Personalized adoptive non-invasive blood glucose monitoring method, system, device and medium

Abstract

The invention discloses a personalized adoptive non-invasive blood glucose monitoring method, system and device and a medium, and the method comprises the steps: obtaining a blood glucose sample and a physiological parameter sample of a target individual, and calibrating a trained universal artificial intelligence prediction model according to the blood glucose sample and the physiological parameter sample of the target individual, a personalized artificial intelligence prediction model is obtained; the universality artificial intelligence prediction model is obtained by training according to a blood glucose sample and a physiological parameter sample of a group; acquiring physiological parameters of the target individual, and monitoring blood glucose of the target individual according to the physiological parameters of the target individual and the personalized artificial intelligence prediction model; the physiological parameters are obtained through non-creative measurement. The embodiment of the invention can noninvasively, accurately and continuously monitor the blood glucose of the target individual, and can be widely applied to the technical field of blood glucose detection.

Description

Technical Field

[0001] The present invention relates to the technical field of blood glucose detection, and in particular to a personalized adoptive non-invasive blood glucose monitoring method, system, device and medium. Background Art

[0002] Diabetes is a metabolic disorder characterized by chronic hyperglycemia, which seriously endangers personal health and affects social economy. The potential pathogenesis of diabetes usually stems from poor insulin secretion or reduced responsiveness of target cells to insulin, resulting in a continuous increase in blood glucose (BG) levels. Effective BG monitoring is crucial for early detection, disease management, and complication prevention. Traditional blood glucose monitoring methods rely on multiple finger pricks per day, which are invasive, cumbersome, painful, have an infection risk, and may reduce patient compliance. In addition, these discrete BG measurements have serious monitoring blind spots and cannot continuously track and comprehensively reflect the overall BG fluctuations. To address these issues, continuous glucose monitoring (CGM) technology has been developed to record blood glucose levels over several days and comprehensively reflect their dynamic trends. Through invasive, non-invasive, or minimally invasive methods, CGM systems fix the monitoring device on the skin or tissue and continuously record the glucose concentration in body fluids (such as interstitial fluid (ISF), tears, and saliva), thereby inferring real-time blood glucose changes. These continuous methods are desirable for tracking the progression of diabetes, revealing the details of the condition, and guiding the adjustment of clinical treatment. Clinical CGM is mainly minimally invasive blood glucose monitoring devices. However, the enzymes on the electrodes degrade over time, which hinders the long-term monitoring of diabetic patients. According to the latest technology, the lifespan of minimally invasive CGM is usually limited to <15 days. Non-invasive CGM technology uses non-invasive electrochemical or physical methods to monitor BG levels. Electrochemical-based CGM detects the glucose concentration in tears, sweat, etc., and then derives the actual blood glucose level. However, the correlation between the glucose concentration in tears and sweat and the actual blood glucose level is limited. Therefore, CGM based on tear or sweat sensors has not yet achieved the accuracy required clinically, hindering its clinical approval and commercialization. Physical detection method-based CGM mainly includes mid-infrared absorption spectroscopy, Raman scattering spectroscopy, and metabolic heat conformation (MHC). However, physical detection-based CGM still does not meet clinical standards, showing low accuracy and insufficient specific recognition of glucose.

[0003] With the rapid development of artificial intelligence (AI), non-invasive blood glucose monitoring can collect a large amount of human physiological data and use an AI model to fit and predict the current BG of the human body. However, due to the large individual differences in the complex physiological characteristics of the human body, such as different tissue characteristics and blood components, currently, AI models ignore individual heterogeneity, which limits the accuracy of AI models. Summary of the Invention

[0004] In view of this, the purpose of the embodiments of the present invention is to provide a personalized adoptive non-invasive blood glucose monitoring method, system, device and medium, which can non-invasively, accurately and continuously monitor the blood glucose of the target individual.

[0005] On the one hand, the embodiments of the present invention provide a personalized adoptive non-invasive blood glucose monitoring method, including:

[0006] Obtaining a blood glucose sample and a physiological parameter sample of the target individual, and calibrating the trained general artificial intelligence prediction model according to the blood glucose sample and the physiological parameter sample of the target individual to obtain a personalized artificial intelligence prediction model; the general artificial intelligence prediction model is trained according to the blood glucose samples and physiological parameter samples of the population;

[0007] Obtaining the physiological parameters of the target individual, and monitoring the blood glucose of the target individual according to the physiological parameters of the target individual and the personalized artificial intelligence prediction model; the physiological parameters are measured non-invasively.

