Neurological Event Detection Using Time-Windowed Signal Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current techniques for diagnosing epilepsy, such as conventional EEG, face challenges in accurately identifying and labeling seizure onset zones due to discontinuity in brainwave characteristics, leading to misjudgment of seizure occurrences.
Innovation Solution
A method and system for neurological event detection that involves obtaining neural oscillation signals, extracting features, using a classification model to determine the occurrence of neurological events, and displaying final determination results within preset time windows, enhancing accuracy by reducing misjudgment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional EEG techniques are used to detect neurological events, then the detection process is simple and fast, but the accuracy of identifying seizure onset zones is low due to discontinuity in brainwave characteristics
Solution Approach 1:
The patent divides the continuous brainwave signal into multiple discrete time windows for independent analysis. Each time window is evaluated separately to determine seizure occurrence, allowing the system to handle discontinuous brainwave characteristics effectively. This segmentation approach enables accurate identification of seizure onset zones by analyzing localized patterns rather than relying on continuous signal interpretation.
Solution Approach 2:
The system performs preliminary feature extraction and classification on divided time windows before making final seizure determination. By pre-processing each segment with feature extraction and classification models, the system prepares data in advance, enabling more accurate final judgment while maintaining computational efficiency.
2Reliability
If discrete brainwave analysis is performed on intermittent onset brainwaves, then the analysis speed is fast, but multiple seizure states are erroneously determined for a single continuous seizure
Solution Approach 1:
The patent implements a feedback mechanism where classification results from multiple time windows are integrated to determine final seizure states. The system uses the classification results from segmented analysis as input for final determination, allowing correction of erroneous discrete judgments by considering the broader temporal context and relationships between adjacent time windows.
Solution Approach 2:
The system merges the results from multiple discrete time window analyses to form a unified seizure determination. By combining classification results across time windows and applying final determination logic, the system consolidates fragmented analyses into accurate seizure state labels, preventing erroneous multiplication of seizure states.
3Measurement precision
If spike detection is performed on brainwaves without seizures, then transient abnormal discharges are detected, but false positive seizure judgments occur
Solution Approach 1:
The patent applies different analysis strategies to different local patterns in the brainwave signal. Instead of using a uniform detection threshold for all signals, the system adapts its analysis to local characteristics of each time window, allowing it to distinguish between pathological spikes indicating seizures and benign transient abnormalities based on their specific local features.
Solution Approach 2:
The system performs preliminary feature extraction and classification on each time window before making final seizure determination. This preliminary analysis allows the system to identify and filter out characteristic patterns of transient abnormal discharges that are not associated with seizures, reducing false positives while maintaining sensitivity to true seizure events.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
A method for neurological event detection is provided and include: obtaining a neural oscillation signal and extracting a plurality of features from the neural oscillation signal; obtaining a plurality of classification results corresponding to a plurality of times according to the plurality of features by using a classification model constructed based on a plurality of training data, wherein whether a neurological event occurs is known in each training data; and calculating a final determination result corresponding to each of a plurality of evaluation time windows according to a preset time window width and the plurality of classification results and displaying the final determination result. The final determination result indicates whether the neurological event occurs. Each evaluation time window corresponds to a plurality of classification results and one single final determination result. In addition, a neurological event detection system using the method for neurological event detection is also provided.