Heuvelmans MA, van Ooijen PMA, Ather S, et al. - Researchers intended to retrospectively verify the Lung Cancer Prediction Convolutional Neural Network (LCP-CNN) on an independent dataset of indeterminate nodules in an European multicenter trial to rule out benign nodules maintaining a high lung cancer sensitivity. The LCP-CNN was trained on US screening data. Experts used area-under-the-ROC-curve analysis (AUC) to assess overall performance per validation site. Across the European centers, an overall AUC of 94.5% was generated. In 22.1% of the nodules, cancer could be ruled out with a high sensitivity of 99.0%; consequently follow-up scans could be avoided in 18.5% of patients. Findings showed excellent performance of LCP-CNN, which was trained on participants with lung nodules from the US National Lung Screening Trial dataset, on detection of benign lung nodules in a multi-center external dataset, as well as its high accuracy in ruling out malignancy in approximately one fifth of the patients with 5-15 mm nodules.