RPS 2104 - Pulmonary nodules and lung cancer screening

Author Block: Y-C. Chang, Y-S. Huang, H-H. Hsu, A. M-F. Yen, H. J. Chiou, W. P. Chan, P-C. Yang; Taipei/TW
Purpose: This study aimed to describe the preliminary experience in quality assurance of LDCT in national lung cancer screening in Taiwan.
Methods or Background: From 1st July 2022, the National Lung Cancer Screening (LCS) program in Taiwan was launched on a biennial basis. Enrollment criteria include heavy smokers (>30 py, 50-74 years old) and subjects with a family history of lung cancer (male 50-74 years and female 45-74 years old) according to international guidelines and evidence from TALENT (Taiwan Lung Cancer Screening for Never Smoker Trial) study. Modified Lung-RADS (Version 1.1) was used for categorising the LDCT interpretation results after adjusting the size criteria of non-solid nodules >20mm diameter as Category 3 (probably benign findings). This is the first national lung cancer screening to include non-smoker subjects with a family history in the world. Quality evaluation parameters included radiation exposure dose, interpretation results based on modified Lung-RADS, cancer detection rate, positive predication rate (PPV) in different groups, and proven lung cancer stages in 2 groups.
Results or Findings: Until 30th June 2023 (data estimated in June 2023), there were a total of 48,372 subjects (male 56.27%, female 43.73%) receiving LDCT LCS in 163 hospitals. 531 lung cancers (1.11%) were identified (data from the Health Promotion Administration, Ministry of Health and Welfare). The majority of lung cancers detected were in the early stage (stage 0 12.5%, stage I 72.58%). Quality assurance (QA) was performed according to the result of positive findings of LDCT (category 3, 4) (11%), positive evaluation after visiting chest specialists (9.23%), cancer detection rate, radiation exposure, etc.
Conclusion: The preliminary results of QA showed the importance of screening subjects with a family history of lung cancer in addition to heavy smokers in Taiwan.
Limitations: Data analysis is subjected to change because of different time for statistical estimation.
Funding for this study: Funding for this study was received from the Health Promotion Administration, Ministry of Health and Welfare, Taiwan.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information: No information provided by the submitter.

Author Block: D. Han¹, M. A. Heuvelmans¹, H. L. Lancaster¹, J. W. C. Gratama², M. Silva³, J. Field⁴, M. Oudkerk¹; ¹Groningen/NL, ²Apeldoorn/NL, ³Parma/IT, ⁴Liverpool/UK
Purpose: The primary objective of this study is to assess the false positive rate in lung cancer screening attributable to PFNs, at the participant level. Perifissural nodules (PFNs) have been definitively established as benign in lung cancer screening (LCS) trials. Their prevalence, accounting for 20-30% of all nodules, could significantly impact the false positive rate of lung cancer screening, potentially leading to an unnecessary number of follow-ups. This issue is particularly relevant when AI systems are employed as the primary readers, as they currently struggle with accurate classification of PFNs.
Methods or Background: We selected 1,253 baseline scans from the UK Lung Cancer Screening Trial based on the presence of pulmonary nodules exceeding 15 mm³ in volume. We employed the AI-based software AVIEW to automatically detect and volumetrically quantify solid pulmonary nodules. Subsequently, all AI-detected pulmonary nodules with a volume of ≥30 mm³ underwent visual classification by an experienced reader, distinguishing between PFNs and non-PFNs. Pulmonary nodules measuring <100 mm³ were considered negative, while those ≥100 mm³ were categorized as positive.
Results or Findings: At the nodule level, a total of 375 pulmonary nodules were classified as PFNs, comprising 296 (78.9%) measuring <100 mm³ and 79 (21.1%) measuring ≥100 mm³. At the participant level, out of 1253 participants, 316 (25.2%) were found to have PFNs. Among these, 250 (20.0%) participants had only negative PFNs, while 66 (5.2%) participants had positive PFNs. Notably, 33 (2.6%) participants with positive PFNs did not have concurrent pulmonary nodules measuring ≥100 mm³.
Conclusion: Using AI-based software as the primary reader results in a few false positive PFNs in lung cancer screening.
Limitations: The false positive rate attributable to PFNs may be influenced by the performance of AI.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information: This study was waived by the ethics committee due to the retrospective nature of this study.

