AI Oxygen Saturation Detection from Time-Frequency PPG Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing pulse oximetry methods struggle to accurately determine oxygen saturation levels due to the complexity of processing photoplethysmographic (PPG) signals, particularly when using raw data without adequate preprocessing.
Innovation Solution
An AI-based method employing time-frequency transforms, such as wavelet transforms, to generate input feature matrices from red and infrared PPG signals, followed by training a neural network, specifically a deep learning model like CNN, to enhance the determination of oxygen saturation levels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional pulse oximetry methods are used to process PPG signals, then the processing is simpler, but the accuracy of oxygen saturation level determination deteriorates
Solution Approach 1:
The patent applies preliminary action by performing time-frequency transforms (wavelet transforms, short-time Fourier transforms, Wigner-Ville transforms) on PPG signals before neural network processing. This preprocessing step extracts temporal and spectral features in advance, enabling the neural network to focus on pattern recognition rather than raw signal processing, thereby improving measurement precision while managing complexity through structured feature extraction
Solution Approach 2:
The patent introduces an intermediary neural network model that acts as a mediator between raw PPG signals and oxygen saturation measurements. The neural network, trained on preprocessed signal features, serves as an intelligent intermediary that learns complex non-linear relationships, improving measurement accuracy while encapsulating processing complexity within a trained model rather than requiring complex real-time processing algorithms
2Measurement precision
If raw PPG data is used without preprocessing, then the processing is faster, but the accuracy of oxygen saturation predictions deteriorates
Solution Approach 1:
The patent performs preliminary time-frequency analysis and feature extraction from PPG signals before neural network inference. By preprocessing signals offline or in advance to extract temporal and spectral features, the system improves prediction accuracy while reducing the computational burden during actual measurement, thus minimizing time loss during critical monitoring moments
Solution Approach 2:
The patent employs dynamic feature extraction through time-frequency transforms that adapt to the temporal characteristics of PPG signals. The neural network dynamically processes extracted features rather than raw signals, allowing the system to balance processing speed and accuracy by focusing computational resources on the most informative signal characteristics rather than processing all raw data uniformly
3Loss of information
If time-frequency transforms are applied to PPG signals, then the feature extraction is more comprehensive, but the computational complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the complex signal processing task into distinct stages: time-frequency transform, feature extraction, and neural network classification. Each stage processes specific aspects of the signal independently, ensuring comprehensive feature extraction while managing computational complexity through modular processing rather than attempting to analyze all signal aspects simultaneously
Solution Approach 2:
The patent extracts only the most relevant temporal and spectral features from PPG signals through time-frequency transforms, rather than processing the entire raw signal spectrum. By selectively extracting informative features and discarding redundant information, the system achieves comprehensive feature extraction for accurate oxygen saturation measurement while reducing unnecessary computational complexity
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
Improves the accuracy and precision of oxygen saturation level predictions by effectively capturing subtle signal features, outperforming traditional methods by providing more precise temporal and frequency analysis.
Implementation Method 1
Oximetry is an optical method for measuring oxygenated hemoglobin in blood. Oximetry is based on the ability of different forms of hemoglobin to absorb light of different wavelengths.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Implementations described herein disclose an artificial intelligence (AI) based method for generating an oxygen saturation level output signal using the trained neural network. In one implementation, the method includes receiving a photopl ethy smographi c (PPG) signal, the PPG signal including a red PPG signal and an infrared PPG signal, generating an input feature matrix by performing time-frequency transform of the PPG signal, training a neural network using the input feature matrix and an oxygen saturation level input signal, and generating an oxygen saturation level output signal using the trained neural network.