920 / 1082
S904
ESTRO 36 2017
_______________________________________________________________________________________________
5
Fundación Instituto Valenciano de Oncología, Servicio de
Radioterapia, Valencia, Spain
Purpose or Objective
To assess if dual energy computed tomography (DECT)
quantitative imaging can distinguish necrotic tumours in
lung cancer.
Material and Methods
From July 2013 to June 2016, 83 patients who underwent
a DECT study were reviewed for their lung tumour necrosis
status (33 positive; 50 negative).
Lesion size varied considerably: the mean lesion volume
was 15 cm
3
(range 0.05-138 cm
3
). Malignant lesions were
predominantly adenocarcinoma (77.1%), squamous cell
carcinoma (13.2%) and metastases (7.2%).
DECT examination was performed on a Discovery CT 750
HD scanner (GE Healthcare, WI, USA). Patients were
injected with 1.35 ml/kg of body weight of non-ionic
iodinated contrast material at 4 ml/s (Iopamidol, 300
mg/ml; Bracco, Italy). A Gemstone Spectral Imaging (GSI)
DECT exam of the entire chest was performed at arterial
phase.
Lesion volume was semi-automatically segmented using
Dexus lung nodule function (ADW4.6; GE Healthcare, USA)
by two radiologists. Images for quantitative iodine content
ρ
I
(mg/cm
3
) and effective atomic number (Z
eff
) were
reconstructed. Maximum, mean and standard deviation
values were recorded for both parameters and for
conventional HU image. Lesion volume and diameter were
also registered. Inter- and intra-observer intraclass
correlation coefficient (ICC) was studied.
Bilateral statistical analysis was performed using the
Mann-Whitney U test. Due to multiple comparisons,
Bonferroni adjustment was made and significance was set
at p < 0.007. Receiver operating characteristic (ROC)
curves were generated and diagnostic capability was
determined by calculating the area under the ROC curve
(AUC). The licensed statistical software package SPSS 20
(IBM, Somers, NY, USA) was used.
Results
Reproducibility of intraobserver lung lesion the ICC was
0.95 (CI 95% 0.80–0.98) and interobserver ICC was 0.92 (CI
95% 0.70–0.98).
The bivariate analysis for distinguishing necrotic from non-
necrotic lesions revealed statistically significant
differences. Larger lesions presented more necrosis than
smaller ones, as previously known in the literature. Values
for p, AUC and its 95% confidence level interval are shown
in
Table 1.
Box-whisker and ROC plots are displayed in Fig. 1 for mean
Z
eff
variable, which presented highest AUC (0.890). Mean
Z
eff
presented a 84.0% sensitivity and 81.8% specificity for
a threshold of 8.96 in ROC curves.
Conclusion
DECT imaging gives information on tumour necrosis.
Quantitative parameters (ρ
I
and Z
eff
) showed better
sensibility and specificity compared to standard HU
imaging. Mean Z
eff
showed better correlation with necrosis
status, due to necrotic core absorbs less iodine contrast.
Our approach has some advantages. Whole tumour semi-
automatic contouring had excellent reproducibility. No
cases were excluded due to geometry or mediastinal
contact.
This method could be a solid approach to assess necrosis
condition. However, we have not studied relationship
with the actual location of necrosis, so it would not be
useful for dose-painting protocols at necrotic core.
EP-1681 [C11]Choline PET/MRI for Prostate Cancer:
Identify Imaging Characteristics Predicting Metastasis
J.R. Tseng
1
, L.Y. Yang
2
, H.Y. Chang
2
, T.C. Yen
1
1
Chang Gung Memorial Hospital at Linkou, Nuclear
Medicine and Molecular Imaging Center, Kwei-Shan-
Taoyuan City, Taiwan
2
Chang Gung Memorial Hospital at Linkou, Biostatistics
Unit- Clinical Trial Center of Chang Gung Memorial
Hospital, Kwei-Shan- Taoyuan City, Taiwan
Purpose or Objective
Intergraded PET/MRI is a powerful imaging modality for
prostate cancer (Pca) in several aspects, from cancer
detection, primary staging, to staging of recurrent Pca.
The goal of primary staging is to detect metastatic spread
from the main tumor. In high risk Pca patients (PSA >20
ng/ml, or Gleason score of 8–10, or clinical stage T3a),
intergraded PET/MRI imaging may have great potential to
change clinical management. In the current study, we
aimed to identify imaging characteristics of main tumor
which can significantly predict distant metastasis.
Material and Methods
This prospective clinical study was approved by the Ethics
Committee (approval 102-3271A and 103-4561C). Since
January 2015 to June 2016, total 30 Pca patients
committed high risk criteria were enrolled to conduct
whole body integrated [C11]Choline PET/MRI (biograph
mMR, Simens). The PET and MRI imaging was interpreted
independently by one clinically-experienced nuclear
medicine physician and radiologist. In the PET imaging
analysis, main tumors were segmented using PMOD 3.3
software package. The borders of volumes of interest were
set by manual adjustment to avoid physiological
[C11]Choline uptake in the urine or intestine. The tumor
boundaries were automatically contoured based on the
thresholds of SUV 2.65. The gray-level run length encoding
matrix (GLRLM) was used for assessing the regional texture
features. In the MRI imaging analysis, anatomic (T2-
weighted MRI) and functional (diffusion-weighted MRI)
imaging features were documented. Multivariate
classification and regression tree analysis was used to
determine the best combination of variables and the
related cutoffs to predict risk for distant metastasis.
Results
The mean age is 70.1±6.2 years, and the mean PSA level
is 91.6 ± 139.4 ng/ml. In these 30 patients, 26 (87%) are
categorized as clinical stage IV, 4 (13%) as stage III. Fifteen
(50%) patients have distant metastasis, including 7 (23%)
non-regional lymph nodes metastasis, 11 (36%) bone
metastasis, 1 (3%) visceral organ metastasis. The