CN116548935A

CN116548935A - Blood pressure measurement system based on flexible organic light detector and deep learning algorithm

Info

Publication number: CN116548935A
Application number: CN202310441323.3A
Authority: CN
Inventors: 李嘉宝; 俞祝良
Original assignee: South China University of Technology SCUT
Current assignee: South China University of Technology SCUT
Priority date: 2023-04-23
Filing date: 2023-04-23
Publication date: 2023-08-08

Abstract

The invention discloses a blood pressure measurement system based on a flexible organic light detector and a deep learning algorithm, which comprises a data acquisition module, a data processing module and a blood pressure calculation module, wherein the data acquisition module is used for acquiring physiological signal data and transmitting the physiological signal data to the data processing module; and after the data processing module preprocesses the data, the data is transmitted to the signal quality analysis module to evaluate the quality of the preprocessed signal, and the evaluated signal is transmitted to the blood pressure calculation module to calculate the blood pressure value. The invention can realize continuous monitoring of the blood pressure value based on the PPG signal and the ECG signal of the human body, reflects the blood circulation condition of the human body, and the ECG signal reflects the electric activity condition of the heart.

Description

Blood pressure measurement system based on flexible organic light detector and deep learning algorithm

Technical Field

The invention belongs to the technical field of human blood pressure measurement and the field of intelligent equipment, and particularly relates to a blood pressure measurement system based on a flexible organic light detector and a deep learning algorithm.

Background

With the rapid increase of the aged population, the prevention and treatment of hypertension-related diseases becomes an increasingly important public health problem. Hypertension increases the risk of various diseases such as heart disease and cerebral apoplexy. However, continuous long-term monitoring of blood pressure has long been difficult to achieve, and the most commonly used blood pressure measuring devices are still of the cuff type. The blood pressure measuring device is more portable and easy to use, can help to realize long-term continuous blood pressure monitoring and better health management functions, and is very important for improving the life quality of people.

In the traditional cuff type blood pressure measurement method, the cuff is required to be accurately worn, and meanwhile, the arm is inflated and pressurized to obtain a blood pressure measurement result; mercury sphygmomanometers also require an experienced physician to use properly. The blood pressure measurement mode has long measurement time, complicated process and poor measurement experience, and the measurement result is easily influenced by other factors and cannot continuously monitor the blood pressure of a human body.

The invasive arterial pressure monitoring method is a current gold standard measuring method. The blood pressure value of the human body can be continuously obtained, but the blood pressure value is required to be obtained by arranging a catheter with a sensor in an artery of the human body, so that the blood pressure value can be damaged, and the blood pressure value is generally limited to shock, serious patients or life-threatening operation patients, and is not suitable for daily blood pressure monitoring. In terms of blood pressure calculation, the existing blood pressure calculation method generally determines parameters of a model based on an empirical formula and a statistical method, so that errors of blood pressure measurement results are different for different people.

There is currently no sophisticated sleeveless continuous blood pressure measuring device. Recent developments in deep learning technology have provided new ideas for sleeveless continuous blood pressure measurement. The blood pressure calculation method based on deep learning can be trained through a large amount of physiological signal data, so that a more accurate blood pressure calculation model is obtained. The most promising application is now the method of calculating blood pressure based on PPG and ECG. Training based on a large amount of physiological data also enables the generalization performance of the model to be better, and the model can be suitable for a wider population.

The existing blood pressure measuring device has complicated measuring process and needs to be pressurized on the arm, so that the measuring process is uncomfortable and easy to cause discomfort or pain (CN 202310088356.4). When the traditional cuff type sphygmomanometer is used, the cuff is required to be worn accurately, otherwise errors can be caused, and the arm can cause larger pressing feeling and even damage to the skin under the condition that the cuff pressure is too high. Meanwhile, the measurement result of this method is easily affected by other factors such as body posture, emotion, and other physical conditions. The traditional blood pressure measurement method needs to wear sleeves on arms and apply pressure to measure blood pressure values, so that the traditional blood pressure measurement method is poor in use experience and is not suitable for repeated measurement and continuous monitoring. The traumatic blood pressure monitoring can realize continuous monitoring of blood pressure, but the method can cause harm to human bodies, is generally only used for occasions needing high attention after operation and the like and having medical staff to ensure safety, and is not suitable for daily monitoring scenes.

