MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Intratumoral and Peritumoral CT Radiomic Modeling to Predict Treatment Failure of Early Stage Non-Small Cell Lung Cancers

K Lafata1*, Y Gao2 , Y Chang3 , C Wang4 , C Kelsey5 , F Yin6 , (1) Duke University Medical Center, Durham, NC, (2) Duke University Medical Physics Graduate Program, Durham, NC, (3) Duke University Medical Center, Durham, NC, (4) Duke University Medical Center, Durham, NC, (5) Duke University, Durham, ,(6) Duke University Medical Center, Durham, NC

Presentations

(Tuesday, 7/16/2019) 7:30 AM - 9:30 AM

Room: Stars at Night Ballroom 2-3

Purpose: To investigate the association between CT radiomics data and treatment failure of early-stage non-small cell lung cancers (NSCLC). We hypothesized that (1) intratumoral radiomic features are associated with local recurrence; and (2) peritumoral radiomic features are associated with distant metastasis.

Methods: Seventy patients who received stereotactic body radiotherapy for NSCLC were retrospectively identified as part of an IRB-approved clinical trial, where robust follow-up data was collected. The gross-tumor-volume (GTV) was segmented on pre-treatment CT images, from which 61 intratumoral radiomic features were extracted as potential biomarkers for local control. The peritumoral region encapsulating the GTV was then defined as a uniform expansion relative to – but not including – gross disease. Peritumoral volumes were segmented at radial distances of 3mm, 6mm, 9mm, and 12mm, from which 55 radiomic features were extracted as potential biomarkers for metastatic disease. LASSO regularized logistic regression modeling was implemented to quantify the multivariate relationship between: (1) intratumoral features and local recurrence; and (2) peritumoral features and metastasis. Model performance was based on Receiver Operating Characteristic (ROC) curve analysis, and generalization was evaluated using stratified Monte Carlo cross-validation. Intratumoral and peritumoral models were developed in parallel, with identical hyper-parameterization, to compare the relative performance between each technique.

Results: Intratumoral radiomics data was more predictive of local recurrence (AUC=0.76±0.03) than distant metastasis (AUC=0.57±0.03, p<0.001). The 3mm peritumoral radiomics data was the most predictive of distant metastasis (AUC=0.72±0.03), which was a statistically significant improvement over the intratumoral approach (AUC=0.57±0.03, p<0.001). The most important radiomic feature selected via LASSO for the intratumoral and peritumoral approach, respectively, was Long Run High Gray-Level Emphasis and Gray-Level Non-uniformity.

Conclusion: Quantitative results suggest that dense tumors with a homogeneous coarse texture and a heterogeneous peritumoral shell were associated with an increased risk of treatment failure in this cohort of early-stage NSCLC patients.

Keywords

Modeling, Quantitative Imaging, Radiation Therapy

Taxonomy

IM- CT: Quantitative imaging/analysis

Contact Email