Classification and mutation prediction from non–small cell lung cancer histopathology images using deep learning

Subjects

Abstract

Visual inspection of histopathology slides is one of the main methods used by pathologists to assess the stage, type and subtype of lung tumors. Adenocarcinoma (LUAD) and squamous cell carcinoma (LUSC) are the most prevalent subtypes of lung cancer, and their distinction requires visual inspection by an experienced pathologist. In this study, we trained a deep convolutional neural network (inception v3) on whole-slide images obtained from The Cancer Genome Atlas to accurately and automatically classify them into LUAD, LUSC or normal lung tissue. The performance of our method is comparable to that of pathologists, with an average area under the curve (AUC) of 0.97. Our model was validated on independent datasets of frozen tissues, formalin-fixed paraffin-embedded tissues and biopsies. Furthermore, we trained the network to predict the ten most commonly mutated genes in LUAD. We found that six of them—STK11, EGFR, FAT1, SETBP1, KRAS and TP53—can be predicted from pathology images, with AUCs from 0.733 to 0.856 as measured on a held-out population. These findings suggest that deep-learning models can assist pathologists in the detection of cancer subtype or gene mutations. Our approach can be applied to any cancer type, and the code is available at https://github.com/ncoudray/DeepPATH.

Access options

Rent or Buy article

Get time limited or full article access on ReadCube.

from$8.99

All prices are NET prices.

Fig. 1: Data and strategy.
Fig. 2: Classification of presence and type of tumor on alternative cohorts.
Fig. 3: Gene mutation prediction from histopathology slides give promising results for at least six genes.
Fig. 4: Spatial heterogeneity of predicted mutations.

Data availability

All relevant data used for training during the current study are available through the Genomic Data Commons portal (https://gdc-portal.nci.nih.gov). These datasets were generated by TCGA Research Network (http://cancergenome.nih.gov/), and they have made them publicly available. Other datasets analyzed during the current study are available from the corresponding author on reasonable request.

References

  1. 1.

    Travis, W. D. et al. International Association for the Study of Lung Cancer/American Thoracic Society/European Respiratory Society international multidisciplinary classification of lung adenocarcinoma. J. Thorac. Oncol. 6, 244–285 (2011).

    Article  PubMed Central  PubMed  Google Scholar 

  2. 2.

    Hanna, N. et al. Systemic therapy for stage IV non–small-cell lung cancer: American Society of Clinical Oncology clinical practice guideline update. J. Clin. Oncol. 35, 3484–3515 (2017).

    Article  Google Scholar 

  3. 3.

    Chan, B. A. & Hughes, B. G. Targeted therapy for non–small cell lung cancer: current standards and the promise of the future. Transl. Lung Cancer Res. 4, 36–54 (2015).

    CAS  PubMed  PubMed Central  Google Scholar 

  4. 4.

    Parums, D. V. Current status of targeted therapy in non–small cell lung cancer. Drugs Today (Barc) 50, 503–525 (2014).

    CAS  Article  Google Scholar 

  5. 5.

    Terra, S. B. et al. Molecular characterization of pulmonary sarcomatoid carcinoma: analysis of 33 cases. Mod. Pathol. 29, 824–831 (2016).

    CAS  Article  Google Scholar 

  6. 6.

    Blumenthal, G. M. et al. Oncology drug approvals: evaluating endpoints and evidence in an era of breakthrough therapies. Oncologist 22, 762–767 (2017).

    Article  PubMed Central  PubMed  Google Scholar 

  7. 7.

    Pérez-Soler, R. et al. Determinants of tumor response and survival with erlotinib in patients with non–small-cell lung cancer. J. Clin. Oncol. 22, 3238–3247 (2004).

    Article  CAS  Google Scholar 

  8. 8.

    Jänne, P. A. et al. Selumetinib plus docetaxel for KRAS-mutant advanced non-small-cell lung cancer: a randomised, multicentre, placebo-controlled, phase 2 study. Lancet Oncol. 14, 38–47 (2013).

    Article  CAS  Google Scholar 

  9. 9.

    Thunnissen, E., van der Oord, K. & den Bakker, M. Prognostic and predictive biomarkers in lung cancer. A review. Virchows Arch. 464, 347–358 (2014).

    CAS  Article  Google Scholar 

  10. 10.

