Deep Learning Versus Classical Regression for Brain Tumor Patient Survival Prediction

  • Yannick SuterEmail author
  • Alain Jungo
  • Michael Rebsamen
  • Urspeter Knecht
  • Evelyn Herrmann
  • Roland Wiest
  • Mauricio Reyes
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 11384)


Deep learning for regression tasks on medical imaging data has shown promising results. However, compared to other approaches, their power is strongly linked to the dataset size. In this study, we evaluate 3D-convolutional neural networks (CNNs) and classical regression methods with hand-crafted features for survival time regression of patients with high-grade brain tumors. The tested CNNs for regression showed promising but unstable results. The best performing deep learning approach reached an accuracy of \(51.5\%\) on held-out samples of the training set. All tested deep learning experiments were outperformed by a Support Vector Classifier (SVC) using 30 radiomic features. The investigated features included intensity, shape, location and deep features.

The submitted method to the BraTS 2018 survival prediction challenge is an ensemble of SVCs, which reached a cross-validated accuracy of \(72.2\%\) on the BraTS 2018 training set, \(57.1\%\) on the validation set, and \(42.9\%\) on the testing set.

The results suggest that more training data is necessary for a stable performance of a CNN model for direct regression from magnetic resonance images, and that non-imaging clinical patient information is crucial along with imaging information.


Brain tumor Survival prediction Regression 3D-Convolutional Neural Networks 



We gladly acknowledge the support of the Swiss Cancer League (grant KFS-3979-08-2016) and the Swiss National Science Foundation (grant 169607). We are grateful for the support of the NVIDIA corporation for the donation of a Titan Xp GPU. Calculations were partly performed on UBELIX, the HPC cluster at the University of Bern.