Author Block: M. Vonder¹, X. Yang¹, H. Groen¹, M. Oudkerk¹, Z. Ye², M. Dorrius¹, G. De Bock¹, Z. Nie¹; ¹Groningen/NL, ²Tianjin/CN
Purpose: This study aimed to assess the prevalence of lung CT findings in a general Chinese population. In lung cancer CT screening, other lung findings like bronchiectasis, airway wall thickening and emphysema are associated with more exacerbations and hospitalisations, as well as increased mortality rate.
Methods or Background: This study included Nelcin-B3 participants aged 40-74 years in China who received low-dose CT lung cancer screening. Baseline characteristics of participants were described. Fleischner criteria were applied to assess bronchiectasis, airway wall thickening, and emphysema (at least mild). The prevalence and combined prevalence for lung findings were determined. Multivariable logistic regression analysis was performed to examine factors associated with the prevalence of these lung CT findings.
Results or Findings: In total, 978 participants (mean age 61.3 years ±6.8; 54.6% women) were included. Bronchiectasis was identified in 9.2% of participants, 35.7% showed airway wall thickening, and 19.9% had emphysema. 2.1% of participants showed all three CT findings. 50% of participants with emphysema were more likely to be current smokers. Multivariable logistic regression showed age (OR=1.04; CI: 1.01-1.07), smoking (OR=3.03; CI: 1.87- 4.93), bronchiectasis (OR=1.68; CI: 1.00-2.83) and airway wall thickening (OR=2.06; CI: 1.46-2.92) were positively associated with the presence of emphysema.
Conclusion: In the general Chinese population, at least 48% have one lung CT finding for lung cancer screening. Only 2% have all three lung CT findings. Smoking is the strongest predictor for the presence of emphysema. The relevance of these CT findings should be considered in future lung cancer screening guidelines.
Limitations: Firstly, no detailed clinical information could be provided. Secondly, CT diagnosis of mild bronchiectasis and airway wall thickening remains challenging which potentially leads to false-positive diagnoses. Lastly, it only analysed patient data collected at one singular medical centre, which limits generalisability to the whole Chinese population.
Funding for this study: Funding for this study was received from The Royal Netherlands Academy of Arts and Sciences and the Ministry of Science and Technology of the People’s Republic of China.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This study was approved by the Ethics of Biomedicine Research of Second Military Medical University.

Author Block: R. Dinnessen¹, K. V. Venkadesh¹, D. Peeters¹, H. A. Piggelen-Gietema², E. Scholten¹, C. M. Schaefer-Prokop³, C. Jacobs¹; ¹Nijmegen/NL, ²Maastricht/NL, ³Amersfoort/NL
Purpose: An AI algorithm for malignancy risk estimation was developed and validated on screen-detected pulmonary nodules. We aimed to test the AI algorithm in clinical data and compare the results to the Brock model.
Methods or Background: A size-matched dataset of solid incidentally detected pulmonary nodules with a diameter range between 5-15 mm was collected, consisting of 53 malignant nodules from CT scans performed at least two months prior to a lung cancer diagnosis, and 53 benign nodules. Differences in patient and nodule characteristics between the malignant and benign groups were assessed. AUCs and 95% confidence intervals were determined and compared using the DeLong method. Sensitivity and specificity were determined at a 10% malignancy risk threshold for the AI algorithm and Brock model, according to the British Thoracic Society guidelines.
Results or Findings: No statistical difference in size was detected between the malignant and benign nodules (median [range]: 10.8 [5.8, 15.4]; 10.4 [5.8, 15.1]; respectively). Cases with malignant nodules had a significantly lower number of nodules (p=0.001). The AI algorithm significantly outperformed the Brock model (p<0.001). AUC [95% CI] of the AI algorithm and Brock model were 0.87 [0.80-0.94] and 0.59 [0.48-0.69], respectively. The AI algorithm had a higher sensitivity (0.60 [0.46-0.74]) and specificity (0.87 [0.75-0.95]) than the Brock model (0.42 [0.28-0.56]; 0.75 [0.62-0.86]; respectively).
Conclusion: The AI algorithm outperformed the Brock model in a clinical dataset with a more heterogeneous population than a screening population. The AI algorithm demonstrated the potential for nodule risk stratification in a clinical setting, which can aid clinicians in decisions in nodule management, thereby potentially reducing unnecessary follow-up.
Limitations: This is a retrospective validation on a single-centre dataset. More research is needed to test the performance in larger and multi-centre data.
Funding for this study: Funding was provided by the Dutch Cancer Society (KWF Kankerbestrijding, project number 14113).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This study included data collected retrospectively from one university medical centre. The local IRB board waived the need for informed consent because of the retrospective design and the use of anonymized data in this study.

