. The activity distributions of
11
C and
15
O within the irradiated volume can be separated from
each other because the half-life of
11
C and
15
O is 20