(Sunday, 7/12/2020) [Eastern Time (GMT-4)]
Room: AAPM ePoster Library
Purpose: compare effect of the radiomic features extracted from on non-small cell lung cancer(NSCLC) target of PET imagines delineated under different standardized uptake values(SUV).
Methods: selected 31 patients with NSCLC performed PET/CT scan from Shandong Cancer Hospital from 2016 to 2019. Firstly, the SUV threshold method was used to delineate the gross tumor volume (GTV) on non-small cell lung cancer PET images. and then extracted the radiomics features. Calculate the absolute value of the Spearman correlation coefficient between different SUVs and features in general , each feature was performed horizontal and vertical analysis. Compared the coefficient of variation (%COV) of each feature in the longitudinal direction and compare the intra-class correlation coefficient(ICC)under different SUV values in the horizontal direction. One-way ANOVA was performed for each characteristic ICC value under different SUVs. P<0.05 was statistically significant.
Results: total of 68 effective radiomics features were extracted, The Spearman coefficient indicates that the number of weak correlation, moderate correlation, strong correlation and strong correlation are 0 (0.00%), 0 (0.00%), and 14 (20.59%), 29 (44.62%), 25 (36.76%). The low-variation features, medium-variation features and large-variation features were 19 (27.94%), 14 (20.59%), and 14 (51.47%), respectively. There are 25 (36.76%) with high consistency characteristics. A total of 5 image features with high consistency and stability were obtained.
Conclusion: was a correlation between radiomic features and the selection of SUV values. Different SUV thresholds delineated the radiomic features extracted from the NSCLC target area with different variability, and the features exected from nearly SUV values usually showed high consistency.
Keywords
PET, Radiation Therapy, Segmentation
Taxonomy
IM- PET : Quantitative imaging/analysis
Contact Email
