Cuff Waveform Imaging for Noise-Robust Blood Pressure Measurement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Pulse waves obtained from cuff pressure often include noise due to body motion or cuff positioning, which degrades the accuracy of blood pressure measurement.
Innovation Solution
A sphygmomanometer that uses a cuff to measure cuff pressure and generate a waveform image with both global and local features of the pulse waveform, utilizing machine learning to construct a calculation model that outputs accurate blood pressure despite noise, by indicating cuff pressure through line or background features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If pulse wave measurement is performed using cuff pressure in the oscillometric method, then blood pressure can be measured non-invasively, but measurement accuracy is degraded when noise is included in the pulse wave
Solution Approach 1:
The patent transforms the one-dimensional time-series pulse wave signal into a two-dimensional waveform image, adding a visual dimension that enables simultaneous representation of global waveform characteristics and local features. This dimensional transformation allows the system to capture both overall pulse wave patterns and subtle local variations that may be obscured in traditional time-series analysis, thereby improving measurement accuracy despite noise presence.
Solution Approach 2:
The waveform image is segmented to preserve both global characteristics (overall waveform shape, amplitude, and timing) and local features (specific waveform segments, inflection points, and fine structures). This segmentation approach allows the machine learning model to analyze different portions of the pulse wave independently while maintaining contextual relationships, enabling robust blood pressure estimation even when certain segments contain noise.
2Measurement precision
If traditional oscillometric method is used for blood pressure measurement, then the measurement process is simple, but accuracy is insufficient when noise is present
Solution Approach 1:
The patent replaces traditional mechanical signal processing methods with machine learning-based image analysis. Instead of using conventional algorithms to process time-series pressure signals, the system generates waveform images and applies trained machine learning models to estimate blood pressure. This substitution enables the system to automatically learn complex patterns and relationships from data, improving accuracy while maintaining relative simplicity in implementation.
Solution Approach 2:
The patent changes the representation parameters of the pulse wave from time-series amplitude values to spatial-image parameters. By encoding the pulse wave as a waveform image with visual features, the system transforms the problem from signal processing to image analysis, enabling the use of powerful computer vision techniques and pre-trained image recognition models for blood pressure estimation.
3Measurement precision
If waveform image with both global and local features is generated, then blood pressure measurement accuracy is improved, but data processing complexity increases
Solution Approach 1:
The patent performs preliminary actions by pre-processing the time-series pulse wave data into waveform images before analysis. This preprocessing step organizes the data into a standardized visual format that preserves both global and local features, making subsequent machine learning analysis more efficient. By preparing the data in advance in an optimized format, the system reduces the computational burden during actual blood pressure measurement and estimation.
Data Source
Figure 1
Figure 2
Figure 3A
AI summary
The sphygmomanometer includes: a cuff wound around an arm of a measurement subject; a control unit configured to control a pressure applied to the arm by the cuff; a detection unit configured to detect a cuff pressure; a generation unit configured to generate a first waveform image that indicates the cuff pressure; a calculation model constructed by machine learning in which a second waveform image that includes a second image indicating a change in pressure brought by pulsation of a blood flow flowing through an artery of a subject, and that indicates the cuff pressure with features of the second image is used as an explanatory variable, and a blood pressure of the subject is used as a response variable; and an acquisition unit configured to acquire a blood pressure of the measurement subject from the calculation model by inputting the first waveform image to the calculation model.