7 mei 2023: zie ook dit artikel: https://kanker-actueel.nl/ademtest-via-electronische-neus-ruikt-niet-kleincellige-longkanker-blijkt-uit-nieuwe-studie-van-sharina-kort-aan-het-medisch-spectrum-twente-mst.html

7 mei 2023: Bron: The Lancet

Uit een studie met kunstmatige intelligentie (AI-algoritmes) blijkt dat het checken van grote longknobbeltjes (15 tot 30 mm.) op kwaadaardigheid via kunstmatige intelligentie in combinatie met de veel gebruikte Herder methode veel patiënten eerder behandeld zouden kunnen worden. Uit de studie bij longknobbeltjes van 15 tot 30 mm en een Herder score van 10 tot 70 procent zou maar liefst 82 procent in aanmerking komen voor een behandeling van beginnende longkanker. 

Het totale aandeel goedaardige versus kwaadaardige knobbeltjes was respectievelijk 37,5 versus 62,5%. De dataset omvatte een mix van vaste (70,6%), subvaste (22,8%) en geslepen glasopaciteiten (6,6%), met verhoudingen die goed in evenwicht waren tussen de trainings- en testsets.

Kunstmatige intelligentie zou veel kunnen toevoegen aan wetenschappelijk onderzoek, diagnostiek en behandeling van kanker aldus Jonas Teuwen die onderzoekt hoe algoritmen op allerlei manieren kunnen bijdragen aan kankeronderzoek. Jonas Teuwen leidt sinds 2020 zijn eigen onderzoeksgroep in het Nederlands Kanker Instituut, het onderzoeksinstituut van het Antoni van Leeuwenhoek. 

Uit de LIBRA studie kwamen de volgende resultaten:

Tussen juli 2020 en april 2022 werden 502 patiënten uit vijf Britse ziekenhuizen gerekruteerd voor de groep met grote longknobbeltjes van de retrospectieve LIBRA-studie.
838 CT-scans werden gebruikt voor modelontwikkeling, opgesplitst in trainings- en testsets (respectievelijk 70% en 30%). Een nnUNet-model werd getraind om de segmentatie van longknobbels te automatiseren. Er is een radiomics-handtekening ontwikkeld om knobbeltjes te classificeren op basis van maligniteitsrisico. De prestaties van het radiomics-model, de large-nodule radiomics predictive vector (LN-RPV) genoemd, werden vergeleken met die van drie radiologen en de Brock- en Herder-scores.

Resultaten:

499 patiënten hadden technisch evalueerbare scans (gemiddelde leeftijd 69 ± 11, 257 mannen, 242 vrouwen). In de testset van 252 scans behaalde de nnUNet een DICE-score van 0,86 en de LN-RPV een AUC van 0,83 (95% BI 0,77-0,88) voor maligniteitsclassificatie.
De prestaties waren beter dan die van de mediane radioloog (AUC 0,75 [95% BI 0,70-0,81], DeLong p = 0,03). LN-RPV was robuust voor automatische segmentatie (ICC 0.94).
Voor vaste knobbeltjes bij aanvang in de testset (117 patiënten) had LN-RPV een AUC van 0,87 (95% BI 0,80-0,93) vergeleken met 0,67 (95% BI 0,55-0,76, DeLong p = 0,002) voor de Brock-score en 0,83 (95% BI 0,75–0,90, DeLong p = 0,4) voor de Herder-score.
In de internationale externe testset (n = 151) handhaafde LN-RPV een AUC van 0,75 (95% BI 0,63-0,85). 18 van de 22 (82%) kwaadaardige knobbeltjes in de Herder 10-70% categorie in de testset werden door de beslissingsondersteunende tool als hoog risico geïdentificeerd en zijn mogelijk doorverwezen voor eerdere interventie.