In addition, the method can not continuously monitor the blood pressure of the human body, only one measurement result can be obtained in each measurement, and the dynamic change of the blood pressure can not be monitored in real time. Multiple measurements can cause injury to the human body and are inconvenient, especially in a scene where continuous night monitoring is required, which can cause great trouble to rest of the human body.

The existing continuous noninvasive blood pressure measurement schemes need to calculate model parameters according to the characteristics of each person, and different ginseng numbers can be greatly different, so that the scheme of the traditional noninvasive blood pressure measurement method is poor in adaptability and cannot be widely applied directly.

Disclosure of Invention

In order to overcome the defects of the prior art, the invention provides a blood pressure measurement system based on a flexible organic light detector and a deep learning algorithm, and relates to physiological signal acquisition equipment and an intelligent blood pressure calculation method.

The invention is realized at least by one of the following technical schemes.

The blood pressure measurement system based on the flexible organic light detector and the deep learning algorithm comprises a data acquisition module, a data processing module and a blood pressure calculation module, wherein the data acquisition module is used for acquiring physiological signal data signals and transmitting the physiological signal data signals to the data processing module; and after the data processing module preprocesses the data, the data is transmitted to the signal quality analysis module to evaluate the quality of the preprocessed signal, and the evaluated signal is transmitted to the blood pressure calculation module to calculate the blood pressure value.

Further, the data acquisition module acquires PPG signals by using a flexible organic photodetector, and acquires ECG signals by using an ECG module.

Further, the data acquisition module transmits the data to the data processing module in a wired or wireless mode.

Further, the data processing module performs filtering and noise reduction processing on the signals.

Further, the signal quality analysis module evaluates the PPG signal by a self-encoder.

Further, the self-encoder is trained using the normal PPG signal by minimizing the mean square error of the output, and the quality of the signal is evaluated by reconstructing the similarity of the signal to the input signal.

Further, the signal quality analysis module evaluates the ECG signal by detecting QRS peaks, which signal is normal if the number of peaks within the threshold is detected, otherwise abnormal.

Further, different signal characteristics are extracted through different encoders according to the blood pressure calculation module, the different signal characteristics are fused, and finally, an ABP waveform is obtained through prediction, and the values of the systolic pressure SBP, the diastolic pressure DBP and the mean arterial pressure MBP are extracted from the ABP waveform.

Further, the physiological signal data signals acquired by the data acquisition module are stored in a local data storage module.

Further, after the user agrees with the authorization, the data storage module provides a function of synchronizing to the cloud database, so that the subsequent checking or the physiological signal data analysis is facilitated.

Compared with the prior art, the invention has the beneficial effects that:

according to the blood pressure measuring method based on the human PPG signal and the ECG signal, excessive pressure is not required to be applied, only a flexible organic light detector is required to be worn on a wrist part or other body parts suitable for measurement, and a plurality of electrodes for measuring the ECG signal are required to be worn on the body, so that the blood pressure measuring method is more comfortable in use experience and suitable for continuously, noninvasively and sleevelessly monitoring the blood pressure value of a human body.

The invention can realize continuous monitoring of the blood pressure value based on the PPG signal and the ECG signal of the human body. The PPG signal reflects the blood circulation condition of a human body, the ECG signal reflects the electric activity condition of the heart, and the two signals are combined by using a deep learning method to accurately predict the blood pressure value.

In addition, the flexible sensor used in the invention is softer, lighter and thinner, comfortable to wear, can be easily fixed on the skin of a human body, is convenient to use, and can realize long-term monitoring in daily life. Compared with the traditional blood pressure measuring method, the method provided by the invention has higher measuring precision and stability, accords with the human engineering principle better and is more comfortable and convenient to use.

Drawings

The invention is further described below with reference to the drawings and embodiments:

FIG. 1 is a block diagram of a blood pressure measurement system based on a flexible organic photodetector and a deep learning algorithm according to an embodiment;

FIG. 2 is a flow chart of a specific data processing of the data processing module according to an embodiment;

FIG. 3 is a flowchart illustrating an embodiment of a signal quality analysis module for analyzing signal quality;

FIG. 4 is a flowchart showing a specific calculation of a blood pressure value according to two collected physiological signals by the blood pressure calculation module according to the embodiment;

FIG. 5 is a flowchart illustrating the operation of a blood pressure measurement system based on a flexible organic photodetector and a deep learning algorithm according to an embodiment.

