The critical dynamics of hippocampal seizures

Nature Communications, Год журнала: 2024, Номер 15(1)

Опубликована: Авг. 13, 2024

Abstract Epilepsy is defined by the abrupt emergence of harmful seizures, but nature these regime shifts remains enigmatic. From perspective dynamical systems theory, such critical transitions occur upon inconspicuous perturbations in highly interconnected and can be modeled as mathematical bifurcations between alternative regimes. The predictability represents a major challenge, theory predicts appearance subtle signatures on verge instability. Whether measured before impending seizures uncertain. Here, we verified that predictions applied to onset hippocampal providing concordant results from silico modeling, optogenetics experiments male mice intracranial EEG recordings human patients with epilepsy. Leveraging pharmacological control over neural excitability, showed boundary physiological excitability inferred passively recorded or actively probed circuits. Of importance for design future neurotechnologies, active probing surpassed passive recording decode underlying levels notably when assessed network propagating responses. Our findings provide promising approach predicting preventing based sound understanding their dynamics.

Язык: Английский

---

Removing artefacts and periodically retraining improve performance of neural network-based seizure prediction models

Scientific Reports, Год журнала: 2023, Номер 13(1)

Опубликована: Апрель 11, 2023

Abstract The development of seizure prediction models is often based on long-term scalp electroencephalograms (EEGs) since they capture brain electrical activity, are non-invasive, and come at a relatively low-cost. However, suffer from major shortcomings. First, EEG usually highly contaminated with artefacts. Second, changes in the signal over long intervals, known as concept drift, neglected. We evaluate influence these problems deep neural networks using time series shallow widely-used features. Our patient-specific were tested 1577 hours continuous EEG, containing 91 seizures 41 patients temporal lobe epilepsy who undergoing pre-surgical monitoring. results showed that cleaning data, previously developed artefact removal method convolutional networks, improved performance. also found retraining reduced false predictions. Furthermore, show although processing less susceptible to alarms, may need more data surpass feature-based methods. These findings highlight importance robust denoising periodic adaptation models.

Язык: Английский

---

Hippocampal network activity forecasts epileptic seizures

Nature Medicine, Год журнала: 2024, Номер 30(10), С. 2787 - 2790

Опубликована: Июль 12, 2024

Язык: Английский

---

Forecasting epileptic seizures with wearable devices: A hybrid short‐ and long‐horizon pseudo‐prospective approach

Epilepsia, Год журнала: 2025, Номер unknown

Опубликована: Май 24, 2025

Abstract Objective Seizure unpredictability can be debilitating and dangerous for people with epilepsy. Accurate seizure forecasters could improve quality of life those epilepsy but must practical long‐term use. This study presents the first validation a seizure‐forecasting system using ultra‐long‐term, non‐invasive wearable data. Methods Eleven participants were recruited continuous monitoring, capturing heart rate step count via wrist‐worn devices seizures electroencephalography (average recording duration 337 days). Two hybrid models—combining machine learning cycle‐based methods—were proposed to forecast at both short (minutes) long (up 44 days) horizons. Results The Warning System (SWS), designed forecasting near‐term seizures, Risk (SRS), risk, outperformed traditional models. In addition, SRS reduced high‐risk time by 29% while increasing sensitivity 11%. Significance These improvements mark significant advancement in making more effective.

Язык: Английский

---

Forecasting seizure likelihood from cycles of self-reported events and heart rate: a prospective pilot study

EBioMedicine, Год журнала: 2023, Номер 93, С. 104656 - 104656

Опубликована: Июнь 16, 2023

BackgroundSeizure risk forecasting could reduce injuries and even deaths in people with epilepsy. There is great interest using non-invasive wearable devices to generate forecasts of seizure risk. Forecasts based on cycles epileptic activity, times or heart rate have provided promising results. This study validates a method multimodal recorded from devices.MethodSeizure were extracted 13 participants. The mean period data smartwatch was 562 days, 125 self-reported seizures smartphone app. relationship between onset time phases investigated. An additive regression model used project cycles. results cycles, combination both compared. Forecasting performance evaluated 6 participants prospective setting, long-term collected after algorithms developed.FindingsThe showed that the best achieved area under receiver-operating characteristic curve (AUC) 0.73 for 9/13 showing above chance during retrospective validation. Subject-specific AUC 0.77 4/6 chance.InterpretationThe this demonstrate detected can be combined within single, scalable algorithm provide robust performance. presented enabled estimated an arbitrary future generalised across range types. In contrast earlier work, current prospectively, subjects blinded their outputs, representing critical step towards clinical applications.FundingThis funded by Australian Government National Health & Medical Research Council BioMedTech Horizons grant. also received support Epilepsy Foundation America's 'My Seizure Gauge'

Язык: Английский

---

User experience of a seizure risk forecasting app: A mixed methods investigation

Epilepsy & Behavior, Год журнала: 2024, Номер 157, С. 109876 - 109876

Опубликована: Июнь 7, 2024

Over recent years, there has been a growing interest in exploring the utility of seizure risk forecasting, particularly how it could improve quality life for people living with epilepsy. This study reports on user experiences and perspectives forecaster app, as well potential impact mood adjustment to

Язык: Английский

---

Seizure forecasting: Where do we stand?

