ABSTRACT
Mapping the human brain, or understanding how certain brain regions relate to specific aspects of cognition, has been and remains an active area of neuroscience research. Functional magnetic resonance imaging (fMRI) data---in the form of images, time series or graphs---are central in this research, but pose many challenges in phenotype prediction tasks (e.g., noisy, small training samples). Standardly employed handcrafted models and newly proposed neural network methods pose limitations in the expressive power and interpretability, respectively, in this context. In this work focusing on fMRI-derived brain graphs, a modality that partially handles some challenges of fMRI data, we propose a grouping-based interpretable neural network model, GroupINN, that effectively classifies cognitive performance with 85% fewer model parameters than baseline deep models, while also identifying the most predictive brain subnetworks within several task-specific contexts. Our method incorporates the idea of node grouping into the design of the neural network. That way, unlike other methods that employ clustering as a preprocessing step to reorder nodes, GroupINN learns the node grouping and extracts graph features jointly. Experiments on task-based fMRI datasets show that our method is $2.6-69\times$ faster than other deep models, while achieving comparable or better accuracy and providing interpretability.
- Ariana Anderson, Pamela K Douglas, Wesley T Kerr, Virginia S Haynes, Alan L Yuille, Jianwen Xie, Ying Nian Wu, Jesse A Brown, and Mark S Cohen. Nonnegative matrix factorization of multimodal mri, fmri and phenotypic data reveals differential changes in default mode subnetworks in adhd. NeuroImage, 102:207-- 219, 2014.Google ScholarCross Ref
- Karsten M. Borgwardt and Hans-Peter Kriegel. Shortest-path kernels on graphs. In ICDM, pages 74--81. IEEE, 2005. Google ScholarDigital Library
- Ed Bullmore and Olaf Sporns. Complex brain networks: graph theoretical analysis of structural and functional systems. Nature Reviews Neuroscience, 10(3):186, 2009.Google ScholarCross Ref
- Edward T Bullmore and Danielle S Bassett. Brain graphs: graphical models of the human brain connectome. Annu Rev Clin Psychol, 7:113--140, 2011.Google ScholarCross Ref
- Jessica R Cohen and Mark D'Esposito. The segregation and integration of distinct brain networks and their relationship to cognition. J Neurosci, 36(48):12083--12094, 2016.Google ScholarCross Ref
- Michael W Cole, Danielle S Bassett, Jonathan D Power, Todd S Braver, and Steven E Petersen. Intrinsic and task-evoked network architectures of the human brain. Neuron, 83(1):238--251, 2014.Google ScholarCross Ref
- MichaelWCole andWalter Schneider. The cognitive control network: integrated cortical regions with dissociable functions. Neuroimage, 37(1):343--360, 2007.Google ScholarCross Ref
- Elizabeth N Davison, Kimberly J Schlesinger, Danielle S Bassett, Mary-Ellen Lynall, Michael B Miller, Scott T Grafton, and Jean M Carlson. Brain network adaptability across task states. PLoS comput biol, 11(1):e1004029, 2015.Google Scholar
- Ian J Deary, Lars Penke, and Wendy Johnson. The neuroscience of human intelligence differences. Nat rev neurosci, 11(3):201, 2010.Google Scholar
- Michaël Defferrard, Xavier Bresson, and Pierre Vandergheynst. Convolutional neural networks on graphs with fast localized spectral filtering. In NeurIPS, 2016. Google ScholarDigital Library
- Bruce Fischl, David H Salat, Evelina Busa, Marilyn Albert, Megan Dieterich, Christian Haselgrove, Andre Van Der Kouwe, Ron Killiany, David Kennedy, Shuna Klaveness, et al. Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain. Neuron, 33(3):341--355, 2002.Google ScholarCross Ref
