Abstract 6392: Predicting malignancy in indeterminate pulmonary nodules using quantitative CT imaging

Abstract

Abstract With an increasing number of positive lung cancer screening trials and the growing utilization of low dose CT screening, the detection of indeterminate pulmonary nodules is an important clinical problem. A biomarker that will be able to differentiate between benign andmalignant nodules would help to accelerate diagnosis and reduce unnecessary and invasive procedures. Here, we investigated the utility of quantitative radiomics to predict malignancy of small pulmonary nodules. Data including cancer status, age, gender, smokingstatus and CT radiomics data, from 242 patients accrued from the Detection of Early Lung Cancer Among Military Personnel (DECAMP1) consortium was used. DECAMP1 is a prospective study of 500 current or former smokers with indeterminate pulmonary nodules (0.7-3.0 cm). A total of 446 quantitative radiomic features were extracted from CT images of each patient. For the prediction models, various feature selection methods and machine learning algorithms including Random Forest, Gradient Boosting Machine and Support Vector Machine were used. Data was split into train and test set with tenfold cross-validation.Models were evaluated using; Accuracy, Sensitivity, Specificity and Area under the curve - a receiver operating characteristic curve (AUC-ROC). In terms of predictive performance, the AUC value was 0.64 (95% CI: 0.50-0.78) for the clinical model and 0.77 (95% CI: 0.64-0.90) for the radiomics model. Adding clinical features to the radiomics model (gender, age, smoking status) did not improve the model. We have also investigated the slice thickness and scanner (PET vs PET-CT) variables on the models’ performance and saw a modest improvement of the AUC values when using more homogenous data. In this study, we showed the potential of radiomics for lung cancer prediction in DECAMP1.For future work, we are planning to test whether incorporating bronchial gene expression data improves lung cancer detection and to validate these findings in separate, independent cohorts. Citation Format: Kate M. Bloch, Xingyi Shi, Fenghai Duan, George R. Washko, Avrum Spira, Denise R. Aberle, Raul S. Estepar, Ehab Billatos, Marc E. Lenburg, on behalf of the DECAMP Investigators. Predicting malignancy in indeterminate pulmonary nodules using quantitative CT imaging [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 6392.