Virtual epileptic patient cohort: generation and evaluation
Presented During:
Thursday, June 27, 2024: 11:30 AM - 12:45 PM
COEX
Room:
Grand Ballroom 101-102
Poster No:
402
Submission Type:
Abstract Submission
Authors:
Borana Dollomaja1, Huifang WANG2, Maxime Guye3, Fabrice Bartolomei4, Viktor Jirsa5
Institutions:
1Institut de Neurosciences des Systemes UMR1106, Marseille, Marseille, 2AMU, INS, INSERM U1106, Marseille, PACA, 3Aix Marseille Université, Marseille, PACA, 4AMU, INS, INSERMU1106, Marseille, PACA, 5nstitut de Neurosciences des Systèmes, Marseille, N/A
First Author:
Co-Author(s):
Introduction:
Epilepsy is one of the most common brain diseases, affecting 1% of the world's population. Drug-resistant epilepsy (DRE) in particular affects 1 in 3 epileptic people. Recurrent seizures which characterize the disorder, occur due to sudden abnormal activity in the brain. This activity is generated in the so-called epileptogenic zone (EZ) network. A precise detection of the epileptogenic zone is crucial to treat DRE. Seizure recordings are used by clinicians to estimate the EZ network. In addition, brain stimulation is used to induce seizures (George 2020). By varying stimulation parameters via trial and error, clinicians aim to pinpoint regions responsible for seizure activity. In this work, we built a virtual epileptic patient cohort and evaluated this modeling framework for capturing empirical SEEG data.
Methods:
In our study, we collected brain imaging data from 30 DRE patients, consisting of T1-MRI, diffusion-weighted MRI (DW-MRI) and recordings from implanted stereoelectroencephalography (SEEG) electrodes. The SEEG recordings consist of spontaneous seizures, stimulated seizures and interictal spikes. For each patient, we built virtual brain copies based on personalized whole-brain models (Wang 2023). We combined T1-MRI and DW-MRI to build brain network models. Brain regions are parcellated according to a brain atlas and represented as nodes in the network. The white matter fiber connections are represented as edges in the network. Brain region activity is simulated using a model which captures spatiotemporal seizure dynamics (Jirsa 2014). This model is placed in each node of the brain network model. In particular, the model's parameter x0 determines the node's epileptogenicity. We set epileptogenic nodes based on the EZ clinical hypothesis. Epileptogenic nodes generate seizure dynamics autonomously which propagate following edge connections from white matter tracts. Simulations are performed on the whole-brain level. Implanted SEEG electrodes are reconstructed in 3D from the post-implantation CT scan. Using euclidean distances between brain network nodes and electrode contact locations, we map the simulated brain activity onto the SEEG electrodes.
Results:
Based on this workflow, we built the virtual epileptic patient cohort in BIDS format (Appelhoff et al. 2019). We set up metrics to compare patient-specific empirical and simulated SEEG seizures, such as binary overlap and correlation of the spatio-temporal time series. We built a control cohort where the EZ hypothesis was chosen randomly and show that patient-specific EZ hypothesis captures empirical SEEG data features significantly better than a randomly chosen one.
For stimulated seizures, we also interrogated the stimulation parameters used to generate seizures. We built control cohorts where the same virtual model is used and only stimulation parameters vary, such as stimulation amplitude and stimulation location. We showed that as we get further away from the empirical stimulation amplitude or location, the simulated SEEG time series fail to capture the empirical SEEG seizure dynamics.
Finally, we compared interictal spike time series by measuring spike rate for each SEEG channel. We showed that patient-specific EZ hypothesis captures spike rate better than a randomly chosen one, however not to the same extent as seizure dynamics are captured. This result confirms previous works (Bartolomei 2016, Luders 2006), where the interictal spike network is not identical to the EZ network, but overlaps with it.
For stimulated seizures, we also interrogated the stimulation parameters used to generate seizures. We built control cohorts where the same virtual model is used and only stimulation parameters vary, such as stimulation amplitude and stimulation location. We showed that as we get further away from the empirical stimulation amplitude or location, the simulated SEEG time series fail to capture the empirical SEEG seizure dynamics.
Finally, we compared interictal spike time series by measuring spike rate for each SEEG channel. We showed that patient-specific EZ hypothesis captures spike rate better than a randomly chosen one, however not to the same extent as seizure dynamics are captured. This result confirms previous works (Bartolomei 2016, Luders 2006), where the interictal spike network is not identical to the EZ network, but overlaps with it.
Conclusions:
In conclusion, we provided a virtual epileptic cohort and demonstrated the capacity of personalized whole brain models in capturing empirical brain activity of drug-resistant epilepsy patients. In particular, our personalized modeling framework captures stimulated seizure dynamics, where patient-specific stimulation parameters reproduce the observed seizures. This approach can be useful in estimating optimal stimulation parameters that induce habitual seizures to diagnose epilepsy.
Brain Stimulation:
Invasive Stimulation Methods Other
Disorders of the Nervous System:
Neurodevelopmental/ Early Life (eg. ADHD, autism) 1
Modeling and Analysis Methods:
Connectivity (eg. functional, effective, structural)
Methods Development
Neuroinformatics and Data Sharing:
Databasing and Data Sharing 2
Keywords:
Computational Neuroscience
Data analysis
Data Organization
Design and Analysis
Epilepsy
Modeling
Open Data
Open-Source Code
Statistical Methods
Workflows
1|2Indicates the priority used for review
Provide references using author date format
Bartolomei, F. et al. (2016), 'What is the concordance between the seizure onset zone and the irritative zone? A SEEG quantified study.' Clinical Neurophysiology, 127(2), 1157-1162.
George, D. D. et al. (2020). Stimulation mapping using stereoelectroencephalography: current and future directions. Frontiers in Neurology, 11, 320.
Jirsa, V. K. et al. (2014), 'On the nature of seizure dynamics', Brain : a journal of neurology, 137(Pt 8), 2210–2230.
Lüders, H. O. et al. (2006), 'The epileptogenic zone: general principles', Epileptic disorders, 8, S1-S9.
Wang, H. E. et al. (2023), Delineating epileptogenic networks using brain imaging data and personalized modeling in drug-resistant epilepsy. Science Translational Medicine, 15(680), eabp8982.