Mapping short- and long-term changes in brain state for a seizure forecasting application
Poster No:
1327
Submission Type:
Abstract Submission
Authors:
Rachel Stirling1, Mona Nasseri2, Philippa Karoly3, Jodie Naim-Feil3, Ewan Nurse4, Pedro Viana5, Jie Cui6, Vaclav Kremen6, Matthias Dümpelmann7, Gregory Worrell6, Dean Freestone3, Mark Richardson5, Benjamin Brinkmann6
Institutions:
1University of Melbourne, Carlton, Victoria, 2University of North Florida, Jacksonville, FL, 3University of Melbourne, Melbourne, Victoria, 4Seer Medical, Melbourne, Victoria, 5King's College London, London, United Kingdom, 6Mayo Institute, Rochester, MN, 7University of Freiburg, Freiburg, Germany
First Author:
Co-Author(s):
Introduction:
The human brain is a dynamic system influenced by intrinsic biological rhythms oscillating across various timescales, including ultradian, circadian and multiday (>2 day) (Lehnertz et al., 2021). In epilepsy, circadian and multiday rhythms in seizure timing (known as "seizure cycles") and physiological biomarkers of seizure risk (such as epileptic brain activity) are well-observed phenomena (Baud et al., 2018; Karoly et al., 2021; Leguia et al., 2021). These multiday rhythms map long-term changes in brain state (on the order of days), are used in state-of-the-art seizure forecasting models (Proix et al., 2021) and can be tracked with long-term electroencephalography (EEG) recording devices (Baud et al., 2018) and wearable devices (Gregg et al., 2023; Karoly et al., 2021). On the other hand, it is possible to detect distinct pre-seizure signatures (known as the "preictal state") in the minutes leading up to seizure onset (Nasseri et al., 2021), suggesting that seizures are also influenced by short-term brain dynamics operating on the order of minutes. To test this theory, we developed a hybrid seizure forecasting system that simultaneously maps long-term multiday rhythms and short-term preictal states and compared the performance to traditional forecasting models that use just one of these approaches.
Methods:
Eleven participants with epilepsy were recruited for continuous monitoring, capturing heart rate and step count via wrist-worn wearable devices (Empatica E4, Fitbit Inspire HR or Charge 3), and seizures via chronically implanted EEG monitoring devices, with an average recording duration of 337 (SD=96) days. Seizures were confirmed by board-certified neurophysiologists after a thorough review of EEG data.
Two hybrid forecasting systems were developed (Figure 1): the Seizure Warning System (SWS), optimised for short-term prediction (minutes before a seizure), and the Seizure Risk System (SRS), optimised for long-term risk forecasting (hours to weeks before a seizure). The models combined two traditional forecasting models: a machine learning (Long Short-Term Memory networks) model for detecting preictal patterns (Nasseri et al., 2021) and a cycle-based model that tracked circadian and multiday trends in seizure timing and physiological signals (Xiong et al., 2023). All analyses were performed on retrospective data in a pseudo-prospective manner, simulating real-world performance.
Two hybrid forecasting systems were developed (Figure 1): the Seizure Warning System (SWS), optimised for short-term prediction (minutes before a seizure), and the Seizure Risk System (SRS), optimised for long-term risk forecasting (hours to weeks before a seizure). The models combined two traditional forecasting models: a machine learning (Long Short-Term Memory networks) model for detecting preictal patterns (Nasseri et al., 2021) and a cycle-based model that tracked circadian and multiday trends in seizure timing and physiological signals (Xiong et al., 2023). All analyses were performed on retrospective data in a pseudo-prospective manner, simulating real-world performance.
Results:
Both hybrid systems outperformed traditional models while reducing time spent in warning or high-risk states (example forecast shown in Figure 2). Hybrid forecasts maintained high accuracy across diverse participants, with more than 80% of the participants' forecasts significantly outperforming chance. Notably, the SRS enabled seizure risk to be accurately predicted up to 44 days in advance, a breakthrough in forecasting horizons.
Conclusions:
This novel hybrid approach to seizure forecasting demonstrates that mapping both short- and long-term changes in brain state via pre-ictal signatures and circadian and multiday biological rhythms leads to superior performance outcomes compared to traditional forecasts. This was achieved using only two physiological signals - heart rate and step counts - both of which are easily detected by most commercially available, affordable wrist-worn wearables.
