Blood pressure monitoring method and system and storage medium

Abstract

The invention discloses a blood pressure monitoring method and system and a storage medium. The method comprises the following steps: acquiring pulse wave data of an object through a distributed nanorod triboelectric sensor; inputting the pulse wave data into a blood pressure monitoring model to obtain a blood pressure monitoring result; the blood pressure monitoring model is obtained through training by the following steps: inputting pulse wave sample data into an original blood pressure monitoring model to obtain a prediction result, and updating parameters of the original blood pressure monitoring model through a loss function according to the prediction result and a real result to obtain a trained original blood pressure monitoring model; after long-term use, secondary migration is conducted on the original blood pressure monitoring model based on a migration learning framework through pulse wave secondary sample data, and the blood pressure monitoring model is obtained. According to the embodiment of the invention, dynamic monitoring of blood pressure can be realized; and the monitoring accuracy in the long-term monitoring process can be improved. The method can be widely applied to the technical field of computers.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a blood pressure monitoring method, system and storage medium. Background Art

[0002] At present, conventional blood pressure monitoring technology still has defects and shortcomings. Although conventional invasive blood pressure monitoring technology is widely used in the medical field, it usually needs to be inserted into the body through a catheter or acupuncture, which may cause infection, bleeding or other complications. Conventional cuff-type blood pressure monitoring technology also has certain limitations, such as auscultation measurement, oscillation measurement and sphygmomanometer method. These blood pressure detection methods not only have problems such as cumbersome operation, discomfort, and inconvenience in carrying, but also due to the working mechanism of cuff pressurization, their application in long-term dynamic monitoring is limited, and it is difficult to provide continuous and real-time blood pressure data, thus affecting the patient's timely diagnosis and accurate treatment.

[0003] Conventional blood pressure monitoring methods based on pulse wave transmission time cannot solve the temporal variability of the patient population. Over time, the physiological status of different patients may change, such as abnormal cardiopulmonary function, changes in blood viscosity, arteriosclerosis, and changes in skin tissue characteristics, which may cause the original model or algorithm to be inaccurate. Summary of the invention

[0004] The purpose of the present invention is to solve one of the technical problems existing in the prior art to at least a certain extent.

[0005] To this end, an object of the present invention is to provide an accurate blood pressure monitoring method, system and storage medium.

[0006] In order to achieve the above technical objectives, one aspect of an embodiment of the present invention provides a blood pressure monitoring method, comprising the following steps: obtaining pulse wave data of an object through a distributed nano-column triboelectric sensor; inputting the pulse wave data into a blood pressure monitoring model to obtain a blood pressure monitoring result; the blood pressure monitoring model is trained through the following steps: inputting pulse wave sample data into the original blood pressure monitoring model to obtain a prediction result, and updating the parameters of the original blood pressure monitoring model through a loss function according to the prediction result and the actual result to obtain a trained original blood pressure monitoring model; after long-term use, the original blood pressure monitoring model is secondary migrated based on a transfer learning architecture through pulse wave secondary sample data to obtain a blood pressure monitoring model; the generation time of the pulse wave secondary sample data is after the generation time of the pulse wave sample data. The embodiment of the present application uses a nano-column triboelectric sensor to quickly capture pulse wave signals to achieve dynamic monitoring of blood pressure; at the same time, the monitoring accuracy during long-term monitoring is improved through the migration processing of the blood pressure monitoring model.

[0007] In some embodiments, in the blood pressure monitoring method of the embodiment of the present invention, the nanocolumn triboelectric sensor is prepared by a series of steps:

[0008] Fixing the micro-nanostructure material on a substrate;

[0009] Spin-coating an intermediate conversion material onto the micro-nano structure material, and after natural cooling, peeling off the intermediate conversion material to obtain a mold having a nanocolumn cavity structure;

[0010] The flexible composite material is spin-coated on the mold and subjected to thermal curing to obtain a flexible nanocolumn substrate; gold is sprayed on both sides of the flexible nanocolumn substrate by magnetron sputtering to obtain a flexible nanocolumn friction layer substrate;

[0011] The friction layer interface of the negative electron affinity thin film material is treated by corona technology, and gold is sprayed on the back of the negative electron affinity thin film material by magnetron sputtering to obtain a flexible thin film friction layer;

[0012] The flexible nano-column friction layer substrate and the flexible thin film friction layer are assembled to obtain a nano-column triboelectric sensor.

[0013] In some embodiments, in one embodiment of the present invention, the flexible nanocolumn friction layer substrate and the flexible thin film friction layer are assembled to obtain a nanocolumn triboelectric sensor, comprising:

[0014] The two ends of the flexible nanocolumn friction layer substrate are surrounded by a packaging material, and the outer layer wraps the flexible thin film friction layer; wherein one side of the flexible thin film friction layer that has been subjected to corona treatment is in contact with the flexible nanocolumn friction layer substrate;

[0015] The outer layer is wrapped with packaging material to obtain a nanocolumn triboelectric sensor.

[0016] In some embodiments, in one embodiment of the present invention, the method further comprises:

[0017] A silicon-based nanocolumn structure substrate is selected as the micro-nano structure material, a photocurable resin is selected as the intermediate conversion material, silver nanowires and polydimethylsiloxane are selected as the flexible composite material, and polytetrafluoroethylene is selected as the negative electron affinity film material.

[0018] In some embodiments, in one embodiment of the present invention, the method further comprises:

[0019] Preprocessing the pulse wave data to extract double pulse signal features;

[0020] A dual pulse wave-blood pressure data set is constructed, and the blood pressure monitoring model is trained based on the dual pulse wave-blood pressure data set.

[0021] In some embodiments, in one embodiment of the present invention, the secondary migration of the original blood pressure monitoring model based on the transfer learning architecture through the secondary pulse wave sample data to obtain the blood pressure monitoring model includes:

[0022] Inputting the pulse wave secondary sample data into the original blood pressure monitoring model for prediction to obtain a prediction error;

[0023] updating data weights according to the prediction error;

[0024] Based on the data weights, the original blood pressure monitoring model is trained to obtain a blood pressure monitoring model.

[0025] In some embodiments, in one embodiment of the present invention, the step of obtaining pulse wave data of an object by using a distributed nano-pillar triboelectric sensor includes:

[0026] The nanorod triboelectric sensor is attached to the radial artery tree branch of the subject to obtain the pulse wave data of the subject.

[0027] On the other hand, an embodiment of the present invention provides a blood pressure monitoring system, including:

[0028] The first module is used to obtain pulse wave data of the subject through a distributed nano-pillar triboelectric sensor;

[0029] The second module is used to input the pulse wave data into a blood pressure monitoring model to obtain a blood pressure monitoring result;

[0030] The third module is used to train the blood pressure monitoring model, including:

[0031] Inputting the pulse wave sample data into the original blood pressure monitoring model to obtain a prediction result, and updating the parameters of the original blood pressure monitoring model through a loss function according to the prediction result and the actual result to obtain a trained original blood pressure monitoring model;

[0032] After long-term use, the original blood pressure monitoring model is secondary migrated based on the transfer learning architecture through the pulse wave secondary sample data to obtain a blood pressure monitoring model; the generation time of the pulse wave secondary sample data is after the generation time of the pulse wave sample data.

[0033] On the other hand, an embodiment of the present invention provides a blood pressure monitoring device, comprising:

[0034] at least one processor;

[0035] at least one memory for storing at least one program;

[0036] When the at least one program is executed by the at least one processor, the at least one processor implements the above-mentioned blood pressure monitoring method.

[0037] On the other hand, an embodiment of the present invention provides a storage medium storing a program executable by a processor, wherein the program executable by the processor is used to implement the above-mentioned blood pressure monitoring method when executed by the processor.

