Diagnostic accuracy of computed tomography imaging for the detection of differences between peripheral small cell lung cancer and peripheral non-small cell lung cancer

Abstract

Background: To evaluate the computed tomography features of peripheral small cell lung cancer and non-small cell lung cancer and to establish a predictive model to conveniently distinguish between them. Materials and methods: We retrospectively reviewed the computed tomography features of 51 patients with peripheral small cell lung cancer and 207 patients with peripheral non-small cell lung cancer after pathological diagnosis. Thirteen computed tomography morphologic findings were included and analyzed statistically. Meaningful features were analyzed by logistic regression for multivariate analysis. We then used β-coefficients as the basis to establish an image scoring prediction model. Result: The meaningful morphologic features for distinguishing between peripheral small cell lung cancer and other tumor types are multinodular shape and lymphadenectasis, with scores of 12 and 11, respectively. The scores ranged from −51 to 23, and the most reasonable cut-off was −24. The available area under the curve was 0.834 (95% confidence interval [CI] 0.783–0.877). Sensitivity and specificity were 86.3% (95% CI 0.737–0.943) and 69.6% (95% CI 0.628–0.758), respectively. Conclusion: The image scoring predictive model that we constructed provides a simple and economical noninvasive method for distinguishing between peripheral small cell lung cancer and peripheral non-small cell lung cancer.