Search

Federated learning enables training models on distributed, privacy-sensitive medical imaging data. However, data heterogeneity across participating institutions leads to reduced model performance and fairness issues, especially for underrepresented datasets. To address these challenges, we propose leveraging the multi-head attention mechanism in Vision Transformers to align the representations of heterogeneous data across clients. By focusing on the attention mechanism as the alignment objective, our approach aims to improve both the accuracy and fairness of federated learning models in medical imaging applications. We evaluate our method on the IQ-OTH/NCCD Lung Cancer dataset, simulating various levels of data heterogeneity using Latent Dirichlet Allocation (LDA). Our results demonstrate that our approach achieves competitive performance compared to state-of-the-art federated learning methods across different heterogeneity levels and improves the performance of models for underrepresented clients, promoting fairness in the federated learning setting. These findings highlight the potential of leveraging the multi-head attention mechanism to address the challenges of data heterogeneity in medical federated learning., Competing Interests: Declaration of competing interest The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Erfan reports financial support was provided by Dutch Cancer Society. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2024 Elsevier B.V. All rights reserved.)

Background and Purpose: The aim of this study was to develop and evaluate a prediction model for 2-year overall survival (OS) in stage I-IIIA non-small cell lung cancer (NSCLC) patients who received definitive radiotherapy by considering clinical variables and image features from pre-treatment CT-scans., Materials and Methods: NSCLC patients who received stereotactic radiotherapy were prospectively collected at the UMCG and split into a training and a hold out test set including 189 and 81 patients, respectively. External validation was performed on 228 NSCLC patients who were treated with radiation or concurrent chemoradiation at the Maastro clinic (Lung1 dataset). A hybrid model that integrated both image and clinical features was implemented using deep learning. Image features were learned from cubic patches containing lung tumours extracted from pre-treatment CT scans. Relevant clinical variables were selected by univariable and multivariable analyses., Results: Multivariable analysis showed that age and clinical stage were significant prognostic clinical factors for 2-year OS. Using these two clinical variables in combination with image features from pre-treatment CT scans, the hybrid model achieved a median AUC of 0.76 [95 % CI: 0.65-0.86] and 0.64 [95 % CI: 0.58-0.70] on the complete UMCG and Maastro test sets, respectively. The Kaplan-Meier survival curves showed significant separation between low and high mortality risk groups on these two test sets (log-rank test: p-value < 0.001, p-value = 0.012, respectively) CONCLUSION: We demonstrated that a hybrid model could achieve reasonable performance by utilizing both clinical and image features for 2-year OS prediction. Such a model has the potential to identify patients with high mortality risk and guide clinical decision making., Competing Interests: Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2023 The Author(s). Published by Elsevier B.V. All rights reserved.)

Objective: To evaluate the performance of a deep learning-based computer-aided detection (DL-CAD) system in a Chinese low-dose CT (LDCT) lung cancer screening program., Materials and Methods: One-hundred-and-eighty individuals with a lung nodule on their baseline LDCT lung cancer screening scan were randomly mixed with screenees without nodules in a 1:1 ratio (total: 360 individuals). All scans were assessed by double reading and subsequently processed by an academic DL-CAD system. The findings of double reading and the DL-CAD system were then evaluated by two senior radiologists to derive the reference standard. The detection performance was evaluated by the Free Response Operating Characteristic curve, sensitivity and false-positive (FP) rate. The senior radiologists categorized nodules according to nodule diameter, type (solid, part-solid, non-solid) and Lung-RADS., Results: The reference standard consisted of 262 nodules ≥ 4 mm in 196 individuals; 359 findings were considered false positives. The DL-CAD system achieved a sensitivity of 90.1% with 1.0 FP/scan for detection of lung nodules regardless of size or type, whereas double reading had a sensitivity of 76.0% with 0.04 FP/scan (P = 0.001). The sensitivity for detection of nodules ≥ 4 - ≤ 6 mm was significantly higher with DL-CAD than with double reading (86.3% vs. 58.9% respectively; P = 0.001). Sixty-three nodules were only identified by the DL-CAD system, and 27 nodules only found by double reading. The DL-CAD system reached similar performance compared to double reading in Lung-RADS 3 (94.3% vs. 90.0%, P = 0.549) and Lung-RADS 4 nodules (100.0% vs. 97.0%, P = 1.000), but showed a higher sensitivity in Lung-RADS 2 (86.2% vs. 65.4%, P < 0.001)., Conclusions: The DL-CAD system can accurately detect pulmonary nodules on LDCT, with an acceptable false-positive rate of 1 nodule per scan and has higher detection performance than double reading. This DL-CAD system may assist radiologists in nodule detection in LDCT lung cancer screening., (Copyright © 2021. Published by Elsevier B.V.)