These results support the hypothesis that seizures are influenced by at least two unique systems simultaneously (Bernard, 2020; Maturana et al., 2020): (1) a slower multiday oscillator that causes cyclical fluctuations in seizure risk, periodically lowering the threshold at which a seizure may occur, and (2) a rapid pre-ictal system, characterised by distinctive neurophysiological changes minutes before seizure onset. Understanding these seizure genesis mechanisms further may help us to find new predictive biomarkers and therapeutic targets for epilepsy.
These results support the hypothesis that seizures are influenced by at least two unique systems simultaneously (Bernard, 2020; Maturana et al., 2020): (1) a slower multiday oscillator that causes cyclical fluctuations in seizure risk, periodically lowering the threshold at which a seizure may occur, and (2) a rapid pre-ictal system, characterised by distinctive neurophysiological changes minutes before seizure onset. Understanding these seizure genesis mechanisms further may help us to find new predictive biomarkers and therapeutic targets for epilepsy.
Modeling and Analysis Methods:
EEG/MEG Modeling and Analysis 1
Novel Imaging Acquisition Methods:
EEG 2
Keywords:
Data analysis
Electroencephaolography (EEG)
Epilepsy
Machine Learning
Modeling
Other - Forecasting
1|2Indicates the priority used for review
Abstract Information
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
Provide references using APA citation style.
Bernard, C. (2020). Circadian/multidien Molecular Oscillations and Rhythmicity of Epilepsy (MORE). Epilepsia, epi.16716.
Gregg, N. M., Pal Attia, T., Nasseri, M., Joseph, B., Karoly, P., Cui, J., Stirling, R. E., Viana, P. F., Richner, T. J., & Nurse, E. S. (2023). Seizure occurrence is linked to multiday cycles in diverse physiological signals. Epilepsia, 64(6), 1627–1639.
Karoly, P. J., Stirling, R. E., Freestone, D. R., Nurse, E. S., Maturana, M. I., Halliday, A. J., Neal, A., Gregg, N. M., Brinkmann, B. H., & Richardson, M. P. (2021). Multiday cycles of heart rate are associated with seizure likelihood: An observational cohort study. EBioMedicine, 72, 103619.
Leguia, M. G., Andrzejak, R. G., Rummel, C., Fan, J. M., Mirro, E. A., Tcheng, T. K., Rao, V. R., & Baud, M. O. (2021). Seizure cycles in focal epilepsy. JAMA Neurology, 78(4), 454–463.
Lehnertz, K., Rings, T., & Bröhl, T. (2021). Time in brain: How biological rhythms impact on EEG signals and on EEG-derived brain networks. Frontiers in Network Physiology, 1, 755016.
Maturana, M. I., Meisel, C., Dell, K., Karoly, P. J., D’Souza, W., Grayden, D. B., Burkitt, A. N., Jiruska, P., Kudlacek, J., Hlinka, J., Cook, M. J., Kuhlmann, L., & Freestone, D. R. (2020). Critical slowing down as a biomarker for seizure susceptibility. Nature Communications, 11(1), 2172.
Nasseri, M., Pal Attia, T., Joseph, B., Gregg, N. M., Nurse, E. S., Viana, P. F., Worrell, G., Dümpelmann, M., Richardson, M. P., & Freestone, D. R. (2021). Ambulatory seizure forecasting with a wrist-worn device using long-short term memory deep learning. Scientific Reports, 11(1), 21935.
Proix, T., Truccolo, W., Leguia, M. G., Tcheng, T. K., King-Stephens, D., Rao, V. R., & Baud, M. O. (2021). Forecasting seizure risk in adults with focal epilepsy: A development and validation study. The Lancet Neurology, 20(2), 127–135.
Xiong, W., Stirling, R. E., Payne, D. E., Nurse, E. S., Kameneva, T., Cook, M. J., Viana, P. F., Richardson, M. P., Brinkmann, B. H., & Freestone, D. R. (2023). Forecasting seizure likelihood from cycles of self-reported events and heart rate: A prospective pilot study. EBioMedicine, 93.