A radiomic low-dose computed tomography (LDCT)-based approach may be promising for characterizing indeterminate screen-detected lung nodules according to an article published by PLoS ONE.
Lung cancer accounts for more cancer deaths in the United States than colon, prostate, and breast cancer combined. The National Lung Screening Trial (NLST) showed a 20% relative reduction in lung cancer death with annual LDCT. While these results generated widespread endorsement of lung cancer screening, 40% of individuals randomly assigned to LDCT screening had at least one pulmonary nodule identified. Ultimately, 96% of those nodules were found to be benign.
Estimates suggest that there are 1.5 million incidentally discovered indeterminate lung nodules each year. With more than 10 million adults in the United States who meet the criteria for screening eligibility, full implementation of LDCT screening would increase this number dramatically. Furthermore, given that invasive methods of determining which nodules are benign and which are malignant are associated with increased mortality and morbidity, finding a noninvasive method of doing so is crucial.
Tobias Peikert, MD, of the Division of Pulmonary and Critical Care Medicine at the Mayo Clinic in Rochester, Minnesota, and colleagues used the NLST dataset to develop independent quantitative variables to assess various radiologic nodule features such as sphericity, flatness, elongation, spiculation, lobulation, and curvature. The investigators selected 726 indeterminate nodules from the NLST dataset that were all ≥7 mm (318 benign and 408 malignant) on which to base their assessment. They performed a multivariate analysis using a least absolute shrinkage and selection operator (LASSO) method.
The investigators selected 8 of the original 57 quantitative radiologic features considered using LASSO multivariate modeling. These 8 features included metrics capturing spatial location of the nodule, nodule size, bulk metrics, radiodensity metrics, nodule texture/density metrics, texture/density nodule surrounding lung metrics, metrics capturing the nodule surface descriptors, and metrics capturing the distribution of the nodule surface characteristics exemplars.
The study findings were limited by the heterogeneous nature of the NLST CT dataset sample, the relatively small number of cases, and the semi-automatic segmentation technique used, which could introduce operator-associated variability. In addition, the model must be externally validated before clinical use.
The investigators noted that their radiomic CT-based approach to classifying lung nodules is promising and may lower the risks associated with lung cancer screening by reducing avoidable mortality, morbidity, radiation exposure, patient anxiety, and healthcare costs.
Peikert T, Duan F, Rajagopalan S, et al. Novel high-resolution computed tomography-based radiomic classifier for screen-identified pulmonary nodules in the National Lung Screening Trial [published online May 14, 2018]. PLoS ONE. doi:10.1371/journal.pone.0196910