Page 218 - MEGIN Book Of Abstracts - 2023

instead of epileptic foci, play a key role in the complex ity: 91.60% - 99.96%; f1 score: 91.70% - 99.48%; and
pathophysiological mechanisms of insular epilepsy. The area under the curve: 0.9688 - 0.9998).
dysfunction of HFO networks may prove to be a novel
promising biomarker and the cause of interictal brain IEEE transactions on medical imaging (2020), Vol. 39, No. 6
dysfunctions in insular epilepsy. (31831410) (10 citations)

Keywords: Effective connectivity, Graph theory, Insular
epilepsy, Magnetoencephalography, Ripples The impact of MEG results on surgical outcomes in
patients with drug-resistant epilepsy associated
Clinical neurophysiology: official journal of the with focal encephalomalacia: a single-center
International Federation of Clinical Neurophysiology experience (2020)
(2020), Vol. 131, No. 2 (31865139) (10 citations)
He, Xinghui; Zhou, Jian; Teng, Pengfei; Wang, Xiongfei;
Guan, Yuguang; Zhai, Feng; Li, Tianfu; Luan, Guoming
EMS-Net: A Deep Learning Method for
Autodetecting Epileptic Magnetoencephalography Department of Neurosurgery, Epilepsy Center, Beijing Sanbo
Spikes (2020) Brain Hospital, Capital Medical University, Xiangshan Yike-
song 50, Haidian District, Beijing, 100093, China; Beijing In-
Zheng, Li; Liao, Pan; Luo, Shen; Sheng, Jingwei; Teng, stitute for Brain Disorders, Capital Medical University, Beijing,
Pengfei; Luan, Guoming; Gao, Jia-Hong China; Beijing Institute for Brain Disorders, Capital Medical
University, Beijing, China. [email protected]
ABSTRACT Epilepsy is a neurological disorder charac-
terized by sudden and unpredictable epileptic seizures, PURPOSE To analyze the impact of magnetoencepha-
which incurs significant negative impacts on patients' lography (MEG) results on surgical outcomes in patients
physical, psychological and social health. A practical with drug-resistant epilepsy secondary to encephalo-
approach to assist with the clinical assessment and malacia.
treatment planning for patients is to process magneto-
encephalography (MEG) data to identify epileptogenic METHODS We retrospectively reviewed 121 patients
zones. As a widely accepted biomarker of epileptic foci, with drug-resistant epilepsy associated with encepha-
epileptic MEG spikes need to be precisely detected. lomalacia who underwent MEG followed by resection
Given that the visual inspection of spikes is time con- surgery. Patients were subdivided into concordant
suming, an automatic and efficient system with ade- MEG group and dis-concordant MEG group for analysis
quate accuracy for spike detection is valuable in clinical based on whether the MEG results were in concordance
practice. However, current approaches for MEG spike with epileptogenic zones or not.
autodetection are dependent on hand-engineered
features. Here, we propose a novel multiview Epileptic RESULTS 121 patients were included in the present
MEG Spikes detection algorithm based on a deep learn- study. The MEG spike sources of 73 (60.33%) patients
ing Network (EMS-Net) to accurately and efficiently rec- were in concordance with epileptogenic zones while
ognize the spike events from MEG raw data. The results the MEG spike sources of the other 48 (39.67%) were
of the leave-k-subject-out validation tests for multiple in dis-concordance with epileptogenic zones. Favor-
datasets (i.e., balanced and realistic datasets) showed able seizure outcomes were achieved in 79.45% (58 of
that EMS-Net achieved state-of-the-art classification 73) of patients with concordant MEG results while only
performance (i.e., accuracy: 91.82% - 99.89%; precision: 62.50% (30 of 48) of patients with dis-concordant MEG
91.90% - 99.45%; sensitivity: 91.61% - 99.53%; specific- results were seizure free with a follow-up of 2-10 years.
The differences of seizure-free rate between patients
with concordant MEG results and dis-concordant MEG

 ontents  Index 197
C