[0038] The method provided in the embodiment of the present application includes: obtaining the pulse wave data of the object through a distributed nano-column triboelectric sensor; inputting the pulse wave data into a blood pressure monitoring model to obtain a blood pressure monitoring result; the blood pressure monitoring model is trained through the following steps: inputting the pulse wave sample data into the original blood pressure monitoring model to obtain a predicted result, and updating the parameters of the original blood pressure monitoring model through a loss function according to the predicted result and the actual result to obtain a trained original blood pressure monitoring model; after long-term use, the original blood pressure monitoring model is secondary migrated based on a transfer learning architecture through the pulse wave secondary sample data to obtain a blood pressure monitoring model; the generation time of the pulse wave secondary sample data is after the generation time of the pulse wave sample data. The embodiment of the present application quickly captures the pulse wave signal through the nano-column triboelectric sensor to realize dynamic monitoring of blood pressure; at the same time, the monitoring accuracy in the long-term monitoring process is improved through the migration processing of the blood pressure monitoring model. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the embodiments of the present invention or the drawings of related technical solutions in the prior art are introduced below. It should be understood that the drawings introduced below are only for the convenience of clearly describing some embodiments of the technical solutions of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0040] Figure 1 A schematic diagram of the preparation process of the nanocolumn triboelectric pulse sensor provided by the present invention;

[0041] Figure 2 (a) is a graph showing the relationship between time and voltage in the output performance of the nanocolumn triboelectric pulse sensor provided by the present invention;

[0042] Figure 2 (b) is a graph showing the relationship between frequency and voltage in the output performance of the nano-pillar triboelectric pulse sensor provided by the present invention;

[0043] Figure 2 (c) is a graph showing the relationship between the time and charge amount in the output performance of the nanocolumn triboelectric pulse sensor provided by the present invention;

[0044] Figure 2 (d) is a graph showing the relationship between frequency and charge amount in the output performance of the nano-column triboelectric pulse sensor provided by the present invention;

[0045] Figure 3 A schematic diagram of the pulse signal detection result of the nano-column triboelectric pulse sensor provided by the present invention;

[0046] Figure 4 (a) is a schematic diagram of an embodiment of a blood pressure monitoring model based on a decision tree algorithm provided by the present invention;

[0047] Figure 4 (b) is a schematic diagram of an embodiment of a blood pressure monitoring model based on a random forest algorithm provided by the present invention;

[0048] Figure 4 (c) is a schematic diagram of an embodiment of a blood pressure monitoring model based on a decision tree algorithm provided by the present invention;

[0049] Figure 4 (d) is a schematic diagram of an embodiment of a blood pressure monitoring model based on a decision tree algorithm provided by the present invention;

[0050] Figure 4 (e) is a schematic diagram of an embodiment of a blood pressure monitoring model based on a decision tree algorithm provided by the present invention;

[0051] Figure 4 (f) is a schematic diagram of an embodiment of a blood pressure monitoring model based on a decision tree algorithm provided by the present invention;

[0052] Figure 4 (g) is a schematic diagram of an embodiment of a blood pressure monitoring model based on a decision tree algorithm provided by the present invention;

[0053] Figure 4 (h) is a schematic diagram of an embodiment of a blood pressure monitoring model based on a decision tree algorithm provided by the present invention;

[0054] Figure 5 (a) is a flow chart of an embodiment of a blood pressure monitoring process provided by the present invention;

[0055] Figure 5 (b) is a flow chart of another embodiment of the blood pressure monitoring process provided by the present invention. DETAILED DESCRIPTION

[0056] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and are not to be construed as limitations of the present invention. For the step numbers in the following embodiments, they are only provided for the convenience of explanation, and the order between the steps is not limited in any way, and the execution order of each step in the embodiment can be adaptively adjusted according to the understanding of those skilled in the art.

[0057] Cardiovascular diseases (CVDs) are one of the diseases with the highest number of deaths worldwide. Blood pressure, as an important physiological indicator parameter of the cardiovascular system, can effectively evaluate and reflect the functional state of the cardiovascular system. Continuous dynamic monitoring of blood pressure is of vital importance for the early diagnosis, disease management and clinical treatment of patients with cardiovascular diseases. At present, the clinic mainly relies on traditional cuff blood pressure measurement methods to monitor the blood pressure values ​​of CVDs patients, such as auscultation measurement, oscillation measurement and sphygmomanometer method. These blood pressure detection methods not only have the problems of cumbersome operation, discomfort and inconvenience in carrying, but also due to the working mechanism of cuff pressurization, their application in long-term dynamic monitoring is limited, and it is difficult to provide continuous and real-time blood pressure data, thus affecting the timely diagnosis and accurate treatment of patients. With the growth of medical needs in modern society, there is an increasing demand for portable, accurate and long-term cardiovascular system monitoring. However, the development of long-term dynamic continuous blood pressure monitoring technology still faces many challenges, and the existing technology has not yet provided reliable and accurate wearable blood pressure detection equipment.

[0058] With the continuous advancement of sensing technology, non-invasive dynamic blood pressure continuous monitoring technology based on the principle of pulse wave conduction velocity has achieved significant development and recognition. Various types of wearable pulse sensors, such as ultrasound, bioimpedance, photoplethysmography (PPG), piezoresistive and triboelectric sensors, have been widely used in pulse wave monitoring. Among them, triboelectric sensors have shown unique advantages in pulse wave detection due to their high dynamic sensitivity, convenient data acquisition, self-powered capability and simple and diverse structure. They can accurately, wirelessly and in real time monitor the pulse signal of the cardiovascular system and analyze the characteristic index of the pulse wave. On the other hand, with the continuous advancement of artificial intelligence technology, machine learning, as a powerful tool, has shown unique advantages in dealing with the inference of complex relationships. The machine learning model is combined with the wearable pulse sensor to realize the derivation and monitoring of the pulse wave signal to the blood pressure signal. A large number of pulse wave signals are collected by the wearable pulse sensor, and then the complex relationship between the dual pulse wave and blood pressure in the cardiovascular system is mapped through iterative calculation of the machine learning model, and an accurate dual pulse wave-blood pressure inference model can be obtained. This mode of combining advanced sensing technology with artificial intelligence helps to reveal the complex relationship between pulse wave and blood pressure, and promotes the development of "pulse wave-blood pressure" sensing technology. However, traditional universal models and algorithms fail to fully consider the physiological differences between individuals, so it is difficult to solve the problem of individual variability in practical applications, which greatly limits the accuracy and practicality of pulse wave-blood pressure monitoring devices.

[0059] Due to the natural differences in physiological characteristics between different patients, such as heart rate, blood pumping speed, vascular wall elasticity, blood viscosity, and skin tissue characteristics, the correlation between pulse wave and blood pressure also has significant diversity and differences in different individuals. At present, many systems that combine pulse sensors with artificial intelligence technology mainly rely on a single algorithm or universal model to derive the blood pressure of all patients, but this method often ignores the differences between individual patients. Although some studies have taken individual factors into consideration, they usually only perform simple and inefficient individual calibration based on parameters such as age, gender, height, and weight. Such practices may have large uncertainties and differences and cannot fully reflect the individual physiological state of patients. On the other hand, over time, the patient's physiological state may change, such as abnormal cardiopulmonary function, changes in blood viscosity, arteriosclerosis, and changes in skin tissue characteristics, which may lead to inaccuracy of the original model or algorithm. Long-term use of pulse wave-blood pressure monitoring devices may require re-collecting data and training new models based on changes in the patient's physiological state, which is not conducive to the long-term use and maintenance of the equipment. In addition, the secondary construction and training of personalized blood pressure monitoring models usually cannot fully utilize the previously existing universal data sets and personalized data sets, which places high demands on the collection of pulse wave-blood pressure data. Therefore, the physiological differences, temporal variability of individual patients, and the acquisition of secondary training data sets have severely limited the widespread application and commercialization potential of personalized wearable pulse wave-blood pressure monitoring devices.

[0060] In short, the current conventional blood pressure monitoring technology still has defects and shortcomings. Although conventional invasive blood pressure monitoring technology is widely used in the medical field, it usually needs to be inserted into the body through a catheter or acupuncture, which may cause infection, bleeding or other complications. In addition, invasive blood pressure monitoring technology has defects such as insufficient comfort, high cost of use, and high technical requirements, and is rarely used in daily health monitoring. Conventional cuff blood pressure monitoring technology also has certain limitations, such as auscultation measurement, oscillation measurement, and sphygmomanometer method. These blood pressure detection methods not only have problems such as cumbersome operation, discomfort, and inconvenience in carrying, but also due to the working mechanism of cuff pressurization, its application in long-term dynamic monitoring is limited, and it is difficult to provide continuous and real-time blood pressure data, which affects the patient's timely diagnosis and precise treatment.

[0061] Conventional blood pressure monitoring methods based on pulse wave transmission time cannot solve the individual differences of patient groups. Conventional blood pressure monitoring methods based on pulse wave transmission time often use traditional universal blood pressure inference models or algorithms. Although traditional universal machine learning models show good results when facing large-scale patient groups, they often ignore the impact of individual physiological differences and are difficult to solve individual differences in practical applications, thus limiting the accuracy and practicality of "double pulse wave-blood pressure" monitoring. At present, most studies usually only consider basic information such as height, weight, gender and age when performing individual optimization. However, this information cannot fully reflect the individual physiological state of the patient. In the process of double pulse wave-blood pressure derivation, due to differences in heart beat frequency, blood pumping speed, vascular wall elasticity, blood density and subcutaneous tissue characteristics among different patients, there are significant individual differences in their double pulse wave-blood pressure models. There is a large uncertainty and error in individual optimization calibration based solely on height, weight, gender and age.

