Matched signal detection on graphs: Theory and application to brain imaging data classification
Graphical abstract
Introduction
Matched subspace detection is a classic tool that determines whether a multidimensional signal lies in a given linear subspace or not (Scharf and Friedlander, 1994). It has achieved great success in applications such as radar, hyperspectral imaging (Manolakis and Shaw, 2002) and medical imaging (Li et al., 2009). The subspace is either known from the physical system that generates the signal, or can be inferred from training data. Subspace learning is a natural way of data dimension reduction and can be achieved by principal component analysis (PCA), which projects the original data to a linear subspace spanned by the leading eigenvectors of the covariance matrix (Jolliffe, 2005). While a common assumption of PCA is that the data come from a linear subspace, many real data are lying in or close to a nonlinear manifold, which is a topological space that resembles Euclidean space around each point (Belkin and Niyogi, 2003). Examples of the latter case include brain images (Liu et al., 2013), genetic data (Lee et al., 2008), social network records, and sensor network measurements. In this setting, the low-dimensional subspace that best preserves the intrinsic geometry of the data can be effectively learned by graph spectral methods, e.g., isomap, locality linear embedding (LLE), Laplacian eigenmaps (Belkin and Niyogi, 2003, Roweis and Saul, 2000, Saul et al., 2006, Tenenbaum et al., 2000).
In neuroimaging, as more and more nonlinear data are collected by multiple imaging modalities, there is a need for classifying data with complex intrinsic structures. For instance, the analysis and classification of positron emission tomography (PET) images or functional magnetic resonance imaging (fMRI) data may facilitate the prediction and early detection of Alzheimer's disease (AD). Concurrently, an emerging area of signal processing on graphs is developed for handling these challenging data through the combination of algebraic and spectral graph theoretic concepts with computational harmonic analysis (Shuman et al., 2013). Signals are assumed to reside on vertices of weighted graphs which are often naturally defined by the application. The weight associated with a certain edge in the graph represents the similarity between the two vertices joined by the edge. We refer to graph supported data as graph-signals, to differentiate them from conventional signals in Euclidean spaces. In the brain imaging classification, we could view the PET/fMRI data as graph-signals on weighted graphs describing the affinity between each pair of brain regions.
Motivated by the above data classification requirement, we are interested in developing a detection framework for graph-signals. Specifically, we formulate several hypotheses to decide which graph structure is more likely to match a given signal. Moreover, we exploit the matched subspace detection technique and propose different types of graph-signal models to make our framework generic to deal with a variety of real situations. The subspace for graph-signal is formed by eigenvectors of the Laplacian matrix L of the graph. The graph Laplacian matrix encodes the structure of the graph concisely. From a graph signal processing point-of-view, eigenvectors of L could be treated as the generalization of the basis of conventional Fourier transform (Agaskar and Lu, 2013, Sandryhaila and Moura, 2013, Shuman et al., 2013). Based on spectral graph theory, we can define the variation of graph-signals. It follows that the variation of an eigenvector of L is equal to the associated eigenvalue (Hu et al., 2015a). When we decompose a signal containing instantaneous fMRI measurements into linear combination of eigenvectors of Laplacian matrix associated with the brain-region-affinity graph, we might deem that the components in those eigenvectors with larger eigenvalues as being noisier if the true fMRI signal is assumed to be bandlimited on the graph (Gadde et al., 2013, Kim et al., 2013, Meyer and Shen, 2014).
Our first hypothesis test model simply assumes that the signal lies in a subspace spanned by the first few Laplacian eigenvectors corresponding to smaller eigenvalues. The traditional matched subspace detection could be applied directly to this case. Furthermore, we consider two categories of graph-signal models: deterministic signals with constraints and probabilistic signals with prior distributions. For deterministic signals, we impose a bounded variation on the signal with respect to the graph. The penalized maximum likelihood estimator (MLE) of the true signal is derived by solving a constrained optimization problem. We find that the test is a weighted energy detector. For probabilistic signals, when we choose a certain degenerate Gaussian distribution as the prior of the projection coefficients of the signal onto the graph Laplacian eigenvectors, the decision ends up comparing the signal variations on the two hypothetic graphs in a noise-free case.
