University of Groningen Pulmonary Nodules: 2D versus 3D evaluation in lung cancer screening Han, Daiwei

(1)

Pulmonary Nodules: 2D versus 3D evaluation in lung cancer screening

Han, Daiwei

DOI:

10.33612/diss.172563513

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.

Document Version

Publisher's PDF, also known as Version of record

Publication date: 2021

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA):

Han, D. (2021). Pulmonary Nodules: 2D versus 3D evaluation in lung cancer screening. University of Groningen. https://doi.org/10.33612/diss.172563513

Copyright

Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

Take-down policy

If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.

(2)

559233-l-bw-Han 559233-l-bw-Han 559233-l-bw-Han 559233-l-bw-Han Processed on: 20-5-2021 Processed on: 20-5-2021 Processed on: 20-5-2021

Processed on: 20-5-2021 PDF page: 135PDF page: 135PDF page: 135PDF page: 135

General discussion

(3)

9

138

Computed tomography (CT) lung cancer screening has been shown to reduce lung cancer related mortality by multiple large randomized controlled trials worldwide. The clinical relevance of the use of semi-automatic volumetry in lung cancer screening and the benignity of perifissural nodules have been discussed in previous chapters. This chapter will provide a more general discussion of the main findings in this thesis and methodological considerations will be discussed. Furthermore, potential clinical implications and future research directions will be addressed.

MAIN FINDINGS

Volume versus diameter-based nodule assessment

Interest in lung cancer screening has been increasing since the publication of the National Lung Screening Trial (NLST) results for lung cancer screening (1). The NLST has reported 20% lung cancer mortality reduction using low-dose CT compared to screening by chest radiography. Based on the results of NLST translated into recommendation by the U.S. Preventive Services Task Force (USPSTF) and is covered by Medicare for high-risk population (2,3). Following the positive results from European lung cancer screening trials, experts in lung cancer screening and medical societies have put forth joint statements recommending LDCT screening in Europe (4,5).

A shortcoming of CT screening encountered by the NLST trial is the large number of small to intermediate sized nodules and high false positivity rate. European studies used semi-automatic volumetry for the assessment of pulmonary nodules. Instead of a positive or negative nodule management strategy used in the NLST, the NELSON trial included an additional intermediate nodule size group that required a short follow-up scan and used volumetry for lung cancer risk stratification. The NELSON study has reported far lower false positive rates (1.2% versus 24%), but comparable sensitivity (92.5% versus 93.5%) when compared to the NLST trials (6,7).

An accurate and precise measurement method is crucial for the management of a lung nodule, as it allows reliable lung cancer risk prediction. The measurement needs to accurately represent the true size of an object. A nodule has infinite diameters but a single volume. In clinical practice lung nodules are rarely symmetrical. Therefore, on theoretical grounds semiautomatic volumetry is more favorable for nodule size and growth assessment than diameter measurements.

It can be hypothesized that diameter measurements perform especially poorly for nodules with non-smooth margin, while these nodules have the highest uncertainty

(4)

9

139

for malignancy. However, the comparison between diameter measurements and semiautomatic volumetry is intricate. Not only because the two measurements are in two different dimensions, but also each has its own nodule management algorithms. The comparison is further complicated by the absolute and relative reference values used in the two nodule management algorithms. Therefore, indirect comparison between volume and diameter measurements was used throughout this thesis. Inter- and intra-reader variability for diameter measurements and semiautomatic volumetry were compared in measuring lung nodules with smooth and non-smooth margins. The growth cutoffs from Lung CT Screening Reporting & Data Systems (Lung-RADS v. 1.0) (±1.5 mm in mean diameter) and British Thoracic Society guidelines for pulmonary nodules (25% in volume) were used for this comparison (8,9). We have found semiautomatic volumetry to be more reliable for non-smooth margin nodules than diameter measurements. In a phantom study, Petrick et al. reported 3D volumetry achieved smaller measurement variability compared to manual maximum diameter measurement, especially for complex nodule shapes (10). The relative standard deviation of maximum diameter measurement for spiculated and elliptical nodules were larger when compared to spherical nodules (20.3% and 16.4% versus 5.7%, respectively). For volumetric measurement, the standard deviation did not exceed 10% for any nodule shapes. However, readers need to note that a 10% error in diameter measurement does not equal to a 10% error in volumetry. Using an 8 mm diameter (268 mm3_{) spherical}