    Zachara-Szczakowski, S., Verdun, T. & Churg, A. Accuracy of classifying poorly differentiated non–small cell lung carcinoma biopsies with commonly used lung carcinoma markers. Hum. Pathol. 46, 776–782 (2015).

    CAS  Article  Google Scholar 

  11. 11.

    Luo, X. et al. Comprehensive computational pathological image analysis predicts lung cancer prognosis. J. Thorac. Oncol. 12, 501–509 (2017).

    Article  Google Scholar 

  12. 12.

    Yu, K.-H. et al. Predicting non–small cell lung cancer prognosis by fully automated microscopic pathology image features. Nat. Commun. 7, 12474 (2016).

    CAS  Article  PubMed Central  PubMed  Google Scholar 

  13. 13.

    Khosravi, P., Kazemi, E., Imielinski, M., Elemento, O. & Hajirasouliha, I. Deep convolutional neural networks enable discrimination of heterogeneous digital pathology images. EBioMedicine 27, 317–328 (2018).

    Article  Google Scholar 

  14. 14.

    Sozzi, G. et al. Quantification of free circulating DNA as a diagnostic marker in lung cancer. J. Clin. Oncol. 21, 3902–3908 (2003).

    CAS  Article  Google Scholar 

  15. 15.

    Terry, J. et al. Optimal immunohistochemical markers for distinguishing lung adenocarcinomas from squamous cell carcinomas in small tumor samples. Am. J. Surg. Pathol. 34, 1805–1811 (2010).

    Article  Google Scholar 

  16. 16.

    Schmidhuber, J. Deep learning in neural networks: an overview. Neural Netw. 61, 85–117 (2015).

    Article  Google Scholar 

  17. 17.

    Greenspan, H., Ginneken, Bv & Summers, R. M. Guest editorial deep learning in medical imaging: overview and future promise of an exciting new technique. IEEE Trans. Med. Imaging 35, 1153–1159 (2016).

    Article  Google Scholar 

  18. 18.

    Qaiser, T., Tsang, Y.-W., Epstein, D. & RajpootEma, N. Tumor segmentation in whole slide images using persistent homology and deep convolutional features. In Medical Image Understanding and Analysis: 21st Annual Conference on Medical Image Understanding and Analysis. (Eds. Valdes Hernandez, M. & González-Castro, V.) 320–329 (Springer International Publishing, New York, 2018).

  19. 19.

    Shen, D., Wu, G. & Suk, H.-I. Deep learning in medical image analysis. Annu. Rev. Biomed. Eng. 19, 221–248 (2017).

    CAS  Article  PubMed Central  PubMed  Google Scholar 

  20. 20.

    Xing, F., Xie, Y. & Yang, L. An automatic learning-based framework for robust nucleus segmentation. IEEE Trans. Med. Imaging 35, 550–566 (2016).

    Article  Google Scholar 

  21. 21.

    de Bel, T. et al. Automatic segmentation of histopathological slides of renal tissue using deep learning. In Medical Imaging 2018: Digital Pathology Vol. 10581 (Eds. Tomaszewski, J. E. & Gurcan, M. N.) 1058112 (International Society for Optics and Photonics, Bellingham, WA, USA, 2018).

  22. 22.

    Simon, O., Yacoub, R., Jain, S., Tomaszewski, J. E. & Sarder, P. Multi-radial LBP features as a tool for rapid glomerular detection and assessment in whole slide histopathology images. Sci. Rep. 8, 2032 (2018).

    Article  PubMed Central  CAS  PubMed  Google Scholar 

  23. 23.

    Cheng, J.-Z. et al. Computer-aided diagnosis with deep learning architecture: applications to breast lesions in US images and pulmonary nodules in CT scans. Sci. Rep. 6, 24454 (2016).

    CAS  Article  PubMed Central  PubMed  Google Scholar 

  24. 24.

    Cruz-Roa, A. et al. Accurate and reproducible invasive breast cancer detection in whole-slide images: a deep learning approach for quantifying tumor extent. Sci. Rep. 7, 46450 (2017).

    CAS  Article  PubMed Central  PubMed  Google Scholar 

  25. 25.

    Sirinukunwattana, K. et al. Locality sensitive deep learning for detection and classification of nuclei in routine colon cancer histology images. IEEE Trans. Med. Imaging 35, 1196–1206 (2016).

