Authors :
Presenting Author: Chenyang Zhao, MD – Department of Neurology, West China Hospital of Sichuan University
Ping Jiang, PhD – West China Hospital, Sichuan University
Jiani Chen, MD – West China Hospital, Sichuan University
Siqi Yang, PhD – Chengdu University of Information Technology
Yi Liang, MM – Department of Neurology, West China Hospital of Sichuan University
Yue Li, MM – West China Hospital, Sichuan University
Xintong Wu, MD – Department of Neurology, West China Hospital of Sichuan University
Yingying Tang, MD – West China Hospital, Sichuan University
Qiyong Gong, PhD – West China Hospital, Sichuan University
Dong Zhou, MD – West China Hospital of Sichuan University, Sichuan, China
Rationale:
Focal to bilateral tonic-clonic seizures (FBTCS) are a severe form of epileptic seizures, often associated with poor prognosis, increased risk of injury, and sudden unexpected death in epilepsy. Although some resting-state functional MRI (rs-fMRI) studies provided insights into FBTCS-related mechanisms, they were limited by prior assumptions and seed region selection, potentially overlooking other relevant brain areas and introducing bias. This study aimed to compare whole-brain resting-state neural activity among patients with FBTCS (FBTCS+), patients without FBTCS (FBTCS-), and healthy controls (HCs), and using significant clusters from FBTCS+ vs. FBTCS- comparisons as seeds to investigate functional integration changes underlying FBTCS.
Methods:
T1-weighted and rs-fMRI images were acquired from 47 FBTCS- patients, 84 FBTCS+ patients, and 81 matched HCs. We flipped images to uniform the left side as the ipsilateral side for epileptogenesis and the right side as the contralateral side. The amplitude of low-frequency fluctuations (ALFF), regional homogeneity (ReHo), and degree centrality (DC) were calculated across the whole brain, and significant clusters identified between the FBTCS+ and FBTCS- groups were used as seeds for resting-state functional connectivity (rs-FC) analysis. Using these significant functional metrics as features, four classic machine learning models were employed to classify FBTCS+ and FBTCS- patients at the individual level, with leave-one-out cross-validation used to assess classification performance.
Results:
Compared to HCs, both FBTCS+ and FBTCS- patients exhibited bilateral, diffuse alterations in ALFF, ReHo, and DC, showing similar brain distribution patterns. Notably, these changes were more widespread and pronounced in FBTCS+ patients. Importantly, direct comparisons between the patient groups revealed a significant decrease in DC in the ipsilateral temporal pole in FBTCS- patients compared to FBTCS+ patients. Rs-FC analysis, using the ipsilateral temporal pole as a seed, showed reduced connectivity in the ipsilateral lingual gyrus, contralateral superior temporal gyrus, and contralateral temporal pole in FBTCS- patients relative to FBTCS+ patients (Fig 1). Leveraging these identified significant DC and rs-FC abnormalities as features, four machine learning classifiers demonstrated stable performance, achieving an average AUC of 0.76 (Fig 2).
Conclusions:
Our findings suggest that FBTCS+ patients showed more widespread regional neural activity abnormalities, indicating broader functional network involvement. Reduced DC and rs-FC in specific brain regions might explain the lower likelihood of FBTCS in FBTCS- patients. Machine learning models based on these metrics effectively distinguished FBTCS+ from FBTCS- patients, demonstrating their potential as biomarkers for FBTCS, although further validation in larger cohorts is needed.
Funding: This study was supported by the National Natural Science Foundation of China (Grant No. 82471484), the Natural Science Foundation of Sichuan Province (Grant No. 2024NSFSC1649), Sichuan Science and Technology program (Grant No. 2024NSFTD0043), National Key Research and Development Program of China (Grant No. 2022YFC2503804)