Exploring Heterogeneity in Vestibular Migraine Using Individualized Differential Structural Covariance Network Analysis

Abstract

Background: The high heterogeneity in vestibular migraine (VM) complicates understanding its precise pathophysiological mechanisms and identifying potential biomarkers. This study investigated the heterogeneity in VM using a newly proposed method called Individualized Differential Structural Covariance Network (IDSCN) analysis.

Methods: Structural T1-weighted MRI scans were performed on 55 patients with VM and 65 healthy controls, and an IDSCN was constructed for each patient. We studied the extent of heterogeneity in the IDSCNs, summarized the distribution of differential edges, and clustered the patients into subtypes with the shared differential edges. Imaging-clinical association analyses were conducted on both the subtype classification and the differential edges exhibiting significant inter-subtype differences.

Results: Patients with VM demonstrated notable heterogeneity in the number of significantly altered IDSCN edges, while sharing several common differential connections that were mainly distributed among the parietal, subcortical, and cerebellar regions. Two robust and distinct neuroanatomical subtypes of VM were identified, which were associated with headache frequency. The differential edge between the left paracentral lobule and right pallidum was associated with both headache frequency and occurrence.

Conclusions: These findings indicate the importance of considering individual differences in VM research and may offer insights for precise diagnosis and individualized treatment of the disease.

Keywords: gray matter; heterogeneity; individualized analysis; magnetic resonance imaging; structural covariance network; vestibular migraine.

FIGURE 1

Summary of the study design and analysis pipeline. (1) The top purple panel shows the IDSCN construction workflow. A reference SCN was initially built using the GM density data from all controls. The integration of VM k into the HC group generated a new SCN, known as the perturbed network. IDSCN for VM k was determined by calculating the Z‐score of the difference between the perturbed network and reference network. (2) The middle green panel shows the clustering process. The VM cohort was clustered into two subgroups (i.e., VM I and II) using the k‐means method, with the top 80 differential edges treated as features. (3) The bottom red panel illustrates the data analysis procedures, including investigations on the heterogeneity of IDSCN, inter‐subtype differences in differential edges, and auxiliary analyses (i.e., imaging–clinical association, functional annotation, subgrouping, and reproducibility analyses). GM, gray matter; HC, healthy control; IDSCN, individualized differential structural covariance network; SCN, structural covariance network; VM, vestibular migraine; VM I, subtype 1 of VM; VM II, subtype 2 of VM.

FIGURE 2

Heterogeneity of IDSCN in VM. (A) In the VM cohort, the number of significantly altered edges exhibited a large variation among patients. (B) The two subtypes of VM had different general patterns of IDSCN. The illustrated general patterns were obtained by calculating the mean maps for each group, where the axes represent the indexes of the brain regions. IDSCN, individualized differential structural covariance network; VM, vestibular migraine; VM I, subtype 1 of VM; VM II, subtype 2 of VM.

FIGURE 3

Illustrations of differential edges with significant differences between the two VM subtypes. (A) The identified edges mapped on the Ch2 template can be classified into two classes with opposite directions of change between subtypes. Subtype 1 demonstrated higher Z‐scores than subtype 2 in the red edges and lower Z‐scores than subtype 2 in the green edges. The thickness of the lines is proportional to the weight of the connections. CAU.R and PCL.L are the core nodes in the identified subnetwork (with a degree of ≥ 2). (B) The connectogram provides a comprehensive representation of the nodes and edges in the subnetwork (red chords: Subtype 1 > subtype 2; green chords: Subtype 1 < subtype 2). Sex, age, education level, and TIV were controlled as confounding covariates. AMYG, amygdala; CAU, caudate; DCG, median cingulate and paracingulate gyri; HES, Heschl gyrus; L, left; MOG, middle occipital gyrus; ORBinf, orbital part of the inferior frontal gyrus; PAL, pallidum; PCL, paracentral lobule; PoCG, postcentral gyrus; PUT, putamen; R, right; ROL, Rolandic operculum; SMG, supramarginal gyrus; TIV, total intracranial volume; VM, vestibular migraine.