[0062] Conventional blood pressure monitoring methods based on pulse wave transmission time cannot solve the temporal variability of patient groups. Over time, the physiological status of different patients may change, such as abnormal cardiopulmonary function, changes in blood viscosity, arteriosclerosis, and changes in skin tissue properties, which may lead to inaccuracy of the original model or algorithm. Long-term use of pulse wave-blood pressure monitoring devices may require re-collecting data and training new models based on changes in the patient's physiological state, which is not conducive to the long-term use and maintenance of the equipment. In addition, the secondary construction and training of personalized blood pressure monitoring models usually cannot make full use of previously existing universal data sets and individualized data sets, which puts high requirements on the collection of pulse wave-blood pressure data, severely limiting the widespread application and commercialization potential of personalized wearable pulse wave-blood pressure monitoring equipment.

[0063] In response to the challenges and shortcomings of the above-mentioned existing conventional blood pressure monitoring technology, the present invention proposes a long-term blood pressure monitoring method based on an individualized transfer learning architecture to enhance sensors, which can effectively solve the individual differences and time variability of patient groups, as well as the problems of personalized data reuse and difficulty in obtaining, and aims to achieve long-term continuous blood pressure monitoring of different patient groups.

[0064] The blood pressure monitoring method and system according to the embodiment of the present invention will be described in detail below with reference to the accompanying drawings. First, the blood pressure monitoring method according to the embodiment of the present invention will be described with reference to the accompanying drawings.

[0065] A blood pressure monitoring method is provided in an embodiment of the present invention. The blood pressure monitoring method in the embodiment of the present invention can be applied to a terminal, a server, or software running in a terminal or a server. The terminal can be a tablet computer, a laptop computer, a desktop computer, etc., but is not limited to this. The server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The blood pressure monitoring method in the embodiment of the present invention mainly includes the following steps:

[0066] S100: Obtaining pulse wave data of the subject through a distributed nano-pillar triboelectric sensor;

[0067] S200: inputting the pulse wave data into a blood pressure monitoring model to obtain a blood pressure monitoring result;

[0068] The blood pressure monitoring model is trained by the following steps:

[0069] Inputting the pulse wave sample data into the original blood pressure monitoring model to obtain a prediction result, and updating the parameters of the original blood pressure monitoring model through a loss function according to the prediction result and the actual result to obtain a trained original blood pressure monitoring model;

[0070] After long-term use, the original blood pressure monitoring model is secondary migrated based on the transfer learning architecture through the pulse wave secondary sample data to obtain a blood pressure monitoring model; the generation time of the pulse wave secondary sample data is after the generation time of the pulse wave sample data.

[0071] In view of the defects and shortcomings of the current conventional blood pressure monitoring technology, this application proposes a dual pulse detection mode of distributed site sensing to achieve the derivation and continuous monitoring of blood pressure. This application obtains a friction interface substrate with a nanocolumn structure through a method of nanocolumn structure morphology replication (including but not limited to silicon-based nanocolumn array substrates, micro-nano cone array substrates, micro-nano hemisphere array substrates, biological sample substrates and other micro-nano structural materials), and prepares a nanocolumn triboelectric pulse sensor with excellent stress response, high sensitivity and excellent durability. The nanocolumn triboelectric pulse sensor can quickly and accurately capture the detailed features of the pulse wave signal. Utilizing the dual pulse detection mode of distributed site sensing, the dual pulse wave signals of the human body are accurately collected, and combined with signal processing, feature extraction and blood pressure derivation model, blood pressure derivation and continuous monitoring based on pulse wave conduction time are achieved.

[0072] At the same time, in response to the individual differences, time variability, and reuse and difficulty in obtaining personalized data of the blood pressure monitoring method based on pulse wave transmission time, this application proposes a blood pressure monitoring model enhanced by an individualized transfer learning architecture. This application uses the dual pulse-blood pressure data of patients with cardiovascular disease as the basic data set, and uses machine learning algorithms (including but not limited to decision trees, random forests, extreme gradient boosting trees, support vector regression, linear regression, ridge regression, LASSO regression, ELASTIC regression and other machine learning algorithms) to train and obtain the original universal blood pressure monitoring model or personalized blood pressure model. After the original blood pressure monitoring model has been used for a long time, in order to solve the individual differences and time variability problems of patients with cardiovascular disease, a small amount of dual pulse-blood pressure data of different patients can be collected again as a migration data set, and the original blood pressure monitoring model can be calibrated for secondary migration using a transfer learning architecture (including but not limited to Two-stage TrAdaBoost.R2, TrAdaBoost.R2, domain adaptation, deep domain adversarial network and other migration learning architectures or algorithms) to regain an accurate blood pressure monitoring model. Exemplarily, the secondary migration can adopt the Two-stage TrAdaBoost.R2 architecture, which is a two-stage weight adjustment optimization transfer learning architecture based on TrAdaBoost, which can fuse the new target data set on the basis of the original data set to realize the weight adjustment and migration of the personalized model. This method effectively solves the problem of inaccurate blood pressure prediction due to time variability, while avoiding the waste of original patient individual data. Performing secondary migration on a limited individual data set can achieve the best model training effect. By completing the personalized object transfer learning enhancement of the blood pressure monitoring model, patients can use the improved personalized blood pressure monitoring model for long-term and stable blood pressure monitoring, thereby improving the accuracy and practicality of blood pressure monitoring. The blood pressure monitoring model enhanced based on the personalized transfer learning architecture can realize personalized continuous monitoring of blood pressure during the long-term monitoring of patients with cardiovascular diseases.

[0073] Therefore, the long-term blood pressure monitoring method based on the personalized transfer learning architecture to enhance the sensor includes: first, a triboelectric pulse sensor based on a nanopillar structure substrate (see Figure 1 ): The nanocolumn triboelectric sensor includes a flexible composite material substrate with a nanocolumn structure on the surface and a high negative electron affinity thin film material after negative charge optimization treatment. The nanocolumn triboelectric pulse sensor can quickly and accurately capture the detailed characteristics of the pulse wave signal, thereby obtaining dual pulse wave-blood pressure data and realizing blood pressure deduction based on the pulse wave transmission time. Figure 2 The output performance of the nano-pillar triboelectric pulse sensor based on the present application is shown. Figure 3The signal detection diagram of the nano-pillar triboelectric pulse sensor is shown in Figure 2. Second, the blood pressure monitoring model based on the individualized transfer learning architecture enhancement is shown in Figure 2. Figure 4 The blood pressure monitoring model under multiple feasible algorithms shown in the figure: Based on the original universal blood pressure monitoring model or personalized blood pressure model, a small amount of dual pulse-blood pressure data of individual patients is used as a migration data set. The original blood pressure monitoring model is recalibrated by using the migration learning architecture to solve the problems of individual differences and time variability, and then an accurate blood pressure monitoring model is obtained again.

[0074] This theoretical approach provides a practical solution for future personalized precision medicine, and solves the problems of individual differences, temporal variability and personalized data reuse for the commercialization of sleeveless blood pressure monitoring devices based on machine learning models, which will help promote the development of personalized blood pressure monitoring medicine and sleeveless blood pressure monitoring technology.

[0075] Optionally, in one embodiment of the present invention, the nanocolumn triboelectric sensor is prepared by a series of steps:

[0076] Fixing the micro-nanostructure material on a substrate;

[0077] Spin-coating an intermediate conversion material onto the micro-nano structure material, and after natural cooling, peeling off the intermediate conversion material to obtain a mold having a nanocolumn cavity structure;

[0078] The flexible composite material is spin-coated on the mold and subjected to thermal curing to obtain a flexible nanocolumn substrate; gold is sprayed on both sides of the flexible nanocolumn substrate by magnetron sputtering to obtain a flexible nanocolumn friction layer substrate;

[0079] The friction layer interface of the negative electron affinity thin film material is treated by corona technology, and gold is sprayed on the back of the negative electron affinity thin film material by magnetron sputtering to obtain a flexible thin film friction layer;

[0080] The flexible nano-column friction layer substrate and the flexible thin film friction layer are assembled to obtain a nano-column triboelectric sensor.