Conclusie: 

De LIBRA-studie heeft een nationale pijplijn opgeleverd voor multi-center longnodule AI-onderzoek, dat is gebruikt om een classificatie-algoritme voor grote nodules te ontwikkelen voor de diagnose van longkanker. Ons model blijkt beter te presteren dan klinisch radiologen en de Brock-score, en vergelijkbaar met de Herder-score. De gemodelleerde beslissingsondersteunende scenario's suggereren dat dit zou kunnen leiden tot eerdere interventie voor kwaadaardige knobbeltjes in de 10-70% Herder-categorie, wat mogelijk levens zou kunnen redden door vroege interventie in de toekomst.

Het volledige studierapport is gratis in te zien of te downloaden. Klik op de titel van het studierapport dat in The Lancet is gepubliceerd:

A radiomics-based decision support tool improves lung cancer diagnosis in combination with the Herder score in large lung nodules

Open AccessPublished:November 09, 2022DOI:https://doi.org/10.1016/j.ebiom.2022.104344


Summary

Background

Large lung nodules (≥15 mm) have the highest risk of malignancy, and may exhibit important differences in phenotypic or clinical characteristics to their smaller counterparts. Existing risk models do not stratify large nodules well. We aimed to develop and validate an integrated segmentation and classification pipeline, incorporating deep-learning and traditional radiomics, to classify large lung nodules according to cancer risk.

Methods

502 patients from five U.K. centres were recruited to the large-nodule arm of the retrospective LIBRA study between July 2020 and April 2022. 838 CT scans were used for model development, split into training and test sets (70% and 30% respectively). An nnUNet model was trained to automate lung nodule segmentation. A radiomics signature was developed to classify nodules according to malignancy risk. Performance of the radiomics model, termed the large-nodule radiomics predictive vector (LN-RPV), was compared to three radiologists and the Brock and Herder scores.

Findings

499 patients had technically evaluable scans (mean age 69 ± 11, 257 men, 242 women). In the test set of 252 scans, the nnUNet achieved a DICE score of 0.86, and the LN-RPV achieved an AUC of 0.83 (95% CI 0.77–0.88) for malignancy classification. Performance was higher than the median radiologist (AUC 0.75 [95% CI 0.70–0.81], DeLong p = 0.03). LN-RPV was robust to auto-segmentation (ICC 0.94). For baseline solid nodules in the test set (117 patients), LN-RPV had an AUC of 0.87 (95% CI 0.80–0.93) compared to 0.67 (95% CI 0.55–0.76, DeLong p = 0.002) for the Brock score and 0.83 (95% CI 0.75–0.90, DeLong p = 0.4) for the Herder score. In the international external test set (n = 151), LN-RPV maintained an AUC of 0.75 (95% CI 0.63–0.85). 18 out of 22 (82%) malignant nodules in the Herder 10–70% category in the test set were identified as high risk by the decision-support tool, and may have been referred for earlier intervention.

Interpretation

The model accurately segments and classifies large lung nodules, and may improve upon existing clinical models.

Funding

This project represents independent research funded by: 1) Royal Marsden Partners Cancer Alliance, 2) the Royal Marsden Cancer Charity, 3) the National Institute for Health Research (NIHR) Biomedical Research Centre at the Royal Marsden NHS Foundation Trust and The Institute of Cancer Research, London, 4) the National Institute for Health Research (NIHR) Biomedical Research Centre at Imperial College London, 5) Cancer Research UK (C309/A31316).

Evidence before this study

The current guidelines for investigating lung nodules rely on clinical risk models, such as the Brock and Herder scores, and most nodules above 15 mm will trigger the 10% threshold for investigation. Many large nodules fall into the 10–70% Herder category, wherein the British Thoracic Society Guidelines suggest a broad range of options, from surveillance to surgery, and methods to improve stratification are needed. In the many years since the Herder model was developed, few studies have investigated how it could integrate with non-invasive radiomics models to improve early cancer diagnosis rates, and no existing studies have looked at large (15–30 mm) nodules only.