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 follow-up scans. The two false-negative results both represented small typical carcinoids., Conclusion: The LCP-CNN, trained on participants with lung nodules from the US NLST dataset, showed excellent performance on identification of benign lung nodules in a multi-center external dataset, ruling out malignancy with high accuracy in about one fifth of the patients with 5-15 mm nodules., (Crown Copyright © 2021. Published by Elsevier B.V. All rights reserved.)

Purpose: Early detection of lung cancer is of importance since it can increase patients' chances of survival. To detect nodules accurately during screening, radiologists would commonly take the axial, coronal, and sagittal planes into account, rather than solely the axial plane in clinical evaluation. Inspired by clinical work, the paper aims to develop an accurate deep learning framework for nodule detection by a combination of multiple planes., Methods: The nodule detection system is designed in two stages, multiplanar nodule candidate detection, multiscale false positive (FP) reduction. At the first stage, a deeply supervised encoder-decoder network is trained by axial, coronal, and sagittal slices for the candidate detection task. All possible nodule candidates from the three different planes are merged. To further refine results, a three-dimensional multiscale dense convolutional neural network that extracts multiscale contextual information is applied to remove non-nodules. In the public LIDC-IDRI dataset, 888 computed tomography scans with 1186 nodules accepted by at least three of four radiologists are selected to train and evaluate our proposed system via a tenfold cross-validation scheme. The free-response receiver operating characteristic curve is used for performance assessment., Results: The proposed system achieves a sensitivity of 94.2% with 1.0 FP/scan and a sensitivity of 96.0% with 2.0 FPs/scan. Although it is difficult to detect small nodules (i.e., <6 mm), our designed CAD system reaches a sensitivity of 93.4% (95.0%) of these small nodules at an overall FP rate of 1.0 (2.0) FPs/scan. At the nodule candidate detection stage, results show that the system with a multiplanar method is capable to detect more nodules compared to using a single plane., Conclusion: Our approach achieves good performance not only for small nodules but also for large lesions on this dataset. This demonstrates the effectiveness of our developed CAD system for lung nodule detection., (© 2020 The Authors. Medical Physics published by Wiley Periodicals LLC on behalf of American Association of Physicists in Medicine.)

Rationale and Objectives: To describe the rational and design of a population-based comparative study. The objective of the study is to assess the screening performance of volume-based management of CT-detected lung nodule in comparison to diameter-based management, and to improve the effectiveness of CT screening for chronic obstructive pulmonary disease (COPD) and cardiovascular disease (CVD), in addition to lung cancer, based on quantitative measurement of CT imaging biomarkers in a Chinese screening setting., Materials and Methods: A population-based comparative study is being performed, including 10,000 asymptomatic participants between 40 and 74 years old from Shanghai urban population. Participants in the intervention group undergo a low-dose chest and cardiac CT scan at baseline and 1 year later, and are managed according to NELCIN-B3 protocol. Participants in the control group undergo a low-dose chest CT scan according to the routine CT protocol and are managed according to the clinical practice. Epidemiological data are collected through questionnaires. In the fourth year from baseline, the diagnosis of the three diseases will be collected., Results: The unnecessary referral rate will be compared between NELCIN-B3 and standard protocol for managing early-detected lung nodules. The effectiveness of quantitative measurement of CT imaging biomarkers for early detection of lung cancer, COPD and CVD will be evaluated., Conclusion: We expect that the quantitative assessment of the CT imaging biomarkers will reduce the number of unnecessary referrals for early detected lung nodules, and will improve the early detection of COPD and CVD in a Chinese urban population., Trial Registration: ClinicalTrials.gov, NCT03988322. Registered on 14 June 2019., (Copyright © 2020 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved.)