Author Block: Y. Sun, L. Jin, C. Li, M. Li; Shanghai/CN
Purpose: The diagnosis of pulmonary ground glass nodules(GGNs) remains a challenge in clinical practice. The growth rate of nodules is heterogeneous, although generally slower than that of other diseases. The purpose of this study is to attempt to predict the long-term stability or growth of GGNs using deep learning based on baseline CT imaging.
Methods or Background: In this retrospective study, 575 GGNs from 456 patients were recruited. Five hundred and seventy-five GGNs were randomized into training (70%) and validation sets (30%). A deep learning-based algorithm was developed and validated using baseline CT imaging and clinical features. The deep learning prediction network model was compared with the traditional radiographic features. Also, the first follow-up imaging was also added as input to improve the performance of the deep learning model.
Results or Findings: The growth and stable groups contained 233 and 342 GGNs, respectively. Traditional radiographic features have limited value in the prediction of growth or long-term stability of GGNs(AUC=0.70 ± 0.06). Size, density, and age were independent predictors of GGN growth. Comparing with traditional radiographic model, our deep learning model yielded a significant higher AUC value of 0.80 ± 0.05 (P < 0.01). The addition of first follow up CT images improved the model performance (AUC=0.84 ± 0.06).
Conclusion: We developed and validated a deep learning model to predict the natural growth pattern of GGNs basing on baseline and first follow up CT imaging. The model achieved good performance and may provide a basis for the improvement of follow-up management of GGNs.
Limitations: First, in this retrospective study, the variety of CT scan protocols may have affected the characteristics of GGNs and deep features. Second, the conclusions require further validation using external large-scale datasets.
Funding for this study: Funding was received from these agencies: National Natural Science Foundation of China 61976238(Ming Li), Science and Technology Planning Project of Shanghai Science and Technology Commission 20Y11902900 (Ming Li), Shanghai “Rising Stars of Medical Talent” Youth Development Program “Outstanding Youth Medical Talents” SHWJRS [2021]-99 (Ming Li), National key research and development program 2022YFF1203301(Ming Li), Cancer Society of Shanghai SACACY21C12 (Yingli Sun), Emerging Talent Program of Huadong Hospital Grant numbers XXRC2213(Ming Li), Leading Talent Program of Huadong Hospital Grant numbers LJRC2202(Ming Li) , and Excellent Academic Leaders of Shanghai 2022XD042(Ming Li).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This study was approved ethically with the approval code: 2019K134.

