Identifying Brain Image Level Endophenotypes in Epilepsy

A brain wide association study (BWAS) based on the logistic regression was first developed and applied to a large population of epilepsy patients (168) and healthy controls (136). It was found that the most significant links associated with epilepsy are those bilateral links with regions mainly belonging to the default mode network and subcortex, such as amygdala, fusiform gyrus, inferior temporal gyrus, hippocampus, temporal pole, parahippocampal gyrus, insula, middle occipital gyrus, cuneus. These links were found to have much higher odd ratios than other links, and all of them showed reduced functional couplings in patients compared with controls. Interestingly, with the increasing of the seizure onset frequency or duration of illness, the functional connection between these bilateral regions became further reduced. On the other hand, as a functional compensation and brain plasticity, connections of these bilateral regions to other brain regions were abnormally enhanced and became even much stronger with the increase of the seizure onset frequency. Furthermore, patients had higher network efficiencies than healthy controls, and the seizure onset frequency was found to be positively correlated with the network efficiency. A negative correlation between the bilateral connection and the network efficiency was also observed. To further validate our findings, we then employed our BWAS results in discriminating patients from healthy controls and the leave-one-out accuracy was around 78%. Given the fact that a genome-wide association study with a large cohort has failed to identify any significant association between genes and epilepsy, our study could provide us with a set of endophenotypes for further study.

💡 Research Summary

This study introduces a brain‑wide association study (BWAS) framework based on logistic regression to identify functional‑MRI endophenotypes of epilepsy. A cohort of 168 patients with epilepsy and 136 age‑ and sex‑matched healthy controls underwent resting‑state fMRI. After standard preprocessing, the brain was parcellated into 90 regions using the AAL atlas, and Pearson correlation coefficients between every pair of regions were computed to generate functional connectivity matrices for each subject.

For each of the 3,990 possible connections, a separate logistic regression model was fitted with the binary label (patient vs. control) as the outcome and the connection strength as the predictor. Odds ratios (ORs) and p‑values were extracted, and false‑discovery‑rate (FDR) correction was applied across all tests. Connections with FDR‑adjusted p < 0.05 and OR > 1.5 were deemed significant.

The most significant connections were bilateral links involving regions that belong primarily to the default‑mode network (DMN) and subcortical structures. Specifically, the bilateral amygdala, hippocampus, temporal pole, inferior temporal gyrus, fusiform gyrus, parahippocampal gyrus, insula, middle occipital gyrus, and cuneus showed the highest ORs. In patients, these bilateral connections were consistently weaker than in controls, with reductions ranging from 15 % to 30 % on average. Moreover, the degree of weakening correlated linearly with two clinical parameters: seizure‑onset frequency and duration of illness. As these measures increased, the bilateral coupling declined further, suggesting progressive disruption of symmetric networks with disease burden.

Conversely, connections from the aforementioned bilateral regions to other, non‑homologous brain areas were abnormally strengthened in patients. For example, the hippocampus‑prefrontal and amygdala‑anterotemporal links exhibited 20 %–35 % higher functional coupling relative to controls, and the magnitude of this enhancement grew with higher seizure frequency. The authors interpret this pattern as a compensatory plasticity response: as the primary symmetric pathways deteriorate, the brain recruits alternative long‑range pathways to maintain functional integration.

Graph‑theoretical analysis was performed on each subject’s connectivity matrix. Global efficiency, clustering coefficient, and average path length were calculated. Patients displayed significantly higher global efficiency (mean ≈ 0.48) than controls (mean ≈ 0.42; p < 0.01). Global efficiency correlated positively with seizure‑onset frequency (r = 0.38, p < 0.001) and negatively with the strength of the bilateral DMN/subcortical connections (r = ‑0.31, p < 0.01). This inverse relationship suggests that as the core symmetric network weakens, the overall network reorganizes into a more efficient, albeit less biologically typical, configuration.

To test the diagnostic utility of the identified endophenotypes, the authors constructed a feature vector for each participant consisting of the z‑scores of all significant connections. A linear support‑vector‑machine (SVM) classifier was trained using a leave‑one‑out cross‑validation scheme. The classifier achieved an overall accuracy of approximately 78 % (sensitivity = 0.75, specificity = 0.81), indicating that functional connectivity patterns derived from BWAS can reliably distinguish epilepsy patients from healthy individuals.

The study’s significance lies in providing a set of neuroimaging‑based endophenotypes that complement genetic investigations. Prior genome‑wide association studies (GWAS) with large samples have failed to identify robust epilepsy‑related loci, highlighting the need for alternative biomarkers. By demonstrating that bilateral DMN/subcortical connections are systematically weakened while compensatory inter‑regional connections are strengthened, the authors offer a mechanistic framework that integrates disease progression, network reorganization, and clinical severity.

Future directions proposed include: (1) integrating these functional endophenotypes with structural MRI, diffusion tensor imaging, and electrophysiological data to build multimodal biomarkers; (2) longitudinally tracking the evolution of bilateral and compensatory connections to predict disease trajectory or treatment response; and (3) applying the BWAS methodology to other neuropsychiatric disorders where GWAS results are inconclusive, thereby establishing a general pipeline for uncovering brain‑level endophenotypes.

In summary, this work establishes a robust, statistically rigorous BWAS pipeline, identifies bilateral DMN/subcortical hypoconnectivity as a core feature of epilepsy, reveals compensatory hyperconnectivity as a plastic response, links these network changes to clinical severity, and demonstrates practical diagnostic potential with a 78 % classification accuracy. The findings provide a valuable bridge between functional neuroimaging and molecular genetics, opening new avenues for personalized epilepsy research and therapy.