Detailed Description

In order that those skilled in the art will better understand the present invention, the following description will be given in detail with reference to the accompanying drawings and detailed description. It will be apparent that the described embodiments are only some, but not all, embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

As shown in fig. 1, the blood pressure measurement system based on the flexible organic photodetector and the deep learning algorithm in this embodiment includes three major parts, namely a data acquisition module, a data processing module and a data storage module.

The data acquisition module is used for acquiring the PPG signal and the ECG signal and transmitting the data to the data processing module through the data transmission module.

In this embodiment, the data acquisition module in the blood pressure measurement system is configured to acquire the PPG signal and the ECG signal synchronously, and transmit the signals to the data processing module in a fixed encoding format.

As a preferred embodiment, the PPG acquisition module uses a novel flexible organic photodetector as a sensor to generate PPG signals by photoplethysmography, and can acquire PPG signals at the wrist of the human body or other suitable acquisition sites.

The flexible organic photodetector can adapt to the shape of the wrist or other parts, which improves the accuracy and stability of signal acquisition. A flexible organic photodetector is used to collect PPG signals and a comfortable electrode is used to collect ECG signals. Therefore, the pressure is not needed to press the blood vessel, and the arm is not needed to be punctured, so that the damage to the human body possibly caused by the traditional measurement method is avoided. Compared with the traditional equipment, the equipment is more comfortable in the use process, and can not bring pain and discomfort to the user. Measuring PPG signals from the radial artery at the wrist location or other body parts can mitigate the effects experienced in these special cases and thus the use of the scene is broader.

The ECG signal may be acquired using I leads, II leads, or other multi-lead means.

The data acquisition module encodes and transmits physiological signal data (PPG signal and ECG signal), which may be wireless or wired. The signal data needs to be transmitted in an encrypted manner if necessary.

The data processing module completes the preprocessing work of the data, then analyzes the quality of the signals, the blood pressure calculation module starts to calculate the blood pressure value, and finally the result display module displays the acquired signals and the calculated results. If necessary, the data storage module completes the storage of physiological signal data and the cloud synchronization of the data.

The data processing module is responsible for processing the collected physiological signal data, including filtering, noise reduction and other operations on the signals, and different processing methods are adopted for different signals. The data processing module shown in fig. 2 is a specific data preprocessing flow. In the module, preprocessing work comprises analysis of PPG signals and ECG signals, filtering of different forms is carried out on the two signals respectively, and the filtered signals are sent to a signal quality analysis module.

When processing physiological signal data, firstly, a data packet with a fixed format needs to be analyzed to obtain specific data information of PPG and ECG. Then, filtering processing is required for the PPG and ECG signals.

And (3) performing time alignment on different signal data, and then splicing or slicing according to a fixed length so as to meet the calculation requirement of a subsequent model.

And a band-pass filter, a notch filter and other methods are used for removing noise in the PPG signal, so that the signal quality is improved.

Noise in the ECG signal is removed by using a notch filter, a morphological filter and other methods, and the signal quality is improved.

The signal quality analysis module evaluates the quality of the pre-processed signal to ensure that the signal is sufficiently reliable and accurate.

Fig. 3 is a specific flow chart of signal quality analysis in the signal quality analysis module. In this module, the quality of the PPG signal is evaluated by a self-encoder network and the quality of the ECG signal is evaluated by detecting the quality of the QRS peak.

Specifically, the processed PPG signal is input into a trained self-encoder for reconstruction, and then the signal quality of the reconstructed PPG signal and the input PPG signal is estimated, so as to give an estimation result. The higher the score, the better the signal quality, and the more reliable the subsequent blood pressure calculation results. The scoring ratio gives early warning in advance when a set threshold value (the threshold value can be adjusted in real time).

The self-encoder is trained in advance using the normal PPG signal, by minimizing the output mean square error. When a normal PPG signal is input to the self-encoder, the error of the reconstructed signal is small, and when an abnormal PPG signal is input, the error of the reconstruction is large. The quality of the signal is assessed by reconstructing the similarity of the signal to the input signal.

The invention uses an intelligent signal quality assessment method and a blood pressure calculation method. The physiological signals can be automatically analyzed and processed, the characteristic values are extracted, and the blood pressure value is calculated according to the characteristic values. Meanwhile, the device can monitor and feed back data in real time.

Methods based on signal quality index characteristics may also be used to evaluate the quality of the signal.

