Deep Neural Network Synthetic ECG from PPG Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Photoplethysmography (PPG) signals used for heart rate estimation are often degraded by motion artifacts, leading to inaccurate heart rate estimation due to the overpowering of heart-beat related components, and existing methods for removing these artifacts can also discard important information.
Innovation Solution
A method involving a deep neural network (DNN) that generates a synthetic electrocardiography (ECG) signal from PPG signals using subject-specific training data, allowing for the conversion of PPG signals to ECG signals without feature extraction, thereby maintaining information and improving heart rate estimation accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If motion artifacts are removed from the PPG signal, then heart rate estimation accuracy is improved, but important information from the PPG signal is also lost
Solution Approach 1:
The patent introduces an intermediary system (deep neural network model) that transforms PPG signals into synthetic ECG signals. This intermediary transformation allows the system to indirectly obtain heart rate information in a form that is less susceptible to motion artifacts, thereby improving measurement precision without directly discarding the original PPG signal information.
Solution Approach 2:
The patent replaces the traditional mechanical signal processing approach (direct PPG analysis or artifact removal) with a computational intelligence approach (deep neural network). This substitution enables the system to learn complex non-linear relationships between PPG and ECG signals, achieving better heart rate estimation accuracy while preserving information through the learned transformation rather than through direct artifact removal.
2Ease of operation
If PPG sensors are used for heart rate monitoring, then device comfort and portability are improved, but motion artifacts deteriorate signal quality
Solution Approach 1:
The patent converts the harmful effect of motion artifacts into a beneficial outcome. Instead of treating motion artifacts as mere noise to be eliminated, the system uses the transformed PPG signal (via synthetic ECG generation) to robustly estimate heart rate even in the presence of motion. The motion conditions that would normally degrade PPG quality are transformed into acceptable input for the neural network model, which learns to extract reliable heart rate information despite the artifacts.
3Device complexity
If feature extraction is performed on the PPG signal, then heart rate determination is simplified, but information in the PPG signal is lost
Solution Approach 1:
The patent creates a copy of the heart rate information by generating a synthetic ECG signal from the PPG signal. Rather than extracting features directly from PPG (which loses information), the system creates a synthetic copy (ECG signal) that preserves the temporal and morphological characteristics needed for accurate heart rate determination, while the original PPG signal information is maintained throughout the transformation process.
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
This approach enables robust heart rate estimation by generating a synthetic ECG signal from PPG signals, reducing the impact of motion artifacts and providing accurate heart rate and variability metrics without the need for additional data, thus enhancing the reliability of heart activity analysis.
Implementation Method 1
Photoplethysmography (PPG) is a technology of interest in monitoring heart activity
Implementation Method 2
using associated pairs of a time-series of the PPG signal and a corresponding time-series of the ECG signal as input to a deep neural network (DNN)
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
A method of generating a model for generating a synthetic electrocardiography (ECG) signal comprises: receiving (102) subject-specific training data for machine learning, said training data comprising a photoplethysmography (PPG) signal acquired from the subject and an ECG signal acquired from the subject, wherein the ECG signal provides a ground truth of the subject for associating the ECG signal with the PPG signal; using (104) associated pairs of a time-series of the PPG signal and a corresponding time-series of the ECG signal as input to a deep neural network, DNN (400); and determining (106; 310), through the DNN (400), a subject-specific model relating the PPG signal of the subject to the ECG signal of the subject for converting the PPG signal to a synthetic ECG signal using the subject-specific model.