Results and Conclusions Methods What is texture analysis

Texture analysis of 18F-FDG PET/CT predicts
local control of stage I NSCLC treated by SBRT.
Kazuya Takeda1*, Kentaro Takanami2, Yuko Shirata1, Takaya Yamamoto1, Noriyoshi Takahashi1, Kengo Ito1, Kei Takase2 and Keiichi Jingu1
1 Department of Radiation Oncology, Tohoku University Graduate School of Medicine, 1-1 Seiryo-machi, Aoba-ku, Sendai, 980-8574, Japan
2 Department of Diagnostic Radiology, Tohoku University Graduate School of Medicine, 1-1 Seiryo-machi, Aoba-ku, Sendai, 980-8574, Japan
*e-mail address: [email protected]
Key words: Lung cancer; stereotactic body radiotherapy (SBRT); texture analysis.
Parameter selection
Introduction and Objectives
Positron emission tomography (PET) is clinically useful in cancer population and may also provide prognostic information. Histogram-based parameters such as maximum standard uptake value (SUV) have been widely investigated but it is controversial whether maximum SUV can predict patients’
prognosis(1,2). Recently, it has been reported that texture analysis of PET
image may provide prognostic information(3).
Conventional parameters
Texture parameters
Maximum SUV
Entropy
Metabolic tumor volume (MTV)
High-intensity large-area emphasis (HILAE)
Total lesion glycolysis (TLG)
Zone percentage
Statistical analysis
Reproducibility evaluation
Local control (LC) evaluation
Dissimilarity
Intraclass correlation coefficients (ICC)
Cox proportional hazards model
Results and Conclusions
What is texture analysis?
Median follow-up period was 30.1 month and 8 (23%) patients occurred
Texture analysis is a method to analyze image data which focuses on
spatial arrangement of voxel values within a tumor. The first step in calculating a texture parameters is preparing a matrix (e.g. coocurrence
matrix, run-length matrix and size-zone matrix) based on spatial arrangement of voxel values. Then each parameter is calculated based
on the matrix.
Conventional parameters
local relapse. Between two observers, six parameters besides zone percentage (ICC value 0.59) showed ICC value ranged between 0.81 and 1.00.
In univariate analysis, there were significant correlations between LC and
tumor diameter>30mm, MTV >= 5.14cm3, TLG >= 59.7, entropy >= -34.3,
dissimilarity >= 2235 and treatment biological equivalent dose>=100Gy, respectively. Maximum SUV>=10.4 was not a significant predictor for LC.
Texture parameters
(SUV, run length) = (4,2) run-length matrix
maximum SUV = 4
mean SUV = 2.1
MTV = 14
TLG = 30 ...
3 conventional parameters and 4 texture parameters
(SUV, area) = (6,3)
size-zone matrix
(SUVleft, SUVright) = (2,4)
cooccurrence
Univariate analysis for local control
Parameter calculation
(SUV) 0 2 4 6 8
female vs. male
≥75 vs. <75
0, 1 vs. 2, 3
>30mm vs. ≤ 30mm
≥5.14 vs. <5.14
≥10.4 vs. <10.4
≥59.7 vs. <59.7
≥-34.3 vs. <-34.3
≥2235 vs. <2235
≥1743 vs. <1743
≥0.46 vs. <0.46
≥100Gy vs. <100Gy
Sex
Age (year)
PS
Tumor diameter
MTV
maximum SUV
TLG
Entropy
Dissimilarity
HILAE
Zone percentage
BED
In this study, we evaluated reproducibility and predictive value of some
texture parameters and existing parameters of 18F-FDG PET/CT image in patients with early stage non-small cell lung cancer (NSCLC) treated by stereotactic body radiation therapy (SBRT).
Methods
ICC value
0.91
1.00
0.99
0.81
0.94
0.91
0.59
-
HR (95% CI)
n.s.
n.s.
n.s.
7.21 (1.40-33.4)
9.38 (1.66-176)
n.s.
5.86 (1.12-27.6)
0.13 (0.01-0.72)
6.87 (1.18-131)
n.s.
n.s.
0.22 (0.04-0.92)
p value
0.50
0.29
0.73
0.02
0.01
0.09
0.04
0.02
0.03
0.07
0.16
0.04
Local control rate plotted with Kaplan-Meier method
Treatment
SBRT with total dose of 40-60Gy in 4-15 fractions.
PET image
18F-FDG PET/CT scan acquired before treatment.
Tumor delineation
PET Edge on MIM (MIM Software Inc., Cleveland).
Parameter calculation CGITA on MATLAB (MathWorks Inc., Natick, MA).
(4)
1.0
Hazard ratio 0.13
0.8
(95% CI 0.01-0.72, p=0.02)
0.6
0.4
≥-34.3
<-34.3
Entropy
0.2
0
Gradient-based semi-
1.0
0
20
40
60
Local control rate
Thirty patients with early stage NSCLC (T1-2N0M0).
Local control rate
Objective
80
Time from treatment (month)
Hazard ratio 6.87
0.8
(95% CI 1.18-131 , p=0.03)
0.6
0.4
0
≥2235
<2235
Dissimilarity
0.2
0
20
40
60
80
Time from treatment (month)
automated delineation
Conclusion: Texture analysis based on gradient-based delineation
Import to CGITA
PET Edge on MIM
Parameter
calculation
Statistical
analysis
ESTRO 2017
EP-1687
larity calculated from co-occurrence matrix are potentially beneficial to predict LC with reproducibility in patients with NSCLC treated by SBRT. To establish utility of texture analysis in 18F-FDG PET/CT image, further study in-
Parameter
selection
References
method has high reproducibility in most parameters. Entropy and dissimi-
cluding prospective trial will be needed.
Disclosure of COI: The authors have no conflict of interest to disclosure.
(1) Takeda A, Sanuki N, Fujii H, et al. Maximum Standardized Uptake Value on FDG-PET Is a Strong Predictor of Overall and Disease-Free Survival for Non–Small-Cell Lung Cancer Patients after Stereotactic Body
Radiotherapy. J Thorac Oncol 2014;9(1):65–73.
(2) Burdick M, Stephans K, Reddy C, et al. Maximum Standardized Uptake Value From Staging FDG-PET/CT Does not Predict Treatment Outcome for Early-Stage Non–Small-Cell Lung Cancer Treated With Stereotactic
Body Radiotherapy. Int J Radiat Oncol Biology Phys 2010;78(4):1033–9.
(3) Pyka T, Bundschuh RA, Andratschke N, et al. Textural features in pre-treatment [F18]-FDG-PET/CT are correlated with risk of local recurrence and disease-specific survival in early stage NSCLC patients receiving primary
stereotactic radiation therapy. Radiat Oncol 2015;10:100.
(4) Fang Y-H, Lin C-Y, Shih M-J, et al. Development and Evaluation of an Open-Source Software Package ‘CGITA’ for Quantifying Tumor Heterogeneity with Molecular Images. Biomed Res Int 2014;2014:1–9.
Physics track: (Quantitative) functional and biological imaging
Kazuya Takeda
DOI: 10.3252/pso.eu.ESTRO36.2017
Poster
presented at: