Neural Network Atrial Fibrillation Detection in Single-Lead ECG
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing ECG devices are not well-suited for short single-channel recordings and require time to adjust processing parameters, making them inefficient for atrial fibrillation detection, and most devices are designed for multi-channel ECGs.
Innovation Solution
A system using a neural network to process a time-frequency representation of ECG signals, generated from single-lead ECG recordings, to classify rhythms including atrial fibrillation, normal sinus rhythm, and noisy recordings, with a signal quality assessment to filter out unacceptably noisy data, allowing for rapid detection and display of results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If existing ECG algorithms are used for short single-channel recordings, then device complexity is reduced, but detection accuracy and reliability deteriorate due to parameter adjustment time requirements
Solution Approach 1:
The patent transforms the ECG signal from the time domain to the frequency domain using Fast Fourier Transform (FFT), changing the representation parameters to enable rapid analysis without requiring time-consuming parameter adjustments. This frequency-domain approach allows immediate processing of short recordings while maintaining detection accuracy.
Solution Approach 2:
The patent replaces traditional time-domain signal processing algorithms with frequency-domain analysis using FFT and neural networks. This substitution eliminates the need for gradual parameter adjustment and enables direct classification of rhythm types from the frequency spectrum, improving both speed and accuracy for short recordings.
2Measurement precision
If multi-channel ECG devices are used, then measurement precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent extracts the essential rhythm classification capability from complex multi-channel ECG systems and applies it to single-channel recordings. By using frequency-domain analysis and neural networks, the system achieves comparable diagnostic accuracy with simplified single-lead devices, removing the necessity for multiple channels.
Solution Approach 2:
The patent creates a universal processing approach that works effectively with both single-channel and multi-channel ECG data. The frequency-domain transformation and neural network classification system can handle various input configurations, making the detection algorithm universally applicable while enabling simplified device design.
3Measurement precision
If ECG processing parameters are adjusted for each recording, then measurement precision is improved, but time consumption increases
Solution Approach 1:
The patent performs preliminary transformation of the ECG signal into the frequency domain using FFT before classification. This pre-processing step prepares the data in an optimized format that enables rapid neural network analysis, eliminating the need for time-consuming parameter adjustments during the actual classification process.
Solution Approach 2:
The patent implements continuous frequency-domain analysis that processes the entire ECG recording simultaneously rather than adjusting parameters incrementally. The neural network continuously evaluates the frequency spectrum to classify rhythm types, maintaining consistent processing speed and precision without interruptions for parameter optimization.
Data Source
AI summary
A non-transitory computer-readable medium stores instructions readable and executable by at least one electronic processor (20) to perform an atrial fibrillation (AF) detection method (100). The method includes: generating a time-frequency representation of an electrocardiogram (ECG) signal acquired over a time interval; processing the time-frequency representation using a neural network (NN) (32) to output probabilities for rhythms of a set of rhythms including at least atrial fibrillation; assigning a rhythm for the ECG signal based on the probabilities for the rhythms of the set of rhythms output by the neural network; and controlling a display device (24) to display the rhythm assigned to the ECG signal.