- Aapo Hyvärinen, Juha Karhunen, and Erkki Oja. Independent component analysis, volume 46. John Wiley & Sons, 2004.Google Scholar
- Rex E Jung and Richard J Haier. The parieto-frontal integration theory (p-fit) of intelligence: converging neuroimaging evidence. Behavioral and Brain Sciences, 30(2):135--154, 2007.Google ScholarCross Ref
- Thomas N. Kipf and Max Welling. Semi-supervised classification with graph convolutional networks. In ICLR, 2017.Google Scholar
- Risi Kondor and Horace Pan. The multiscale laplacian graph kernel. In NeurIPS, 2016. Google ScholarDigital Library
- Danai Koutra and Christos Faloutsos. Individual and Collective Graph Mining: Principles, Algorithms, and Applications. Synthesis Lectures on Data Mining and Knowledge Discovery. Morgan & Claypool Publishers, 2017. Google ScholarDigital Library
- Danai Koutra, Yu Gong, Sephira Ryman, Rex Jung, Joshua Vogelstein, and Christos Faloutsos. Are all brains wired equally. In OHBM, 2013.Google Scholar
- Danai Koutra, Joshua T Vogelstein, and Christos Faloutsos. Deltacon: A principled massive-graph similarity function. In SDM, pages 162--170. SIAM, 2013.Google ScholarCross Ref
- Nils M Kriege, Pierre-Louis Giscard, and Richard Wilson. On valid optimal assignment kernels and applications to graph classification. In NeurIPS, 2016. Google ScholarDigital Library
- Jure Leskovec and Christos Faloutsos. Sampling from large graphs. In KDD, pages 631--636. ACM, 2006. Google ScholarDigital Library
- Martin A Lindquist et al. The statistical analysis of fmri data. Statistical science, 23(4):439--464, 2008.Google ScholarCross Ref
- Huan Liu, Mianzhi Zhang, Xintao Hu, Yudan Ren, Shu Zhang, Junwei Han, Lei Guo, and Tianming Liu. Fmri data classification based on hybrid temporal and spatial sparse representation. In ISBI, pages 957--960. IEEE, 2017.Google ScholarCross Ref
- Yike Liu, Tara Safavi, Abhilash Dighe, and Danai Koutra. Graph summarization methods and applications: A survey. CSUR, 51(3):62, 2018. Google ScholarDigital Library
- Aljoscha C Neubauer, Andreas Fink, and Dietmar G Schrausser. Intelligence and neural efficiency: The influence of task content and sex on the brain--iq relationship. Intelligence, 30(6):515--536, 2002.Google ScholarCross Ref
- Mark EJ Newman. Modularity and community structure in networks. PNAS, 103(23):8577--8582, 2006.Google ScholarCross Ref
- Jonathan D Power, Alexander L Cohen, Steven M Nelson, Gagan S Wig, Kelly Anne Barnes, Jessica A Church, Alecia C Vogel, Timothy O Laumann, Fran M Miezin, Bradley L Schlaggar, et al. Functional network organization of the human brain. Neuron, 72(4):665--678, 2011.Google ScholarCross Ref
- Tara Safavi, Chandra Sripada, and Danai Koutra. Scalable hashing-based network discovery. In ICDM, pages 405--414. IEEE, 2017.Google ScholarCross Ref
- Saman Sarraf and Ghassem Tofighi. Deep learning-based pipeline to recognize alzheimer's disease using fmri data. In FTC, pages 816--820. IEEE, 2016.Google ScholarCross Ref
- Yiru Shen, Chen Feng, Yaoqing Yang, and Dong Tian. Mining point cloud local structures by kernel correlation and graph pooling. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 4548--4557, 2018.Google ScholarCross Ref
- N. Shervashidze, S. V. N. Vishwanathan, T. Petri, K. Mehlhorn, and K. Borgwardt. Efficient graphlet kernels for large graph comparison. In AISTATS, volume 5, pages 488--495. JMLR: W&CP, 2009.Google Scholar
- Stephen M Smith, Aapo Hyvärinen, Gaël Varoquaux, Karla L Miller, and Christian F Beckmann. Group-pca for very large fmri datasets. Neuroimage, 101:738-- 749, 2014.Google ScholarCross Ref
- Olaf Sporns, Giulio Tononi, and Rolf Kötter. The human connectome: a structural description of the human brain. PLoS computational biology, 1(4):e42, 2005.Google Scholar