[0081] In some possible implementations, reference Figure 1As shown, the preparation process of the nanocolumn triboelectric sensor is as follows: (1) Preliminary treatment of micro-nanostructure materials. The present invention utilizes a silicon-based nanocolumn array substrate with a micro-nanostructure to replicate a micro-nanostructure, including but not limited to silicon-based nanocolumn array substrates, micro-nanocone array substrates, micro-nanohemisphere array substrates, biological sample substrates and other micro-nanostructure materials. Taking the silicon-based nanocolumn structure substrate as an example, the sample is cleaned and fixed on a clean glass or metal substrate for subsequent replication operations. (2) Preparation of a micro-nano cavity structure mold. In this technical solution, a photocurable resin is selected as an intermediate conversion material (recessed cavity structure mold material) for the micro-nanostructure replication method. Other types of curable mold materials (soft photolithography materials) can also be selected, including but not limited to Ecoflex, silicone, polyurethane, epoxy resin, h-PDMS, polyacrylate, polyethylene glycol diacrylate (PEGDA) and other curable shaping materials. Taking the photocurable resin as an example, the photocurable resin is evenly spin-coated on the silicon-based nanocolumn structure substrate, and the photocurable resin is photocured and characterized. After the photocurable resin is naturally cooled, it is carefully peeled off from the silicon-based nanocolumn array substrate to obtain a mold with a nanocolumn cavity structure. (3) Preparation of flexible nanocolumn friction layer substrate (positive friction layer interface). The material selected as the flexible nanocolumn friction layer substrate in this technical solution is a flexible composite material of silver nanowires (AgNWs) and polydimethylsiloxane (PDMS). Other types of materials can also be used as substrate doping materials for flexible triboelectric pulse sensors, including but not limited to gold nanowires, single-walled carbon nanotubes, PEDOT:PSS, multi-walled carbon nanotubes, hexagonal boron nitride (h-BN), vertical graphene, MXene and other doped and modified materials. The flexible substrate can be selected from Ecoflex, silicone, thermoplastic polyurethane (TPU) and other curable flexible composite materials. The uncured PDMS composite material doped with AgNWs is evenly dripped and spin-coated on the micro-nano cavity structure mold, and cured by heat treatment to prepare a flexible nanocolumn substrate with a bionic nanocolumn structure. The nanocolumn microstructure of the flexible friction layer substrate effectively increases the actual friction layer interface area of ​​the sensor during the contact electrification process, thereby improving the output performance of the sensor. Finally, magnetron sputtering is performed on both sides of the flexible nanocolumn substrate to deposit a gold layer of about 300 nm, and adhered to a conductive copper tape used as a back electrode, thereby obtaining a positive friction layer interface with a nanocolumn structure. (4) Preparation of a flexible thin film friction layer (negative friction layer interface). In this technical solution, polytetrafluoroethylene (PTFE) is selected as the negative friction layer material of the nanocolumn triboelectric pulse sensor. Other high negative electron affinity thin film materials can also be selected as the negative friction layer of the nanocolumn triboelectric pulse sensor, including but not limited to polyvinyl chloride (PVC), polystyrene (PS), polyamide 66 (PA66), fluorinated ethylene propylene (FEP), polyamide 11 (PA11) and other thin film materials.The friction layer interface of the PTFE film material is treated by corona technology to enhance the negative electrical properties of the material surface, which is conducive to charge exchange during contact electrification. On the back of the PTFE film, a layer of Au is deposited by magnetron sputtering as a back electrode to extract the induced charge, thereby preparing a negative friction layer substrate. (5) Assembly of nanocolumn triboelectric pulse sensor. In this technical solution, Ecoflex and polyimide (PI) tape are selected as the packaging protection layer of the nanocolumn triboelectric pulse sensor. Other types of packaging materials can also be selected to package and protect the sensor, including but not limited to PDMS material, polyester film tape (PET), polyurethane tape (PU), natural rubber, synthetic rubber, etc. Taking Ecoflex and polyimide (PI) tape as examples, the two ends of the positive friction layer substrate are surrounded, and then the negative friction layer substrate is surrounded and coated on the outside of the annular positive friction layer substrate, wherein the corona-treated side is surrounded on the inside, and ensure that the friction interface of the positive friction layer and the negative friction layer is fully in contact. Finally, PI tape and Ecoflex material were used to cover the entire device layer by layer to encapsulate and protect the nanocolumn triboelectric pulse sensor. PI tape and Ecoflex material not only improve the contact interface between the sensor and the skin, but also improve the signal quality of the sensor and increase the stability and durability of the sensor.

[0082] Optionally, in one embodiment of the present invention, the flexible nanocolumn friction layer substrate and the flexible thin film friction layer are assembled to obtain a nanocolumn triboelectric sensor, comprising:

[0083] The two ends of the flexible nanocolumn friction layer substrate are surrounded by a packaging material, and the outer layer wraps the flexible thin film friction layer; wherein one side of the flexible thin film friction layer that has been subjected to corona treatment is in contact with the flexible nanocolumn friction layer substrate;

[0084] The outer layer is wrapped with packaging material to obtain a nanocolumn triboelectric sensor.

[0085] Optionally, in one embodiment of the present invention, the method further includes:

[0086] A silicon-based nanocolumn structure substrate is selected as the micro-nano structure material, a photocurable resin is selected as the intermediate conversion material, silver nanowires and polydimethylsiloxane are selected as the flexible composite material, and polytetrafluoroethylene is selected as the negative electron affinity film material.

[0087] It can be understood that the material selected as the object of nanocolumn structure morphology replication in the present technical scheme is a silicon-based nanocolumn array substrate, and other types of micro-nanostructure morphology materials can also be used, including but not limited to silicon-based nanocolumn array substrates, micro-nanocone array substrates, micro-nanohemisphere array substrates, biological sample substrates and other micro-nanostructure materials.

[0088] The material selected as the flexible friction layer substrate in this technical solution is a flexible composite material of silver nanowires (AgNWs) and polydimethylsiloxane (PDMS). Other types of materials can also be used as substrate doping materials for the flexible triboelectric pulse sensor, including but not limited to gold nanowires, single-walled carbon nanotubes, PEDOT:PSS, multi-walled carbon nanotubes, hexagonal boron nitride (h-BN), vertical graphene, MXene and other doped and modified materials. The flexible substrate can be selected from Ecoflex, silicone, thermoplastic polyurethane (TPU) and other curable flexible composite materials.

[0089] In this technical solution, photocurable resin is selected as the intermediate conversion material (recessed micro-nanostructure cavity mold material) of the micro-nanostructure replication method. Other types of curable mold materials (soft photolithography materials) may also be selected, including but not limited to Ecoflex, silicone, polyurethane, epoxy resin, h-PDMS, polyacrylate, polyethylene glycol diacrylate (PEGDA) and other curable molding materials.

[0090] In this technical solution, polytetrafluoroethylene (PTFE) is selected as the negative friction layer material of the nanocolumn triboelectric pulse sensor. Other thin film materials with high negative electron affinity can also be selected as the negative friction layer of the nanocolumn triboelectric pulse sensor, including but not limited to polyvinyl chloride (PVC), polystyrene (PS), polyamide 66 (PA66), fluorinated ethylene propylene (FEP), polyamide 11 (PA11) and other thin film materials.

[0091] In this technical solution, Ecoflex and polyimide (PI) tape are selected as the packaging protective layer of the nanocolumn triboelectric pulse sensor. Other types of packaging materials can also be selected to encapsulate and protect the sensor, including but not limited to PDMS material, polyester film tape (PET), polyurethane tape (PU), natural rubber, synthetic rubber, etc.

[0092] Optionally, in one embodiment of the present invention, the method further includes:

[0093] Preprocessing the pulse wave data to extract double pulse signal features;

[0094] A dual pulse wave-blood pressure data set is constructed, and the blood pressure monitoring model is trained based on the dual pulse wave-blood pressure data set.

[0095] In some possible embodiments, a nanocolumn triboelectric pulse sensor is used to detect pulse wave signals through distributed sensing. The nanocolumn triboelectric pulse sensor is attached to the branches of the human radial artery tree (~10 cm apart), which can respond quickly and accurately to the pulse wave signal. When the blood vessel expands, the nanocolumn triboelectric pulse sensor is squeezed as the pressure increases, and the positive friction layer contacts the negative friction layer, resulting in contact electrification, causing electrons to transfer from the positive friction layer interface with lower electron affinity to the negative friction layer interface with higher electron affinity. When the blood vessel relaxes, the nanocolumn triboelectric pulse sensor returns to its initial state as the applied pressure decreases, and the charge between the positive friction layer and the negative friction layer generates a potential difference through electrostatic induction, which in turn causes electrons to flow back from the negative friction layer to the positive friction layer through an external circuit. Therefore, when the nanocolumn triboelectric pulse sensor is attached to the surface of the artery, the sensor will produce periodic voltage and current changes as the artery pulsates, thereby reflecting the pulsation state of the artery. The expansion and relaxation of blood vessels correspond to the systolic and diastolic blood pressure of the human body, so the dual pulse wave signal contains rich information about the cardiovascular system and can reflect the blood pressure of the human body. Each set of dual pulse wave data contains two dual pulse wave signals of 10-30s.

[0096] Preprocessing and feature extraction of dual pulse wave data collected by nano-column triboelectric pulse sensor: (1) Preprocessing of dual pulse wave signals. The original dual pulse wave signals collected by the nano-column triboelectric pulse sensor need to be filtered, denoised, baseline removed, normalized and other preliminary data processing to eliminate the interference signal components in the original data. (2) Extraction of dual pulse wave signal features. The dual pulse wave signal needs to extract the systolic peak transmission time, the predicted period peak transmission time, the frequency, the time difference between the systolic peak and the predicted period peak, and the peak height ratio between the systolic peak and the predicted period peak as the characteristic values ​​of the dual pulse wave signal, a total of 8 features. (3) Construct a dual pulse wave-blood pressure data set. The dual pulse wave signal features are constructed together with the corresponding systolic pressure SBP, diastolic pressure DBP and mean arterial pressure MAP into a set of dual pulse-blood pressure data. This application involves the basic training data set, individualized transfer learning data set and final test data set of the blood pressure monitoring model. In this application, in order to preliminarily verify the enhancement effect of the personalized transfer learning architecture, a total of 5 volunteers were recruited, and each of them collected a total of 7 days of dual pulse wave-blood pressure data for more than half a year. Among them, the data from the first four days were used as the basic model training data for the training and preliminary verification of the original dual pulse wave-blood pressure model. After half a year, we again collected dual pulse wave-blood pressure data from these five volunteers for three days. Among them, the first day data set after half a year was used as the model migration training data set, and the next two days were used as the final test set to verify the prediction effect of the blood pressure model enhanced by the personalized transfer learning architecture.

[0097] Build the original universal blood pressure monitoring model and the original personalized blood pressure monitoring model (attached Figure 4 ): (1) Build the original universal blood pressure monitoring model. The universal blood pressure machine learning model uses the double pulse wave-blood pressure data of all volunteers in the first four days as the training set data. The extracted double pulse wave feature value is used as the input value of the model, and the corresponding blood pressure value is used as the output value of the model. Different machine learning algorithms are used to train the universal blood pressure machine learning model. In this technical solution, the random forest algorithm is selected as the machine learning model algorithm of the universal blood pressure monitoring model and the personalized blood pressure monitoring model. Other types of machine learning algorithms can also be selected to realize the construction of the blood pressure model, including but not limited to decision tree, extreme gradient boosting tree, support vector regression, linear regression, ridge regression, LASSO regression, ELASTIC regression and other machine learning algorithms; then, the double pulse wave-blood pressure data of all volunteers in the last two days are used as the test set data, the double pulse wave feature value is input into the universal blood pressure model, and the obtained blood pressure derivation value is compared with the reference blood pressure value to verify the accuracy of the universal blood pressure model. (2) Build the original personalized blood pressure monitoring model. The personalized blood pressure machine learning model uses the personalized double pulse wave-blood pressure data of different volunteers in the first four days as the training set data. The extracted dual pulse wave feature values ​​were used as the input values ​​of the model, and the corresponding blood pressure values ​​were used as the output values ​​of the model. Different machine learning algorithms were used to train blood pressure machine learning models suitable for different individual volunteers. Then, the dual pulse wave-blood pressure data of different volunteers in the last two days were used as test set data, the dual pulse wave feature values ​​were input into the personalized blood pressure model, and the derived blood pressure values ​​were compared with the reference blood pressure values ​​to verify the accuracy of the personalized blood pressure model.

[0098] In this technical solution, the random forest algorithm is selected as the machine learning model algorithm for the universal blood pressure monitoring model and the personalized blood pressure monitoring model. Other types of machine learning algorithms can also be selected to build the blood pressure model, including but not limited to decision trees, extreme gradient boosting trees, support vector regression, linear regression, ridge regression, LASSO regression, ELASTIC regression and other machine learning algorithms.

[0099] In this technical solution, Two-stage TrAdaBoost.R2 is selected as the individualized transfer learning enhancement architecture for blood pressure monitoring model enhancement. Other types of transfer learning architectures or modes can also be selected to enhance the blood pressure monitoring model, including but not limited to TrAdaBoost.R2, domain adaptation, deep domain adversarial network and other transfer learning architectures.

[0100] Optionally, in one embodiment of the present invention, the secondary migration of the original blood pressure monitoring model based on the transfer learning architecture through the secondary pulse wave sample data to obtain the blood pressure monitoring model includes:

[0101] Inputting the pulse wave secondary sample data into the original blood pressure monitoring model for prediction to obtain a prediction error;

[0102] updating data weights according to the prediction error;

[0103] Based on the data weights, the original blood pressure monitoring model is trained to obtain a blood pressure monitoring model.

[0104] In some possible implementations, the construction and enhancement of the individualized object transfer learning architecture (see Appendix Figure 5): After long-term use, a small amount of personalized dual pulse wave-blood pressure data of the patient can be used as a migration data set, and the original blood pressure monitoring model can be secondary migrated through an individualized transfer learning architecture, thereby regaining an accurate personalized object transfer architecture-enhanced blood pressure monitoring model. In this technical solution, Two-stage TrAdaBoost.R2 is selected as the individualized transfer learning enhancement architecture for blood pressure monitoring model enhancement. Other types of transfer learning architectures or modes can also be selected to enhance the blood pressure monitoring model, including but not limited to TrAdaBoost.R2, domain adaptation, deep domain adversarial network and other transfer learning architectures. Two-stage TrAdaBoost.R2 is a two-stage weight adjustment optimization transfer learning architecture based on TrAdaBoost, which can fuse a new target data set on the basis of the original data set to achieve weight adjustment and migration of personalized models. This method effectively solves the problem of inaccurate blood pressure prediction caused by time variability, while avoiding the waste of original patient individual data. Secondary migration on a limited individual data set can achieve the best model training effect. Two-stage TrAdaBoost.R2 is an improved version of the TrAdaBoost.R2 algorithm, which aims to improve the prediction performance of the target domain regression problem, especially when the target domain samples are scarce. The algorithm mainly solves the shortcomings of the traditional TrAdaBoost.R2 in dealing with sample imbalance and distribution differences between domains. Two-stage TrAdaBoost.R2 adopts a two-stage weight adjustment strategy: Phase 1: Gradually reduce the weights of the source data instances until a certain point determined by cross-validation. In this phase, the algorithm gradually adjusts the weights of the source instances, but does not reduce them to zero. Phase 2: Freeze the weights of all source instances, while the weights of the target instances are updated in the normal way of AdaBoost.R2. Only the hypotheses generated in the second phase are stored and used to determine the output of the final model. Using Two-stageTrAdaBoost.R2 to perform secondary training on the original model and change the weights of the source and target instances in the universal blood pressure model or personalized blood pressure model can alleviate the problems of sample imbalance and distribution differences between domains, and obtain a universal blood pressure model or personalized blood pressure model enhanced by an individualized transfer learning architecture. Patients with cardiovascular diseases can use the improved personalized blood pressure monitoring model to perform long-term and stable blood pressure monitoring, thereby improving the accuracy and practicality of blood pressure monitoring.

[0105] The embodiment of the present application first prepares a nano-pillar triboelectric pulse sensor with high sensitivity, fast response and good durability, and then collects the dual pulse wave signals of the human body through distributed sensing, performs signal preprocessing and feature extraction, and constructs a dual pulse wave-blood pressure data set. Finally, the original universal blood pressure monitoring model and the original personalized blood pressure monitoring model are enhanced using an individualized transfer learning architecture, which can solve the problems of individual differences, time variability, and the reuse and difficulty of personalized data, and achieve long-term stable monitoring of blood pressure.

[0106] Optionally, in one embodiment of the present invention, the acquiring of pulse wave data of the object by using a distributed nano-column triboelectric sensor comprises:

[0107] The nanorod triboelectric sensor is attached to the radial artery tree branch of the subject to obtain the pulse wave data of the subject.

