Variational Autoencoder with Truncated Mixture of Gaussians for Functional Connectivity Analysis

  • Qingyu ZhaoEmail author
  • Nicolas Honnorat
  • Ehsan Adeli
  • Adolf Pfefferbaum
  • Edith V. Sullivan
  • Kilian M. Pohl
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 11492)


Resting-state functional connectivity states are often identified as clusters of dynamic connectivity patterns. However, existing clustering approaches do not distinguish major states from rarely occurring minor states and hence are sensitive to noise. To address this issue, we propose to model major states using a non-linear generative process guided by a Gaussian-mixture distribution in a low-dimensional latent space, while separately modeling the connectivity patterns of minor states by a non-informative uniform distribution. We embed this truncated Gaussian-Mixture model in a Variational AutoEncoder framework to obtain a general joint clustering and outlier detection approach, called tGM-VAE. When applied to synthetic data with known ground-truth, tGM-VAE is more accurate in clustering dynamic connectivity patterns than existing approaches. On the rs-fMRI data of 593 healthy adolescents from the National Consortium on Alcohol and Neurodevelopment in Adolescence (NCANDA) study, tGM-VAE identified meaningful major connectivity states. The dwell time of these states significantly correlated with age.



This research was supported in part by NIH grants U24AA021697-06, AA005965, AA013521, AA017168.


  1. 1.
    Buckner, R., Krienen, F., Yeo, B.: Opportunities and limitations of intrinsic functional connectivity MRI. Nat. Neurosci. 16(7), 832–837 (2013)CrossRefGoogle Scholar
  2. 2.
    Allen, E., Damaraju, E., Plis, S., et al.: Tracking whole-brain connectivity dynamics in the resting state. Cerebral Cortex 24(3), 663–676 (2014)CrossRefGoogle Scholar
  3. 3.
    Zalesky, A., Fornito, A., Cocchi, L., Gollo, L.L., Breakspear, M.: Time-resolved resting-state brain networks. PNAS 111(28), 10341–10346 (2014)CrossRefGoogle Scholar
  4. 4.
    Damarajua, E., Allen, E., Belgerc, A., et al.: Dynamic functional connectivity analysis reveals transient states of dysconnectivity in schizophrenia. NeuroImage: Clin. 5, 298–308 (2014)CrossRefGoogle Scholar
  5. 5.
    Yu, Q., Erhardt, E.B., Sui, J., et al.: Assessing dynamic brain graphs of time-varying connectivity in fMRI data: application to healthy controls and patients with schizophrenia. NeuroImage 107, 345–355 (2015)CrossRefGoogle Scholar
  6. 6.
    Taghia, J., Ryali, S., et al.: Bayesian switching factor analysis for estimating time-varying functional connectivity in fMRI. NeuroImage 155, 271–290 (2017)CrossRefGoogle Scholar
  7. 7.
    Nielsen, S., Madsen, K., Røge, R., Schmidt, M.N., Mørup, M.: Nonparametric modeling of dynamic functional connectivity in fMRI data. In: Machine Learning and Interpretation in Neuroimaging Workshop (2015)Google Scholar
  8. 8.
    Blei, D.M., Jordan, M.I.: Variational inference for dirichlet process mixtures. Bayesian Anal. 1(1), 121–144 (2017)MathSciNetCrossRefGoogle Scholar
  9. 9.
    Nalisnick, E., Smyth, P.: Stick-breaking variational autoencoders. In: ICLR (2017)Google Scholar
  10. 10.
    Zhao, Q., Kwon, D., Pohl, K.M.: A riemannian framework for longitudinal analysis of resting-state functional connectivity. In: Frangi, A.F., Schnabel, J.A., Davatzikos, C., Alberola-López, C., Fichtinger, G. (eds.) MICCAI 2018. LNCS, vol. 11072, pp. 145–153. Springer, Cham (2018). Scholar
  11. 11.
    Kingma, D., Welling, M.: Auto-encoding variational bayes. In: ICLR (2013)Google Scholar
  12. 12.
    Kingma, D.P., Rezende, D.J., Mohamed, S., Welling, M.: Semi-supervised learning with deep generative models. In: NIPS (2014)Google Scholar
  13. 13.
    Dilokthanakul, N., Mediano, P.A., Garnelo, M.: Deep unsupervised clustering with Gaussian mixture variational autoencoder (2017, preprint).
  14. 14.
    Jiang, Z., Zheng, Y., Tan, H., Tang, B., Zhou, H.: Variational deep embedding: an unsupervised and generative approach to clustering. In: IJCAI (2017)Google Scholar
  15. 15.
    Abbasnejad, E., Dick, A.R., van den Hengel, A.: Infinite variational autoencoder for semi-supervised learning. In: CVPR (2017)Google Scholar
  16. 16.
    Higgins, I., Matthey, L., Pal, A., et al.: Beta-VAE: learning basic visual concepts with a constrained variational framework. In: ICLR (2017)Google Scholar
  17. 17.
    Ebbers, J., et al.: Hidden Markov model variational autoencoder for acoustic unit discovery. In: InterSpeech (2017)Google Scholar
  18. 18.
    Müller-Oehring, E., Kwon, D., Nagel, B., Sullivan, E., et al.: Influences of age, sex, and moderate alcohol drinking on the intrinsic functional architecture of adolescent brains. Cerebral Cortex 28(3), 1049–1063 (2017)CrossRefGoogle Scholar
  19. 19.
    Rohlfing, T., Zahr, N., Sullivan, E., Pfefferbaum, A.: The SRI24 multichannel atlas of normal adult human brain structure. Hum. Brain Mapp. 31(5), 798–819 (2014)CrossRefGoogle Scholar
  20. 20.
    Chen, Y., Wiesel, A., Eldar, Y.C., Hero, A.O.: Shrinkage algorithms for MMSE covariance estimation. IEEE Trans. Sig. Process. 58(10), 5016–5029 (2010)MathSciNetCrossRefGoogle Scholar
  21. 21.
    Pfefferbaum, A., et al.: Altered brain developmental trajectories in adolescents after initiating drinking. Am. J. Psychiatry 175(4), 370–380 (2017)CrossRefGoogle Scholar
  22. 22.
    Chen, Y., Wang, W., et al.: Age-related decline in the variation of dynamic functional connectivity: a resting state analysis. Front Aging Neurosci. 9(23), 1–11 (2017)Google Scholar

Copyright information

© Springer Nature Switzerland AG 2019

Authors and Affiliations

  • Qingyu Zhao
    • 1
    Email author
  • Nicolas Honnorat
    • 2
  • Ehsan Adeli
    • 1
  • Adolf Pfefferbaum
    • 1
    • 2
  • Edith V. Sullivan
    • 1
  • Kilian M. Pohl
    • 1
    • 2
  1. 1.Stanford University School of MedicineStanfordUSA
  2. 2.Center for Health Sciences, SRI InternationalMenlo ParkUSA

Personalised recommendations