  1. 1.
    Awad, A.W., et al.: Impact of removed tumor volume and location on patient outcome in glioblastoma. J. Neuro Oncol. 135(1), 161–171 (2017). Scholar
  2. 2.
    Bakas, S., et al.: Segmentation labels and radiomic features for the pre-operative scans of the TCGA-GBM collection. Cancer Imaging Arch. (2017). Scholar
  3. 3.
    Bakas, S., et al.: Segmentation labels and radiomic features for the pre-operative scans of the TCGA-LGG collection. Cancer Imaging Arch. (2017). Scholar
  4. 4.
    Bakas, S., Reyes, M., et al.: Identifying the best machine learning algorithms for brain tumor segmentation, progression assessment, and overall survival prediction in the BRATS challenge. ArXiv e-prints, November 2018Google Scholar
  5. 5.
    Bakas, S., et al.: Advancing The Cancer Genome Atlas glioma MRI collections with expert segmentation labels and radiomic features. Sci. Data 4, 170117 (2017). Scholar
  6. 6.
    Breiman, L., Friedman, J.H., Olshen, R.A., Stone, C.J.: Classification and regression trees (1984)Google Scholar
  7. 7.
    Cox, D.R.: The regression analysis of binary sequences. J. R. Stat. Society Ser. B (Methodol.), 20(2), 215–242 (1958)MathSciNetzbMATHGoogle Scholar
  8. 8.
    Fischl, B.: Freesurfer. Neuroimage 62(2), 774–781 (2012). Scholar
  9. 9.
    Hastie, T., Friedman, J., Tibshirani, R.: The Elements of Statistical Learning. SSS, vol. 1. Springer, New York (2001). Scholar
  10. 10.
    Gillies, R.J., Kinahan, P.E., Hricak, H.: Radiomics: images are more than pictures, they are data. Radiology 278(2), 563–577 (2015). Scholar
  11. 11.
    van Griethuysen, J.J., et al.: Computational radiomics system to decode the radiographic phenotype. Cancer Res. 77(21), e104–e107 (2017). Scholar
  12. 12.
    Jungo, A., et al.: Towards uncertainty-assisted brain tumor segmentation and survival prediction. In: Crimi, A., Bakas, S., Kuijf, H., Menze, B., Reyes, M. (eds.) BrainLes 2017. LNCS, vol. 10670, pp. 474–485. Springer, Cham (2018). Scholar
  13. 13.
    Kinga, D., Adam, J.B.: A method for stochastic optimization. In: International Conference on Learning Representations (ICLR), vol. 5 (2015)Google Scholar
  14. 14.
    Lampert, C.H., et al.: Kernel methods in computer vision. Found. Trends® Comput. Graph. Vis. 4(3), 193–285 (2009). Scholar
  15. 15.
    Lao, J., et al.: A deep learning-based radiomics model for prediction of survival in glioblastoma multiforme. Sci. Rep. 7(1), 10353 (2017). Scholar
  16. 16.
    Li, Y., Shen, L.: Deep learning based multimodal brain tumor diagnosis. In: Crimi, A., Bakas, S., Kuijf, H., Menze, B., Reyes, M. (eds.) BrainLes 2017. LNCS, vol. 10670, pp. 149–158. Springer, Cham (2018). Scholar
  17. 17.
    Louis, D.N., et al.: The 2016 world health organization classification of tumors of the central nervous system: a summary. Acta Neuropathol. 131(6), 803–820 (2016). Scholar
  18. 18.
    Meier, R., et al.: Automatic estimation of extent of resection and residual tumor volume of patients with glioblastoma. J. Neurosurg. 127(4), 798–806 (2017). Scholar
  19. 19.
    Menze, B.H., Jakab, A., Bauer, S., et al.: The multimodal brain tumor image segmentation benchmark (BRATS). IEEE Trans. Med. Imaging 34(10), 1993–2024 (2015). Scholar
  20. 20.
    Pereira, S., et al.: Enhancing interpretability of automatically extracted machine learning features: application to a RBM-random forest system on brain lesion segmentation. Med. Image Anal. 44, 228–244 (2018). Scholar
  21. 21.
    Pérez-Beteta, J., et al.: Glioblastoma: does the pre-treatment geometry matter? A postcontrast T1 MRI-based study. Eur. Radiol. (2017). Scholar
  22. 22.
    Rathore, S., et al.: Radiomic MRI signature reveals three distinct subtypes of glioblastoma with different clinical and molecular characteristics, offering prognostic value beyond idh1. Sci. Rep. 8(1), 5087 (2018). Scholar
  23. 23.
    Sanai, N., Polley, M.Y., McDermott, M.W., Parsa, A.T., Berger, M.S.: An extent of resection threshold for newly diagnosed glioblastomas. J. Neurosurg. 115(1), 3–8 (2011). Scholar
  24. 24.
    Steed, T.C., et al.: Differential localization of glioblastoma subtype: implications on glioblastoma pathogenesis. Oncotarget 7(18), 24899 (2016). Scholar
  25. 25.
    Tibshirani, R.: Regression shrinkage and selection via the lasso. J. R. Stat. Society Ser. B (Methodol.), 58(1), 267–288 (1996)MathSciNetzbMATHGoogle Scholar
  26. 26.
    Wang, G., Li, W., Ourselin, S., Vercauteren, T.: Automatic brain tumor segmentation using cascaded anisotropic convolutional neural networks. In: Crimi, A., Bakas, S., Kuijf, H., Menze, B., Reyes, M. (eds.) BrainLes 2017. LNCS, vol. 10670, pp. 178–190. Springer, Cham (2018). Scholar

Copyright information

© Springer Nature Switzerland AG 2019

Authors and Affiliations

  • Yannick Suter
    • 1
    Email author
  • Alain Jungo
    • 1
  • Michael Rebsamen
    • 1
  • Urspeter Knecht
    • 1
  • Evelyn Herrmann
    • 2
  • Roland Wiest
    • 3
  • Mauricio Reyes
    • 1
  1. 1.Institute for Surgical Technology and BiomechanicsUniversity of BernBernSwitzerland
  2. 2.University Clinic for Radio-Oncology, Inselspital, Bern University HospitalUniversity of BernBernSwitzerland
  3. 3.Support Center for Advanced Neuroimaging, University Institute of Diagnostic and Interventional Neuroradiology, InselspitalUniversity of BernBernSwitzerland

Personalised recommendations