nodule as an example, a 1.6 mm underestimation (-20% in relative measurement error in diameter) result in almost 50% decrease in nodule volume (137 mm3_{), whereas the}

volume of a 20% underestimation in volumetry is 214 mm3_{. Furthermore, this study has}

not put its results into the context of lung nodule management guidelines, which limits its clinical relevance. In a recent study, Gierada et al. evaluated measurement variability of manual mean diameter measurement, semiautomatic mean diameter measurement and semiautomatic volumetry using 120 lung nodules from CT screening exams (11). Both intraclass correlation and linear-weighted k for Lung-RADS classification showed that CT volumetry was the best of the three measurement methods across three reader pairs. What is also important to note is that the 95% limit of agreement for manual mean diameter measurement method across three reader pairs ranged from ±2.05 mm to ±3.15 mm, 37-110% greater than 1.5 mm growth cutoff. This again shows the poor reliability of manual diameter measurement in detecting nodule growth. In Lung-RADS v.1.1, solid nodules <8 mm in diameter that are growing are considered suspicious and require a 3-month follow-up LDCT. A measurement variability greater than ±1.5 mm would mean many unnecessary follow-up LDCT will be made, leading to unnecessary anxiety of the patient, and increased exposure to harmful radiation.

(5)

9

140

Influence of denoising techniques on nodule detection and volumetry in low and ultralow-dose CT (ULDCT)

Results from multiple randomized controlled trials on CT lung cancer screening have provided undeniable evidence that lung cancer mortality can be significantly reduced (6,7,12,13). With possible massive increase in CT screening, radiation safety has become a concern. The ALARA principle (an acronym for “as low as reasonably achievable”) is a guiding principle for radiation safety to reduce risks caused by radiation exposure. We evaluated a third-generation dual source CT system in combination with Advanced Model Iterative Reconstruction (ADMIRE) for the detection and quantification of pulmonary nodules, in COPD patients. We have found that at 83% reduced radiation dose compared to standard dose CT, ULDCT offers high nodule detectability [95.0 % sensitivity (95%CI: 83.1-99.4)] and without a significant influence in nodule measurement or size classification.

Several studies have investigated the influence major radiation dose reduction using CT on nodule detection, with the use of iterative reconstruction (14–17). In a phantom study that used comparable scanning and reconstruction protocol to ours, Gordic et al. reported similar nodule detectability [94% sensitivity (95%CI:74-99)] compared to standard dose CT on the same CT system (14). In another phantom study that evaluated the effects of ULDCT and ADMIRE on semiautomatic volumetry of pulmonary nodules 4-10 mm in size, Eberhard et al. reported no significant difference between standard dose CT and 1/8th_{and 1/20}th_{of standard dose CT when ADMIRE strength level 3 was}

applied, similar compared to the results from our study (18). However, significant difference was found when ADMIRE strength level 5 was used.

In a in vivo study that evaluated the nodule detection at ULDCT using same third generation dual source CT, Messerli et al. reported 91.2 % sensitivity (95%CI: 88.2-93.8) on ULDCT at similar mean effective dose compared to our study. However, in their study, majority of the included patients received contrast-enhanced CT, which limits its generalizability especially in the lung cancer screening setting where contrast enhanced CT is not used. Studies evaluating nodule detection on ULDCT using other iterative techniques have reported sensitivity ranging 90% to 98%. One study reported the disappearance of several nodules using adaptive statistical reconstruction-V (19). In our study, nodules missed by a radiologist in ULDCT were all retrospectively found. This emphasizes the importance in the evaluation of IR techniques before their use in clinical practice.