Added value of this study

This study developed a radiomics-based cancer prediction model in 15–30 mm lung nodules, which are not stratified well by existing guidelines. The developed model, termed the large-nodule radiomics predictive vector, achieved higher cancer prediction accuracy than the Brock score, and by integrating with the Herder model, would have led to early intervention in 82% of the malignant nodules with Herder scores of 10–70%. Because the model requires fewer variables than the Brock and Herder scores, it could potentially streamline the risk-classification process for clinicians in the future, particularly where PET scanning is not available or will be delayed. The use of a highly-accurate deep learning segmentation pipeline means that the model is not dependent on human nodule segmentation.

Implications of all the available evidence

The large nodule radiomics model improves upon or extends existing clinical models, and integrates with the British Thoracic Society guidelines to provide net-benefit in terms of early cancer intervention. Although prospective evaluation is needed, this tool may aid clinician decision making with regards to large lung nodules in the future.

In summary, the LIBRA study has provided a national pipeline for multi-centre lung nodule AI research, which has been used to develop a large nodule classification algorithm for lung cancer diagnosis. Our model appears to perform better than clinical radiologists and the Brock score, and comparably to the Herder score. The modelled decision-support scenarios suggest it could lead to earlier-intervention for malignant nodules in the 10–70% Herder category, which could potentially save lives through early intervention in the future.

Contributors

B.H.: Study design, data collection, data analysis, manuscript preparation.
M.C., E.A., A.L., B.R.: Radiology reads.
P.R., D.T., N.S., N.N., A.B., S.B., S.K.: Data collection.
K.L-R.: Radiomics pipeline supervision.
L.B., S.J.: Study management and operational oversight.
S.D.: Study design, XNAT pipeline management and data anonymisation.
A.N., S.G., S.K.: Study design, trial-management group.
C.B.: Study statistician.
E.A., A.D., R.W.L.: Study design and oversight, student supervision, manuscript preparation.
The LIBRA study PIs were: S.B., S.V.K, N.N., A.D. and R.W.L.
All authors read and approved the final version of the manuscript. The underlying data were accessed and verified by BH, LB and RL.

Data sharing statement

The anonymised spreadsheets of radiomics features and clinical outcomes used to generate the LN-RPV model are deposited into the Mendeley database under the accession code https://doi.org/10.17632/rz72hs5dvg.1. The R scripts for model development are provided in notebook format at: https://github.com/dr-benjamin-hunter/LIBRA_Large_Nodules. Access to the source images/data will be considered on request to Dr. Richard Lee.

Declaration of interests

Dr Navani is supported by a Medical Research Council Clinical Academic Research Partnership (MR/T02481X/1). This work was partly undertaken at the University College London Hospitals/University College London that received a proportion of funding from the Department of Health’s National Institute for Health Research (NIHR) Biomedical Research Centre’s funding scheme (NN). Dr Navani reports honoraria for educational talks or advisory boards from Amgen, Astra Zeneca, Boehringer Ingelheim, Bristol Myers Squibb, Guardant Health, Janssen, Lilly, Merck Sharp & Dohme, Olympus, OncLive, PeerVoice, Pfizer, and Takeda.
Dr Nair receives research grants from the Department of Health’s NIHR Biomedical Research Centre and GRAIL. He has received consulting fees from Aidence BV, Faculty Science Limited and MSD. He has received a travel bursary from Takeda. He participates on advisory boards for Aidence BV and Faculty Science Limited. He has leadership roles within the British Society of Thoracic Imaging, the British Lung Foundation and the NHS England Targeted Lung Health Checks Programme.
Dr Lee is funded by the Royal Marsden NIHR BRC, and Royal Marsden Cancer charity. RL's institution receives compensation for time spent in a secondment role for the lung health check program and as a National Specialty Lead for the National Institute of Health and Care Research. He has received research funding from CRUK, Innovate UK (co-funded by GE Healthcare, Roche and Optellum), SBRI (co-applicant in grants with QURE.AI), RM Partners Cancer Alliance and NIHR (co-applicant in grants with Optellum). He has received honoraria from CRUK.
Professor Devaraj is employed by and has stocks in Brainomix. He receives consulting fees from Roche and Boehringer Ingelheim.
The other authors report no potential conflict of interest.