    Article  Google Scholar 

  26. 26.

    Ertosun, M. G. & Rubin., D. L. Automated grading of gliomas using deep learning in digital pathology images: a modular approach with ensemble of convolutional neural networks. In AMIA Annual Symposium Proceedings. 1899–1908 (American Medical Informatics Association, Bethesda, MD, USA).

  27. 27.

    Bulten, W., Kaa, C.A.H.-d., Laak, J.d. & Litjens, G.J. Automated segmentation of epithelial tissue in prostatectomy slides using deep learning. In Medical Imaging 2018: Digital Pathology. Vol. 10581 (Eds. Tomaszewski, J. E. & Gurcan, M. N.) 105810S (International Society for Optics and Photonics, Bellingham, WA, USA, 2018).

  28. 28.

    Mishra, R., Daescu, O., Leavey, P., Rakheja, D. & Sengupta, A. Histopathological Diagnosis for Viable and Non-viable Tumor Prediction for Osteosarcoma Using Convolutional Neural Network. In International Symposium on Bioinformatics Research and Applications Vol. 10330 (Eds. Cai, Z., D. Ovidiu, & Li, M.) 12–23 (Springer International Publishing, New York, 2018).

    Google Scholar 

  29. 29.

    Anthimopoulos, M., Christodoulidis, S., Ebner, L., Christe, A. & Mougiakakou, S. Lung Pattern classification for interstitial lung diseases using a deep convolutional neural network. IEEE Trans. Med. Imaging 35, 1207–1216 (2016).

    Article  Google Scholar 

  30. 30.

    Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J. & Wojna, Z. Rethinking the Inception Architecture for Computer Vision. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2818–2826 (Boston, MA, USA, 2015).

  31. 31.

    Szegedy, C. et al. Going Deeper With Convolutions. In The IEEE Conference on Computer Vision and Pattern Recognition. 1–9 (Boston, 2015).

  32. 32.

    Esteva, A. et al. Dermatologist-level classification of skin cancer with deep neural networks. Nature 542, 115–118 (2017).

    CAS  Article  Google Scholar 

  33. 33.

    Gulshan, V. et al. Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs. J. Am. Med. Assoc. 316, 2402–2410 (2016).

    Article  Google Scholar 

  34. 34.

    Grossman, R. L. et al. Toward a shared vision for cancer genomic data. N. Engl. J. Med. 375, 1109–1112 (2016).

    Article  Google Scholar 

  35. 35.

    Abels, E. & Pantanowitz, L. Current state of the regulatory trajectory for whole slide imaging devices in the USA. J. Pathol. Inform. 8, 23 (2017).

    Article  PubMed Central  PubMed  Google Scholar 

  36. 36.

    Sanchez-Cespedes, M. et al. Inactivation of LKB1/STK11 is a common event in adenocarcinomas of the lung. Cancer Res. 62, 3659–3662 (2002).

    CAS  PubMed  Google Scholar 

  37. 37.

    Shackelford, D. B. et al. LKB1 inactivation dictates therapeutic response of non–small cell lung cancer to the metabolism drug phenformin. Cancer Cell 23, 143–158 (2013).

    CAS  Article  PubMed Central  PubMed  Google Scholar 

  38. 38.

    Makowski, L. & Hayes, D. N. Role of LKB1 in lung cancer development. Br. J. Cancer 99, 683–688 (2008).

    CAS  Article  PubMed Central  PubMed  Google Scholar 

  39. 39.

    Morris, L. G. et al. Recurrent somatic mutation of FAT1 in multiple human cancers leads to aberrant Wnt activation. Nat. Genet. 45, 253–261 (2013).

    CAS  Article  PubMed Central  PubMed  Google Scholar 

  40. 40.

    Mogi, A. & Kuwano, H. TP53 mutations in nonsmall cell lung cancer. J. Biomed. Biotechnol. 2011, 583929 (2011).

    Article  PubMed Central  CAS  PubMed  Google Scholar 

  41. 41.

    Kandoth, C. et al. Mutational landscape and significance across 12 major cancer types. Nature 502, 333–339 (2013).

    CAS  Article  PubMed Central  PubMed  Google Scholar 

  42. 42.