[0108] It is understandable that in order to solve the problems of individual differences, time variability, and reuse and difficulty in obtaining personalized data among patients, the present application proposes a blood pressure monitoring model enhanced by an individualized transfer learning architecture, aiming to achieve long-term individualized continuous blood pressure monitoring for different patient groups. The present application prepares a nanocolumn triboelectric pulse sensor with excellent stress response, high sensitivity and excellent durability by replicating the nanocolumn structural morphology (including but not limited to silicon-based nanocolumn substrates, biological sample substrates and other micro-nanostructure materials). The nanocolumn triboelectric pulse sensor can quickly and accurately capture the detailed features of the pulse wave signal, thereby obtaining dual pulse wave-blood pressure data and realizing blood pressure derivation based on the pulse wave conduction time. Furthermore, in order to solve the problems of individual differences, time variability, and reuse and difficulty in obtaining personalized data, the present application proposes a blood pressure monitoring model enhanced by an individualized object transfer learning architecture. The model uses the dual pulse-blood pressure data of patients with cardiovascular diseases as the basic data set, and uses machine learning algorithms (including but not limited to decision trees, random forests, extreme gradient boosting trees, support vector regression, linear regression, ridge regression, LASSO regression, ELASTIC regression and other machine learning algorithms) to train and obtain the original universal blood pressure monitoring model or personalized blood pressure model. After long-term use, in order to solve the problems of individual differences and time variability, a small amount of dual pulse-blood pressure data of individual patients can be collected again as a migration data set, and the original blood pressure monitoring model can be re-calibrated by using a transfer learning architecture (including but not limited to Two-stage TrAdaBoost.R2, TrAdaBoost.R2, domain adaptation, deep domain adversarial network and other transfer learning architectures or algorithms) to re-obtain an accurate blood pressure monitoring model. This method of enhancing sensors with personalized object transfer learning architecture can effectively solve the physiological differences and time variability of different individual objects, and can effectively reduce the measurement error of the conventional original blood pressure monitoring model. This theoretical approach provides a practical solution for future personalized medicine, and solves the problems of individual differences, time variability and personalized data reuse for the commercialization of cuffless blood pressure monitoring devices based on machine learning models, which will help promote the development of personalized blood pressure monitoring medicine.

[0109] The blood pressure monitoring method provided by the present application is introduced with several specific embodiments below:

[0110] Example 1: A silicon-based nanocolumn array substrate is selected as the micro-nanostructure morphology replication object, a photocurable resin material is selected as the recessed cavity structure mold material of the micro-nanostructure replication method, a flexible composite material of AgNWs and PDMS is selected as the material of the flexible nanocolumn friction layer substrate, a PTFE film is selected as the negative friction layer electrode material, a random forest algorithm is selected as the basic machine learning algorithm of the original universal blood pressure monitoring model, and Two-stage TrAdaBoost.R2 is used as the individualized transfer learning enhancement architecture for blood pressure monitoring model enhancement. Using photocurable resin material as the recessed cavity structure mold material, after the nanocolumn array substrate is morphologically replicated, a flexible AgNWs / PDMS nanocolumn substrate is obtained, which together with the PTFE film constitutes a nanocolumn triboelectric pulse sensor. Through the distributed sensing mode, the nanocolumn triboelectric pulse sensor is attached to the radial artery tree branch of the wearer, and the human body double pulse wave signal is collected and data processing and feature extraction are performed to obtain a double pulse wave-blood pressure data set. The original universal blood pressure monitoring model or the original personalized blood pressure monitoring model is obtained by training with the random forest algorithm to achieve preliminary monitoring of blood pressure. After long-term use, in order to solve the problems of individual differences, time variability, and the reuse and difficulty of personalized data, a small amount of patients' double pulse wave-blood pressure data sets were collected again as individualized transfer learning data sets. The original blood pressure monitoring model was migrated twice using Two-stage TrAdaBoost.R2 as the individualized transfer learning architecture, and then an accurate blood pressure monitoring model enhanced by the personalized object transfer architecture was obtained to achieve long-term stable monitoring of blood pressure.

[0111] Example 2: Same as Example 1, only the object used for micro-nano structure morphology replication is replaced, and the micro-nano cone array substrate is selected as the micro-nano structure morphology replication object, and other materials and preparation processes are consistent. Select photocurable resin material as the recessed cavity structure mold material of the micro-nano structure replication method, select the flexible composite material of AgNWs and PDMS as the material of the flexible nano-cone friction layer substrate, select PTFE film as the negative friction layer electrode material, select the random forest algorithm as the basic machine learning algorithm of the original universal blood pressure monitoring model, and use Two-stage TrAdaBoost.R2 as the individualized transfer learning enhancement architecture for blood pressure monitoring model enhancement. Using photocurable resin material as the recessed cavity structure mold material, after the nanocolumn array substrate is morphologically replicated, a flexible AgNWs / PDMS nano-cone substrate is obtained, which together with the PTFE film forms a nano-cone friction electric pulse sensor. Through the distributed sensing mode, the nano-cone friction electric pulse sensor is attached to the radial artery tree branch of the wearer, the human body double pulse wave signal is collected, and data processing and feature extraction are performed to obtain a double pulse wave-blood pressure data set. The original universal blood pressure monitoring model or the original personalized blood pressure monitoring model is obtained by training with the random forest algorithm to achieve preliminary monitoring of blood pressure. After long-term use, in order to solve the problems of individual differences, time variability, and the reuse and difficulty of personalized data, a small amount of patients' double pulse wave-blood pressure data sets are collected again as personalized transfer learning data sets. The original blood pressure monitoring model is migrated twice using Two-stage TrAdaBoost.R2 as the personalized transfer learning architecture, and then an accurate personalized object transfer architecture-enhanced blood pressure monitoring model is obtained to achieve long-term stable monitoring of blood pressure.

[0112] Example 3: Same as Example 1, only the material used as the recessed cavity structure mold is changed, and h-PDMS is selected as the recessed cavity structure mold material for the micro-nano structure replication method. Other materials and preparation processes are consistent. A silicon-based nanocolumn array substrate is selected as the micro-nano structure morphology replication object, a flexible composite material of AgNWs and PDMS is selected as the material of the flexible nanocolumn friction layer substrate, a PTFE film is selected as the negative friction layer electrode material, a random forest algorithm is selected as the basic machine learning algorithm of the original universal blood pressure monitoring model, and Two-stage TrAdaBoost.R2 is used as the individualized transfer learning enhancement architecture for blood pressure monitoring model enhancement. Using h-PDMS material as the recessed cavity structure mold material, after the nanocolumn array substrate is morphologically replicated, a flexible AgNWs / PDMS nanocolumn substrate is obtained, which together with the PTFE film constitutes a nanocolumn triboelectric pulse sensor. Through the distributed sensing mode, the nanocolumn triboelectric pulse sensor is attached to the wearer's radial artery tree branch, the human body double pulse wave signal is collected, and data processing and feature extraction are performed to obtain a double pulse wave-blood pressure data set. The original universal blood pressure monitoring model or the original personalized blood pressure monitoring model is obtained by training with the random forest algorithm to achieve preliminary monitoring of blood pressure. After long-term use, in order to solve the problems of individual differences, time variability, and the reuse and difficulty of personalized data, a small amount of patients' double pulse wave-blood pressure data sets are collected again as personalized transfer learning data sets. The original blood pressure monitoring model is migrated twice using Two-stage TrAdaBoost.R2 as the personalized transfer learning architecture, and then an accurate personalized object transfer architecture-enhanced blood pressure monitoring model is obtained to achieve long-term stable monitoring of blood pressure.

[0113] Example 4: Same as Example 1, only the material of the flexible nanocolumn friction layer substrate is changed, and a flexible composite material of carbon nanotubes CNTs and PDMS is selected as the material used in the second mold-turning process. Other materials and preparation process methods are consistent. A silicon-based nanocolumn array substrate is selected as the micro-nano structure morphology replication object, a photocurable resin material is selected as the recessed cavity structure mold material of the micro-nano structure replication method, a PTFE film is selected as the negative friction layer electrode material, a random forest algorithm is selected as the basic machine learning algorithm of the original universal blood pressure monitoring model, and Two-stageTrAdaBoost.R2 is used as the individualized transfer learning enhancement architecture for blood pressure monitoring model enhancement. Using photocurable resin material as the recessed cavity structure mold material, after the nanocolumn array substrate is morphologically replicated, a flexible CNTs / PDMS nanocolumn substrate is obtained, which together with the PTFE film constitutes a nanocolumn triboelectric pulse sensor. Through the distributed sensing mode, the nanocolumn triboelectric pulse sensor is attached to the radial artery tree branch of the wearer, the human body double pulse wave signal is collected, and data processing and feature extraction are performed to obtain a double pulse wave-blood pressure data set. The original universal blood pressure monitoring model or the original personalized blood pressure monitoring model is obtained by training with the random forest algorithm to achieve preliminary monitoring of blood pressure. After long-term use, in order to solve the problems of individual differences, time variability, and the reuse and difficulty of personalized data, a small amount of patients' double pulse wave-blood pressure data sets are collected again as personalized transfer learning data sets. The original blood pressure monitoring model is migrated twice using Two-stage TrAdaBoost.R2 as the personalized transfer learning architecture, and then an accurate personalized object transfer architecture-enhanced blood pressure monitoring model is obtained to achieve long-term stable monitoring of blood pressure.