Author Block: R. Ramaesh¹, M. P. Engbersen², M. Javidi¹, J. Rodrigues³, E. Van Beek¹, M. Bernabeu¹; ¹Edinburgh/UK, ²Amsterdam/NL, ³Bath/UK
Purpose: Incidental lung nodules at CT provide an opportunity for the timely detection of early-stage lung cancer. Computer-aided detection (CAD) supported by deep learning aims to assist radiologists in the detection and further assessment of nodules. This study investigates the effect of CAD on agreements in management recommendations between the reporting radiologist and an expert peer reviewer.
Methods or Background: In this multicentre implementation study, participating radiologists from four centres reported on chest CTs in adult patients without a history of malignancy or nodules, during routine practice with the CAD as ‘second reader’. The reporters documented management recommendations concerning lung nodules sequentially, without CAD and with CAD. All cases with nodules found either by CAD or the reporter were independently reviewed by a thoracic radiologist with at least eight years experience. Recommendations of the reporter, with and without CAD, were compared to the recommendations of the reviewer. Agreement between recommendations was assessed with quadratic weighted kappa, and the difference was tested with a William’s test.
Results or Findings: Nineteen percent (237/1264) of CTs had lung nodules found and confirmed by the reviewer. In 11.4% (27/237) of nodule cases, the reporting radiologist had changed their recommendation after CAD. The most frequent reason given for this change was an initially missed nodule (8.4%; 20/237). The weighted kappa between the reporters’ and the reviewers’ recommendations was 0.49 and 0.55, unaided and aided by CAD (p = 0.04).
Conclusion: Aided by CAD, reporting radiologists found more incidental lung nodules and provided management recommendations which are more in agreement with an expert reviewer, suggesting a significant contribution of CAD to the clinical management of incidental nodules.
Limitations: Limitations include a lack of reference standard by longitudinal data or a consensus of multiple reviewers.
Funding for this study: This study was funded by the NHS AI in Health and Care Award (grant number: 2119-C25043).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was approved by the EM-REC (Edinburgh Medical School Research Ethics Committee). Reference number: 22-EMREC-015 dd. 14-04-2022.

Author Block: A. A. Peters¹, N. Wiescholek¹, J. Klaus¹, F. Strodka¹, A. Macek¹, E. C. Primetis², D. D. Drakopoulos², A. Christe¹, L. Ebner¹; ¹Bern/CH, ²Muri bei Bern/CH
Purpose: The aim of this study was to evaluate and compare the performance of two DL-CAD systems regarding detection, localisation and classification of pulmonary nodules.
Methods or Background: The main study cohort contained 122 proven T1 tumors of the lung and was extended by 83 cases (subsolid, n=13; solid<6mm, n=40; controls, n=30), resulting in a primary cohort of n=205. Two different DL-CAD systems analyzed all cases. Five independent blinded readers with different experience levels (residents, n=3; seniors, n=2) performed two readout sessions, first stand-alone and then with access to the results of one of the DL-CAD systems. Two readers used software 1 and the three readers used software 2 and scored nodule size, density and localisation. LungRADS categories were calculated and compared.
Results or Findings: After application of the eligibility criteria, the final cohort consisted of 198 subjects with 221 pulmonary nodules. Residents' mean detection rate increased from 64% to 77% (p<0.001) using the respective DL-CAD (table 2), while the seniors’ detection rates did not improve (p=0.25). Regarding the correct localization of the nodules, the residents' rates for lobar (73% vs. 77%; p<0.001) and segmental (64% vs. 68%; p<0.001) nodule localisation improved significantly, the seniors showed no significant benefit. Regarding software comparison, software 2 lead to a slightly higher increase in detection rates (software 1, 80% to 86% and software 2, 67% to 77%; both p<0.001). Both systems showed no significant effect on the rate of correct LungRADS classification.
Conclusion: Less experienced readers have more benefits from using DL-CAD systems regarding detection and localisation of pulmonary nodules. There is no effect on correct LungRADS classification. Both systems performed comparably, software 2 lead to a higher increase in detection rates.
Limitations: Selection bias (high cancer prevalence). Nodule size groups categorized instead of exact measures.
Funding for this study: No funding was obtained for the current study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This study was approved by the Kantonale Ethikkommision (KEK) Bern.