    Zeiler, M.D. & Fergus, R. Visualizing and understanding convolutional networks. In European Conference on Computer Vision. 818–833 (Springer International Publishing, New York, 2015).

  43. 43.

    Maaten, L. J. Pd Accelerating t-SNE using tree-based algorithms. J. Mach. Learn. Res. 15, 3221–3245 (2014).

    Google Scholar 

  44. 44.

    Bonner, R. F. et al. Laser capture microdissection: molecular analysis of tissue. Science 278, 1481–1483 (1997). 1483.

    CAS  Article  Google Scholar 

  45. 45.

    Ninomiya, H. et al. Correlation between morphology and EGFR mutations in lung adenocarcinomas significance of the micropapillary pattern and the hobnail cell type. Lung Cancer 63, 235–240 (2009).

    Article  Google Scholar 

  46. 46.

    Warth, A. et al. EGFR, KRAS, BRAF and ALK gene alterations in lung adenocarcinomas: patient outcome, interplay with morphology and immunophenotype. Eur. Respir. J. 43, 872–883 (2014).

    CAS  Article  Google Scholar 

  47. 47.

    Sequist, L. V. et al. Genotypic and histological evolution of lung cancers acquiring resistance to EGFR inhibitors. Sci. Transl. Med. 3, 75ra26 (2011).

    Article  PubMed Central  PubMed  Google Scholar 

  48. 48.

    Chiang, S. et al. IDH2 mutations define a unique subtype of breast cancer with altered nuclear polarity. Cancer Res. 76, 7118–7129 (2016).

    CAS  Article  PubMed Central  PubMed  Google Scholar 

  49. 49.

    Baas, A. F., Smit, L. & Clevers, H. LKB1 tumor suppressor protein: partaker in cell polarity. Trends Cell Biol. 14, 312–319 (2004).

    CAS  Article  Google Scholar 

  50. 50.

    Gloushankova, N., Ossovskaya, V., Vasiliev, J., Chumakov, P. & Kopnin, B. Changes in p53 expression can modify cell shape of ras-transformed fibroblasts and epitheliocytes. Oncogene 15, 2985–2989 (1997).

    CAS  Article  Google Scholar 

  51. 51.

    Yatabe, Y. et al. EGFR mutation is specific for terminal respiratory unit type adenocarcinoma. Am. J. Surg. Pathol. 29, 633–639 (2005).

    Article  Google Scholar 

  52. 52.

    Yoshida, A. et al. Comprehensive histologic analysis of ALK-rearranged lung carcinomas. Am. J. Surg. Pathol. 35, 1226–1234 (2011).

    Article  Google Scholar 

  53. 53.

    Rodig, S. J. et al. Unique clinicopathologic features characterize ALK-rearranged lung adenocarcinoma in the western population. Clin. Cancer Res. 15, 5216–5223 (2009).

    CAS  Article  PubMed Central  PubMed  Google Scholar 

  54. 54.

    Dearden, S., Stevens, J., Wu, Y.-L. & Blowers, D. Mutation incidence and coincidence in non small-cell lung cancer: meta-analyses by ethnicity and histology (mutMap). Ann. Oncol 24, 2371–2376 (2013).

    CAS  Article  PubMed Central  PubMed  Google Scholar 

  55. 55.

    Yu, J. et al. Mutation-specific antibodies for the detection of EGFR mutations in non-small-cell lung cancer. Clin. Cancer Res. 15, 3023–3028 (2009).

    CAS  Article  Google Scholar 

  56. 56.

    Houang, M. et al. EGFR mutation specific immunohistochemistry is a useful adjunct which helps to identify false negative mutation testing in lung cancer. Pathology 46, 501–508 (2014).

    CAS  Article  Google Scholar 

  57. 57.

    Dimou, A. et al. Standardization of epidermal growth factor receptor (EGFR) measurement by quantitative immunofluorescence and impact on antibody-based mutation detection in non–small cell lung cancer. Am. J. Pathol. 179, 580–589 (2011).

    CAS  Article  PubMed Central  PubMed  Google Scholar 

  58. 58.

    Schaumberg, A. J., Rubin, M. A. & Fuchs, T. J. H&E-stainedwhole slide deep learning predicts spop mutation state in prostate cancer. Preprint at https://doi.org/10.1101/064279 (2016).

  59. 59.

