ECG Wave Segmentation for Artifact-Aware Cardiac Event Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing electrocardiogram (ECG) wave signal normalization techniques are challenged by variations in voltage ranges, noise, and environmental disturbances, leading to inaccuracies such as voltage spikes and false pauses, which distort cardiac event detection.
Innovation Solution
A computing system applies machine learning models, specifically deep neural networks, to classify ECG wave segments as 'clean', 'artifact', or 'pause', excluding artifacts from normalization, and scales the signal to a standard average and standard deviation, ensuring accurate cardiac event detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If ECG wave signals are normalized to improve convergence of machine learning models, then model training efficiency is improved, but measurement precision deteriorates due to artifacts such as voltage spikes and false pauses causing distortion
Solution Approach 1:
The ECG signal is divided into multiple segments, and each segment is independently classified to identify artifacts. This segmentation allows the system to process and normalize only the clean segments, preventing artifact distortion while maintaining overall signal integrity for accurate machine learning model training.
Solution Approach 2:
Artifacts such as voltage spikes and false pauses are identified and extracted from the ECG signal using machine learning classification. By removing these harmful components before normalization, the system ensures that only clean cardiac signal portions are processed, preserving measurement precision while enabling efficient model training on normalized data.
2Adaptability or versatility
If ECG wave signals from different devices with different voltage ranges, gains, and filters are normalized, then adaptability is improved, but device complexity increases due to the need to handle various signal characteristics
Solution Approach 1:
A universal machine learning-based classification system is implemented that can identify and handle artifacts across ECG signals from different devices regardless of their specific voltage ranges, gains, or filters. This universal approach simplifies the normalization process by providing a device-agnostic method to detect and remove artifacts before normalization, enabling compatibility across multiple device types without increasing complexity.
3Reliability
If ECG wave signals are filtered to remove noise, then signal quality is improved, but false pauses are created when attenuated signal sections are completely removed
Solution Approach 1:
The system applies machine learning classification with feedback mechanisms to distinguish between genuine cardiac pauses and false pauses caused by filtering artifacts. By analyzing segmented ECG signals and using classification confidence thresholds, the system can identify when a pause-like pattern is actually an artifact and correct it, preserving true cardiac information while removing false artifacts.
Solution Approach 2:
Artifact classification and identification is performed as a preliminary step before final signal processing and normalization. By preemptively identifying and correcting false pauses through machine learning classification of segmented signals, the system prevents information loss before it occurs, ensuring that genuine cardiac pauses are preserved while filtering noise is removed.
Data Source
AI summary
A method includes classifying, using a machine learning model, a portion of an electrocardiogram measurement as an artifact. The method further includes normalizing the electrocardiogram measurement except the portion of the electrocardiogram measurement classified as the artifact. The method further includes applying the machine learning model to the normalized electrocardiogram measurement to detect a cardiac event.