We evaluate the effectiveness of the matched signal detection (MSD) theory on both synthetic and real data sets. Simulations based on randomly generated graphs demonstrate the feasibility of our approaches even if we do not know the exact probability distributions of the testing signals. Then, we apply the proposed detection algorithms to brain imaging classification tasks of AD. As one of the most prevalent forms of dementia, AD is believed to be a brain network associated disease (Gomez-Ramirez and Wu, 2014, Raj et al., 2012, Sepulcre et al., 2013), and is characterized by progressive impairment of memory and other cognitive capacity. It affects nearly 36 million people worldwide with an expected number of cases to be 65.7 million by 2030 (Brookmeyer et al., 2007). The development of neuroimaging classification techniques may enable us to monitor the functional and anatomical changes of the brain in vivo and discover reliable biomarkers for identifying AD at an early stage. In this study, we have compared a novel MSD approach with other widely used methods including principle component analysis (PCA), support vector machine (SVM) and linear discriminant analysis (LDA) on two data sets: one is PET imaging of brain amyloid using Pittsburgh compound-B (PIB) tracer of AD and normal control (NC) subjects; the other contains resting-state fMRI (R-fMRI) images of early mild cognitive impairment (EMCI) and NC subjects in the Alzheimer's Disease Neuroimaging Initiative (ADNI) database. For the MSD, we compute the similarity between each of two brain regions with the Gaussian radial basis function (RBF) kernel. This simple way of building brain networks avoids estimating network structures by solving inverse problems, which often requires more data; yet the weighted graphs associated with the networks approximate the data manifolds. Experimental results show that when using the MSD on graphs, we can achieve significantly better classification performance than the compared algorithms. The results indicate that our method provides an effective way for brain imaging classification, probably due to the capability of exploiting the manifold structure of the data.
Our contributions in this paper are three-fold: first, we have developed a matched signal detection theory for graph-signals which are ubiquitous in medical imaging applications; second, we keep the framework generic and simple by proposing a variety of signal models and using simple similarity metrics to construct graphs; third, we demonstrate that the detection theory is particularly suitable for neuroimaging classifications.
Section snippets
Theory
To formulate the framework of matched signal detection on graphs, we first introduce the concept of graph-signals. We extend the traditional Fourier transform to a graph Fourier transform and define a notion of graph-signal frequency based upon spectral graph theory. Then, to model different real data, we propose three classes of signal models on graphs. Finally, we derive the signal detection criterion under each signal model.
Numerical simulations
We have evaluated the proposed MSD rules on small-world networks, which are simplified yet effective models of brain connectivity structures (Bassett and Bullmore, 2006, He et al., 2007, Stam et al., 2007). A small-world network is characterized by dense local cliques of connections between neighboring vertices and a short path length between any pair of vertices due to the existence of few long-range connections. Brain anatomical and functional networks have a small-world topology that
Assumption of smooth graph-signals
The introduced MSD schemes were based on the smoothness of the imaging data on brain connectivity networks. We justify this assumption as follows. Through large data analysis, it has been demonstrated that the propagation of disease agents of AD obeys a network diffusion model (Raj et al., 2012, Zhou et al., 2012). By using linear dynamics defined over the brain network, Raj et al. predicted spatially distinct “persistent modes” of different types of dementia accurately. Meanwhile, the
Conclusion
In this paper, we formulated the MSD for graph-structure data to classify brain imaging data. We adopted the bandlimited, constrained, and probabilistic graph-signal models to capture the smoothness of the imaging data on brain connectivity networks. We found that GLRT statistics derived under these three signal models were weighted energy detectors in general. The effectiveness of the MSD was demonstrated through simulations and real experiments. Specifically, we applied it to two
Conflict of interest
We confirm that there are no known conflicts of interest associated with this publication.
Acknowledgments
This research was funded by the National Institutes of Health grants R01 EB013293, 1 K23 EB19023-01, the National Institute on Aging Grants P01 AG036694, R01 AG034556, R01 AG037497, and the Alzheimers Association Zenith Award.
References (62)
- et al.
Whole cortical and default mode network mean functional connectivity as potential biomarkers for mild Alzheimer's disease
Psychiatry Res. Neuroimaging
(2014) - et al.
In vivo mapping of gray matter loss with voxel-based morphometry in mild Alzheimer's disease
NeuroImage
(2001) - et al.
Large-scale brain networks in cognition: emerging methods and principles
Trends Cogn. Sci.
(2010) - et al.
Forecasting the global burden of Alzheimers disease
Alzheimers Dement.
(2007) - et al.
Whole brain high-resolution functional imaging at ultra high magnetic fields: an application to the analysis of resting state networks
NeuroImage
(2011) - et al.
PET imaging of amyloid deposition in patients with mild cognitive impairment
Neurobiol. Aging
(2008) - et al.
Nonlinear responses in fMRI: the balloon model, volterra kernels, and other hemodynamics
NeuroImage
(2000) - et al.
Cingulate cortex hypoperfusion predicts Alzheimer's disease in mild cognitive impairment
BMC Neurol.
(2002) - et al.
Locally linear embedding (LLE) for MRI based Alzheimer's disease classification
NeuroImage
(2013) - et al.
A lipophilic thioflavin-T derivative for positron emission tomography (PET) imaging of amyloid in brain
Bioorg. Med. Chem. Lett.
