Seizure Detection Using Multi-Modal Biomedical Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing seizure detection methods rely heavily on invasive EEG monitoring and are not suitable for long-term, mobile use, providing unreliable predictions that are limited to a few seconds or minutes, and fail to account for the challenges of heart rate monitoring during seizures.
Innovation Solution
A method and device that utilize a combination of electrodermal activity (EDA), accelerometer, and photoplethysmogram (PPG) signals to detect and predict seizures, with EDA and accelerometer signals used for real-time seizure detection and PPG signals used to confirm recent seizure occurrence, allowing for alerts to be triggered based on likelihood thresholds.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If invasive EEG monitoring is used for seizure detection, then detection reliability is improved, but device complexity and ease of operation deteriorate
Solution Approach 1:
The patent segments the seizure detection task into multiple independent monitoring components: EDA sensor for autonomic nervous system activity, accelerometer for movement detection, and PPG sensor for heart rate monitoring. Each sensor type targets specific physiological markers of seizure activity, allowing the system to achieve reliable detection without requiring complex invasive EEG equipment.
Solution Approach 2:
The patent uses intermediate physiological signals as mediators between brain activity and observable symptoms. Instead of directly monitoring brain electrical activity through invasive EEG, the system monitors intermediate markers such as electrodermal activity, heart rate variations, and movement patterns, which reflect seizure activity indirectly but can be captured through non-invasive wearable sensors.
2Measurement precision
If EEG monitoring is used for seizure detection, then detection accuracy is improved, but ease of operation and duration of use deteriorate
Solution Approach 1:
The patent implements a self-service monitoring system where the wearable device autonomously collects physiological data, processes signals through machine learning algorithms, and generates seizure predictions without requiring manual intervention or complex setup. The system automatically adjusts to individual patient baselines and provides continuous monitoring during daily activities, making it easy to operate while maintaining high detection accuracy.
3Extent of automation
If existing seizure prediction methods are used, then prediction capability is improved, but prediction duration and reliability deteriorate
Solution Approach 1:
The patent applies preliminary action by monitoring physiological signals continuously and identifying pre-seizure patterns before the actual seizure occurs. The machine learning model analyzes trends in EDA, heart rate, and movement data to predict seizures up to ten minutes in advance, allowing timely intervention. The system continuously updates predictions based on evolving physiological patterns, maintaining accuracy throughout the prediction window.
4Reliability
If multiple sensor types are combined for seizure detection, then detection reliability is improved, but device complexity increases
Solution Approach 1:
The patent merges multiple sensor types (EDA, accelerometer, PPG) into a single integrated wearable device that simultaneously collects diverse physiological data. The machine learning model combines these multi-modal signals through feature fusion, weighing each sensor's contribution based on its specificity for different seizure phases. This integration achieves high detection reliability while managing device complexity through unified hardware and software architecture.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The method provides reliable, non-invasive seizure detection and prediction up to ten minutes in advance, improving patient safety by enabling timely preparation and response to seizures, with high sensitivity and specificity in detecting seizure events.
Implementation Method 1
utilize a combination of electrodermal activity (EDA), accelerometer, and photoplethysmogram (PPG) signals to detect and predict seizures, with EDA and accelerometer signals used for real-time seizure detection and PPG signals used to confirm recent seizure occurrence
Implementation Method 2
comparing first information regarding an electrodermal activity of the patient to at least one first condition to generate a first comparison result
Implementation Method 3
comparing second information on a movement of a limb of the patient to at least one second condition to generate a second comparison result
Data Source
AI summary
A method of detecting the likelihood of a seizure event in a patient includes at successive expirations of a first time interval, determining a first likelihood that the patient is experiencing a seizure based on electrodermal activity and a movement of a limb of the patient. The method also includes at successive expirations of a second time interval, determining whether the patient experienced a seizure in a second time period preceding the determining based on a heart rate of the patient. In response to determining that the third comparison result satisfies at least a third detection criterion, the method compares electrodermal activity and the movement of a limb of the patient to determine a second likelihood. In response to determining that the second likelihood satisfies a second detection criterion, the method triggers presentation of a second alert regarding a potential seizure.