(New) perifissural nodules, their classification and lung cancer probability

Perifissural nodules (PFN) are a subcategory of solid nodules, with 20-30% incidence in the baseline of lung cancer screening studies (20–22). They are further grouped into two

(6)

9

141

groups, with defined characteristics: “Typical PFNs were defined as fissure-attached, homogeneous, solid nodules with smooth margins and lentiform triangular shape. Atypical PFNs were nodules that either met all features but were not attached to a visible fissure or were fissure-attached nodules that were convex on one side and round on the other side.” Although PFNs may present with rapid growth (volume doubling time of <400 days), studies in lung cancer screening and clinical population have shown PFNs to be benign.

A recent study from the NELSON trial has shown that new solid nodules, detected in the incidence screening rounds, have higher lung cancer probability than nodules from the baseline screening round (23). However, there is a lack of evidence on whether new PFNs from lung cancer screening can be safely excluded from further follow-up. In chapter 7, we evaluated new solid nodules from the NELSON study with respect to the incidence, lung cancer probability, and classification of PFNs. We found that 4% of incident solid nodules were PFNs, which was noticeably lower compared to the baseline incidence of PFNs (20-28%) (20,21,24). Furthermore, none of the malignant nodules were classified as PFNs. Therefore, PFNs detected in incidental screening rounds can be safely ruled out. As pointed out by Morgan et al. (25), fissure attached non-PFN nodules were found to have 25% lung cancer probability and needs to be treated as an indeterminate nodule with short term follow-up, and to be evaluated using semi-automatic volumetry for growth.

A limitation of this study is the limited number of malignant nodules included in this dataset. One can assume that with increased proportion of malignant nodules in a dataset the rate of misclassification increases. In a recent study Schreuder et al. explored the reliability of radiologists to differentiate malignant nodules from PFNs. In a dataset of 316 nodules (5-10 mm in size) of which 70 (22.2%) were malignant nodules. The authors reported that 0.9% of the nodules (5 od 533) and 4.8% (16 of 332) that were classified as typical PFN and atypical PFN were lung cancers, respectively. This study shows that although lung cancer can be misclassified as PFNs, it is not very likely. To put this into perspective, the lung cancer risk of nodules 5 mm to 10 mm in diameter is 1-2% (26,27). Assuming a 2% lung cancer risk for these nodules, only 0.09% of typical PFNs and 0.5% of atypical PFNs would have been lung cancer, comparable to the lung cancer risk of benign solid nodules <6 mm or 100 mm3_{in size.}

Although the benign nature of PFNs has been generally accepted, disagreement on the definition of PFNs exists amongst radiologists. Although the definition for PFNs provided by de Hoop et al. has been used as the standard for the classification of PFNs, variations of the definition for PFNs still exists in the literature. We concluded that agreement on

(7)

9

142

an unambiguous definition for the classification of PFNs is conditional for future studies on PFN.

Automatic classification of PFNs using machine learning algorithms has previously been done with limited success. To our knowledge, only one study has attempted to classify PFNs using machine learning algorithms. In their study Ciompi et al. used a pre-existing neural network, pre-trained on natural images, for the classification of PFNs (28). Both typical and atypical PFNs were grouped together into a single PFN category. In their Receiver Operating Characteristics (ROC) analysis, the authors reported a respectable area under the curve (AUC) of 0.868. However, sensitivity, specificity and predictive values were not provided, likely due to not drawing a hard cutoff on the ROC curve. Therefore, its applicability in the clinical practice could not be accurately assessed. in collaboration with Optellum Ltd. we attempted to train a novel deep-learning based neural network for the classification of PFNs (PFN-CNN). Contrary to the neural network used by Ciompi et al, which was pre-trained on natural images. The neural network in our study was trained using domain specific images. Based on previously reported findings (29), decision was made to only focus on typical PFNs, due to their less ambiguous definition and better inter-reader agreement by expert readers when compared to atypical PFNs. Since none of the nodules in our study were biopsied, two radiologists experienced with thoracic imaging and a radiology resident were set as our reference standard. Our model achieved an AUC of 95.8 (95% CI 93.3-98.4), 95.6% (95% CI 84.9-99.5) sensitivity, and 88.1% (95% CI 81.8-92.8) specificity. Furthermore, the agreement between the PFN-CNN and trained readers was within the range of inter-reader agreement. This study shows the potential of applying deep-learning based algorithms for the rule out of PFNs, assisting radiologists in their decision making, and improving reader efficiency.