Acknowledgements

This project represents independent research funded by: 1) Royal Marsden Partners Cancer Alliance, 2) the Royal Marsden Cancer Charity, 3) the National Institute for Health Research (NIHR) Biomedical Research Centre at the Royal Marsden NHS Foundation Trust and The Institute of Cancer Research, London, 4) the National Institute for Health Research (NIHR) Biomedical Research Centre at Imperial College London, 5) Cancer Research UK. (C309/A31316).

Appendix A. Supplementary data

References

  1. 1.
    • Gould M.K. 
    • Tang T. 
    • Liu I.L.A. 
    • et al.
    Recent trends in the identification of incidental pulmonary nodules.
    Am J Respir Crit Care Med. 2015; 192: 1208-1214
    View in Article
  2. 2.
    • Larici A.R. 
    • Farchione A. 
    • Franchi P. 
    • et al.
    Lung nodules: size still matters.
    Eur Respir Rev. 2017; 26https://doi.org/10.1183/16000617.0025-2017
    View in Article
  3. 3.
    • Callister M.E.J. 
    • Baldwin D.R. 
    • Akram A.R. 
    • et al.
    British thoracic society guidelines for the investigation and management of pulmonary nodules.
    Thorax. 2015; 70 (ii1–54)
    View in Article
  4. 4.
    • Lam S. 
    • Bryant H. 
    • Donahoe L. 
    • et al.
    Management of screen-detected lung nodules: a Canadian partnership against cancer guidance document.
    Can J Respir Crit Care, Sleep Med. 2020; 4: 236-265
    View in Article
  5. 5.
    • Gould M.K. 
    • Donington J. 
    • Lynch W.R. 
    • et al.
    Evaluation of individuals with pulmonary nodules: when is it lung cancer? Diagnosis and management of lung cancer, 3rd ed: American college of chest physicians evidence-based clinical practice guidelines.
    Chest. 2013; 143: e93S
    View in Article
  6. 6.
    • Horeweg N. 
    • van Rosmalen J. 
    • Heuvelmans M.A. 
    • 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; 15: 1332-1341
    View in Article
  7. 7.
    • Lung Rads
    • American College of Radiology
    https://www.acr.org/Clinical-Resources/Reporting-and-Data-Systems/Lung-Rads
    Date: 2020
    Date accessed: October 6, 2020
    View in Article
  8. 8.
    • Zhang R. 
    • Tian P. 
    • Chen B. 
    • Zhou Y. 
    • Li W.
    Predicting lung cancer risk of incidental solid and subsolid pulmonary nodules in different sizes.
    Cancer Manag Res. 2020; 12: 8057-8066
    View in Article
  9. 9.
    • Mendoza D.P. 
    • Petranovic M. 
    • Som A. 
    • et al.
    Lung-RADS category 3 and 4 nodules on lung cancer screening in clinical practice.
    AJR Am J Roentgenol. 2022; : 1-11
    View in Article
  10. 10.
    • Pinsky P.F. 
    • Gierada D.S. 
    • Black W. 
    • et al.
    Performance of lung-RADS in the national lung screening trial: a retrospective assessment.
    Ann Intern Med. 2015; 162: 485-491
    View in Article
  11. 11.
    • McWilliams A. 
    • Tammemagi M.C. 
    • Mayo J.R. 
    • et al.
    Probability of cancer in pulmonary nodules detected on first screening CT.
    N Engl J Med. 2013; 369: 910-919
    View in Article
  12. 12.
    • Herder G.J. 
    • Van Tinteren H. 
    • Golding R.P. 
    • et al.
    Clinical prediction model to characterize pulmonary nodules: validation and added value of 18F-fluorodeoxyglucose positron emission tomography.
