Toward neuroimaging-based network biomarkers for transient ischemic attack

Abstract

Stroke is associated with topological disruptions of large-scale functional brain networks. However, whether these disruptions occur in transient ischemic attack (TIA), an important risk factor for stroke, remains largely unknown. Combining multimodal MRI techniques, we systematically examined TIA-related topological alterations of functional brain networks, and tested their reproducibility, structural, and metabolic substrates, associations with clinical risk factors and abilities as diagnostic and prognostic biomarkers. We found that functional networks in patients with TIA exhibited decreased whole-brain network efficiency, reduced nodal centralities in the bilateral insula and basal ganglia, and impaired connectivity of inter-hemispheric communication. These alterations remained largely unchanged when using different brain parcellation schemes or correcting for micro head motion or for regional gray matter volume, cerebral blood flow or hemodynamic lag of BOLD signals in the patients. Moreover, some alterations correlated with the levels of high-density lipoprotein cholesterol (an index related to ischemic attacks via modulation of atherosclerosis) in the patients, distinguished the patients from healthy individuals, and predicted future ischemic attacks in the patients. Collectively, these findings highlight the emergence of characteristic network dysfunctions in TIA, which may aid in elucidating pathological mechanisms and establishing diagnostic and prognostic biomarkers for the disease.

Keywords: brain networks; grapy theory; inter-hemispheric connectivity; multimodal MRI; transient ischemic attack.

Figure 1

Altered global network metrics in TIA. Compared with the HCs, patients with TIA exhibited significantly increased characteristic path length L_p and normalized characteristic path length ${\widehat{L}}_p$, and decreased clustering coefficient C_p, local efficiency E_loc, global efficiency E_glob and normalized global efficiency ${\widehat{E}}_{\text{glob}}$. TIA, transient ischemic attack; HCs, healthy controls [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 2

Impaired interregional functional connectivity in TIA. (a) A single connected component was identified to exhibit decreased functional connectivity in the patients with TIA when using a cluster defining threshold of p < .01. (b) Using a more rigorous cluster‐defining threshold of p < .001, three smaller components were detected to show decreased functional connectivity in the patients with TIA. Notably, the components were mainly composed of inter‐hemispheric connections. Nodal size denotes nodal degree (i.e., the number of connections) in the connected component (bigger for more) and edge width denotes the significance level of between‐group differences (thicker for more significant). See Table S1 for regional abbreviations. TIA, transient ischemic attack [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 3

Relationship between network measures and clinical/physiological/biochemical variables in TIA. In the TIA patients, significant correlations were observed for the high‐density lipoprotein cholesterol (HDL‐C) with characteristic path length L_p, global efficiency E_glob, and nodal degree k_i and nodal efficiency e_i of the left insula. HDL‐C, high‐density lipoprotein cholesterol; TIA, transient ischemic attack [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 4

TIA‐control classification and prediction of follow‐up ischemic attacks. (a) The mean functional connectivity strength of the identified NBS component (cluster defining threshold of p < .01) distinguished TIA patients from HCs with a high sensitivity and specificity. (b) Nodal degree of the right insula predicted follow‐up ischemic attacks with a high sensitivity and specificity. ROC, receiver operating characteristic; AUC, area under curve; TIA, transient ischemic attack; HCs, healthy controls [Color figure can be viewed at http://wileyonlinelibrary.com]