EEG Signal Segmentation for Delirium Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing delirium assessment methods using single-channel EEG-recordings fail to reliably distinguish between target signal segments indicative of delirium or encephalopathy and non-target signal segments, such as artifacts, leading to false positive detections and incorrect classifications.

Innovation Solution

A system that applies target and non-target parameter sets, determined through wavelet decomposition and machine learning algorithms, to classify EEG signal segments, with a voting process to verify the correctness of classifications based on temporal proximity of time stamps assigned to detected segments.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of manufacture

If single-channel EEG-recordings are used for delirium detection, then the method is simple and cost-effective, but the system cannot reliably distinguish between target signal segments and non-target signal segments

Engineering Contradiction:

Improvesimplicity of methodVSAvoidreliability of distinction between target and non-target segments

Core Design Contradiction:

Ease of manufactureVSReliability

Solution Approach 1:

The EEG signal is divided into segments, and each segment is classified as target or non-target using wavelet decomposition and machine learning algorithms. The system processes segments individually, applying different parameter sets (target parameter set, non-target parameter set) to distinguish between delirium-related signals and artifacts.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system applies different parameter sets to the same signal segments - a target parameter set for identifying delirium-related signals and a non-target parameter set for identifying artifacts. The parameters include wavelet coefficients and other features that are transformed and compared to determine segment classification.

Inventive Principle:

Principle #35Parameter changes

2Measurement precision

If classification algorithms are applied to EEG signal segments, then the distinction between target and non-target segments is improved, but false positive detections and incorrect classifications occur

Engineering Contradiction:

Improveprecision of signal segment classificationVSAvoidaccuracy of classification

Core Design Contradiction:

Measurement precisionVSReliability

Solution Approach 1:

The system uses a voting process where multiple algorithms or parameter sets provide classifications, and these classifications are voted upon to determine the final result. This feedback mechanism allows the system to correct individual algorithm errors and reduce false positives by cross-validating classifications.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system performs preliminary processing of the EEG signal by applying wavelet decomposition and extracting features before classification. Reference target signal segments and reference non-target signal segments are prepared in advance to serve as training data for the machine learning algorithms.

Inventive Principle:

Principle #10Preliminary action

3Reliability

If multiple parameter sets and voting processes are used to improve classification accuracy, then the distinction between target and non-target segments is enhanced, but the system complexity increases

Engineering Contradiction:

Improveaccuracy of signal segment classificationVSAvoidcomplexity of classification system

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The system uses a unified framework that handles both target and non-target signal segments through the same basic processing pipeline - wavelet decomposition, feature extraction, and classification algorithms. The only difference is the application of different parameter sets, which simplifies the overall system architecture despite the multiple classification tasks.

Inventive Principle:

Principle #6Universality (Multi-functionality)

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 system improves the distinction between target and non-target signal segments, reducing false positive detections and incorrect classifications by using wavelet coefficients and machine learning to accurately identify delirium-related signals amidst artifacts.

Implementation Method 1

The target parameter set comprises wavelet coefficients that are determined using wavelet decomposition of the plurality of reference target signal segments

Methodology Applied

Scientific EffectWavelet decomposition:

Data Source

PatentEP4084671B1System for detecting and classifying segments of signals from EEG-recordings

Publication Date: 2023.11.08 PROLIRA BV

Figure 1
Figure 2

AI summary

The invention relates to a data processing method for detecting and classifying a segment of a signal (1) that is obtained from a single-channel EEG-recording as a target signal segment or as a non-target signal segment. The method comprises a voting process to determine whether classification of a first detected segment of the signal as a target signal segment or classification of a second detected segment of the signal as a non-target signal segment is correct. The invention further relates to a device (2) and a system (3) that are configured and arranged to perform the data processing method according to the invention.