Background and Objective: To investigate the effect of the slab thickness in maximum intensity projections (MIPs) on the candidate detection performance of a deep learning-based computer-aided detection (DL-CAD) system for pulmonary nodule detection in CT scans., Methods: The public LUNA16 dataset includes 888 CT scans with 1186 nodules annotated by four radiologists. From those scans, MIP images were reconstructed with slab thicknesses of 5 to 50 mm (at 5 mm intervals) and 3 to 13 mm (at 2 mm intervals). The architecture in the nodule candidate detection part of the DL-CAD system was trained separately using MIP images with various slab thicknesses. Based on ten-fold cross-validation, the sensitivity and the F 2 score were determined to evaluate the performance of using each slab thickness at the nodule candidate detection stage. The free-response receiver operating characteristic (FROC) curve was used to assess the performance of the whole DL-CAD system that took the results combined from 16 MIP slab thickness settings., Results: At the nodule candidate detection stage, the combination of results from 16 MIP slab thickness settings showed a high sensitivity of 98.0% with 46 false positives (FPs) per scan. Regarding a single MIP slab thickness of 10 mm, the highest sensitivity of 90.0% with 8 FPs/scan was reached before false positive reduction. The sensitivity increased (82.8% to 90.0%) for slab thickness of 1 to 10 mm and decreased (88.7% to 76.6%) for slab thickness of 15-50 mm. The number of FPs was decreasing with increasing slab thickness, but was stable at 5 FPs/scan at a slab thickness of 30 mm or more. After false positive reduction, the DL-CAD system, utilizing 16 MIP slab thickness settings, had the sensitivity of 94.4% with 1 FP/scan., Conclusions: The utilization of multi-MIP images could improve the performance at the nodule candidate detection stage, even for the whole DL-CAD system. For a single slab thickness of 10 mm, the highest sensitivity for pulmonary nodule detection was reached at the nodule candidate detection stage, similar to the slab thickness usually applied by radiologists., (Copyright © 2020. Published by Elsevier B.V.)

Accurate pulmonary nodule detection is a crucial step in lung cancer screening. Computer-aided detection (CAD) systems are not routinely used by radiologists for pulmonary nodule detection in clinical practice despite their potential benefits. Maximum intensity projection (MIP) images improve the detection of pulmonary nodules in radiological evaluation with computed tomography (CT) scans. Inspired by the clinical methodology of radiologists, we aim to explore the feasibility of applying MIP images to improve the effectiveness of automatic lung nodule detection using convolutional neural networks (CNNs). We propose a CNN-based approach that takes MIP images of different slab thicknesses (5 mm, 10 mm, 15 mm) and 1 mm axial section slices as input. Such an approach augments the two-dimensional (2-D) CT slice images with more representative spatial information that helps discriminate nodules from vessels through their morphologies. Our proposed method achieves sensitivity of 92.7% with 1 false positive per scan and sensitivity of 94.2% with 2 false positives per scan for lung nodule detection on 888 scans in the LIDC-IDRI dataset. The use of thick MIP images helps the detection of small pulmonary nodules (3 mm-10 mm) and results in fewer false positives. Experimental results show that utilizing MIP images can increase the sensitivity and lower the number of false positives, which demonstrates the effectiveness and significance of the proposed MIP-based CNNs framework for automatic pulmonary nodule detection in CT scans. The proposed method also shows the potential that CNNs could gain benefits for nodule detection by combining the clinical procedure.

Background: There are limited data from randomized trials regarding whether volume-based, low-dose computed tomographic (CT) screening can reduce lung-cancer mortality among male former and current smokers., Methods: A total of 13,195 men (primary analysis) and 2594 women (subgroup analyses) between the ages of 50 and 74 were randomly assigned to undergo CT screening at T0 (baseline), year 1, year 3, and year 5.5 or no screening. We obtained data on cancer diagnosis and the date and cause of death through linkages with national registries in the Netherlands and Belgium, and a review committee confirmed lung cancer as the cause of death when possible. A minimum follow-up of 10 years until December 31, 2015, was completed for all participants., Results: Among men, the average adherence to CT screening was 90.0%. On average, 9.2% of the screened participants underwent at least one additional CT scan (initially indeterminate). The overall referral rate for suspicious nodules was 2.1%. At 10 years of follow-up, the incidence of lung cancer was 5.58 cases per 1000 person-years in the screening group and 4.91 cases per 1000 person-years in the control group; lung-cancer mortality was 2.50 deaths per 1000 person-years and 3.30 deaths per 1000 person-years, respectively. The cumulative rate ratio for death from lung cancer at 10 years was 0.76 (95% confidence interval [CI], 0.61 to 0.94; P = 0.01) in the screening group as compared with the control group, similar to the values at years 8 and 9. Among women, the rate ratio was 0.67 (95% CI, 0.38 to 1.14) at 10 years of follow-up, with values of 0.41 to 0.52 in years 7 through 9., Conclusions: In this trial involving high-risk persons, lung-cancer mortality was significantly lower among those who underwent volume CT screening than among those who underwent no screening. There were low rates of follow-up procedures for results suggestive of lung cancer. (Funded by the Netherlands Organization of Health Research and Development and others; NELSON Netherlands Trial Register number, NL580.)., (Copyright © 2020 Massachusetts Medical Society.)