    Chest. 2005; 128: 2490-2496
    View in Article
  13. 13.
    • Al-Ameri A. 
    • Malhotra P. 
    • Thygesen H. 
    • et al.
    Risk of malignancy in pulmonary nodules: a validation study of four prediction models.
    Lung Cancer. 2015; 89: 27-30
    View in Article
  14. 14.
    • Evison M. 
    • Taylor S. 
    • Grundy S. 
    • Perkins A. 
    • Peake M.
    Promoting early diagnosis and recovering from the COVID-19 pandemic in lung cancer through public awareness campaigns: learning from patient and public insight work.
    BMJ Open Respir Res. 2021; 8e001120
    View in Article
  15. 15.
    • Gallach M. 
    • Mikhail Lette M. 
    • Abdel-Wahab M. 
    • Giammarile F. 
    • Pellet O. 
    • Paez D.
    Addressing global inequities in positron emission tomography-computed tomography (PET-CT) for cancer management: a statistical model to guide strategic planning.
    Med Sci Monit. 2020; 26: e926544
    View in Article
  16. 16.
    • Kang G. 
    • Liu K. 
    • Hou B. 
    • Zhang N.
    3D multi-view convolutional neural networks for lung nodule classification.
    PLoS One. 2017; 12e0188290
    View in Article
  17. 17.
    • Lyu J. 
    • Ling S.H.
    Using multi-level convolutional neural network for classification of lung nodules on CT images.
    in: Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society. EMBS. Institute of Electrical and Electronics Engineers Inc., 2018: 686-689
    View in Article
  18. 18.
    • Shaffie A. 
    • Soliman A. 
    • Fraiwan L. 
    • et al.
    A generalized deep learning-based diagnostic system for early diagnosis of various types of pulmonary nodules.
    Technol Cancer Res Treat. 2018; 17https://doi.org/10.1177/1533033818798800
    View in Article
  19. 19.
    • Ardila D. 
    • Kiraly A.P. 
    • Bharadwaj S. 
    • et al.
    End-to-end lung cancer screening with three-dimensional deep learning on low-dose chest computed tomography.
    Nat Med. 2019; 25: 954-961
    View in Article
  20. 20.
    • Massion P.P. 
    • Antic S. 
    • Ather S. 
    • et al.
    Assessing the accuracy of a deep learning method to risk stratify indeterminate pulmonary nodules.
    Am J Respir Crit Care Med. 2020; 202: 241-249
    View in Article
  21. 21.
    • Baldwin D.R. 
    • Gustafson J. 
    • Pickup L. 
    • et al.
    External validation of a convolutional neural network artificial intelligence tool to predict malignancy in pulmonary nodules.
    Thorax. 2020; 75: 306-312
    View in Article
  22. 22.
    • Hunter B. 
    • Hindocha S. 
    • Lee R.W.
    The role of artificial intelligence in early cancer diagnosis.
    Cancers. 2022; 14https://doi.org/10.3390/cancers14061524
    View in Article
  23. 23.
    • Armato S.G. 
    • McLennan G. 
    • Bidaut L. 
    • et al.
    The lung image database consortium (LIDC) and image database resource initiative (IDRI): a completed reference database of lung nodules on CT scans.
    Med Phys. 2011; 38: 915-931
    View in Article
  24. 24.
    • Armato S.G. 
    • Drukker K. 
    • Li F. 
    • et al.
    LUNGx Challenge for computerized lung nodule classification.vol. 3. 2016044506https://doi.org/10.1117/1JMI34044506
    View in Article
  25. 25.
    • Bakr S. 
    • Gevaert O. 
    • Echegaray S. 
    • et al.
