• Cancer Early Detection and Prevention Initiative, Vanderbilt Ingram Cancer Center, Division of Allergy, Pulmonary and Critical Care Medicine.
• Oxford University Hospitals NHS Foundation Trust, Oxford, United Kingdom.
• Optellum Ltd., Oxford, United Kingdom.
• Faculty of Medicine, Masaryk University, Brno, Czech Republic.
• Department of Biostatistics, and.
• Department of Electrical Engineering, Vanderbilt University, Nashville, Tennessee; and.
• Department of Radiology, Wake Forest Baptist Health, Winston Salem, North Carolina.
• Department of Radiology, Vanderbilt University School of Medicine, Nashville, Tennessee.

RATIONALE: The management indeterminate pulmonary nodules (IPNs) remains challenging, resulting in invasive procedures and delays in diagnosis and treatment. Strategies to decrease the rate of unnecessary invasive procedures and to optimize surveillance regimens are needed.

OBJECTIVES: Develop and validate a deep learning method to improve the management of IPNs.

METHODS: A Lung Cancer Prediction Convolutional Neural Network (LCP-CNN) model was trained using CT images of IPNs from the National Lung Screening Trial (NLST), internally validated, and externally tested on cohorts from two academic institutions.

MEASUREMENTS AND MAIN RESULTS: The area under the receiver operating characteristic curve in the external validation cohorts were 83.5% (95%CI:75.4-90.7%) and 91.9% (95%CI:88.7-94.7%) compared with 78.1% (95%CI:68.7-86.4%) and 81.9 (95%CI:76.1-87.1%) respectively for a commonly used clinical risk model for incidental nodules. Using 5% and 65% malignancy thresholds defining low and high-risk categories, the overall net reclassification in the validation cohorts for cancers and benign nodules compared to the Mayo model was 0.34 (Vanderbilt) and 0.30 (Oxford) as a rule-in test, and 0.33 (Vanderbilt) and 0.58 (Oxford) as a rule-out test. The LCP-CNN compared to traditional risk prediction models was associated with an improved accuracy in the predicted likelihood of disease at each threshold of management and in our external validation cohorts.

CONCLUSIONS: This study demonstrates that this deep learning algorithm can correctly reclassify IPNs into low or high-risk categories in over a third of cancers and benign nodules when compared to conventional risk models, potentially reducing the number of unnecessary invasive procedures and delays in diagnosis. This article is open access and distributed under the terms of the Creative Commons Attribution Non-Commercial No Derivatives License 4.0 (http://creativecommons.org/licenses/by-nc-nd/4.0/).