CLINICAL IMPLICATIONS AND FUTURE PERSPECTIVES

Adoption of volumetry in LDCT lung cancer screening

Although semiautomatic volumetry is the go-to nodule assessment method in Europe. Worldwide, diameter measurement is still the most used nodule assessment method. In lung cancer screening low-risk and intermediate risk nodules, with either 1 year or 3 months follow-up screens (4,8,9). The risk stratification at follow-up screens should be based on nodule growth (4,30,31). We have found manual diameter measurement to be unreliable, especially for nodules with non-smooth margins. With the recently updated Lung Imaging Reporting and Data Systems (Lung-RADS ver. 1.1) screening

(8)

9

143

guidelines (32), cutoff values for volumetry has been added, allowing future transition from diameter measurements to semiautomatic volumetry. In this version of Lung-RADS, while diameter retains its 1.5 mm growth cutoff, the cutoff for volumetry has been set at 2 mm3_{, contrary to the 25% in volume or volume doubling time of 400 days}

from previous literature. Furthermore, the nodule volume standards the newest version of Lung-RADS is a direct conversion from the given diameter cutoffs. Therefore, further optimizations for volumetric nodule management in Lung-RADS is needed.

A reliable semiautomatic volumetry software package is crucial to volume-based lung cancer screening. In the NELSON study, the semiautomatic volumetry software package was kept as a constant, this was done to avoid possible measurement differences introduced by software updates. It is known that variation exists in nodule volumetry between different semiautomatic volumetry software packages (33,34). This difference may influence the overall recall rate of a lung cancer screening (33). Therefore, it is imperative to benchmark semiautomatic volumetry packages before implementing them into lung cancer screening.

Ultralow-dose lung cancer screening

Compared to standard dose CT, ULDCT at 83% reduced radiation dose offers high nodule detectability and without significant difference in nodule measurement. However, in our study subsolid nodules were not evaluated. Although subsolid nodules are considerably less prevalent than solid nodules in the western population (35,36), they are known to have a significantly higher risk of malignancy compared to solid nodules and are particularly prevalent in Asian populations (35). Since subsolid nodule have a lower density than solid nodules, the noise tolerability for the detection of subsolid nodules may be lower than solid nodules (37). Further studies are required to optimize CT scanning and reconstruction protocol for the detection of subsolid nodules.

ULDCT may not only be useful in reducing radiation exposure from the serial imaging in lung cancer screening. Shared risk factors of the Big-3 diseases (lung cancer, COPD and coronary artery disease) and visualization of lungs and heart in a single thorax CT examination, inspired the idea of combining the detection of Big-3 diseases within one chest CT acquisition, potentially reducing radiation dose by many folds. Further studies can be focused on the optimization of ULDCT for each of the Big-3 diseases, and eventually integrating them into a single acquisition protocol.

Towards automated lung cancer screening programs

New PFNs detected in incidental screening rounds are benign nodules. This has allowed the generalizability of the benign nature of PFNs in screening, clinical setting and in

(9)

9

144

incidental screening rounds (20,21,24). We have found in the literature various definitions of PFNs and concluded that an unambiguous definition of PFNs should be a prerequisite for future PFN research. Nevertheless, discussion is ongoing whether juxtapleural and peripleural nodules can also be safely excluded from further evaluation (38,39). Before an AI program such as PFN-CNN classifier can be implemented in the clinical practice, its ability to differentiate PFNs from malignant nodules needs to be carefully evaluated. Furthermore, it needs to be evaluated with an external dataset.