(2002)
Perturbation of the eigenvectors of the graph laplacian: application to image denoising
Appl. Comput. Harmon. Anal.
A network diffusion model of disease progression in dementia
Neuron
Decoding brain states from fMRI connectivity graphs
NeuroImage
Whole-brain anatomical networks: does the choice of nodes matter?
NeuroImage
Predicting regional neurodegeneration from the healthy brain functional connectome
Neuron
A spectral graph uncertainty principle
IEEE Trans. Inf. Theory
Patterns of interregional correlations of cerebral glucose metabolic rates in patients with dementia of the Alzheimer type
Neurodegeneration
Small-world brain networks
Neuroscientist
Laplacian eigenmaps for dimensionality reduction and data representation
Neural Comput.
Convex Optimization
Development of Alzheimer-related neurofibrillary changes in the neocortex inversely recapitulates cortical myelogenesis
Acta Neuropathol.
DPARSF: a MATLAB toolbox for “pipeline” data analysis of resting-state fMRI
Front. Syst. Neurosci.
Classification of Alzheimer disease, mild cognitive impairment, and normal cognitive status with large-scale network analysis based on resting-state functional MR imaging
Radiology
Spectral graph theory
Am. Math. Soc.
An Introduction to Support Vector Machines and Other Kernel-based Learning Methods
Fitting a graph to vector data. Proceedings of the 26th Annual International Conference on Machine Learning
High resolution pet imaging of tau using an MR-based information theoretic anatomical prior
J. Nucl. Med.
Bilateral Filter: Graph Spectral Interpretation and Extensions
Revisiting the Nystrom Method for Improved Large-scale Machine Learning
Network-based biomarkers in Alzheimer's disease: review and future directions
Front. Aging Neurosci.
Small-world anatomical networks in the human brain revealed by cortical thickness from MRI
Cereb. Cortex
Cited by (40)
Graph learning from band-limited data by graph Fourier transform analysis
2023, Signal ProcessingObserving and tracking bandlimited graph processes from sampled measurements
2020, Signal ProcessingCitation Excerpt :Sampling theory for graph signals assumes the latter are often sparse (or bandlimited) in the graph frequency domain. This assumption is satisfied in temperature measurements in sensor networks [8], fMRI data in brain networks [12], and user ratings in recommendation networks [13], to name a few. The works in [8,10,11,14,15] exploit signal bandlimitedness to propose sampling and reconstruction strategies for graph signals.
Patterns of cortical interactivity supporting speech production and lexical retrieval: A graph signal processing approach at the individual level
2020, Journal of NeurolinguisticsCitation Excerpt :The analysis of the entire graph and signals on the graph is called GSP. GSP has gained significant interest in the signal processing community with potential applications in many different areas including neuroscience (Hu et al., 2016; Jabłoński, 2017; Ménoret, Farrugia, Pasdeloup, & Gripon, 2017; Sandryhaila & Moura, 2014). One of the important concepts in GSP is Graph Learning (GL), where the goal is to find the underlying graph connections and their weights from the observed signals at the vertices.
Graph-time spectral analysis for atrial fibrillation
2020, Biomedical Signal Processing and ControlCitation Excerpt :Graph-time signal processing: Graphs are natural tools to model data living in high-dimensional and irregular domains [17]. Graph signal processing provides a harmonic analysis for signals residing on the graph vertices and has been applied to brain signal analysis, Alzheimer classification, and body motion [18–24]. However, despite showing promise, graph signal processing is still unexplored for heart-related problems.
Distance disintegration delineates the brain connectivity failure of Alzheimer's disease
2020, Neurobiology of AgingCitation Excerpt :Using graph theory–based methods, previous rs-fMRI studies have shown AD-related alterations in measures of network segregation (Brier et al., 2014b; Deng et al., 2016; Kim et al., 2015; Sun et al., 2014; Supekar et al., 2008; Toussaint et al., 2014; Xiang et al., 2013), integration (Deng et al., 2016; Kim et al., 2015; Minati et al., 2014; Sanz-Arigita et al., 2010; Wang et al., 2013; Xiang et al., 2013), modularity (Brier et al., 2014b; Sun et al., 2014), and centrality (Dai et al., 2015; Kim et al., 2015; McCarthy et al., 2014; Toussaint et al., 2014). Furthermore, graph theory metrics have been demonstrated as strong classifier variables for distinguishing individuals across the AD spectrum such as distinguishing MCI individuals who progress to AD (Hojjati et al., 2017; Hu et al., 2016; Khazaee et al., 2015). However, most graph theory metrics investigating how brain communication is broken down in AD are often based on the direct strengths or the shortest paths connecting nodes, overlooking the many indirect routes by which information flow is spread in the brain.