INTRODUCTION

Numerous neuroimaging studies of epilepsy have been conducted for decades, but results have been inconsistent (1, 2). This may be due to neurobiological heterogeneity in the illness, which could preclude clinical translations and genetic research. While different imaging modalities were used previously, structural magnetic resonance imaging (MRI) is of fundamental importance to the diagnosis and treatment of epilepsy and strongly recommended (2). Therefore, studying brain structure in epilepsy patients and using anatomical brain measures to differentiate discrete patient subgroups biologically and establishing imaging biomarkers associated with seizure onset is an important step to both facilitate mechanistic understanding of the illness and to develop individualized care strategies to improve outcomes.

METHODS

This study was approved by the ethics committee of West China Second University Hospital, Sichuan University. All study participants and their legal guardians provided written informed consent/assent after study procedures were fully explained. Seventy-two patients, as well as 39 healthy controls, were consecutively recruited from Neurology Clinics in the Division of West China Second Hospital, Sichuan University according to International League Against Epilepsy (ILAE) criteria (https://www.epilepsydiagnosis.org) for the diagnosis of generalized epilepsy and focal epilepsy. All patients underwent electroencephalogram (EEG), the Structured Clinical Interview of Diagnostic and Statistical Manual of Mental Disorders (5th Edition, DSM-V) for children with attention-Deficit/Hyperactivity Disorder (ADHD) exhibition, and assessment with children’s sleep habits questionnaire (CSHQ). High-resolution T1-weighted images of all participants were acquired on a 3.0 T scanner (Siemens, Germany) with a 16-channel head coil, and cortical modeling and segmentation of structural MRI data were performed with FreeSurfer software (version 6.0, http://surfer.nmr.mgh.harvard.edu/). We selected the mean values of cortical thickness across the 34 cortical regions in each hemisphere, that is 68 structural features, from the Desikan/Killiany Atlas (3) for the cluster analysis. A data-driven approach of agglomerative clustering (4) was employed to identify potential discrete homogeneous subgroups of pediatric patients with epilepsy according to their neuroanatomic imaging measures, and Euclidean distance was used as the distance metric between subjects. The cluster quality and optimal cluster number were also determined.

RESULTS

The hierarchical clustering analysis of cortical thickness data indicated a two-cluster pattern of cortical mapping, and the combination of dendrogram and heat map illustrations were showed in Figure 1. Visual observation of the two patient subgroups from the dendrogram and heat map illustrates that patients within subgroup 1 (56 patients, 77.8% of patients) had greater cortical thickness than patients within subgroup 2 (16 patients, 22.2% of patients). To evaluate the quality of the cluster result and the optimal number of clusters, the dendrograms from 2 to 10 cluster solutions were evaluated with Silhouette, Dunn, Calinski-Harabasz, Krzanowski-Lai and Homogeneity & Separation indices, as well as The Jaccard coefficient. All parameters reached their maximum in the two-cluster solution. Therefore, the optimal and most stable number of discrete data structures that best represent the data for this patient subgroup is two. We found that, patients within subgroup 1 had significant higher seizure onset rate compared with that in subgroup 2 (41.1% vs 14.3, p<0.05). Patients within subgroup 1 also had significant higher sleep anxiety scores than patients in subgroup 2 (24.1% vs 0, p<0.05). The rate of EEG abnormalities or DSM abnormalities did not significantly different between patient subgroups (both with p>0.05). In identifying the cortical differences that discriminated participant groups among the two patient subgroups and controls, we found that, patients within subgroup 1 showed significant increased cortical thickness in bilateral precuneus, bilateral cuneus, left rostral middle frontal gyrus, left superior parietal gyrus, left fusiform gyrus, right lateral orbitofrontal gyrus, right frontal pole, right pericalcarine cortex and right lingual gyrus, whereas patients in subgroup 2 only showed regional cortical thinning in right superior temporal gyrus (all with Bonferroni corrected p<0.05).

DISCUSSION

Using biological imaging measures and data-driven method, we observed two different patterns of cortical morphometric features of gray matter in the early stage of pediatric patients with epilepsy. In contrast to healthy controls, one group of patients had widespread increase of the cortical thickness in neocortex, while the other group of patients only showed regional cortical thinning superior temporal cortex. While the two patient subgroups did not differ in demographic or clinical ratings, the seizure relapse rate after typical treatment was significantly different between them: patients in subgroup 1 did show a higher rate of seizure onset than patients within subgroup 2. In epilepsy patients early in their illness course, our findings suggest the potential existence of two distinct subgroups neurobiologically within pediatric patients with epilepsy who were seizure-free or not after treatment. The distinct patterns of cortical gray matter alteration might be associated with multiple mechanisms that remain to be determined. More importantly, as in previous efforts that defined patients with neurobiological features (5), we identified two groups of epilepsy patients based on their biological cortical gray matter features, and demonstrated its clinical relevance by observing different seizure onset rate in these two patient subgroups who shared similar demographic and clinical characteristics.

CONCLUSION

The neuroanatomic measures of gray matter may provide clinically useful predictors of differential rate of seizure onset in pediatric patients with epilepsy after treatment, which is useful in monitoring illness progression and developing individualized treatment.

Acknowledgements

This work is supported by the Fundamental Research Funds for the Central Universities (Grant No. 2020SCU12053), Sichuan Science and Technology Program (Grant No. 2020YJ0018), the Science and Technology Project of the Health Planning Committee of Sichuan (Grant No. 20PJ010), and Postdoctoral Interdisciplinary Research Project of Sichuan University (Grant No. 0040204153248) .