Pro-ictal State Classifier Using Thalamocortical EEG Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current seizure prediction technologies are limited in detecting pro-ictal states, which represent periods of heightened seizure risk, and often focus solely on cortical EEG, failing to identify these states minutes or hours prior to seizure onset, thereby limiting the effectiveness of seizure-preventative therapies.
Innovation Solution
A deep neural network-based classifier is used to analyze thalamocortical EEG features, such as power, nonlinear synchrony, Lempel-Ziv complexity, weighted phase lag index, and phase-amplitude coupling, to distinguish pro-ictal states from inter-ictal states, enabling the prediction of seizure onset hours in advance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional EEG classification methods are used, then the system is simple to implement, but the ability to detect pro-ictal states is insufficient
Solution Approach 1:
The patent segments the detection task by dividing EEG signal analysis into multiple independent feature dimensions (power, nonlinear synchrony, Lempel-Ziv complexity, weighted phase lag index, phase-amplitude coupling). Each feature is computed separately and then integrated by the classifier, allowing complex detection to be broken down into manageable computational steps that improve reliability without overwhelming system complexity.
Solution Approach 2:
The patent transitions from traditional single-dimensional EEG analysis to multi-dimensional feature space analysis. By computing five distinct feature types across multiple EEG channels and frequency bands, the system projects neural activity patterns into a higher-dimensional space where pro-ictal states can be more reliably distinguished from inter-ictal states, significantly improving detection accuracy.
2Loss of time
If cortical EEG only is analyzed, then the recording system is simple, but pro-ictal states cannot be detected hours before seizure onset
Solution Approach 1:
The patent extends the spatial dimension of EEG recording from cortical surfaces to include thalamic structures. By placing electrodes in both cortical and thalamic regions, the system captures deep brain oscillatory activity that precedes seizure onset by hours, dramatically extending the prediction time horizon beyond what cortical EEG alone can achieve.
3Measurement precision
If multiple EEG features are analyzed, then the detection accuracy improves, but the computational load increases
Solution Approach 1:
The patent segments computational processing into offline and online phases. Complex feature extraction (power, synchrony, complexity metrics) is performed offline during periods when computational resources are abundant, producing precomputed feature vectors. During online deployment, the classifier only needs to evaluate precomputed features, significantly reducing real-time computational energy consumption while maintaining high classification accuracy.
Solution Approach 2:
The patent performs preliminary computation of complex EEG features during offline processing periods. By precomputing power spectra, synchrony measures, and complexity metrics from historical EEG data, the system prepares feature vectors in advance, so that real-time classification requires minimal computational energy, resolving the contradiction between measurement precision and energy consumption.
Data Source
AI summary
The present disclosure describes various embodiments of systems, apparatuses, and methods for predicting an onset of a seizure by identifying a pro-octal state in advance of the seizure, potentially hours prior to seizure onset. One such method comprise acquiring, by a computing device, electroencephalography (EEG)-based features from brain activity electrical recordings of an individual; inputting, by the computing device, the EEG-based features into a deep neural network-based classifier; classifying, by the computing device using the deep neural network-based classifier, the EEG-based features to a real-valued principal dimension; and/or based on a value of the real-valued principal dimension, generating, by the computing device, a prediction of a seizure onset pro-ictal event.