Purpose: To improve outcomes for lung cancer through low-dose computed tomography (LDCT) early lung cancer detection. The International Association for the Study of Lung Cancer is developing the Early Lung Imaging Confederation (ELIC) to serve as an open-source, international, universally accessible environment to analyze large collections of quality-controlled LDCT images and associated biomedical data for research and routine screening care., Methods: ELIC is an international confederation that allows access to efficiently analyze large numbers of high-quality computed tomography (CT) images with associated de-identified clinical information without moving primary imaging/clinical or imaging data from its local or regional site of origin. Rather, ELIC uses a cloud-based infrastructure to distribute analysis tools to the local site of the stored imaging and clinical data, thereby allowing for research and quality studies to proceed in a vendor-neutral, collaborative environment. ELIC's hub-and-spoke architecture will be deployed to permit analysis of CT images and associated data in a secure environment, without any requirement to reveal the data itself (ie, privacy protecting). Identifiable data remain under local control, so the resulting environment complies with national regulations and mitigates against privacy or data disclosure risk., Results: The goal of pilot experiments is to connect image collections of LDCT scans that can be accurately analyzed in a fashion to support a global network using methodologies that can be readily scaled to accrued databases of sufficient size to develop and validate robust quantitative imaging tools., Conclusion: This initiative can rapidly accelerate improvements to the multidisciplinary management of early, curable lung cancer and other major thoracic diseases (eg, coronary artery disease and chronic obstructive pulmonary disease) visualized on a screening LDCT scan. The addition of a facile, quantitative CT scanner image quality conformance process is a unique step toward improving the reliability of clinical decision support with CT screening worldwide.

Lung cancer, chronic obstructive pulmonary disease (COPD), and coronary artery disease (CAD) are expected to cause most deaths by 2050. State-of-the-art computed tomography (CT) allows early detection of lung cancer and simultaneous evaluation of imaging biomarkers for the early stages of COPD, based on pulmonary density and bronchial wall thickness, and of CAD, based on the coronary artery calcium score (CACS), at low radiation dose. To determine cut-off values for positive tests for elevated risk and presence of disease is one of the major tasks before considering implementation of CT screening in a general population. The ImaLife (Imaging in Lifelines) study, embedded in the Lifelines study, is designed to establish the reference values of the imaging biomarkers for the big three diseases in a well-defined general population aged 45 years and older. In total, 12,000 participants will undergo CACS and chest acquisitions with latest CT technology. The estimated percentage of individuals with lung nodules needing further workup is around 1-2%. Given the around 10% prevalence of COPD and CAD in the general population, the expected number of COPD and CAD is around 1000 each. So far, nearly 4000 participants have been included. The ImaLife study will allow differentiation between normal aging of the pulmonary and cardiovascular system and early stages of the big three diseases based on low-dose CT imaging. This information can be finally integrated into personalized precision health strategies in the general population.