QRS peak detection is performed on the ECG signal, and the number of peaks within the threshold is detected as normal, otherwise abnormal. The method can also be used for detecting whether the abnormal value exists in the ECG signal, whether the missing value exists or not and the like to assist in signal quality detection.

And inputting the PPG signal and the ECG signal subjected to quality evaluation into a blood pressure calculation module for blood pressure calculation.

Fig. 4 is a specific calculation flow of the blood pressure calculation module to calculate a blood pressure value according to two acquired physiological signals. In this module, the PPG signal and the ECG signal go through two different encoders (encoder 1, encoder 2) respectively, resulting in two different sets of features; and then fusing the characteristics of the two signals, and finally predicting to obtain an ABP waveform. Values of systolic pressure SBP, diastolic pressure DBP and mean arterial pressure MBP can be extracted from the ABP waveform. These results are fed into a results display module.

The encoder 1 is used to extract features of the PPG signal. The encoder 2 is used to extract features of the ECG signal.

The encoder may be a conventional convolutional neural network, a modified convolutional neural network (e.g., resnet), an LSTM network, a transducer, or the like.

The features of the PPG and ECG are fused in a feature fusion module.

And sending the data after feature fusion into a blood pressure prediction model to predict continuous ABP waveforms.

And calculating indexes related to blood pressure such as systolic pressure SBP, diastolic pressure DBP, mean arterial pressure MBP and the like according to the ABP waveform.

The specific calculation formula is as follows:

SBP＝max(ABP)

DBP＝min(ABP)

SBP and DBP can also be obtained by first obtaining the peak value and the estimated value of ABP and respectively averaging.

And sending the result to a result display module, and displaying the results of the PPG waveform, the ECG waveform, the ABP waveform, the blood pressure index and the like by a display terminal.

The blood pressure prediction model was trained in advance using a dataset containing PPG, ECG, ABP signals.

As one example, each sample is 1024 data points in length during training. The blood pressure prediction model extracts characteristics of the PPG signal and the ECG signal through two encoders respectively, then performs characteristic fusion at a decoder, and finally aims at synthesizing an ABP signal.

The blood pressure prediction model is trained by minimizing the mean square error of the output.

The blood pressure calculation method used by the invention is trained through a large amount of data, can be suitable for wider crowds, and has strong universality. The method can be self-adaptive according to different human body characteristics, and the accuracy and stability of the measurement result are improved.

The PPG and ECG signals acquired by the blood pressure measurement system may be stored in a local data storage module, if desired. The data storage module provides a function of synchronizing to the cloud database, so that the follow-up check or the physiological signal data analysis can be conveniently carried out.

The cloud database provides more functions for storing, managing and analyzing data.

Fig. 5 is a flowchart of the operation of the preferred embodiment, and the specific operation steps are as follows:

and (5) starting up the equipment and performing self-checking: and after the device is started, the hardware detection and self-checking program of the device is automatically carried out, the normal operation of the device is ensured, and a repair suggestion or an error code is provided when necessary, so that a user is helped to quickly locate the abnormal position of the device, and the device is timely adjusted and restored to the normal working state of the device.

Device ADC initialization configuration: and initializing ADC (analog-to-digital converter) and other related circuits of the configuration device, such as configuring proper sampling rate, data transmission interface, data format and the like, so that the device can accurately acquire PPG signals and ECG signals of the wrist part, carry out digital processing and finally completely transmit the data to a data processing module.

Wearing a wrist PPG measurement device: the PPG measurement device is worn correctly at the user's wrist or other measurement location and ensures that the device fits well with the user's skin to obtain an accurate PPG signal.

Wearing an ECG measurement device: the ECG measurement device is correctly worn at different lead locations on the chest of the user and ensures that the device fits well to the user's skin to obtain an accurate ECG signal.

Detecting whether wearing is correct: and collecting signals, wherein the signal quality analysis module judges whether the equipment is correctly worn on the user through the signal quality. If the wearing position is incorrect, the device prompts the user to adjust.

And (3) signal filtering: the data processing module carries out digital filtering on the acquired physiological signals to remove noise and interference so as to obtain clean signals.

Evaluating signal quality: the signal quality analysis module carries out quality assessment on the filtered signal to judge whether the signal is reliable, if the signal quality is bad, prompt is timely given, and a user judges whether further adjustment is needed.

Outputting a prediction result: the data processing module is used for processing the collected PPG and ECG signals, the blood pressure prediction model is used for calculating the blood pressure value, and the result display module is used for displaying the predicted blood pressure value.

