Cascaded Binary Classifier for Single-Lead ECG Rhythm Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional systems for detecting atrial fibrillation (AF) in single-lead electrocardiogram (ECG) signals face challenges such as real-time deployment limitations, noise in ECG signals, inadequate dataset sizes, and misclassification due to the absence of consideration for non-AF abnormal rhythms.
Innovation Solution
The implementation of a processor-based method using a cascaded binary classifier that applies a spectrogram-based noisy data removal technique to extract clean ECG signals, followed by multi-layer cascaded binary classification with optimum feature selection at each layer to identify normal, AF, and other abnormal rhythms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional binary classification methods are used for AF detection, then the classification process is simple, but misclassification occurs due to non-AF abnormal rhythms being ignored
Solution Approach 1:
The patent divides the classification task into multiple binary classifiers arranged in a cascade structure. Each classifier handles a specific differentiation task (e.g., AF vs. non-AF, normal vs. abnormal), breaking down the complex multi-class problem into simpler binary decisions that improve overall accuracy while managing complexity through modular design.
Solution Approach 2:
The patent transitions from a single binary classification dimension to multiple classification dimensions by introducing a cascade of binary classifiers. This multi-dimensional approach allows the system to distinguish between AF, non-AF abnormal rhythms, and normal sinus rhythm through sequential binary decisions, effectively adding classification depth without proportionally increasing complexity.
2Measurement precision
If traditional AF detection methods are applied on clean selected data, then detection accuracy is improved, but real-time deployment becomes difficult due to noise in practical ECG signals
Solution Approach 1:
The patent applies noise removal preprocessing before the classification stage. By cleaning the ECG signal in advance through spectral analysis and noise filtering, the system prepares the data for more accurate classification while maintaining real-time capability through efficient signal processing algorithms that can operate on incoming streams.
Solution Approach 2:
The patent introduces an intermediate noise removal processing stage between signal acquisition and classification. This intermediary step uses spectral analysis and filtering techniques to eliminate noise components, creating a cleaner intermediate signal that facilitates more accurate subsequent classification without requiring the entire system to be redesigned for real-time operation.
3Reliability
If multi-class classification including non-AF abnormal rhythms is implemented, then misclassification is reduced, but the classification task becomes more challenging
Solution Approach 1:
The patent segments the multi-class classification problem into multiple binary classification tasks. Instead of attempting to classify all rhythm types simultaneously in a single complex model, the cascade structure divides the problem into sequential binary decisions (e.g., first distinguish AF from non-AF, then differentiate normal from abnormal), reducing the complexity of each individual classification task while achieving comprehensive multi-class discrimination.
Solution Approach 2:
The patent resolves the complexity issue by transforming the multi-class classification problem into multiple binary classification dimensions. Each binary classifier operates on a specific dimension of rhythm differentiation, and the cascade structure combines these dimensional decisions to achieve complete multi-class classification. This approach reduces the computational difficulty of each individual classifier while maintaining overall classification accuracy.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Current technologies analyze electrocardiogram (ECG) signals for a long duration, which is not always a practical scenario. Moreover the current scenarios perform a binary classification between normal and Atrial Fibrillation (AF) only, whereas there are many abnormal rhythms apart from AF. Conventional systems/methods have their own limitations and may tend to misclassify ECG signals, thereby resulting in an unbalanced multi-label classification problem. Embodiments of the present disclosure provide systems and methods that are robust and more efficient for classifying rhythms for example, normal, AF, other abnormal rhythms and noisy ECG recordings by implementing a spectrogram based noise removal that obtains clean ECG signal from an acquired single-lead ECG signal, an optimum feature selection at each layer of classification that selects optimum features from a pool of extracted features, and a multi-layer cascaded binary classifier that identifies rhythms in the clean ECG signal at each layer of the classifier.