Background: The US guidelines recommend low-dose CT (LDCT) lung cancer screening for high-risk individuals. New solid nodules after baseline screening are common and have a high lung cancer probability. Currently, no evidence exists concerning the risk stratification of non-resolving new solid nodules at first LDCT screening after initial detection., Methods: In the Dutch-Belgian Randomized Lung Cancer Screening (NELSON) trial, 7295 participants underwent the second and 6922 participants the third screening round. We included participants with solid nodules that were registered as new or <15 mm³ (study detection limit) at previous screens and received additional screening after initial detection, thereby excluding high-risk nodules according to the NELSON management protocol (nodules ≥500 mm 3 )., Results: Overall, 680 participants with 1020 low-risk and intermediate-risk new solid nodules were included. A total of 562 (55%) new solid nodules were resolving, leaving 356 (52%) participants with a non-resolving new solid nodule, of whom 25 (7%) were diagnosed with lung cancer. At first screening after initial detection, volume doubling time (VDT), volume, and VDT combined with a predefined ≥200 mm 3 volume cut-off had high discrimination for lung cancer (VDT, area under the curve (AUC): 0.913; volume, AUC: 0.875; VDT and ≥200 mm 3 combination, AUC: 0.939). Classifying a new solid nodule with either ≤590 days VDT or ≥200 mm 3 volume positive provided 100% sensitivity, 84% specificity and 27% positive predictive value for lung cancer., Conclusions: More than half of new low-risk and intermediate-risk solid nodules in LDCT lung cancer screening resolve. At follow-up, growth assessment potentially combined with a volume limit can be used for risk stratification., Trial Registration Number: ISRCTN63545820; pre-results., Competing Interests: Competing interests: KtH reports grants and non-financial support from NELSON-Netherlands-Leuven Lung Cancer Screening, during the conduct of the study, and grants from the University of Zurich, non-financial support from the International Association for the Study of Lung Cancer Strategic Screening Advisory Committee, and grants from Sunnybrook Health Sciences, outside the submitted work. CMvdA reports grants from Symposium in Thoracic Oncology, grants from the American Thoracic Society, and grants from Lancet Respiratory Medicine, outside the submitted work. KN reports grants from Flemish League against Cancer and grants from the Belgian Foundation against Cancer, during the conduct of the study. HJMG reports grants from Eli Lilly, Roche, MSD, BMS and Novartis, outside the submitted work. HJdK took part in a 1-day advisory meeting on biomarkers organised by MD Anderson/Health Sciences during the 16th World Conference on Lung Cancer, outside the submitted work. All other authors declare no competing interests over the last 36 months., (© Author(s) (or their employer(s)) 2019. No commercial re-use. See rights and permissions. Published by BMJ.)

Background: New nodules are regularly found after the baseline round of low-dose computed tomography (LDCT) lung cancer screening. The relationship between a participant's number of new nodules and lung cancer probability is unknown., Methods: Participants of the ongoing Dutch-Belgian Randomized Lung Cancer Screening (NELSON) Trial with (sub)solid nodules detected after baseline and registered as new by the NELSON radiologists were included. The correlation between a participant's new nodule count and the largest new nodule size was assessed using Spearman's rank correlation. To evaluate the new nodule count as predictor for new nodule lung cancer together with largest new nodule size, a multivariable logistic regression analysis was performed., Results: In total, 705 participants with 964 new nodules were included. In 48% (336/705) of participants no nodule had been found previously during baseline screening and in 22% (154/705) of participants >1 new nodule was detected (range 1-12 new nodules). Eventually, 9% (65/705) of the participants had lung cancer in a new nodule. In 100% (65/65) of participants with new nodule lung cancer, the lung cancer was the largest or only new nodule at initial detection. The new nodule lung cancer probability did not differ significantly between participants with 1 (10% [56/551], 95%CI 8-13%) or >1 new nodule (6% [9/154], 95%CI 3-11%, P = .116). An increased number of new nodules positively correlated with a participant's largest nodule size (P < 0.001, Spearman's rho 0.177). When adjusted for largest new nodule size, the new nodule count had a significant negative association with lung cancer (odds ratio 0.59, 0.37-0.95, P = .03)., Conclusion: A participant's new nodule count alone only has limited association with lung cancer. However, a higher new nodule count correlates with an increased largest new nodule size, while the lung cancer probability remains equivalent, and may improve lung cancer risk prediction by size only., (Copyright © 2018 The Author(s). Published by Elsevier B.V. All rights reserved.)