Author Block: H. S. Ryu, H. N. Lee, J. I. Kim, J. K. Ryu, Y. J. Lim; Seoul/KR
Purpose: Patients with breast cancer have a higher risk of developing lung cancer than the general population. The study aimed to evaluate the incidence of GGN and risk factors for GGN growth in patients with breast cancer and to evaluate the incidence and pathologic features of lung cancer.
Methods or Background: We retrospectively reviewed the clinical data and chest CTs of 1384 patients diagnosed with breast cancer who underwent chest CT between January 2008 and December 2022. We evaluated the incidence and size change of GGNs during follow-up and identified independent risk factors for their growth using multivariate analysis. Furthermore, the incidence and pathologic features of lung cancer were also evaluated.
Results or Findings: We detected persistent GGNs in 69 of 1384 (5.0%) patients. The initial diameter of GGNs was 6.3 ± 3.6 mm on average, with primarily (85.5%) pure GGNs. Among them, 27 (39.1%) exhibited interval growth with a median volume doubling time of 1006.0 days (interquartile range, 622.0–1528.0 days) during the median 959 days (interquartile range, 612.0–1645.0 days) follow-up period. Older age (P = 0.026), part-solid nodules (P = 0.006), and total number of GGNs (≥2) (P = 0.007) were significant factors for GGN growth. Lung cancer was confirmed in 13 of 1384 patients (0.9%), all with adenocarcinoma, including one case of minimally invasive adenocarcinoma. The cancers demonstrated a high rate of epidermal growth factor receptor mutation (69.2%).
Conclusion: Persistent GGNs in breast cancer patients with high-risk factors should be monitored for early detection and treatment of lung cancer.
Limitations: This retrospective study was conducted at a single centre with a small sample size, manual measurement of GGNs was subject to errors, and chest CT was not dedicated to automated volume measurement.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was approved by institutional review board of Kyung Hee University Hospital at Gangdong (2023-02-016) and informed consent was waived owing to the retrospective nature of the study.

Author Block: Q. C. Meng, P. Gao; Zhengzhou/CN
Purpose: Pure ground-glass nodule (pGGN) invisibility drives clinical intervention. Radiomics and radiogenomics aid pGGN diagnosis but lack of standardised acquisition parameters, reproducibility and inconsistent methods. We aim to evaluate a risk score system for diagnosing invasive adenocarcinoma presenting as pGGN .
Methods or Background: Seven hundred and seventy-two pGGNs from 707 individuals were grouped into training (509 patients/558 observations) and validation (198 patients/214 observations) sets. A test set with 143 observations was also analysed. The quantitative parameters were obtained using AI. The positive pGGN cutoff score was ≥3. Risk score systems3 were calculated as the history of carcinoma*1+chronic obstructive pulmonary disease (COPD)*1 + long diameters*1 + volume of nodule*1 + mean CT values*1 + type II vascular supply sign*1 or type III*2 + other variables of radiographic characteristics*1. The risk score system and AI model were evaluated using areas under the receiver operating characteristics curve (AUCs), accuracy, sensitivity, specificity, and positive predictive values.
Results or Findings: Risk score system 3 (AUC, 0.840) performed better than the AI model (AUC, 0.553), risk score system 1 (AUC, 0.802, and risk score system 2 (AUC, 0.816), with 88.0% (0.850–0.904) accuracy, 95.6% (0.932–0.972) PPV, 89.6% (0.864–0.920) sensitivity, and 80.6% (0.717–0.872) specificity in the training sets. Risk score system 3 yielded the best performance in the validation and test set, with AUCs of 0.769 and 0.801.
Conclusion: The risk scoring system based on AI-based quantitative image parameters, clinical features, and radiographic characteristics can effectively predict the invasive adenocarcinoma of pulmonary pGGNs.
Limitations: First, the study was retrospective, with some selection bias. Second, the improved AI algorithm may enhance the diagnosis of the invasiveness of pGGNs in clinical practice, and iterative upgrading of the algorithm is needed in the future.
Funding for this study: Funding of this study was received from the Key project of Medical science and Technology of Henan Province in China (No: SBGJ202102057).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This multi-centre, retrospective cohort study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the The Affiliated Cancer Hospital of Zhengzhou University & Henan Cancer Hospital Medical and the The Affiliated Cancer Hospital of Zhengzhou University & Henan Cancer Hospital Ethics Committee (2021-KY-0022).