Three Strikes Algorithm for High Frequency Oscillation Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for identifying high frequency oscillations (HFOs) in brain recordings are labor-intensive, prone to false positives due to ringing artifacts, and often require supervised training or heuristic approaches, limiting their accuracy and efficiency.
Innovation Solution
A three strikes algorithm that detrends candidate HFO signals and tests them for amplitude, rhythmicity, and the absence of ringing artifacts, using event-specific criteria compared to critical values, without relying on heuristics or labeled training samples.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If visual inspection method is used to identify HFOs, then detection accuracy is improved, but labor time and operational complexity increase significantly
Solution Approach 1:
The patent replaces manual visual inspection with an automated computational algorithm that analyzes EEG signals. The system uses objective mathematical criteria (amplitude threshold, duration, frequency) to detect HFOs automatically, substituting the mechanical process of human visual examination with an automated signal processing system that maintains high detection accuracy while dramatically reducing time requirements
Solution Approach 2:
The detection system performs self-validation through automated criteria assessment without requiring human intervention. The algorithm independently evaluates each signal segment against predefined HFO characteristics and generates detections autonomously, enabling the system to serve itself rather than relying on external human expertise for each analysis
2Measurement precision
If high frequency filtering is applied to detect HFOs, then HFO detection capability is improved, but ringing artifacts increase causing false positives
Solution Approach 1:
The patent acknowledges that high-frequency filtering inevitably produces ringing artifacts, but converts this harmful effect into a useful diagnostic feature. The system detects HFOs by identifying characteristic ringing patterns that follow specific mathematical criteria, transforming the previously problematic artifact into a recognizable signature of genuine HFO events when they meet amplitude, duration, and frequency thresholds
Solution Approach 2:
The detection algorithm applies different evaluation criteria to different signal characteristics locally. Rather than uniformly accepting all high-frequency content, the system selectively identifies HFOs by examining specific local properties (amplitude relative to baseline, duration between 5-150ms, frequency 80-500Hz) while ignoring other high-frequency components that do not meet these localized criteria, thereby distinguishing true HFOs from artifact
3Measurement precision
If supervised training methods are used to improve HFO detection accuracy, then detection precision is improved, but device complexity and data requirements increase
Solution Approach 1:
The patent achieves high detection precision by carefully optimizing specific detection parameters (amplitude threshold set at 5 standard deviations above baseline, duration window of 5-150 milliseconds, frequency range of 80-500 Hz) rather than using complex trained models. This parameter-based approach maintains simplicity while achieving accuracy comparable to or exceeding supervised methods, avoiding the need for large training datasets and complex neural network architectures
Data Source
AI summary
A method of identifying high frequency oscillations (HFOs) in neural signal data involves detrending the neural signal data, and identifying HFOs through one or more objective and training-free strike tests selected from the group consisting of (i) amplitude, (ii) rhythmicity, and (iii) ringing.