    A radiogenomic dataset of non-small cell lung cancer.
    Sci Data. 2018; 5https://doi.org/10.1038/SDATA.2018.202
    View in Article
  26. 26.
    • Isensee F. 
    • Jaeger P.F. 
    • Kohl S.A.A. 
    • Petersen J. 
    • Maier-Hein K.H.
    nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation.
    Nat Methods. 2021; 18: 203-211
    View in Article
  27. 27.
    • Chung K. 
    • Mets O.M. 
    • Gerke P.K. 
    • et al.
    Brock malignancy risk calculator for pulmonary nodules: validation outside a lung cancer screening population.
    Thorax. 2018; 73: 857-863
    View in Article
  28. 28.
    • Ma Y. 
    • Byrne S.C. 
    • Gange C. 
    • et al.
    Radiologic-pathologic correlation for nondiagnostic CT-guided lung biopsies performed for the evaluation of lung cancer.
    AJR. 2020;https://doi.org/10.2214/AJR.19.22244
    View in Article
  29. 29.
    • Li X. 
    • Guindani M. 
    • Ng C.S. 
    • Hobbs B.P.
    Spatial Bayesian modeling of GLCM with application to malignant lesion characterization.
    J Appl Stat. 2018; 46: 230-246
    View in Article
  30. 30.
    • Tang C. 
    • Hobbs B. 
    • Amer A. 
    • et al.
    Development of an immune-pathology informed radiomics model for non-small cell lung cancer.
    Sci Rep. 2018; 8https://doi.org/10.1038/S41598-018-20471-5
    View in Article
  31. 31.
    • Sanduleanu S. 
    • Jochems A. 
    • Upadhaya T. 
    • et al.
    Non-invasive imaging prediction of tumor hypoxia: a novel developed and externally validated CT and FDG-PET-based radiomic signatures.
    Radiother Oncol. 2020; 153: 97-105
    View in Article
  32. 32.
    • Chen X. 
    • Feng B. 
    • Chen Y. 
    • et al.
    A CT-based radiomics nomogram for prediction of lung adenocarcinomas and granulomatous lesions in patient with solitary sub-centimeter solid nodules.
    Cancer Imag. 2020; 20: 1-13
    View in Article
  33. 33.
    • Choi W. 
    • Oh J.H. 
    • Riyahi S. 
    • et al.
    Radiomics analysis of pulmonary nodules in low-dose CT for early detection of lung cancer.
    Med Phys. 2018; 45: 1537-1549
    View in Article
  34. 34.
    • Feng B. 
    • Chen X. 
    • Chen Y. 
    • et al.
    Radiomics nomogram for preoperative differentiation of lung tuberculoma from adenocarcinoma in solitary pulmonary solid nodule.
    Eur J Radiol. 2020; 128109022
    View in Article
  35. 35.
    • Hawkins S. 
    • Wang H. 
    • Liu Y. 
    • et al.
    Predicting malignant nodules from screening CT scans.
    J Thorac Oncol. 2016; 11: 2120-2128
    View in Article
  36. 36.
    • Peikert T. 
    • Duan F. 
    • Rajagopalan S. 
    • et al.
    Novel high-resolution computed tomography-based radiomic classifier for screen-identified pulmonary nodules in the National Lung Screening Trial.
    PLoS One. 2018; 13https://doi.org/10.1371/journal.pone.0196910
    View in Article
  37. 37.
    • Liu A. 
    • Wang Z. 
    • Yang Y. 
    • et al.
    Preoperative diagnosis of malignant pulmonary nodules in lung cancer screening with a radiomics nomogram.
    Cancer Commun. 2020; 40: 16-24
    View in Article
  38. 38.
    • Vickers A.J. 
    • Cronin A.M. 
    • Begg C.B.
    One statistical test is sufficient for assessing new predictive markers.
    BMC Med Res Methodol. 2011; 11: 13
    View in Article

