Daiwei Han1, Marjolein Heuvelmans2,3, Mieneke Rook1,4, Monique Dorrius1, Luutsen van Houten1, Noah Waterfield Price5, Lyndsey C Pickup5, Petr Novotny5, Matthijs Oudkerk6,7, Jerome Declerck5, Fergus Gleeson8, Peter van Ooijen9, Rozemarijn Vliegenthart1. 1. University Medical Center Groningen, Department of Radiology, University of Groningen, Groningen, The Netherlands. 2. University Medical Center Groningen, Department of Epidemiology, University of Groningen, Groningen, The Netherlands. m.a.heuvelmans@umcg.nl. 3. Department of Pulmonology, Medisch Spectrum Twente, Enschede, The Netherlands. m.a.heuvelmans@umcg.nl. 4. Department of Radiology, Martini Ziekenhuis, Groningen, The Netherlands. 5. Optellum Ltd, Oxford, UK. 6. Faculty of Medical Sciences, University of Groningen, Groningen, the Netherlands. 7. Institute for Diagnostic Accuracy, Groningen, The Netherlands. 8. National Consortium of Intelligent Medical Imaging, Oxford University, Oxford, Great Britain, UK. 9. University Medical Center Groningen, Department of Radiotherapy, University of Groningen, Groningen, The Netherlands.
Abstract
OBJECTIVES: To evaluate the performance of a novel convolutional neural network (CNN) for the classification of typical perifissural nodules (PFN). METHODS: Chest CT data from two centers in the UK and The Netherlands (1668 unique nodules, 1260 individuals) were collected. Pulmonary nodules were classified into subtypes, including "typical PFNs" on-site, and were reviewed by a central clinician. The dataset was divided into a training/cross-validation set of 1557 nodules (1103 individuals) and a test set of 196 nodules (158 individuals). For the test set, three radiologically trained readers classified the nodules into three nodule categories: typical PFN, atypical PFN, and non-PFN. The consensus of the three readers was used as reference to evaluate the performance of the PFN-CNN. Typical PFNs were considered as positive results, and atypical PFNs and non-PFNs were grouped as negative results. PFN-CNN performance was evaluated using the ROC curve, confusion matrix, and Cohen's kappa. RESULTS: Internal validation yielded a mean AUC of 91.9% (95% CI 90.6-92.9) with 78.7% sensitivity and 90.4% specificity. For the test set, the reader consensus rated 45/196 (23%) of nodules as typical PFN. The classifier-reader agreement (k = 0.62-0.75) was similar to the inter-reader agreement (k = 0.64-0.79). Area under the ROC curve was 95.8% (95% CI 93.3-98.4), with a sensitivity of 95.6% (95% CI 84.9-99.5), and specificity of 88.1% (95% CI 81.8-92.8). CONCLUSION: The PFN-CNN showed excellent performance in classifying typical PFNs. Its agreement with radiologically trained readers is within the range of inter-reader agreement. Thus, the CNN-based system has potential in clinical and screening settings to rule out perifissural nodules and increase reader efficiency. KEY POINTS: • Agreement between the PFN-CNN and radiologically trained readers is within the range of inter-reader agreement. • The CNN model for the classification of typical PFNs achieved an AUC of 95.8% (95% CI 93.3-98.4) with 95.6% (95% CI 84.9-99.5) sensitivity and 88.1% (95% CI 81.8-92.8) specificity compared to the consensus of three readers.
OBJECTIVES: To evaluate the performance of a novel convolutional neural network (CNN) for the classification of typical perifissural nodules (PFN). METHODS: Chest CT data from two centers in the UK and The Netherlands (1668 unique nodules, 1260 individuals) were collected. Pulmonary nodules were classified into subtypes, including "typical PFNs" on-site, and were reviewed by a central clinician. The dataset was divided into a training/cross-validation set of 1557 nodules (1103 individuals) and a test set of 196 nodules (158 individuals). For the test set, three radiologically trained readers classified the nodules into three nodule categories: typical PFN, atypical PFN, and non-PFN. The consensus of the three readers was used as reference to evaluate the performance of the PFN-CNN. Typical PFNs were considered as positive results, and atypical PFNs and non-PFNs were grouped as negative results. PFN-CNN performance was evaluated using the ROC curve, confusion matrix, and Cohen's kappa. RESULTS: Internal validation yielded a mean AUC of 91.9% (95% CI 90.6-92.9) with 78.7% sensitivity and 90.4% specificity. For the test set, the reader consensus rated 45/196 (23%) of nodules as typical PFN. The classifier-reader agreement (k = 0.62-0.75) was similar to the inter-reader agreement (k = 0.64-0.79). Area under the ROC curve was 95.8% (95% CI 93.3-98.4), with a sensitivity of 95.6% (95% CI 84.9-99.5), and specificity of 88.1% (95% CI 81.8-92.8). CONCLUSION: The PFN-CNN showed excellent performance in classifying typical PFNs. Its agreement with radiologically trained readers is within the range of inter-reader agreement. Thus, the CNN-based system has potential in clinical and screening settings to rule out perifissural nodules and increase reader efficiency. KEY POINTS: • Agreement between the PFN-CNN and radiologically trained readers is within the range of inter-reader agreement. • The CNN model for the classification of typical PFNs achieved an AUC of 95.8% (95% CI 93.3-98.4) with 95.6% (95% CI 84.9-99.5) sensitivity and 88.1% (95% CI 81.8-92.8) specificity compared to the consensus of three readers.
