Dynamical Brain Network Model for Seizure Onset Inference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current surgical interventions for epilepsy have low success rates due to incomplete localization of epileptogenic zones, primarily caused by insufficient spatial sampling of whole-brain seizure propagation patterns, which existing methods fail to adequately address.
Innovation Solution
A method that uses computerized brain networks and dynamical models to infer the onset time and excitability of brain regions not observed during seizures by training on data from a cohort of patients, employing magnetic resonance imaging, intracranial electroencephalography, and Bayesian inference to predict seizure activity in hidden regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If implanted depth electrodes or subdural electrode grids are used for pre-surgical evaluation, then regional seizure activity can be observed, but whole-brain seizure propagation patterns cannot be adequately sampled
Solution Approach 1:
The patent creates a virtual brain model that copies and simulates the physical brain's seizure propagation patterns. This virtual model allows comprehensive whole-brain observation without requiring physical electrodes in every brain region, thus achieving complete spatial sampling without proportional increase in device complexity
Solution Approach 2:
The patent introduces a computational model as an intermediary between the limited electrode recordings and the complete brain network. This model infers unobserved seizure propagation patterns from observed data, acting as a mediator that bridges the gap between limited measurements and comprehensive understanding
2Adaptability or versatility
If manual tuning of model settings is performed to model whole-brain network activity, then the model can be adapted to patient-specific data, but the process cannot be fully automated and requires clinical expert input
Solution Approach 1:
The patent implements self-service through automated machine learning algorithms that train the computational model using patient-specific seizure data. The system automatically adjusts model parameters and learns propagation patterns without requiring manual tuning by clinicians, achieving full automation while maintaining patient-specific adaptability
Solution Approach 2:
The patent employs feedback mechanisms where the model is trained iteratively using observed seizure data from electrodes. The model continuously refines its parameters based on the difference between predicted and actual seizure propagation patterns, enabling automated adaptation to patient-specific characteristics
Data Source
AI summary
The method for determining an onset time and an excitability of a brain region that is not observed as recruited or not recruited in a seizure activity of an epileptic patient brain includes: providing a dynamical model of a propagation of an epileptic seizure in the brain networks; providing a statistical model which defines the probability of generating sets of observations of the state of the brain networks by said dynamical model; training the dynamical model of the propagation of an epileptic seizure using the statistical model and the data set of observations of the training cohort; and inverting the trained dynamical model and inferring the onset time and excitability of a third region from the onset time that is observed for the first and second regions using the statistical model.