A Universal and Efficient Method to Compute Maps from Image-Based Prediction Models

  • Mert R. Sabuncu
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 8675)


Discriminative supervised learning algorithms, such as Support Vector Machines, are becoming increasingly popular in biomedical image computing. One of their main uses is to construct image-based prediction models, e.g., for computer aided diagnosis or “mind reading.” A major challenge in these applications is the biological interpretation of the machine learning models, which can be arbitrarily complex functions of the input features (e.g., as induced by kernel-based methods). Recent work has proposed several strategies for deriving maps that highlight regions relevant for accurate prediction. Yet most of these methods either rely on strong assumptions about the prediction model (e.g., linearity, sparsity) and/or data (e.g., Gaussianity), or fail to exploit the covariance structure in the data. In this work, we propose a computationally efficient and universal framework for quantifying associations captured by black box machine learning models. Furthermore, our theoretical perspective reveals that examining associations with predictions, in the absence of ground truth labels, can be very informative. We apply the proposed method to machine learning models trained to predict cognitive impairment from structural neuroimaging data. We demonstrate that our approach yields biologically meaningful maps of association.


Machine learning image-based prediction 


  1. 1.
    Breiman, L.: Statistical modeling: The two cultures (with comments and a rejoinder by the author). Statistical Science 16(3), 199–231 (2001)CrossRefzbMATHMathSciNetGoogle Scholar
  2. 2.
    Kawasaki, Y., et al.: Multivariate voxel-based morphometry successfully differentiates schizophrenia patients from healthy controls. Neuroimage 34(1) (2007)Google Scholar
  3. 3.
    Davatzikos, C., et al.: Individual patient diagnosis of ad and ftd via high-dimensional pattern classification of mri. Neuroimage 41(4), 1220–1227 (2008)CrossRefGoogle Scholar
  4. 4.
    Klöppel, S., et al.: Automatic detection of preclinical neurodegeneration presymptomatic huntington disease. Neurology 72(5), 426–431 (2009)CrossRefGoogle Scholar
  5. 5.
    Zhang, D., et al.: Multimodal classification of alzheimer’s disease and mild cognitive impairment. Neuroimage 55(3), 856–867 (2011)CrossRefGoogle Scholar
  6. 6.
    Plant, C., et al.: Automated detection of brain atrophy patterns based on mri for the prediction of alzheimer’s disease. Neuroimage 50(1), 162–174 (2010)CrossRefGoogle Scholar
  7. 7.
    Ashburner, J., Friston, K.J.: Voxel-based morphometry–the methods. Neuroimage 11(6), 805–821 (2000)CrossRefGoogle Scholar
  8. 8.
    Sabuncu, M.R., Van Leemput, K.: The Relevance Voxel Machine (RVoxM): A self-tuning bayesian model for informative image-based prediction. IEEE Transactions on Medical Imaging (2012)Google Scholar
  9. 9.
    Gaonkar, B., Davatzikos, C.: Analytic estimation of statistical significance maps for support vector machine based multi-variate image analysis and classification. NeuroImage 78, 270–283 (2013)CrossRefGoogle Scholar
  10. 10.
    Konukoglu, E., Ganz, M., Van Leemput, K., Sabuncu, M.R.: On feature relevance in image-based prediction models: An empirical study. In: Wu, G., Zhang, D., Shen, D., Yan, P., Suzuki, K., Wang, F. (eds.) MLMI 2013. LNCS, vol. 8184, pp. 171–178. Springer, Heidelberg (2013)CrossRefGoogle Scholar
  11. 11.
    Haufe, S., et al.: On the interpretation of weight vectors of linear models in multivariate neuroimaging. NeuroImage 87, 96–110 (2014)CrossRefGoogle Scholar
  12. 12.
    Golland, P.: Discriminative direction for kernel classifiers. Advances in Neural Information Processing Systems 1, 745–752 (2002)Google Scholar
  13. 13.
    Strobl, C., et al.: Conditional variable importance for random forests. BMC Bioinformatics 9(1), 307 (2008)CrossRefGoogle Scholar
  14. 14.
    Zien, A., Krämer, N., Sonnenburg, S., Rätsch, G.: The feature importance ranking measure. In: Buntine, W., Grobelnik, M., Mladenić, D., Shawe-Taylor, J. (eds.) ECML PKDD 2009, Part II. LNCS (LNAI), vol. 5782, pp. 694–709. Springer, Heidelberg (2009)CrossRefGoogle Scholar
  15. 15.
    Meinshausen, N., Bühlmann, P.: Stability selection. Journal of the Royal Statistical Society: Series B (Statistical Methodology) 72(4), 417–473 (2010)CrossRefMathSciNetGoogle Scholar
  16. 16.
    Goldstein, A., et al.: Peeking inside the black box: Visualizing statistical learning with plots of individual conditional expectation. J. of Comp. and Graph. Stat. (2014)Google Scholar
  17. 17.
    Dosenbach, N.U.F., et al.: Prediction of individual brain maturity using fMRI. Science 329(5997), 1358 (2010)CrossRefGoogle Scholar
  18. 18.
    Sonnenburg, S., et al.: Poims: positional oligomer importance matricesunderstanding support vector machine-based signal detectors. Bioinformatics 24(13), i6–i14 (2008)Google Scholar
  19. 19.
    Zheng, S., et al.: Generalized measures of correlation for asymmetry, nonlinearity, and beyond. J. of the American Statistical Association 107(499), 1239–1252 (2012)CrossRefzbMATHGoogle Scholar
  20. 20.
    Dickerson, B., et al.: The cortical signature of alzheimer’s disease. Cerebral Cortex 19(3), 497–510 (2009)CrossRefGoogle Scholar

Copyright information

© Springer International Publishing Switzerland 2014

Authors and Affiliations

  • Mert R. Sabuncu
    • 1
  1. 1.A.A. Martinos Center for Biomedical Imaging, Massachusetts General HospitalHarvard Medical SchoolCharlestownUSA

Personalised recommendations