The advancement in AI in the past decade has allowed great progress in automated segmentation, detection, and classification of pulmonary nodules. Before the implementation of these algorithms in routine clinical care, their performance should be evaluated using external data. Moreover, the implementation of AI tools in routine clinical care also faces many ethical challenges, such as lack of transparency in their decisions, discriminatory outcomes, and issues of responsibility. Further research is needed in the field of bioethics for using AI in health care.

CONCLUSION

To conclude, we have gathered evidence to show that semiautomatic volumetry should be used for the assessment of pulmonary nodules in lung cancer screening, instead of manual diameter measurements. At 83% radiation dose reduction the nodule detectability in ULDCT remained comparable to HRCT, without significant trade-off in nodule measurement and size classification. While new fissure-attached PFNs detected in incidental screening rounds are benign, and can be safely ruled out from further follow-ups, new fissure-attached non-PFNs have high lung cancer probability and require short term follow-up. Last but not least, the performance of deep learning-based PFN classifier, in the classification of typical PFNs, is similar when compared to trained readers.

(10)

9

145

BIBLIOGRAPHY

1. Reduced Lung-Cancer Mortality with Low-Dose Computed Tomographic Screening. N Engl J Med. 2011 Aug 4;365(5):395–409.

2. Recommendation: Lung Cancer: Screening | United States Preventive Services Taskforce [Internet]. [cited 2020 Nov 20]. Available from: https://www. uspreventiveservicestaskforce.org/uspstf/recommendation/lung-cancer-screening 3. Lung Cancer Screening Coverage [Internet]. [cited 2020 Dec 28]. Available from:

https://www.medicare.gov/coverage/lung-cancer-screenings

4. Oudkerk M, Devaraj A, Vliegenthart R, Henzler T, Prosch H, Heussel CP, et al. European position statement on lung cancer screening. Lancet Oncol. 2017 Dec 1;18(12):e754–66.

5. Kauczor H-U, Baird A-M, Blum TG, Bonomo L, Bostantzoglou C, Burghuber O, et al. ESR/ERS statement paper on lung cancer screening. Eur Respir J [Internet]. 2020 Feb 1 [cited 2020 Nov 15];55(2). Available from: https://erj.ersjournals.com/ content/55/2/1900506

6. Reduced Lung-Cancer Mortality with Low-Dose Computed Tomographic Screening | NEJM [Internet]. [cited 2019 May 27]. Available from: https://www.nejm.org/doi/ full/10.1056/NEJMoa1102873

7. de Koning HJ, van der Aalst CM, de Jong PA, Scholten ET, Nackaerts K, Heuvelmans MA, et al. Reduced Lung-Cancer Mortality with Volume CT Screening in a Randomized Trial. N Engl J Med. 2020 Feb 6;382(6):503–13.

8. Callister MEJ, Baldwin DR, Akram AR, Barnard S, Cane P, Draffan J, et al. British Thoracic Society guidelines for the investigation and management of pulmonary nodules: accredited by NICE. Thorax. 2015 Aug 1;70(Suppl 2):ii1-ii54.

9. Lung Rads [Internet]. [cited 2020 Nov 30]. Available from: https://www.acr.org/ Clinical-Resources/Reporting-and-Data-Systems/Lung-Rads

10. Petrick N, Kim HJG, Clunie D, Borradaile K, Ford R, Zeng R, et al. Comparison of 1D, 2D, and 3D Nodule Sizing Methods by Radiologists for Spherical and Complex Nodules on Thoracic CT Phantom Images. Acad Radiol. 2014 Jan 1;21(1):30–40. 11. Gierada DS, Rydzak CE, Zei M, Rhea L. Improved Interobserver Agreement on

Lung-RADS Classification of Solid Nodules Using Semiautomated CT Volumetry. Radiology [Internet]. 2020 Sep 15 [cited 2020 Dec 17]; Available from: https://pubs.rsna.org/ doi/abs/10.1148/radiol.2020200302

12. Pastorino U, Silva M, Sestini S, Sabia F, Boeri M, Cantarutti A, et al. Prolonged lung cancer screening reduced 10-year mortality in the MILD trial: new confirmation of lung cancer screening efficacy. Ann Oncol. 2019 Jul 1;30(7):1162–9.

