Atrial Fibrillation Signal Recognition Using Multi-Scale CNNs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for recognizing atrial fibrillation signals in electrocardiograms suffer from low reliability and accuracy, especially in dynamic signals from wearable devices, due to imbalanced data and the need for manual processing and long signal durations.
Innovation Solution
The method employs a pre-established atrial fibrillation signal recognition model using synthetic minority oversampling technique (SMOTE) to balance data and multiple convolutional neural networks with varying receptive fields for comprehensive signal recognition, allowing for automatic and efficient detection of atrial fibrillation signals without manual intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional methods detect variability of P wave or RR interval, then atrial fibrillation signal can be recognized, but recognition reliability and accuracy are insufficient
Solution Approach 1:
The patent divides the recognition task into multiple parallel convolutional neural networks, each with different receptive fields (e.g., 7, 11, 15, 19, 23 samples). This segmentation allows the system to analyze signal features at multiple scales simultaneously, improving both reliability and accuracy by capturing different temporal patterns of atrial fibrillation.
Solution Approach 2:
The patent combines multiple convolutional neural networks with different receptive fields into a composite recognition system. Each network processes the same input signal but with different temporal windows, and their results are integrated to produce a final recognition outcome, similar to how composite materials combine different properties to achieve superior overall performance.
2Reliability
If machine learning method is used for atrial fibrillation recognition, then recognition capability is improved, but long electrocardiogram signal and high data processing requirements are needed
Solution Approach 1:
The patent uses partial action by employing multiple CNNs with different receptive fields that collectively cover the necessary temporal range. Instead of requiring one extremely large receptive field (excessive action), the system uses several smaller fields that together provide comprehensive coverage, reducing the required signal duration while maintaining recognition capability.
Solution Approach 2:
The patent transforms the temporal dimension problem into a parallel processing dimension. Rather than extending signal duration to capture all necessary features, it creates multiple parallel processing paths (different CNNs) that simultaneously analyze the signal at different temporal scales, effectively adding a processing dimension that reduces temporal requirements.
3Reliability
If machine learning method is used for atrial fibrillation recognition, then recognition capability is improved, but manual extraction of signal characteristics and noise reduction processing are required
Solution Approach 1:
The convolutional neural networks perform automatic feature extraction and noise filtering without manual intervention. The CNN architecture inherently learns to identify relevant signal characteristics and suppress noise during training, making the system self-sufficient for preprocessing tasks that traditionally required manual expert input, thereby reducing operational complexity.
Solution Approach 2:
The patent replaces manual mechanical processes (hand-crafted feature extraction and noise reduction algorithms) with an automated neural network system. The CNN automatically learns optimal feature representations and filtering strategies from data, substituting the need for manual signal processing expertise and reducing the complexity of data preparation.
4Measurement precision
If multiple convolutional neural networks with different receptive fields are used, then recognition accuracy is improved, but model complexity increases
Solution Approach 1:
The patent uses multiple CNNs with different receptive fields, where each network serves a universal purpose (atrial fibrillation detection) but with specialized temporal coverage. This multi-functional approach allows the system to maintain relatively simple individual network structures while achieving high accuracy through their combined operation, as each network remains computationally efficient despite the ensemble complexity.
Data Source
AI summary
The present disclosure provides an atrial fibrillation signal recognition method, apparatus and device. The method comprises: obtaining an electrocardiogram signal to be recognized; inputting the electrocardiogram signal to be recognized to a pre-established atrial fibrillation signal recognition model, and outputting an atrial fibrillation signal recognition result, where the atrial fibrillation signal recognition model is established in the following way: obtaining a specified number of electrocardiogram sample signals and corresponding identifier information; balancing, according to the number of normal signals, atrial fibrillation signals by means of SMOTE; establishing a network structure of multiple convolutional neural networks, each of the convolutional neural networks being provided with a specific receptive field for recognizing the atrial fibrillation signals of a corresponding granularity; and inputting the normal signals and the balanced atrial fibrillation signals to the network structure for training to generate an atrial fibrillation signal recognition model.