Hierarchical Adversarial Connectomic Domain Alignment for Target Brain Graph Prediction and Classification from a Source Graph

Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 11843)


Recently, deep learning methods have been widely used for medical data synthesis. However, existing deep learning frameworks are mainly designed to predict Euclidian structured data (i.e., image), which causes them to fail when handling geometric data (e.g., brain graphs). Besides, these do not naturally account for domain fracture between training source and target data distributions and generally ignore any hierarchical structure that might be present in the data, which causes a flat domain alignment. To address these limitations, we unprecedentedly propose a Hierarchical Graph Adversarial Domain Alignment (HADA) framework for predicting a target brain graph from a source brain graph. We first propose to align the source domain to the target domain by learning their successive embeddings using training samples. In this way, we are optimally learning a hierarchical alignment of both domains as we are learning a graph embedding using the previously aligned source-to-target embedding. To predict the target brain graph of a testing subject, we learn a source embedding using training and testing samples. Second, we learn a connectomic manifold for each of the resulting embeddings (i.e., the hierarchical embedding and the source embedding). Next, we select the nearest neighbors to the testing subject in the source manifold in order to average their corresponding target graphs. Finally, using the source and predicted target graphs by our HADA method, we train a random forest classifier to distinguish between disordered and healthy subjects. Our HADA framework outperformed comparison methods in predicting target brain graph and yields the best classification accuracy in comparison with using only the source graphs.


  1. 1.
    Li, R., et al.: Deep learning based imaging data completion for improved brain disease diagnosis. In: Golland, P., Hata, N., Barillot, C., Hornegger, J., Howe, R. (eds.) MICCAI 2014. LNCS, vol. 8675, pp. 305–312. Springer, Cham (2014). Scholar
  2. 2.
    Ben-Cohen, A., et al.: Cross-modality synthesis from CT to PET using FCN and GAN networks for improved automated lesion detection. Eng. Appl. Artif. Intell. 78, 186–194 (2019)CrossRefGoogle Scholar
  3. 3.
    Goodfellow, I., et al.: Generative adversarial nets. Advances in Neural Information Processing Systems, pp. 2672–2680 (2014)Google Scholar
  4. 4.
    Arslan, S., Ktena, S.I., Glocker, B., Rueckert, D.: Graph saliency maps through spectral convolutional networks: Application to sex classification with brain connectivity. arXiv preprint arXiv:1806.01764 (2018)
  5. 5.
    Ktena, S.I., et al.: Distance metric learning using graph convolutional networks: application to functional brain networks. In: Descoteaux, M., Maier-Hein, L., Franz, A., Jannin, P., Collins, D.L., Duchesne, S. (eds.) MICCAI 2017. LNCS, vol. 10433, pp. 469–477. Springer, Cham (2017). Scholar
  6. 6.
    Kipf, T.N., Welling, M.: Semi-supervised classification with graph convolutional networks. arXiv preprint arXiv:1609.02907 (2016)
  7. 7.
    Olut, S., Sahin, Y.H., Demir, U., Unal, G.: Generative adversarial training for MRA image synthesis using multi-contrast MRI. arXiv preprint arXiv:1804.04366 (2018)
  8. 8.
    Yang, H., et al.: Unpaired brain MR-to-CT synthesis using a structure-constrained CycleGAN. In: Stoyanov, D., et al. (eds.) DLMIA/ML-CDS -2018. LNCS, vol. 11045, pp. 174–182. Springer, Cham (2018). Scholar
  9. 9.
    Soussia, M., Rekik, I.: A review on image-and network-based brain data analysis techniques for Alzheimer’s disease diagnosis reveals a gap in developing predictive methods for prognosis. arXiv preprint arXiv:1808.01951 (2018)
  10. 10.
    Pan, S., Hu, R., Long, G., Jiang, J., Yao, L., Zhang, C.: Adversarially regularized graph autoencoder. arXiv preprint arXiv:1802.04407 (2018)
  11. 11.
    Mahjoub, I., Mahjoub, M.A., Rekik, I.: Brain multiplexes reveal morphological connectional biomarkers fingerprinting late brain dementia states. Sci. Rep. 8, 4103 (2018) CrossRefGoogle Scholar
  12. 12.
    Lisowska, A., Rekik, I.: Alzheimer’s disease neuroimaging initiative and others: joint pairing and structured mapping of convolutional brain morphological multiplexes for early dementia diagnosis. Brain Connect. 9, 22–36 (2018)CrossRefGoogle Scholar
  13. 13.
    Raeper, R., Lisowska, A., Rekik, I.: Cooperative correlational and discriminative ensemble classifier learning for early dementia diagnosis using morphological brain multiplexes. IEEE Access 6, 43830–43839 (2018)CrossRefGoogle Scholar
  14. 14.
    Soussia, M., Rekik, I.:Unsupervised manifold learning using high-order morphological brain networks derived from T1-w MRI for autism diagnosis. Front. Neuroinf. 12 (2018)Google Scholar
  15. 15.
    Wang, B., Ramazzotti, D., De Sano, L., Zhu, J., Pierson, E., Batzoglou, S.: SIMLR: a tool for large-scale single-cell analysis by multi-kernel learning. bioRxiv p. 118901 (2017)Google Scholar
  16. 16.
    Breiman, L.: Random forests. Mach. Learn. 45, 5–32 (2001)CrossRefGoogle Scholar
  17. 17.
    Fischl, B.: Freesurfer. Neuroimage 62, 774–781 (2012)CrossRefGoogle Scholar

Copyright information

© Springer Nature Switzerland AG 2019

Authors and Affiliations

  1. 1.LATIS Lab, ISITCOMUniversity of SousseSousseTunisia
  2. 2.BASIRA Lab, Faculty of Computer and InformaticsIstanbul Technical UniversityIstanbulTurkey

Personalised recommendations