Article info

Publication history

Published: November 09, 2022
Accepted: October 18, 2022
Received in revised form: October 12, 2022
Received: August 8, 2022

Identification

DOI: https://doi.org/10.1016/j.ebiom.2022.104344

Copyright

© 2022 The Authors. Published by Elsevier B.V.

User license

Creative Commons Attribution (CC BY 4.0) | 
How you can reuse Information Icon

ScienceDirect

Access this article on ScienceDirect

Figures

  • Figure thumbnail gr1
    Fig. 1Study recruitment diagram. The numbers of scans are shown in parentheses. Three patients could not be analysed for technical reasons, leading to a final subset of 499 patients and 838 scans. Abbreviations: NUH, Nottingham University Hospital; RBH, The Royal Brompton Hospital; RMH, The Royal Marsden Hospital; UCLH, University College London Hospitals; LIDC, Lung-image database consortium; RG, Non-small cell lung cancer Radiogenomics study.
  • Figure thumbnail gr2
    Fig. 2Radiomics decision-support tool. The large-nodule radiomics-predictive vector (LN-RPV) is used to prompt earlier intervention in patients with intermediate (10–70%) Herder scores but a high-risk radiomics score.
  • Figure thumbnail gr3
    Fig. 3Radiologist Performance Benchmarking (n = 252). a) Malignancy-prediction ROC curves for the three radiologists in the test set. The AUCs were: R1: 0.75 (95% CI 0.70–0.81), R2: 0.74 (95% CI 0.67–0.79) and R3: 0.77 (95% CI 0.71–0.82). b) Malignancy prediction performance metrics for the radiologist with the median AUC (R1) after selecting the optimum cut-point to maximise the Youden index (4 – Probably malignant). The radiologist achieved an accuracy of 65% (95% CI 59–71%). Abbreviations: AUC, Area under the curve; PPV, Positive predictive value; NPV, Negative predictive value; CI, Confidence interval.
  • Figure thumbnail gr4
    Fig. 4The Large-Nodule Radiomics Predictive Vector (LN-RPV). a) The LASSO regularisation plot, lambda plot and the regression weights for the two selected features are shown. b) ROC-AUC curves for malignancy prediction were used to select cut-offs based on the training-set Youden index. In the test set, the model achieved an AUC of 0.83 (95% CI 0.77–0.88) and an accuracy of 76% (95% CI 70–81%). Abbreviations: AUC, Area under the curve; PPV, Positive predictive value; NPV, Negative predictive value, CI, Confidence interval.
  • Figure thumbnail gr5
    Fig. 5Performance comparison between the LN-RPV and the Brock and Herder Scores for baseline solid nodules in the test set (117 patients). Performance metrics are reported using training set cut-offs to maximise the Youden index. For accuracy, 95% CIs are given in parentheses. Abbreviations: AUC, Area under the curve; PPV, Positive predictive value; NPV, Negative predictive value; CI, Confidence interval.
  • Figure thumbnail gr6
    Fig. 6Reproducibility of the LN-RPV using auto-segmentation. a) Scatterplot showing high correlation between the manual and automated LN-RPV values (r = 0.95). b) Intra-class correlation co-efficients for each radiomics feature comprising the LN-RPV, and the LN-RPV alone. P values are reported for the null hypothesis that there is no-correlation between the two methods. Abbreviation: ICC, Intra-class correlation coefficient.

Tables




AI-algoritmen, Kunstmatige intelligentie, The Lancet, kankerknobbeltjes, longen


Plaats een reactie ...

Reageer op "AI-algoritmen (Kunstmatige intelligentie) ontdekken kankerknobbeltjes op de longen die radiologen nog niet kunnen waarnemen en kan belangrijk zijn om patienten eerder en succesvol te behandelen"


Gerelateerde artikelen
 

Gerelateerde artikelen

Bevolkingsonderzoek via CT-scan >> AI-algoritmen (Kunstmatige >> Ademtest via electronische >> Luchtvervuiling lijkt voornaamste >> Westers dieet met voeding >> Rauwe knoflook zou de de kans >> Preventie longkanker: Regelmatige >> Tiotropium Respimat zachte >> Longkanker: D-Alpha-tocopherol >> Longkanker: Zelfs minder roken >>