Entities:
Keywords:
Deep learning; Solitary pulmonary nodule; Tomography, X-ray computed
Authors: Myeong I Ahn; Tadhg G Gleeson; Ida H Chan; Annette M McWilliams; Sharyn L Macdonald; Stephen Lam; Sukhinder Atkar-Khattra; John R Mayo Journal: Radiology Date: 2010-03 Impact factor: 11.105
Authors: M E J Callister; D R Baldwin; A R Akram; S Barnard; P Cane; J Draffan; K Franks; F Gleeson; R Graham; P Malhotra; M Prokop; K Rodger; M Subesinghe; D Waller; I Woolhouse Journal: Thorax Date: 2015-08 Impact factor: 9.139
Authors: Daiwei Han; Marjolein A Heuvelmans; Carlijn M van der Aalst; Lisa H van Smoorenburg; Monique D Dorrius; Mieneke Rook; Kristiaan Nackaerts; Joan E Walter; Harry J M Groen; Rozemarijn Vliegenthart; Harry J de Koning; Matthijs Oudkerk Journal: J Thorac Oncol Date: 2019-10-10 Impact factor: 15.609
Authors: Harry J de Koning; Carlijn M van der Aalst; Pim A de Jong; Ernst T Scholten; Kristiaan Nackaerts; Marjolein A Heuvelmans; Jan-Willem J Lammers; Carla Weenink; Uraujh Yousaf-Khan; Nanda Horeweg; Susan van 't Westeinde; Mathias Prokop; Willem P Mali; Firdaus A A Mohamed Hoesein; Peter M A van Ooijen; Joachim G J V Aerts; Michael A den Bakker; Erik Thunnissen; Johny Verschakelen; Rozemarijn Vliegenthart; Joan E Walter; Kevin Ten Haaf; Harry J M Groen; Matthijs Oudkerk Journal: N Engl J Med Date: 2020-01-29 Impact factor: 91.245
Authors: Marjolein A Heuvelmans; Joan E Walter; Robin B Peters; Geertruida H de Bock; Uraujh Yousaf-Khan; Carlijn M van der Aalst; Harry J M Groen; Kristiaan Nackaerts; Peter Ma van Ooijen; Harry J de Koning; Matthijs Oudkerk; Rozemarijn Vliegenthart Journal: Lung Cancer Date: 2017-09-01 Impact factor: 5.705
Authors: David R Baldwin; Jennifer Gustafson; Lyndsey Pickup; Carlos Arteta; Petr Novotny; Jerome Declerck; Timor Kadir; Catarina Figueiras; Albert Sterba; Alan Exell; Vaclav Potesil; Paul Holland; Hazel Spence; Alison Clubley; Emma O'Dowd; Matthew Clark; Victoria Ashford-Turner; Matthew Ej Callister; Fergus V Gleeson Journal: Thorax Date: 2020-03-05 Impact factor: 9.139
Authors: U Pastorino; M Silva; S Sestini; F Sabia; M Boeri; A Cantarutti; N Sverzellati; G Sozzi; G Corrao; A Marchianò Journal: Ann Oncol Date: 2019-07-01 Impact factor: 32.976
Authors: Onno M Mets; Kaman Chung; Ernst Th Scholten; Wouter B Veldhuis; M Prokop; Bram van Ginneken; Cornelia M Schaefer-Prokop; Pim A de Jong Journal: Eur Radiol Date: 2017-10-06 Impact factor: 5.315