[0114] Example 5: Same as Example 1, only the PTFE material used as the negative friction layer electrode is replaced, and the PVC film is selected as the negative friction layer electrode material. Other selected materials and preparation process methods are consistent. A silicon-based nanocolumn array substrate is selected as the micro-nano structure morphology replication object, a photocurable resin material is selected as the recessed cavity structure mold material of the micro-nano structure replication method, a flexible composite material of AgNWs and PDMS is selected as the material of the flexible nanocolumn friction layer substrate, and a random forest algorithm is selected as the basic machine learning algorithm of the original universal blood pressure monitoring model. Two-stage TrAdaBoost.R2 is used as the individualized transfer learning enhancement architecture for blood pressure monitoring model enhancement. Using photocurable resin material as the recessed cavity structure mold material, after the nanocolumn array substrate is morphologically replicated, a flexible AgNWs / PDMS nanocolumn substrate is obtained, which together with the PTFE film constitutes a nanocolumn triboelectric pulse sensor. Through the distributed sensing mode, the nanocolumn triboelectric pulse sensor is attached to the wearer's radial artery tree branch, the human body's double pulse wave signal is collected, and data processing and feature extraction are performed to obtain a double pulse wave-blood pressure data set. The original universal blood pressure monitoring model or the original personalized blood pressure monitoring model is obtained by training with the random forest algorithm to achieve preliminary monitoring of blood pressure. After long-term use, in order to solve the problems of individual differences, time variability, and the reuse and difficulty of personalized data, a small amount of patients' double pulse wave-blood pressure data sets are collected again as personalized transfer learning data sets. The original blood pressure monitoring model is migrated twice using Two-stage TrAdaBoost.R2 as the personalized transfer learning architecture, and then an accurate personalized object transfer architecture-enhanced blood pressure monitoring model is obtained to achieve long-term stable monitoring of blood pressure.

[0115] Example 6: Same as Example 1, only the random forest algorithm used to build a universal blood pressure monitoring model is replaced, and support vector regression is selected as the basic machine learning algorithm of the original universal blood pressure monitoring model. Other materials and preparation process methods are consistent. A silicon-based nanocolumn array substrate is selected as the micro-nano structure morphology replication object, a photocurable resin material is selected as the recessed cavity structure mold material of the micro-nano structure replication method, a flexible composite material of AgNWs and PDMS is selected as the material of the flexible nanocolumn friction layer substrate, and a PTFE film is selected as the negative friction layer electrode material. Two-stageTrAdaBoost.R2 is used as an individualized transfer learning enhancement architecture for blood pressure monitoring model enhancement. Using photocurable resin material as the recessed cavity structure mold material, after the nanocolumn array substrate is morphologically replicated, a flexible AgNWs / PDMS nanocolumn substrate is obtained, which together with the PTFE film constitutes a nanocolumn triboelectric pulse sensor. Through the distributed sensing mode, the nanocolumn triboelectric pulse sensor is attached to the radial artery tree branch of the wearer, the human body double pulse wave signal is collected, and data processing and feature extraction are performed to obtain a double pulse wave-blood pressure data set. The original universal blood pressure monitoring model or the original personalized blood pressure monitoring model is trained using the support vector regression algorithm to achieve preliminary blood pressure monitoring. After long-term use, in order to solve the problems of individual differences, time variability, and the reuse and difficulty of personalized data, a small amount of patients' double pulse wave-blood pressure data sets are collected again as personalized transfer learning data sets. The original blood pressure monitoring model is secondary migrated using Two-stage TrAdaBoost.R2 as the personalized transfer learning architecture, and then the accurate personalized object transfer architecture enhanced blood pressure monitoring model is regained to achieve long-term stable blood pressure monitoring.

[0116] Example 7: Same as Example 1, only the architecture for individualized transfer learning enhancement is replaced, and the deep domain adversarial network is selected as the individualized transfer learning enhancement architecture for blood pressure monitoring model enhancement. Other materials and preparation process methods are consistent. A silicon-based nanocolumn array substrate is selected as the micro-nano structure morphology replication object, a photocurable resin material is selected as the recessed cavity structure mold material of the micro-nano structure replication method, a flexible composite material of AgNWs and PDMS is selected as the material of the flexible nanocolumn friction layer substrate, a PTFE film is selected as the negative friction layer electrode material, and a random forest algorithm is selected as the basic machine learning algorithm of the original universal blood pressure monitoring model. Using a photocurable resin material as a recessed cavity structure mold material, after the nanocolumn array substrate is morphologically replicated, a flexible AgNWs / PDMS nanocolumn substrate is obtained, which together with the PTFE film constitutes a nanocolumn triboelectric pulse sensor. Through the distributed sensing mode, the nanocolumn triboelectric pulse sensor is attached to the radial artery tree branch of the wearer, the human body double pulse wave signal is collected, and data processing and feature extraction are performed to obtain a double pulse wave-blood pressure data set. The original universal blood pressure monitoring model or the original personalized blood pressure monitoring model is trained using the random forest algorithm to achieve preliminary blood pressure monitoring. After long-term use, in order to solve the problems of individual differences, time variability, and the reuse and difficulty of personalized data, a small amount of patients' double pulse wave-blood pressure data sets are collected again as personalized transfer learning data sets. The original blood pressure monitoring model is secondary migrated using the deep domain adversarial network as the personalized transfer learning architecture, and then an accurate personalized object transfer architecture-enhanced blood pressure monitoring model is obtained to achieve long-term stable blood pressure monitoring.

[0117] In summary, facing the challenges and shortcomings of the current continuous blood pressure monitoring technology, this technology combines the blood pressure monitoring model enhanced by the individualized transfer learning architecture with distributed site dual pulse sensing, solving the problems of individual differences, time variability, and personalized data reuse and difficulty in obtaining in different patient groups, and achieving accurate long-term blood pressure monitoring.

[0118] Aiming at the defects and shortcomings of conventional blood pressure monitoring technology. This technology adopts a dual pulse detection mode of distributed site sensing to achieve the derivation and continuous monitoring of blood pressure. This application obtains a friction interface substrate with a nanocolumn structure through a method of replicating the structure morphology of a nanocolumn, and prepares a nanocolumn triboelectric pulse sensor with excellent stress response, high sensitivity and excellent durability. The nanocolumn triboelectric pulse sensor can quickly and accurately capture the detailed features of the pulse wave signal. The dual pulse detection mode of distributed site sensing is used to accurately collect the dual pulse wave signals of the human body, and combined with signal processing, feature extraction and blood pressure derivation model, blood pressure derivation and continuous monitoring based on pulse wave conduction time are achieved.

[0119] Aiming at the individual differences and time variability problems of patient groups. This technology adopts a blood pressure monitoring model enhanced by an individualized transfer learning architecture. This application uses the dual pulse-blood pressure data of patients with cardiovascular disease as the basic data set, and uses machine learning algorithms to train to obtain the original universal blood pressure monitoring model or personalized blood pressure model. After long-term use, in order to solve the individual differences and time variability problems of patients with cardiovascular disease, a small amount of dual pulse-blood pressure data from different patients is collected again as a migration data set, and the original blood pressure monitoring model is re-calibrated using the transfer learning architecture to regain an accurate blood pressure monitoring model. Compared with the original non-transferred enhanced universal blood pressure model and personalized blood pressure model, the universal blood pressure model and personalized blood pressure model enhanced based on the individualized transfer learning architecture have an overall improvement of up to 27.05% and 26.15%, which can realize personalized continuous monitoring of blood pressure during the long-term monitoring of patients with cardiovascular disease.

[0120] The present invention combines the blood pressure monitoring model enhanced by the personalized transfer learning architecture with the distributed site dual pulse sensing to solve the individual differences and time variability of different patient groups. This theoretical method provides a practical solution for future personalized medicine, and solves the problems of individual differences, time variability and personalized data reuse for the commercialization of sleeveless blood pressure monitoring equipment based on machine learning models, which helps to promote the development of personalized blood pressure monitoring medicine.

