LSTM-Based AF Classification from Short Single-Lead ECG
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting Atrial Fibrillation (AF) from single lead electrocardiogram (ECG) recordings of short duration face challenges in accuracy due to vulnerability to noise and difficulty in automatic detection of P-R and P-Q intervals, leading to compromised performance in classification.
Innovation Solution
A processor-implemented method using a recurrent neural network architecture, specifically Long short-term memory (LSTM) networks and Bidirectional LSTM (BiLSTM) networks, combined with handcrafted statistical features, to analyze R-R intervals and P wave regions, creating a composite feature set for accurate classification of AF.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If single lead ECG recordings of short duration are used for AF classification, then the ease of operation and data acquisition is improved, but the measurement precision and classification accuracy deteriorates
Solution Approach 1:
The patent segments the ECG signal into multiple characteristic intervals (P-wave interval, QRS interval, T-wave interval, and post-T interval) and processes each segment separately through dedicated LSTM networks. This segmentation allows the system to extract specific features from each interval independently, improving classification accuracy even with short-duration single-lead ECG recordings by focusing computational attention on diagnostically relevant segments.
Solution Approach 2:
The patent transforms the temporal signal data into a multi-dimensional feature space by extracting multiple types of features (temporal features from LSTM networks, statistical features, frequency domain features) and combining them into a composite feature vector. This dimensional transformation enables the classifier to capture complex patterns across different feature dimensions, compensating for the limited temporal duration of the input signal.
2Measurement precision
If atrial analysis based approaches are used to detect AF, then the measurement precision is improved, but the reliability deteriorates due to vulnerability to background noise
Solution Approach 1:
The patent merges multiple detection approaches by combining atrial analysis (P-wave detection) with ventricular response analysis (R-R interval analysis) in a unified LSTM-based framework. The system processes both P-wave interval sequences and R-R interval sequences through separate LSTM networks and integrates their outputs, creating a robust classification system that maintains high precision while improving reliability through multi-modal feature fusion that is resilient to noise in any single modality.
Solution Approach 2:
The patent introduces LSTM networks as intermediary components that process raw ECG interval sequences and transform them into robust temporal features before classification. The LSTM networks act as intermediaries that filter out noise and extract meaningful temporal patterns, bridging the gap between raw noisy signals and reliable classification decisions, thereby maintaining measurement precision while improving reliability.
3Measurement precision
If manual detection of intermittent AF episodes is performed, then the measurement precision is improved, but the loss of time increases making it impractical
Solution Approach 1:
The patent replaces the manual mechanical inspection process with an automated computational system based on LSTM networks and machine learning classifiers. The system automatically processes ECG recordings, extracts features, and performs classification without human intervention, maintaining high measurement precision through sophisticated temporal pattern recognition while eliminating the time loss associated with manual review of long-duration recordings.
Solution Approach 2:
The patent implements a self-service automated detection system that processes ECG recordings independently without requiring cardiologist intervention. The LSTM-based system performs feature extraction, pattern recognition, and classification automatically, enabling the system to serve itself in detecting AF episodes while maintaining precision comparable to manual detection but with significantly reduced time loss.
Data Source
AI summary
Conventionally, Atrial Fibrillation (AF) has been detected using atrial analyses which is vulnerable to background noise. Again there is a dependency on statistical features which are extracted from R-R intervals of long ECG recordings. The present disclosure addresses AF detection from single lead short ECG recordings of less than one minute wherein automatic detection of P-R and P-Q intervals is difficult, which introduces error in feature computing from the segregated intervals and compromises the performance of the classifier. In the present disclosure, a Recurrent Neural Network (RNN) based architecture comprising two Long Short Term Memory (LSTM) networks is provided for temporal analysis of R-R intervals and P wave regions in an ECG signal respectively. Output sates of the two LSTM networks are merged at a dense layer along with a set of hand-crafted statistical features to create a composite feature set for classification of the AF.