Program Information
CT-Based Volumetric Features Are Associated with Somatic Mutations in Lung Cancer
T Coroller*, P Grossmann , Y Hou , S Lee , R Mak , H Aerts , Brigham and Women s Hospital, Dana Farber Cancer Institute, Harvard Medical school, Boston, MA
Presentations
SU-E-J-246 (Sunday, July 12, 2015) 3:00 PM - 6:00 PM Room: Exhibit Hall
Purpose: Subsets of non-small cell lung cancer (NSCLC) are driven by mutations in key oncogenes, with unique biology including susceptibility to targeted treatment. Additionally, those mutations could lead to phenotypic differences of the primary tumor that can be assess with quantitative imaging. In this study, we investigated whether somatic mutation are associated with, and hence can be predicted by CT tumor volume-based features of NSCLC patients.
Methods: We included 117 NSCLC patients with treatment-planning CT scans in our analysis and clinical genotyping for the epidermal growth factor receptor (EGFR) and Kirsten rat sarcoma viral oncogene homolog (KRAS) oncogenes. We extracted four volumetric features describing volume and diameters (x/y axis and 3D) of the primary tumor. Volumetric differences between mutant and wild-type tumors were assessed using Wilcoxon test. Predictive value of the volumetric features for mutation status was assessed using the area under the curve (AUC).
Results: Genotype distribution included 14 (12%) EGFR mutant, 35 (30%) KRAS mutant, and 68 (58%) wild-type tumors. All volumetric features for EGFR mutant were significantly (p-value <0.05) lower than for KRAS mutant and Wild-Type. No volumetric features were significantly different between KRAS and Wild-Type. The median (Q1-Q3) for volume was 10.2(6.1-29.6), 39.3(14.3-89.7) and 49(10.7-119) for EGFR, KRAS and Wild-Type respectively. All volumetric features were also significant predictive features for EGFR mutation with median (range) AUC of 0.69(0.67-0.70) and all p-value<0.05. However, the AUC was only 0.51(0.50-0.51) for KRAS mutation.
Conclusion: EGFR mutant primary tumors were significantly smaller (for all volumetric features) than KRAS or Wild-Type. Moreover, all volumetric features were significantly predictive for EGFR. KRAS and Wild-type could not be discriminated only based on volumetric features. A larger set of imaging features (e.g. Radiomics) would help find more predictive biomarkers for tumor mutation status.
Contact Email: