MobileNet ECG Classification via Wavelet Transform
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for diagnosing heart diseases using ECG data are cumbersome, requiring extensive data analysis and large hardware, making early detection challenging and costly, especially when performed without a personal computer or large device.
Innovation Solution
An apparatus and method utilizing a MobileNet with wavelet transformation to classify ECG signals in a frequency domain, allowing for classification of heart conditions like AFIB, LBBB, NSR, and PVC on smartphones or smartwatches, preventing hardware performance deterioration and ensuring sufficient training data through algorithms like matching pursuit.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a deep learning structure with significantly large size is used for ECG analysis, then classification accuracy may be improved, but hardware performance is deteriorated
Solution Approach 1:
The patent segments the deep learning model into a lightweight MobileNet architecture specifically designed for mobile devices. This segmentation allows the system to maintain high classification accuracy for heart disease detection while reducing the computational burden and hardware requirements compared to larger models
Solution Approach 2:
The patent changes the parameters of the neural network by selecting MobileNet with its specific hyperparameters (128 hidden units, ReLU activation, dropout rate of 0.5) optimized for mobile deployment. This parameter optimization enables the model to achieve high accuracy while maintaining low computational complexity suitable for smartphone processors
2Measurement precision
If ECG is measured for a long period of time to secure sufficient data, then detection accuracy is improved, but data analysis complexity and storage requirements increase
Solution Approach 1:
The patent applies preliminary action by using data augmentation techniques during the training phase to artificially expand the training dataset. This includes adding noise, scaling, and other transformations to existing ECG data, which allows the model to learn from diverse patterns without requiring actual long-term continuous ECG monitoring from patients
Solution Approach 2:
The patent uses copying by creating synthetic ECG data through augmentation methods. Instead of collecting large amounts of real patient data over extended periods, the system creates copies and variations of existing ECG signals, maintaining data diversity and training effectiveness while avoiding the complexity of long-term data collection and management
3Measurement precision
If traditional ECG measurement methods are used in hospitals, then diagnostic accuracy is improved, but patient convenience and accessibility are reduced
Solution Approach 1:
The patent enables self-service by deploying the heart disease detection system directly on patient-owned smartphones and smartwatches. Users can independently perform ECG monitoring and receive diagnostic results without requiring hospital visits or professional operator intervention, making healthcare accessible at home
Solution Approach 2:
The patent achieves universality by designing a mobile application that can be used by any patient with a smartphone or smartwatch, regardless of location or medical background. The system performs multiple functions including ECG signal acquisition, preprocessing, classification, and result display within a single portable platform, replacing the need for specialized hospital equipment
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables efficient, periodic cardiovascular health management on mobile devices with high classification accuracy, preventing hardware degradation and securing a sufficient training dataset, achieving 92.65% classification accuracy using 12-fold cross validation.
Implementation Method 1
a wavelet transformation unit configured to transform the ECG signal in the time domain into an ECG signal in a frequency domain
Data Source
AI summary
An apparatus for classifying heart disease using a MobileNet according to an embodiment of the present invention may comprise: an input unit for receiving a time-domain electrocardiogram signal; a wavelet transform unit for transforming the timedomain electrocardiogram signal into a frequency-domain electrocardiogram signal by using a wavelet transform; and a neural network for classifying the frequency-domain electrocardiogram signal as one of atrial fibrillation (AFIB), left bundle branch block beat (LBBB), normal sinus rhythm (NSR), or premature ventricular contraction (PVC), wherein the neural network may be a MobileNet trained using a training data set.