The PPG waveform is shown: the filtered waveform of the acquired raw PPG signal is displayed on a device screen for reference and analysis by the user.

The ECG waveform is shown: the filtered waveform of the acquired raw ECG signal is displayed on the device screen for reference and analysis by the user.

Displaying the blood pressure calculation result: and displaying the blood pressure prediction result calculated by the blood pressure prediction module on a device screen for reference and analysis by a user.

Abnormality alert: if the predicted result of the blood pressure is abnormal, the wearing of the equipment is abnormal or the inside of the equipment is abnormal, the equipment gives an alarm in time to prompt the user to re-measure, and the user adjusts or seeks medical help according to the prompt.

Physiological data preservation: the measured physiological data (such as PPG and ECG signals) are saved in the device or in a local memory for later analysis and study. If additional personal information related content needs to be saved, the sub-files are additionally saved in a fixed format.

Synchronizing the data cloud: and uploading the physiological data stored in the equipment or the local memory to a cloud server for storage and backup.

The traditional cuff type blood pressure measurement method only can provide one measurement result for each measurement, but the scheme of the invention can continuously measure and continuously provide more detailed blood pressure monitoring data for a long time, can know the change trend of human blood pressure more accurately, can be used for realizing long-term human health condition monitoring, and in addition, the equipment supports data storage and cloud data analysis, so that a user can know the health condition of the user more conveniently.

The preferred embodiments of the invention disclosed above are intended only to assist in the explanation of the invention. The preferred embodiments are not exhaustive or to limit the invention to the precise form disclosed. Obviously, many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the invention and the practical application, to thereby enable others skilled in the art to best understand and utilize the invention. The invention is limited only by the claims and the full scope and equivalents thereof.

Claims

1. Blood pressure measurement system based on flexible organic light detector and degree of deep learning algorithm, its characterized in that: the system comprises a data acquisition module, a data processing module and a blood pressure calculation module, wherein the data acquisition module is used for acquiring physiological signal data signals and transmitting the physiological signal data signals to the data processing module; and after the data processing module preprocesses the data, the data is transmitted to the signal quality analysis module to evaluate the quality of the preprocessed signal, and the evaluated signal is transmitted to the blood pressure calculation module to calculate the blood pressure value.

2. The blood pressure measurement system based on a flexible organic photodetector and deep learning algorithm of claim 1, wherein: the data acquisition module acquires PPG signals by adopting a flexible organic photodetector, and acquires ECG signals by using an ECG module.

3. The blood pressure measurement system based on a flexible organic photodetector and deep learning algorithm of claim 1, wherein: the data acquisition module transmits the data to the data processing module in a wired or wireless mode.

4. The blood pressure measurement system based on a flexible organic photodetector and deep learning algorithm of claim 1, wherein: the data processing module carries out filtering and noise reduction processing on the signals.

5. The blood pressure measurement system based on a flexible organic photodetector and deep learning algorithm of claim 1, wherein: the signal quality analysis module evaluates the PPG signal by means of a self-encoder.

6. The blood pressure measurement system based on a flexible organic photodetector and deep learning algorithm of claim 5, wherein: the self-encoder is trained using the normal PPG signal by minimizing the output mean square error, and the quality of the signal is assessed by reconstructing the similarity of the signal to the input signal.

7. The blood pressure measurement system based on a flexible organic photodetector and deep learning algorithm of claim 1, wherein: the signal quality analysis module evaluates the ECG signal by detecting QRS peaks, which signal is normal if the number of peaks within the threshold is detected, otherwise abnormal.

8. The blood pressure measurement system based on a flexible organic photodetector and deep learning algorithm according to any of claims 1 to 7, wherein: the blood pressure calculation module extracts different signal characteristics through different encoders, fuses the different signal characteristics, predicts to obtain an ABP waveform finally, and extracts the values of the systolic pressure SBP, the diastolic pressure DBP and the mean arterial pressure MBP from the ABP waveform.

9. The blood pressure measurement system based on a flexible organic photodetector and deep learning algorithm of claim 1, wherein: the physiological signal data signals acquired by the data acquisition module are stored in a local data storage module.

10. The blood pressure measurement system based on a flexible organic photodetector and deep learning algorithm of claim 1, wherein: after the user agrees with the authorization, the data storage module provides a function of synchronizing to the cloud database, so that the subsequent check or the physiological signal data analysis are convenient.