Intracranial EEG Sensor Selection for Sleep Stage Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for classifying and tracking human brain behavioral states, particularly sleep stages, using intracranial EEG recordings are limited by the need for supervised learning and reliance on scalp EEG data, and are not suitable for next-generation implantable devices with limited computational power and electrode numbers.
Innovation Solution
An automated, unsupervised method for classifying wake, REM, and non-REM sleep stages using subscalp, epidural, and intracranial EEG data, which selects appropriate sensors based on signal thresholds and calculates measured values to classify behavioral states, enabling efficient behavioral state classification and modulation in implantable devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If supervised learning methods are used for brain state classification, then classification accuracy can be improved, but the computational complexity and need for extensive training data increase
Solution Approach 1:
The patent extracts and utilizes only the most discriminative features from intracranial EEG signals (such as spectral power in specific frequency bands, Hjorth parameters, and entropy measures) rather than processing the entire signal spectrum. This feature extraction approach maintains classification accuracy while significantly reducing computational complexity for implantable devices
Solution Approach 2:
The patent transforms the raw EEG signals into derived parameters (spectral power, Hjorth parameters, entropy) that capture essential brain state information in a compressed form. This parameter transformation enables accurate classification with reduced computational burden by working with condensed feature representations rather than raw signals
2Measurement precision
If multiple electrodes are used for recording, then measurement precision and reliability improve, but device complexity and power consumption increase
Solution Approach 1:
The patent makes each intracranial electrode multi-functional by using the same electrode both for stimulation delivery and for recording local field potentials. This eliminates the need for separate recording electrodes, reducing the total electrode count while maintaining both stimulation and monitoring capabilities
Solution Approach 2:
The patent combines the stimulation and recording functions into a single integrated system using the same electrode contacts. The electrode serves dual purposes: delivering therapeutic stimulation and capturing neural signals for brain state classification, thereby reducing device complexity and electrode quantity
3Reliability
If continuous monitoring of brain states is implemented, then therapeutic effectiveness improves, but power consumption and data processing requirements increase
Solution Approach 1:
The patent implements periodic sampling of EEG signals at optimized intervals rather than continuous high-rate acquisition. The system monitors brain states at sufficient frequency to detect transitions and adjust therapy, but at lower rates that conserve battery power while maintaining therapeutic effectiveness
Solution Approach 2:
The patent processes only the essential features needed for brain state classification (spectral power in key bands, Hjorth parameters) rather than performing exhaustive signal analysis. This partial processing approach provides sufficient information for reliable classification while minimizing computational power consumption
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
A behavioral state of a brain is classified by automatically selecting one or more sensors based on the signals received from each sensor and one or more selection criteria using one or more processors, calculating at least one measured value from the signal(s) of the selected sensor(s), classifying the behavioral state as: (a) an awake state whenever the measured value(s) for the selected sensor(s) is lower than a first threshold value, (b) a sleep state (N2) whenever the measured value(s) for the selected sensor(s) is equal to or greater than the first threshold value and the measured value(s) is not greater than a second threshold value, or (c) a slow wave sleep state (N3) whenever the measured value(s) from the selected sensor(s) is greater than the first threshold value and the measured value(s) is greater than the second threshold value, and providing a notification of the classified behavioral state.