[0121] In summary, the method provided in the embodiment of the present application includes: obtaining the pulse wave data of the object through a distributed nano-column triboelectric sensor; inputting the pulse wave data into a blood pressure monitoring model to obtain a blood pressure monitoring result; the blood pressure monitoring model is trained through the following steps: inputting the pulse wave sample data into the original blood pressure monitoring model to obtain a prediction result, and updating the parameters of the original blood pressure monitoring model through a loss function according to the prediction result and the actual result to obtain a trained original blood pressure monitoring model; after long-term use, the original blood pressure monitoring model is secondary migrated based on a transfer learning architecture through the pulse wave secondary sample data to obtain a blood pressure monitoring model; the generation time of the pulse wave secondary sample data is after the generation time of the pulse wave sample data. The embodiment of the present application quickly captures the pulse wave signal through a nano-column triboelectric sensor to achieve dynamic monitoring of blood pressure; at the same time, the monitoring accuracy during long-term monitoring is improved through the migration processing of the blood pressure monitoring model.

[0122] Secondly, an embodiment of the present invention provides a blood pressure monitoring system, the system specifically comprising:

[0123] The first module is used to obtain pulse wave data of the subject through a distributed nano-pillar triboelectric sensor;

[0124] The second module is used to input the pulse wave data into a blood pressure monitoring model to obtain a blood pressure monitoring result;

[0125] The third module is used to train the blood pressure monitoring model, including:

[0126] Inputting the pulse wave sample data into the original blood pressure monitoring model to obtain a prediction result, and updating the parameters of the original blood pressure monitoring model through a loss function according to the prediction result and the actual result to obtain a trained original blood pressure monitoring model;

[0127] After long-term use, the original blood pressure monitoring model is secondary migrated based on the transfer learning architecture through the pulse wave secondary sample data to obtain a blood pressure monitoring model; the generation time of the pulse wave secondary sample data is after the generation time of the pulse wave sample data.

[0128] It can be seen that the contents of the above method embodiments are all applicable to the present system embodiments, the functions specifically implemented by the present system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0129] An embodiment of the present invention provides a blood pressure monitoring device, comprising:

[0130] at least one processor;

[0131] at least one memory for storing at least one program;

[0132] When the at least one program is executed by the at least one processor, the at least one processor implements the blood pressure monitoring method.

[0133] Similarly, the contents of the above method embodiments are all applicable to the present device embodiments. The functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0134] An embodiment of the present invention further provides a computer-readable storage medium, in which a program executable by a processor is stored. The program executable by the processor is used to perform the above-mentioned blood pressure monitoring method when executed by the processor.

[0135] Similarly, the contents of the above method embodiments are all applicable to the present storage medium embodiments. The functions specifically implemented by the present storage medium embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0136] In some selectable embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the present invention is provided by way of example, for the purpose of providing a more comprehensive understanding of technology. The disclosed method is not limited to the operation and logic flow presented herein. Selectable embodiments are expected, wherein the order of various operations is changed and the sub-operation of a part for which is described as a larger operation is performed independently.

[0137] In addition, although the present invention is described in the context of functional modules, it should be understood that, unless otherwise specified, one or more of the functions and / or features can be integrated into a single physical device and / or software module, or one or more functions and / or features can be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the present invention. More specifically, in view of the properties, functions and internal relationships of the various functional modules in the device disclosed herein, the actual implementation of the module will be understood within the conventional skills of the engineer. Therefore, those skilled in the art can implement the present invention set forth in the claims without excessive experimentation using ordinary techniques. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0138] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several programs to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0139] The logic and / or steps represented in the flowchart or otherwise described herein, for example, may be considered as an ordered list of executable programs for implementing the logical functions, and may be embodied in any computer-readable medium for use by a program execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch and execute a program from a program execution system, device or apparatus), or in conjunction with such program execution systems, devices or apparatuses. For purposes of this specification, a "computer-readable medium" may be any device that can contain, store, communicate, propagate or transmit a program for use by a program execution system, device or apparatus, or in conjunction with such program execution systems, devices or apparatuses.

[0140] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0141] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable program execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0142] In the above description of this specification, the description with reference to the terms "one embodiment / example", "another embodiment / example" or "certain embodiments / examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0143] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

[0144] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the embodiments. Those skilled in the art may make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of the present invention.

Claims

1. A blood pressure monitoring method, characterized in that: The following steps are involved: Obtain the subject’s pulse wave data through distributed nanopillar triboelectric sensors; Inputting the pulse wave data into a blood pressure monitoring model to obtain a blood pressure monitoring result; The blood pressure monitoring model is trained by the following steps: Inputting the pulse wave sample data into the original blood pressure monitoring model to obtain a prediction result, and updating the parameters of the original blood pressure monitoring model through a loss function according to the prediction result and the actual result to obtain a trained original blood pressure monitoring model; After long-term use, the original blood pressure monitoring model is secondary migrated based on the transfer learning architecture through the pulse wave secondary sample data to obtain a blood pressure monitoring model; the generation time of the pulse wave secondary sample data is after the generation time of the pulse wave sample data.

2. The blood pressure monitoring method according to claim 1, characterized in that: The nanocolumn triboelectric sensor is prepared by a series of steps: Fixing the micro-nanostructure material on a substrate; Spin-coating an intermediate conversion material onto the micro-nano structure material, and after natural cooling, peeling off the intermediate conversion material to obtain a mold having a nanocolumn cavity structure; The flexible composite material is spin-coated on the mold and subjected to thermal curing to obtain a flexible nanocolumn substrate; gold is sprayed on both sides of the flexible nanocolumn substrate by magnetron sputtering to obtain a flexible nanocolumn friction layer substrate; The friction layer interface of the negative electron affinity thin film material is treated by corona technology, and gold is sprayed on the back of the negative electron affinity thin film material by magnetron sputtering to obtain a flexible thin film friction layer; The flexible nano-column friction layer substrate and the flexible thin film friction layer are assembled to obtain a nano-column triboelectric sensor.

3. The blood pressure monitoring method according to claim 2, characterized in that: The method of assembling the flexible nanocolumn friction layer substrate and the flexible thin film friction layer to obtain a nanocolumn triboelectric sensor comprises: The two ends of the flexible nanocolumn friction layer substrate are surrounded by a packaging material, and the outer layer wraps the flexible thin film friction layer; wherein one side of the flexible thin film friction layer that has been subjected to corona treatment is in contact with the flexible nanocolumn friction layer substrate; The outer layer is wrapped with packaging material to obtain a nanocolumn triboelectric sensor.

4. The blood pressure monitoring method according to claim 3, characterized in that: The method further comprises: A silicon-based nanocolumn structure substrate is selected as the micro-nano structure material, a photocurable resin is selected as the intermediate conversion material, silver nanowires and polydimethylsiloxane are selected as the flexible composite material, and polytetrafluoroethylene is selected as the negative electron affinity film material.

5. The blood pressure monitoring method according to claim 1, characterized in that: The method further comprises: Preprocessing the pulse wave data to extract double pulse signal features; A dual pulse wave-blood pressure data set is constructed, and the blood pressure monitoring model is trained based on the dual pulse wave-blood pressure data set.

6. The blood pressure monitoring method according to claim 1, characterized in that: The blood pressure monitoring model is obtained by performing secondary migration on the original blood pressure monitoring model based on the migration learning architecture through the secondary pulse wave sample data, including: Inputting the pulse wave secondary sample data into the original blood pressure monitoring model for prediction to obtain a prediction error; updating data weights according to the prediction error; Based on the data weights, the original blood pressure monitoring model is trained to obtain a blood pressure monitoring model.

7. The blood pressure monitoring method according to claim 1, characterized in that: The method of obtaining pulse wave data of an object by using a distributed nanocolumn triboelectric sensor includes: The nanorod triboelectric sensor is attached to the radial artery tree branch of the subject to obtain the pulse wave data of the subject.

8. A blood pressure monitoring system, characterized in that: include: The first module is used to obtain pulse wave data of the subject through a distributed nano-pillar triboelectric sensor; The second module is used to input the pulse wave data into a blood pressure monitoring model to obtain a blood pressure monitoring result; The third module is used to train the blood pressure monitoring model, including: Inputting the pulse wave sample data into the original blood pressure monitoring model to obtain a prediction result, and updating the parameters of the original blood pressure monitoring model through a loss function according to the prediction result and the actual result to obtain a trained original blood pressure monitoring model; After long-term use, the original blood pressure monitoring model is secondary migrated based on the transfer learning architecture through the pulse wave secondary sample data to obtain a blood pressure monitoring model; the generation time of the pulse wave secondary sample data is after the generation time of the pulse wave sample data.

9. A blood pressure monitoring device, characterized in that: include: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the blood pressure monitoring method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a program executable by a processor, characterized in that: The processor-executable program is used to implement the blood pressure monitoring method according to any one of claims 1 to 7 when executed by the processor.

Images