Authors :
Presenting Author: Chunyao Zhou, MBBS – Department of Neurosurgery, Xiangya Hospital, Central South University, Changsha, China
LI feng, MBBS – Department of Neurology, Xiangya Hospital, Central South University, Changsha 410008, PR China
Rationale:
Epilepsy is increasingly considered a network-based disorder. The default mode network (DMN) was identified as the hub that facilitate the spread of epileptiform activity, especially in the cases of temporal lobe epilepsy (TLE). Previous works demonstrated that correlation between structural connectivity (SC) and functional connectivity (FC) at DMN is associated with the occurrence and development of epilepsy. Thus, we hypothesize that the SC-FC coupling features at DMN can delineate patterns of post-surgical seizure facilitation, and further predict the long-term surgical outcome.Methods:
Here, we retrospectively analyzed diffusion and functional MRI data recorded from 71 surgically treated mesial temporal lobe epilepsy patients and 48 healthy controls. Patients were categorized as seizure free (SF, n=48) or non-seizure free (nSF, n=23) according to their status of postoperative seizure recurrence. By segmenting every individual’s connectivity matrix into 16 functional modules, we extracted four types of features (modular SC, FC, SC-FC coupling and nodal SC-FC coupling) and compared the features between groups. We sorted these features into five datasets (DMN all; DMN coupling, all features, SC and FC, all coupling), and incorporated those datasets into machine-learning models to classify the surgical outcomes.
Results:
Robust inter group differences between HC, SF, and nSF were observed at all types of features. The outcome-related features were mostly SC-FC coupling features at DMN, where nSF showed significantly higher values than SF. Among all five datasets, the classifier trained by DMN related dataset showed best predictive performance, achieving the AUROC of 0.86 (Sensitivity=0.90, specificity=0.65).
Conclusions:
These results suggest that the hyper SC-FC coupling of the DMN is related to poor surgical outcome in TLE patients. Our work firstly combined the modular interaction and SC-FC coupling features by machine learning models to classify surgical outcome of TLE.
Funding:
This study has received funding by Grant No. 2022YFC2503804 from National Key Program of China, Grant Nos. 82071461, 82271503 from National Natural Science Foundation of China, and Grant No. 2021JJ31060 from Natural Science Foundation of Hunan Province.