13. Lung cancer mortality reduction by LDCT screening—Results from the randomized German LUSI trial - Becker - 2020 - International Journal of Cancer - Wiley Online

(11)

9

146

Library [Internet]. [cited 2020 Dec 28]. Available from: https://onlinelibrary.wiley. com/doi/abs/10.1002/ijc.32486

14. Gordic S, Morsbach F, Schmidt B, Allmendinger T, Flohr T, Husarik D, et al. Ultralow-dose chest computed tomography for pulmonary nodule detection: first performance evaluation of single energy scanning with spectral shaping. Invest Radiol. 2014 Jul;49(7):465–73.

15. Martini K, Higashigaito K, Barth BK, Baumueller S, Alkadhi H, Frauenfelder T. Ultralow-dose CT with tin filtration for detection of solid and sub solid pulmonary nodules: a phantom study. Br J Radiol. 2015 Oct 22;88(1056):20150389.

16. Messerli M, Kluckert T, Knitel M, Wälti S, Desbiolles L, Rengier F, et al. Ultralow dose CT for pulmonary nodule detection with chest x-ray equivalent dose – a prospective intra-individual comparative study. Eur Radiol. 2017 Aug 1;27(8):3290–9.

17. Sui X, Meinel FG, Song W, Xu X, Wang Z, Wang Y, et al. Detection and size measurements of pulmonary nodules in ultra-low-dose CT with iterative reconstruction compared to low dose CT. Eur J Radiol. 2016 Mar;85(3):564–70.

18. Eberhard M, Stocker D, Milanese G, Martini K, Nguyen-Kim TDL, Wurnig MC, et al. Volumetric assessment of solid pulmonary nodules on ultralow-dose CT: a phantom study. J Thorac Dis [Internet]. 2019 Aug [cited 2021 Jan 17];11(8). Available from: https://jtd.amegroups.com/article/view/30854

19. Ye K, Zhu Q, Li M, Lu Y, Yuan H. A feasibility study of pulmonary nodule detection by ultralow-dose CT with adaptive statistical iterative reconstruction-V technique. Eur J Radiol. 2019 Oct;119:108652.

20. Ahn MI, Gleeson TG, Chan IH, McWilliams AM, MacDonald SL, Lam S, et al. Perifissural Nodules Seen at CT Screening for Lung Cancer. Radiology. 2010 Feb 8;254(3):949–56.

21. de Hoop B, van Ginneken B, Gietema H, Prokop M. Pulmonary Perifissural Nodules on CT Scans: Rapid Growth Is Not a Predictor of Malignancy. Radiology. 2012 Nov 1;265(2):611–6.

22. Mets OM, Chung K, Scholten ET, Veldhuis WB, Prokop M, van Ginneken B, et al. Incidental perifissural nodules on routine chest computed tomography: lung cancer or not? Eur Radiol. 2018 Mar;28(3):1095–101.

23. Walter JE, Heuvelmans MA, Jong PA de, Vliegenthart R, Ooijen PMA van, Peters RB, et al. Occurrence and lung cancer probability of new solid nodules at incidence screening with low-dose CT: analysis of data from the randomised, controlled NELSON trial. Lancet Oncol. 2016 Jul 1;17(7):907–16.

24. Mets OM, Chung K, Scholten ET, Veldhuis WB, Prokop M, van Ginneken B, et al. Incidental perifissural nodules on routine chest computed tomography: lung cancer or not? Eur Radiol. 2018;28(3):1095–101.