- Chandra Sripada, Mike Angstadt, Saige Rutherford, Daniel Kessler, Yura Kim, Mike Yee, and Liza Levina. Fundamental units of inter-individual variation in resting state connectomes. bioRxiv, page 326082, 2018.Google Scholar
- Cornelis J Stam and Jaap C Reijneveld. Graph theoretical analysis of complex networks in the brain. Nonlinear biomedical physics, 1(1):3, 2007.Google Scholar
- Wei Tang, Zhengdong Lu, and Inderjit S Dhillon. Clustering with multiple graphs. In ICDM, pages 1016--1021. IEEE, 2009. Google ScholarDigital Library
- Kamil Ugurbil and David Van Essen. Hcp young adult. https://www. humanconnectome.org/study/hcp-young-adult, 2017.Google Scholar
- Heather L Urry, Carien M van Reekum, Tom Johnstone, and Richard J Davidson. Individual differences in some (but not all) medial prefrontal regions reflect cognitive demand while regulating unpleasant emotion. Neuroimage, 47(3):852-- 863, 2009.Google ScholarCross Ref
- Martijn P van den Heuvel, Cornelis J Stam, René S Kahn, and Hilleke E Hulshoff Pol. Efficiency of functional brain networks and intellectual performance. J Neurosci, 29(23):7619--7624, 2009.Google ScholarCross Ref
- S. V. N. Vishwanathan, Nicol N. Schraudolph, Risi Imre Kondor, and Karsten M. Borgwardt. Graph kernels. JMLR, 11:1201--1242, 2010. Google ScholarDigital Library
- Qi Wang, Mengying Sun, Liang Zhan, Paul Thompson, Shuiwang Ji, and Jiayu Zhou. Multi-modality disease modeling via collective deep matrix factorization. In KDD, pages 1155--1164. ACM, 2017. Google ScholarDigital Library
- Shen Wang, Lifang He, Bokai Cao, Chun-Ta Lu, Philip S. Yu, and Ann B. Ragin. Structural deep brain network mining. In KDD, pages 475--484. ACM, 2018. Google ScholarDigital Library
- Yujun Yan, Mark Heimann, Di Jin, and Danai Koutra. Fast flow-based random walk with restart in a multi-query setting. In Proceedings of the 2018 SIAM International Conference on Data Mining, pages 342--350. SIAM, 2018.Google ScholarCross Ref
- Yaoqing Yang, Chen Feng, Yiru Shen, and Dong Tian. Foldingnet: Point cloud auto-encoder via deep grid deformation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 206--215, 2018.Google ScholarCross Ref
- Zhitao Ying, Jiaxuan You, Christopher Morris, Xiang Ren, Will Hamilton, and Jure Leskovec. Hierarchical graph representation learning with differentiable pooling. In NeurIPS, pages 4800--4810, 2018. Google ScholarDigital Library
- Muhan Zhang, Zhicheng Cui, Marion Neumann, and Yixin Chen. An end-to-end deep learning architecture for graph classification. In AAAI, 2018.Google ScholarCross Ref
Index Terms
- GroupINN: Grouping-based Interpretable Neural Network for Classification of Limited, Noisy Brain Data
Recommendations
PET and fMRI studies of the neural basis of speech perception
Special issue on the nature of speech perception (the psychophysics of speech perception III)Functional imaging (PET and fMRI) has made great advances in our understanding of the neural basis of auditory processing and speech perception. Here we review perceptual and sensory processing aspects of speech and hearing, as revealed by functional ...
Beyond topographic representation: Decoding visuospatial attention from local activity patterns in the human frontal cortex
The ability to detect where a person is attending is fundamental for brain-computer-interfaces. We explore how the attentional focus can be decoded from brain signals noninvasively acquired with functional magnetic resonance imaging (fMRI). Several ...
Decoding visual brain states from fMRI using an ensemble of classifiers
Decoding perceptual or cognitive states based on brain activity measured using functional magnetic resonance imaging (fMRI) can be achieved using machine learning algorithms to train classifiers of specific stimuli. However, the high dimensionality and ...
Comments