Currently, lung cancer screening by low-dose chest CT is implemented in the United States for high-risk persons. A disadvantage of lung cancer screening is the large number of small-to-intermediate sized lung nodules, detected in around 50% of all participants, the large majority being benign. Accurate estimation of nodule size and growth is essential in the classification of lung nodules. Currently, manual diameter measurements are the standard for lung cancer screening programs and routine clinical care. However, European screening studies using semi-automated volume measurements have shown higher accuracy and reproducibility compared to diameter measurements. In addition to this, with the optimization of CT scan techniques and reconstruction parameters, as well as advances in segmentation software, the accuracy of nodule volume measurement can be improved even further. The positive results of previous studies on volume and diameter measurements of lung nodules suggest that manual measurements of nodule diameter may be replaced by semi-automated volume measurements in the (near) future.

Introduction: Deep Learning has been proposed as promising tool to classify malignant nodules. Our aim was to retrospectively validate our Lung Cancer Prediction Convolutional Neural Network (LCP-CNN), which was trained on US screening data, on an independent dataset of indeterminate nodules in an European multicentre trial, to rule out benign nodules maintaining a high lung cancer sensitivity. Methods: The LCP-CNN has been trained to generate a malignancy score for each nodule using CT data from the U.S. National Lung Screening Trial (NLST), and validated on CT scans containing 2106 nodules (205 lung cancers) detected in patients from from the Early Lung Cancer Diagnosis Using Artificial Intelligence and Big Data (LUCINDA) study, recruited from three tertiary referral centers in the UK, Germany and Netherlands. We pre-defined a benign nodule rule-out test, to identify benign nodules whilst maintaining a high sensitivity, by calculating thresholds on the malignancy score that achieve at least 99 % sensitivity on the NLST data. Overall performance per validation site was evaluated using Area-Under-the-ROC-Curve analysis (AUC). Results: The overall AUC across the European centers was 94.5 % (95 %CI 92.6-96.1). With a high sensitivity of 99.0 %, malignancy could be ruled out in 22.1 % of the nodules, enabling 18.5 % of the patients to avoid followup scans. The two false-negative results both represented small typical carcinoids.

Nodule assessment based on manual diameter measurements is susceptible to nodule margin. This effect is almost neglectable for semi-automated volume measurements. The larger interreader variability for manual diameter measurement results in inaccurate lung nodule growth detection and size classification.