Epilepsia, Год журнала: 2023, Номер 64(S3)

Опубликована: Фев. 13, 2023

A lot of mileage has been made recently on the long and winding road toward seizure forecasting. Here we briefly review some selected milestones passed along way, which were discussed at International Conference for Technology Analysis Seizures-ICTALS 2022-convened University Bern, Switzerland. Major impetus was gained from wearable implantable devices that record not only electroencephalography, but also data motor behavior, acoustic signals, various signals autonomic nervous system. This multimodal monitoring can be performed ultralong timescales covering months or years. Accordingly, features metrics extracted these now assess dynamics with a greater degree completeness. Most prominently, this allowed confirmation long-suspected cyclical nature interictal epileptiform activity, risk, seizures. The cover daily, multi-day, yearly cycles. Progress fueled by approaches originating interdisciplinary field network science. Considering epilepsy as large-scale disorder yielded novel perspectives pre-ictal evolving epileptic brain. In addition to discrete predictions will take place in specified prediction horizon, community broadened scope probabilistic forecasts risk continuously time. shift gears triggered incorporation additional quantify performance forecasting algorithms, should compared chance constrained stochastic null models. An imminent task utmost importance is find optimal ways communicate output seizure-forecasting algorithms patients, caretakers, clinicians, so they have socioeconomic impact improve patients' well-being.

Язык: Английский

---

The goal of explaining black boxes in EEG seizure prediction is not to explain models’ decisions

Epilepsia Open, Год журнала: 2023, Номер 8(2), С. 285 - 297

Опубликована: Апрель 19, 2023

Many state-of-the-art methods for seizure prediction, using the electroencephalogram, are based on machine learning models that black boxes, weakening trust of clinicians in them high-risk decisions. Seizure prediction concerns a multidimensional time-series problem performs continuous sliding window analysis and classification. In this work, we make critical review which explanations increase models' decisions predicting seizures. We developed three methodologies to explore their explainability potential. These contain different levels model transparency: logistic regression, an ensemble 15 support vector machines, convolutional neural networks. For each methodology, evaluated quasi-prospectively performance 40 patients (testing data comprised 2055 hours 104 seizures). selected with good poor explain Then, grounded theory, how these helped specialists (data scientists working epilepsy) understand obtained dynamics. four lessons better communication between clinicians. found goal is not system's but improve system itself. Model transparency most significant factor explaining decision prediction. Even when intuitive features, it hard brain dynamics relationship models. achieve understanding by developing, parallel, several systems explicitly deal signal changes help develop complete formulation.

Язык: Английский

---

Comparison between epileptic seizure prediction and forecasting based on machine learning

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Март 7, 2024

Epilepsy affects around 1% of the population worldwide. Anti-epileptic drugs are an excellent option for controlling seizure occurrence but do not work one-third patients. Warning devices employing prediction or forecasting algorithms could bring patients new-found comfort and quality life. These would attempt to detect a seizure's preictal period, transitional moment between regular brain activity seizure, relay this information user. Over years, many studies using Electroencephalogram-based methodologies have been developed, triggering alarm when detecting period. Recent suggested shift in view from forecasting. Seizure takes probabilistic approach problem question instead crisp prediction. In field study, triggered symbolize detection period is substituted by constant risk assessment analysis. The present aims explore capable establish comparison with results. Using 40 EPILEPSIAE database, we developed several patient-specific different classifiers (a Logistic Regression, 15 Support Vector Machines ensemble, Shallow Neural Networks ensemble). Results show increase sensitivity relative up 146% number that displayed improvement over chance 300%. results suggest methodology may be more suitable warning than one.

Язык: Английский

---

Learning to generalize seizure forecasts

Epilepsia, Год журнала: 2022, Номер 64(S4)

Опубликована: Сен. 8, 2022

Epilepsy is characterized by spontaneous seizures that recur at unexpected times. Nonetheless, using years-long electroencephalographic (EEG) recordings, we previously found patient-reported consistently occur when interictal epileptiform activity (IEA) cyclically builds up over days. This multidien (multiday) interictal-ictal relationship, which shared across patients, may bear phasic information for forecasting seizures, even if individual patterns of seizure timing are unknown. To test this rigorously in a large retrospective dataset, pretrained algorithms on data recorded from group and forecasted other, unseen patients.We used long-term participants (N = 159) the RNS System clinical trials, including intracranial EEG recordings (icEEG), two UNEEG Medical trial subscalp system (sqEEG). Based IEA detections, extracted instantaneous phases trained generalized linear models (GLMs) recurrent neural networks (RNNs) to forecast probability occurrence 24-h horizon.With GLMs RNNs, could be above chance 79% 81% subjects with median discrimination area under curve (AUC) .70 .69 Brier skill score (BSS) .07 .08. In direct comparison, individualized had similar performance (AUC .67, BSS .08), but fewer (60%). Moreover, calibration maintained accommodate different rates subjects.Our findings suggest based cycles can generalize drastically reduce amount needed issue forecasts individuals who recently started collecting chronic data. addition, show generalization independent method record (patient-reported vs. electrographic) or (icEEG sqEEG).

Язык: Английский

---