(12)

9

147

cancer screening: more practical implications from the NELSON study? Transl Lung Cancer Res [Internet]. 2020 Jul 7 [cited 2020 Jul 30];0(0). Available from: http://tlcr. amegroups.com/article/view/41430

26. Lung Rads [Internet]. [cited 2021 Jan 17]. Available from: https://www.acr.org/ Clinical-Resources/Reporting-and-Data-Systems/Lung-Rads

27. Horeweg N, Rosmalen J van, Heuvelmans MA, Aalst CM van der, Vliegenthart R, Scholten ET, et al. Lung cancer probability in patients with CT-detected pulmonary nodules: a prespecified analysis of data from the NELSON trial of low-dose CT screening. Lancet Oncol. 2014 Nov 1;15(12):1332–41.

28. Ciompi F, de Hoop B, van Riel SJ, Chung K, Scholten ET, Oudkerk M, et al. Automatic classification of pulmonary peri-fissural nodules in computed tomography using an ensemble of 2D views and a convolutional neural network out-of-the-box. Med Image Anal. 2015 Dec;26(1):195–202.

29. Schreuder A, van Ginneken B, Scholten ET, Jacobs C, Prokop M, Sverzellati N, et al. Classification of CT Pulmonary Opacities as Perifissural Nodules: Reader Variability. Radiology. 2018 Jul 3;288(3):867–75.

30. MacMahon H, Naidich DP, Goo JM, Lee KS, Leung ANC, Mayo JR, et al. Guidelines for Management of Incidental Pulmonary Nodules Detected on CT Images: From the Fleischner Society 2017. Radiology. 2017 Feb 23;284(1):228–43.

31. Callister MEJ, Baldwin DR, Akram AR, Barnard S, Cane P, Draffan J, et al. British Thoracic Society guidelines for the investigation and management of pulmonary nodules: accredited by NICE. Thorax. 2015 Aug 1;70(Suppl 2):ii1-ii54.

32. Dyer SC, Bartholmai BJ, Koo CW. Implications of the updated Lung CT Screening Reporting and Data System (Lung-RADS version 1.1) for lung cancer screening. J Thorac Dis. 2020 Nov;12(11):6966–77.

33. Soo E, Edey AJ, Mak SM, Moser J, Mohammadi S, Rodrigues T, et al. Impact of choice of volumetry software and nodule management guidelines on recall rates in lung cancer screening. Eur J Radiol. 2019 Nov 1;120:108646.

34. Comparison of three software systems for semi-automatic volumetry of pulmonary nodules on baseline and follow-up CT examinations - Ying Ru Zhao, Peter MA van Ooijen, Monique D Dorrius, Marjolein Heuvelmans, Geertruida H de Bock, Rozemarijn Vliegenthart, Matthijs Oudkerk, 2014 [Internet]. [cited 2020 Dec 28]. Available from: https://journals-sagepub-com.proxy-ub.rug.nl/doi/10.1177/028 4185113508177?url_ver=Z39.88-2003&rfr_id=ori:rid:crossref.org&rfr_dat=cr_ pub%20%200pubmed

35. Kakinuma R, Muramatsu Y, Kusumoto M, Tsuchida T, Tsuta K, Maeshima AM, et al. Solitary Pure Ground-Glass Nodules 5 mm or Smaller: Frequency of Growth. Radiology. 2015 Apr 23;276(3):873–82.

(13)

9

148

nodules in CT lung cancer screening. Expert Rev Respir Med. 2018 Jan 2;12(1):1–3. 37. Martini K, Higashigaito K, Barth BK, Baumueller S, Alkadhi H, Frauenfelder T. Ultralow-dose CT with tin filtration for detection of solid and sub solid pulmonary nodules: a phantom study. Br J Radiol. 2015;88(1056):20150389.

(14)

9

(15)