Objective:: To evaluate the influence of nodule margin on inter- and intrareader variability in manual diameter measurements and semi-automatic volume measurements of solid nodules detected in low-dose CT lung cancer screening., Methods:: 25 nodules of each morphological category (smooth, lobulated, spiculated and irregular) were randomly selected from 93 participants of the Dutch-Belgian Randomized Lung Cancer Screening Trial (NELSON). Semi-automatic volume measurements were performed using Syngo LungCARE® software (Version Somaris/5 VB10A-W, Siemens, Forchheim, Germany). Three radiologists independently measured mean diameters manually. Impact of nodule margin on interreader variability was evaluated based on systematic error and 95% limits of agreement. Interreader variability was compared with the nodule growth cut-off as used in Lung CT Screening Reporting and Data System (LungRADS; +1.5-mm diameter) and the Dutch-Belgian Randomized Lung Cancer Screening Trial(acronym: NELSON) /British Thoracic Society (+25% volume)., Results:: For manual diameter measurements, a significant systematic error (up to 1.2 mm) between readers was found in all morphological categories. For semi-automatic volume measurements, no statistically significant systematic error was found. The interreader variability in mean diameter measurements exceeded the 1.5-mm cut-off for nodule growth for all morphological categories [smooth: ±1.9 mm (+27%), lobulated: ±2.0 mm (+33%), spiculated: ±3.5 mm (+133%), irregular: ±4.5 mm (+200%)]. The 25% vol growth cut-off was exceeded slightly for spiculated [28% (+12%)] and irregular [27% (+8%)] nodules., Conclusion:: Lung nodule sizing based on manual diameter measurement is affected by nodule margin. Interreader variability increases especially for nodules with spiculated and irregular margins, and causes substantial misclassification of nodule growth. This effect is almost neglectable for semi-automated volume measurements. Semi-automatic volume measurements are superior for both size and growth determination of pulmonary nodules., Advances in Knowledge:: 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.

Background: US guidelines now recommend lung cancer screening with low-dose CT for high-risk individuals. Reports of new nodules after baseline screening have been scarce and are inconsistent because of differences in definitions used. We aimed to identify the occurrence of new solid nodules and their probability of being lung cancer at incidence screening rounds in the Dutch-Belgian Randomized Lung Cancer Screening Trial (NELSON)., Methods: In the ongoing, multicentre, randomised controlled NELSON trial, between Dec 23, 2003, and July 6, 2006, 15 822 participants who had smoked at least 15 cigarettes a day for more than 25 years or ten cigarettes a day for more than 30 years and were current smokers, or had quit smoking less than 10 years ago, were enrolled and randomly assigned to receive either screening with low-dose CT (n=7915) or no screening (n=7907). From Jan 28, 2004, to Dec 18, 2006, 7557 individuals underwent baseline screening with low-dose CT; 7295 participants underwent second and third screening rounds. We included all participants with solid non-calcified nodules, registered by the NELSON radiologists as new or smaller than 15 mm(3) (study detection limit) at previous screens. Nodule volume was generated semiautomatically by software. We calculated the maximum volume doubling time for nodules with an estimated percentage volume change of 25% or more, representing the minimum growth rate for the time since the previous scan. Lung cancer diagnosis was based on histology, and benignity was based on histology or stable size for at least 2 years. The NELSON trial is registered at trialregister.nl, number ISRCTN63545820., Findings: We analysed data for participants with at least one solid non-calcified nodule at the second or third screening round. In the two incidence screening rounds, the NELSON radiologists registered 1222 new solid nodules in 787 (11%) participants. A new solid nodule was lung cancer in 49 (6%) participants with new solid nodules and, in total, 50 lung cancers were found, representing 4% of all new solid nodules. 34 (68%) lung cancers were diagnosed at stage I. Nodule volume had a high discriminatory power (area under the receiver operating curve 0·795 [95% CI 0·728-0·862]; p<0·0001). Nodules smaller than 27 mm(3) had a low probability of lung cancer (two [0·5%] of 417 nodules; lung cancer probability 0·5% [95% CI 0·0-1·9]), nodules with a volume of 27 mm(3) up to 206 mm(3) had an intermediate probability (17 [3·1%] of 542 nodules; lung cancer probability 3·1% [1·9-5·0]), and nodules of 206 mm(3) or greater had a high probability (29 [16·9%] of 172 nodules; lung cancer probability 16·9% [12·0-23·2]). A volume cutoff value of 27 mm(3) or greater had more than 95% sensitivity for lung cancer., Interpretation: Our study shows that new solid nodules are detected at each screening round in 5-7% of individuals who undergo screening for lung cancer with low-dose CT. These new nodules have a high probability of malignancy even at a small size. These findings should be considered in future screening guidelines, and new solid nodules should be followed up more aggressively than nodules detected at baseline screening., Funding: Zorgonderzoek Nederland Medische Wetenschappen and Koningin Wilhelmina Fonds Kankerbestrijding., (Copyright © 2016 Elsevier Ltd. All rights reserved.)