Three-way Rhythm Classifier for Ambulatory ECG Signal Confidence
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current low-cost wireless single-lead ECG monitoring systems struggle to accurately detect heart rates in patients with atrial fibrillation due to motion artifacts, leading to false alerts and failure to trigger necessary notifications, as they often classify abnormal rhythms as noise, which is dangerous given the severity of atrial fibrillation.
Innovation Solution
A computer-implemented method that determines a heart rate from an ECG signal and enhances confidence estimation by classifying segments into noise, normal rhythm, and abnormal rhythm, using a three-way classification approach to discriminate between noise and abnormal rhythms, allowing for reliable identification of atrial fibrillation and triggering alerts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If motion artifact rejection algorithms are used to improve signal quality, then false alerts are reduced, but abnormal rhythms like atrial fibrillation are incorrectly classified as noise
Solution Approach 1:
The ECG signal is divided into segments, and each segment is independently classified into one of three categories: noise, normal rhythm, or abnormal rhythm. This segmentation allows the system to process and evaluate different portions of the signal separately, improving the accuracy of abnormal rhythm detection while maintaining motion artifact rejection capabilities.
Solution Approach 2:
Different classification criteria are applied to different segments of the ECG signal based on their local characteristics. The system evaluates features such as RR interval variability, amplitude, and morphology in each segment to determine whether it represents noise, normal rhythm, or abnormal rhythm, allowing for localized optimization of detection accuracy.
2Measurement precision
If confidence level thresholds are used to filter heart rate measurements, then measurement precision is improved, but false negatives in abnormal rhythm detection increase
Solution Approach 1:
The system changes the parameter used for classification from a single confidence level threshold to a three-category classification system. Instead of simply accepting or rejecting measurements based on a threshold, the system evaluates multiple features (RR intervals, amplitude, morphology) to classify segments into noise, normal rhythm, or abnormal rhythm categories, thereby improving both precision and reliability.
3Device complexity
If single-lead ECG monitoring is used to reduce device complexity, then cost and portability are improved, but motion artifact rejection capability deteriorates
Solution Approach 1:
The single-lead ECG system performs self-diagnosis and self-correction by analyzing the signal's own characteristics. The algorithm evaluates features such as RR interval variability, signal amplitude, and waveform morphology to automatically distinguish between motion artifacts and genuine cardiac rhythms, eliminating the need for additional reference leads or complex external correction systems.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method of determining a confidence level in a heart rate obtained from an ECG signal of a patent comprises determining a heart rate from a segment of an ECG signal, the segment having a pre-determined length, and determining a first confidence level in the determined heart rate. If the first confidence level exceeds a predetermined value, the heart rate is output. Otherwise, the method further comprises determining a set of features from the segment of the ECG signal; and classifying the segment of the ECG signal, based on one or more features selected from the determined set of features and the heart rate, into one of at least three classes: noise, an abnormal sinus rhythm and a normal sinus rhythm. The present invention makes it possible to discriminate between an abnormal rhythm and noise (as well as between a normal rhythm), whereas conventional methods can only discriminate between noise and a normal heart rhythm. The invention thus makes it possible adequately to process the abnormal heart rate value resulting from an aperiodic heart rhythm such as Atrial Fibrillation and trigger a notification/alert during its onset.