Hyper-connectivity of functional networks for brain disease diagnosis

Abstract

Exploring structural and functional interactions among various brain regions enables better understanding of pathological underpinnings of neurological disorders. Brain connectivity network, as a simplified representation of those structural and functional interactions, has been widely used for diagnosis and classification of neurodegenerative diseases, especially for Alzheimer's disease (AD) and its early stage - mild cognitive impairment (MCI). However, the conventional functional connectivity network is usually constructed based on the pairwise correlation among different brain regions and thus ignores their higher-order relationships. Such loss of high-order information could be important for disease diagnosis, since neurologically a brain region predominantly interacts with more than one other brain regions. Accordingly, in this paper, we propose a novel framework for estimating the hyper-connectivity network of brain functions and then use this hyper-network for brain disease diagnosis. Here, the functional connectivity hyper-network denotes a network where each of its edges representing the interactions among multiple brain regions (i.e., an edge can connect with more than two brain regions), which can be naturally represented by a hyper-graph. Specifically, we first construct connectivity hyper-networks from the resting-state fMRI (R-fMRI) time series by using sparse representation. Then, we extract three sets of brain-region specific features from the connectivity hyper-networks, and further exploit a manifold regularized multi-task feature selection method to jointly select the most discriminative features. Finally, we use multi-kernel support vector machine (SVM) for classification. The experimental results on both MCI dataset and attention deficit hyperactivity disorder (ADHD) dataset demonstrate that, compared with the conventional connectivity network-based methods, the proposed method can not only improve the classification performance, but also help discover disease-related biomarkers important for disease diagnosis.

Keywords: Alzheimer's disease; Classification; Functional MR imaging; Hyper-network.

Fig. 1

Flowchart of the proposed method.

Fig. 2

Hyper-graph vs. graph. Left: A conventional graph in which two nodes are connected together by an edge. Middle: a hyper-graph in which each hyper-edge can connect more than two nodes. Right: The incidence matrix for the hyper-graph in the middle.

Fig. 3

ROC curves of the compared methods for MCI classification.

Fig. 4

The important ROIs selected by the proposed method for MCI classification.

Fig. 5

The average hyper-edges for NC (left) and MCI (right) groups based on 8 ROIs listed in Table 4 with λ=0.3. Here, each sub-figure denotes a hyper-edge constructed based on the corresponding ROI, where all nodes in each sub-figure form a hyper-edge, the red node (i.e., centroid node linked by other nodes) in each sub-figure represents the ROI used for constructing the hyper-edge, and the green nodes (i.e., nodes lying between left hemisphere and right hemisphere) represent the corresponding ROIs coming from cerebellum.

Fig. 6

Visualization on p-values on connection between ROIs. Here, color denotes the corresponding p-value. (L.MFG = left Middle frontal gyrus, R.MFG = right Middle frontal gyrus, L.ORBinf = left Orbitofrontal cortex (inferior), L.OLF = left Olfactory, R.PHG = right ParaHippocampal gyrus, R.AMYG = right Amygdala, R.LING = right Lingual gyrus, R.FFG = right Fusiform gyrus, L.PCL = left Paracentral lobule, R.PUT = right Putamen, L.TPOmid = left Temporal pole (middle), L.ITG = left Inferior temporal, L.CIICH = Left crus II of cerebellar hemisphere, R.LIVVCH = Right lobule IV, V of cerebellar hemisphere, LVIV = Lobule VI of vermis.).

Fig. 7

Average connectivity weights for MCI and NC groups and their differences. Colors in (a) and (b) represent the average connectivity weight of MCI and NC groups respectively, while colors in (c) represent difference of connectivity weights between MCI and NC.

Fig. 8

The R values for all subjects.

Fig. 9

ROC curves of six different methods on ADHD classification.

Fig. 10

Classification accuracy w.r.t. the use of different number of λ values.

Fig. 11

The p-value on three clustering coefficients from hyper-network with λ= 0.1 and other groups of λ values.

Fig. 12

The classification accuracy w.r.t. the selections of β and γ values.

Fig. 13

Classification results of proposed method when using different combinations of coefficient weights.

Fig. 14

The computation cost of each step of our proposed method with different numbers of functional atlases.

Fig. 15

Changes of ICC values of three types of clustering coefficients w.r.t. the different number of λ values.