Seizure Prediction: From Patient Perspectives to Advanced Signal Processing and Machine Learning Algorithms
Bibliographic record
Abstract
RÉSUMÉ: L'épilepsie est une maladie neurologique chronique qui touche environ 60 millions de personnes de tous âges dans le monde. Caractérisée par des crises récurrentes qui ont pour origine des décharges neuronales excessives dans le cerveau et qui se manifestent par une variété de symptômes physiques invalidants, l'épilepsie a été qualifiée de « priorité de santé publique » par l'Organisation mondiale de la santé. Malgré la disponibilité de médicaments anticrises et la chirurgie de l'épilepsie, plus d'un tiers des personnes atteintes d'épilepsie (PAE) souffrent de crises incontrôlées. La prédiction ou la prévision de la survenue des crises est un domaine de recherche important et interdisciplinaire qui fait progresser notre compréhension de la génération des crises et vise à réduire le fardeau des crises incontrôlées en proposant de nouvelles technologies aux PAE. Alors que les crises sont souvent considérées comme spontanées et imprévisibles, des décennies de recherche sur les enregistrements d'électroencéphalographie intracrânienne (iEEG) ont révélé des preuves d'une période de transition entre l'activité interictale (état cérébral normal) et la période ictale (état de crise) appelée l'état préictal. Cependant, malgré plusieurs avancées scientifiques et le développement de nombreux algorithmes de prédiction des crises basés sur des approches computationnelles, une bonne compréhension de l'état préictal échappe encore aux chercheurs et aux cliniciens, notamment en raison de la nature complexe et non stationnaire de la dynamique cérébrale épileptique. Les algorithmes de pointe nécessitent de combiner plusieurs caractéristiques extraites de l'iEEG avec des classificateurs sophistiqués et se sont révélés prometteurs pour l'identification automatique de l'état préictal. Malheureusement, la complexité de calcul de ces algorithmes a limité leur translation à un dispositif clinique réalisable. De nouvelles approches basées sur des biomarqueurs plus informatifs et des réseaux de neurones artificiels (ANNs) pourraient aider à minimiser l'extraction de caractéristiques et catalyser la traduction de ces algorithmes de prédiction personnalisés en dispositifs cliniques. ABSTRACT: Epilepsy is a chronic neurological condition which affects about 60 million people of all ages worldwide. Characterized by recurrent seizures which originate as excessive neuronal discharges in the brain and manifest as a variety of disabling physical symptoms, epilepsy has been termed a “public health imperative" by the World Health Organization. Despite the availability of antiseizure medications and epilepsy surgery to remove the brain area initiating seizures, over a third of people with epilepsy (PWE) suffer from uncontrolled seizures. The prediction or forecasting of seizure occurrence is an important and interdisciplinary area of research which advances our understanding of seizure generation and aims to reduce the burden of uncontrolled seizures by offering novel technologies to PWE. While seizures are often considered spontaneous and unpredictable, decades of research on intracranial electroencephalography (iEEG) recordings of seizures have revealed evidence of a transition period between interictal activity (normal brain state) and ictal events (seizure state) termed the preictal state. However, despite several scientific advances and the development of numerous seizure prediction algorithms based on computational approaches, a proper understanding of the preictal state still eludes researchers and clinicians, namely due to the complex and nonstationary nature of epileptic brain dynamics. State-of-the-art algorithms require combining multiple features extracted from iEEG with sophisticated classifiers and have shown promise for the automatic identification of the preictal state. Unfortunately, the computational complexity of such algorithms has limited their translation to a feasible real-time clinical device. New approaches based on more informative features and artificial neural networks (ANNs) could help minimize feature extraction and catalyze the translation of personalized real-time forecasting. The overarching goal of this thesis was to improve our understanding of preictal mechanisms and predict epileptic seizures using personalized machine learning algorithms. More specifically, the first objective was to assess the ability of PWE and caregivers to predict oncoming seizures based on epileptic prodromes which are visible and subjective warning symptoms. A survey conducted in 346 PWE and caregivers across Canada showed that both patient and caregiver populations support the idea that there exists a distinct group of PWE for whom seizures are predictable at least five minutes before clinical onset. Results show that this population of patients who experience warnings symptoms between five minutes and more than 24 hours preceding seizures (prodromes) represents around 12% of PWE and can be distinguished by a significantly longer recovery period of postictal confusion and/or impaired awareness.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.013 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.003 | 0.002 |
| Open science | 0.000 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.004 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".