    Donovan, M. J. et al. A systems pathology model for predicting overall survival in patients with refractory, advanced non-small-cell lung cancer treated with gefitinib. Eur. J. Cancer 45, 1518–1526 (2009).

    CAS  Article  Google Scholar 

  60. 60.

    Cancer Genome Atlas Research Network. Comprehensive genomic characterization of squamous cell lung cancers. Nature 489, 519–525 (2012).

    Article  CAS  Google Scholar 

  61. 61.

    Cancer Genome Atlas Research Network. Comprehensive molecular profiling of lung adenocarcinoma. Nature 511, 543–550 (2014).

    Article  CAS  Google Scholar 

  62. 62.

    Hershey, S. et al. CNN architectures for large-scale audio classification. In 2017 IEEE International Conference on Acoustics, Speech and Signal Processing, 131–135 (New Orleans, LA, USA, 2017).

  63. 63.

    Hanley, J. A. & McNeil, B. J. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology 143, 29–36 (1982).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  64. 64.

    Pedregosa, F. et al. Scikit-learn: machine learning in Python. J. Mach. Learn. Res. 12, 2825–2830 (2011).

    Google Scholar 

  65. 65.

    Efron, B. & Tibshirani, R. J. An Introduction to the Bootstrap 56 (CRC Press, Boca Raton, FL, USA, 1994).

    Google Scholar 

  66. 66.

    Cohen, J. A coefficient of agreement for nominal scales. Educ. Psychol. Meas. 20, 37–46 (1960).

    Article  Google Scholar 

  67. 67.

    McHugh, M. L. Interrater reliability: the kappa statistic. Biochem. Med. (Zagreb) 22, 276–282 (2012).

    Article  Google Scholar 

Download references

Acknowledgements

We would like to thank the Applied Bioinformatics Laboratories (ABL) at the NYU School of Medicine for providing bioinformatics support and helping with the analysis and interpretation of the data. The Applied Bioinformatics Laboratories are a Shared Resource, partially supported by the Cancer Center Support Grant, P30CA016087 (A.T.), at the Laura and Isaac Perlmutter Cancer Center (A.T.). For this work, we used computing resources at the High-Performance Computing Facility (HPC) at NYU Langone Medical Center. The slide images and the corresponding cancer information were uploaded from the Genomic Data Commons portal (https://gdc-portal.nci.nih.gov) and are in whole or in part based upon data generated by the TCGA Research Network (http://cancergenome.nih.gov/). These data were publicly available without restriction, authentication or authorization necessary. We thank the GDC help desk for providing assistance and information regarding the TCGA dataset. For the independent cohorts, we only used whole-slide images; the NYU dataset we used consists of slide images without identifiable information and therefore does not require approval according to both federal regulations and the NYU School of Medicine Institutional Review Board. For this same reason, written informed consent was not necessary. We thank C. Dickerson, from the Center for Biospecimen Research and Development (CBRD), for scanning the whole-slide images from the NYU Langone Medical Center. We also thank T. Papagiannakopoulos, H. Pass and K.-K. Wong or their valuable and constructive suggestions.

Author information

Affiliations

Authors

Contributions

N.C. performed the experiments; N.C., A.T. and N.R. designed the experiments; N.C. and T.S. wrote the code to achieve different tasks; T.S. gathered the mutation information and contributed to their analysis; M.S. helped identify cases validated by next-generation sequencing; A.L.M. and P.S.O. collected and labeled the independent cohorts. A.L.M, P.S.O and N.N. manually labeled the TCGA dataset; N.C., A.L.M., P.S.O., N.R. and A.T. contributed to the analysis of the data; D.F., N.R. and A.T. conceived and directed the project; N.C., A.T., N.R., A.L.M. and P.S.O. wrote the manuscript with the assistance and feedback of all the other co-authors.

Corresponding authors

Correspondence to Narges Razavian or Aristotelis Tsirigos.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Text and Figures

Supplementary Figures 1–10 and Supplementary Tables 1–7.

Reporting Summary

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Coudray, N., Ocampo, P.S., Sakellaropoulos, T. et al. Classification and mutation prediction from non–small cell lung cancer histopathology images using deep learning. Nat Med 24, 1559–1567 (2018). https://doi.org/10.1038/s41591-018-0177-5

Download citation

Further reading