PPG Blood Pressure Model Training for Subject-Independent Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing blood pressure estimation methods using PPG signals struggle to accurately estimate highly variable blood pressure due to subject dependency and lack of reliability, particularly in cuff-less wearable devices.
Innovation Solution
A blood pressure estimation model learning system utilizing a convolutional neural network (CNN) that preprocesses PPG data and blood pressure signals with subject-independent variance, including steps to remove abnormalities, down-sample, segment, and normalize data, using two 1D-CNNs and a fully connected layer to construct a reliable model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If subject-dependent learning models are used for blood pressure estimation, then the model can be trained with available data, but the estimation reliability deteriorates when blood pressure varies significantly in individual subjects
Solution Approach 1:
The patent transforms the learning model from subject-dependent to subject-independent by incorporating data from multiple subjects with diverse blood pressure characteristics. The model learns universal patterns of blood pressure variation across different individuals rather than adapting to specific subject characteristics, enabling reliable estimation for highly variable blood pressure in any subject.
Solution Approach 2:
The patent changes the fundamental parameter of data selection criteria from subject-specific to subject-independent. By selecting data based on blood pressure variation characteristics rather than subject identity, the model learns to recognize patterns applicable across different subjects, improving reliability for highly variable blood pressure estimation.
2Measurement precision
If cuff-based blood pressure measurement is used, then measurement accuracy can be maintained, but device portability and convenience deteriorate due to the necessary inclusion of a cuff
Solution Approach 1:
The patent replaces the mechanical cuff-based measurement system with an optical PPG sensor-based system. Instead of using mechanical pressure application and detection through a cuff, the invention uses optical sensors to detect blood volume changes and employs machine learning to estimate blood pressure from these optical signals, eliminating the need for a mechanical cuff while maintaining measurement capability.
Solution Approach 2:
The patent introduces PPG signals as an intermediary between the optical sensor and blood pressure measurement. Rather than directly measuring blood pressure mechanically, the system uses PPG signals as an intermediate indicator that correlates with blood pressure, allowing indirect measurement through optical detection and computational estimation.
3Ease of operation
If existing PPG-based blood pressure estimation models are used, then portability and convenience are improved, but estimation accuracy deteriorates for highly variable blood pressure due to subject dependency
Solution Approach 1:
The patent enhances the portability advantage of PPG-based systems while correcting the accuracy limitation by making the learning model universal rather than subject-specific. The model incorporates data from multiple subjects with diverse blood pressure characteristics, enabling the portable device to accurately estimate blood pressure even when individual subject variation is high, thus maintaining both convenience and accuracy.
Solution Approach 2:
The patent changes the data selection parameter from subject-dependent to subject-independent, allowing the portable PPG device to leverage diverse population data for training. This enables the device to generalize better across different subjects and blood pressure conditions, improving accuracy for highly variable blood pressure while maintaining the portability benefits of PPG-based measurement.
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
The system enables highly reliable blood pressure estimation even in unstable environments by leveraging subject-independent data, improving accuracy and portability in wearable devices.
Implementation Method 1
research using PPG (Photoplethysmography) optical sensors has been increasing recently, because vascular elasticity information can be found using characteristic values such as percussion wave, tidal wave, etc. in the PPG signal
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
An embodiment of the present invention relates to a method and system for learning a blood pressure estimation model using photoplethysmography signal (PPG). The blood pressure estimation model learning method includes the steps of: preprocessing raw data including blood pressure signal data as well as PPG data collected from subjects; and constructing a blood pressure estimation model based on the preprocessed blood pressure signal data and PPG data. The raw data includes blood pressure signal data, a variance level of which is greater than or equal to a predetermined value.