MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Predicting Lung SBRT Clinical Outcomes Using Planning-CT Radiomics

K Lafata1*, R Geng2 , B Ackerson3 , C Kelsey4 , J Torok5 , F Yin6 , (1) Duke University, Durham, NC, (2) ,Durham, NC, (3) Duke University Medical Center, Durham, NC, (4) Duke University Medical Center, Durham, North Carolina, (5) Duke University Medical Center, Durham, NC, (6) Duke University Medical Center, Durham, NC

Presentations

(Tuesday, 7/31/2018) 4:30 PM - 5:30 PM

Room: Exhibit Hall | Forum 2

Purpose: To investigate the relationship between radiomic features extracted from pre-treatment CT images and clinical outcomes following stereotactic body radiation therapy (SBRT) for non-small-cell lung cancer (NSCLC).

Methods: Seventy patients who received lung SBRT for stage-1 NSCLC were retrospectively identified as part of an IRB-approved clinical trial. The gross-tumor-volume (GTV) was contoured on pre-treatment free-breathing CT images, from which 43 quantitative radiomic features were extracted to collectively capture tumor morphology, intensity, fine-texture, and coarse-texture. Clinical outcome endpoints were defined as follows. Local failure (LF): cancer recurrence within 2 cm of the GTV; Regional failure (RF): cancer recurrence within regional lymph nodes; and Distant failure (DF): development of metastatic disease. Endpoints were scored based on follow-up CT, PET/CT, and pathological confirmation. The statistical association between the radiomic features and each endpoint was first analyzed using ANOVA, and p-values were corrected for multiple hypotheses testing using the Bonferroni method. Artificial Neural Networks (ANNs) were then developed to model each endpoint based on the radiomic features. The ANN architecture included a 43-dimensional input layer, a single hidden layer with 15 nodes, and a classification output layer. Weight optimization was achieved via backpropagation with scaled conjugate gradient minimization. Model generalization was evaluated using a 10-fold cross-validation technique and Receiver Operating Characteristic (ROC) curve analysis.

Results: Following Bonferroni correction, two features demonstrated a statistically significant association with LF: Homogeneity (p=0.022) and Long-Run-High-Gray-Level-Emphasis (p=0.048). These results indicate that relatively dense tumors with a homogenous coarse texture might be linked to higher rates of local recurrence. No such relationship was observed for either RF or DF. The ANN models produced AUC values of 0.68, 0.61 and 0.56 for LF, RF, and DF, respectively.

Conclusion: Quantitative radiomic features may be more predictive of LF than either RF or DF following lung SBRT for NSCLC.

Funding Support, Disclosures, and Conflict of Interest: This research is partially supported by a grant from Varian Medical Systems.

Keywords

Not Applicable / None Entered.

Taxonomy

Not Applicable / None Entered.

Contact Email