[0008] Optionally, the general artificial intelligence prediction model is trained by the following method:

[0009] Obtaining blood glucose samples and physiological parameter samples of several populations, and the several populations include hyperglycemic populations, hypoglycemic populations and normoglycemic populations;

[0010] Dividing the blood glucose samples and physiological parameter samples of several of the populations into a training set and a test set, training a preset artificial intelligence prediction model according to the training set, and testing the trained preset artificial intelligence prediction model according to the test set until the preset requirements are met, to obtain a trained general artificial intelligence prediction model.

[0011] Optionally, the personalized adoptive non-invasive blood glucose monitoring method further includes:

[0012] During the monitoring process, updating the personalized artificial intelligence prediction model according to the feedback information.

[0013] On the other hand, the embodiments of the present invention provide a personalized adoptive non-invasive blood glucose monitoring system, including:

[0014] A first module, configured to obtain a blood glucose sample and a physiological parameter sample of the target individual, and calibrate the trained general artificial intelligence prediction model according to the blood glucose sample and the physiological parameter sample of the target individual to obtain a personalized artificial intelligence prediction model; the general artificial intelligence prediction model is trained according to the blood glucose samples and physiological parameter samples of the population;

[0015] A second module, configured to obtain physiological parameters of the target individual, and monitor blood glucose of the target individual according to the physiological parameters of the target individual and the personalized artificial intelligence prediction model; the physiological parameters are measured non-invasively.

[0016] On the other hand, an embodiment of the present invention provides a personalized adoptive non-invasive blood glucose monitoring device, including:

[0017] At least one processor;

[0018] At least one memory, configured to store at least one program;

[0019] When the at least one program is executed by the at least one processor, the at least one processor implements the above method.

[0020] On the other hand, an embodiment of the present invention provides a computer-readable storage medium, in which a processor-executable program is stored, and the processor-executable program is used to execute the above method when executed by a processor.

[0021] On the other hand, an embodiment of the present invention provides a personalized adoptive non-invasive blood glucose monitoring system, including a microneedle blood glucose sensing module, a non-invasive blood glucose sensing module, and a computer device, where the computer device is connected to the microneedle blood glucose sensing module and the non-invasive blood glucose sensing module; wherein,

[0022] The microneedle blood glucose sensing module is configured to measure a blood glucose value and send the blood glucose value to the computer device;

[0023] The non-invasive blood glucose sensing module is configured to measure physiological parameters and send the physiological parameters to the computer device;

[0024] The computer device includes:

[0025] At least one processor;

[0026] At least one memory, configured to store at least one program;

[0027] When the at least one program is executed by the at least one processor, the at least one processor implements the above method.

[0028] Optionally, the non-invasive blood glucose sensing module includes a metabolic heat sensor and a cardiovascular sensor; wherein;

[0029] The metabolic heat sensor is configured to measure metabolic heat-related parameters;

[0030] The cardiovascular sensor is configured to measure a pulse wave signal of the cardiovascular system.

[0031] Optionally, the metabolic heat sensor includes a surface temperature sensor, an ambient temperature and humidity sensor, and an infrared radiation sensor; wherein,

[0032] The surface temperature sensor is used to measure the surface temperature of the target individual;

[0033] The environmental temperature and humidity sensor is used to measure the temperature and humidity of the detection environment;

[0034] The infrared radiation sensor is used to measure the thermal radiation temperature of the target individual.

[0035] Optionally, the cardiovascular sensor includes a plurality of photoplethysmography sensors, and the photoplethysmography sensors are used to measure pulse wave signals at preset locations.

[0036] The implementation of the embodiments of the present invention includes the following beneficial effects: first, a universal artificial intelligence prediction model is obtained by training with blood glucose samples and physiological parameter samples of a group. The universal artificial intelligence prediction model constructs the intrinsic relationship between blood glucose and physiological parameters. Then, the trained universal artificial intelligence prediction model is calibrated with the blood glucose samples and physiological parameter samples of the target individual to obtain a personalized artificial intelligence prediction model. The personalized artificial intelligence prediction model takes into account the personalized differences of the target individual, thereby realizing personalized migration calibration, that is, converting minimally invasive + non-invasive blood glucose sensing into a non-invasive device based on the artificial intelligence model, namely "adoptive non-invasive CGM". Finally, the blood glucose of the target individual is monitored according to the physiological parameters measured non-invasively by the target individual and the personalized artificial intelligence prediction model, thereby realizing non-invasive, accurate and continuous monitoring of the blood glucose of the target individual. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a schematic flow chart of the steps of a personalized adoptive non-invasive blood glucose monitoring method provided by an embodiment of the present invention;

[0038] Figure 2 This is a structural block diagram of a personalized adoptive non-invasive blood glucose monitoring system provided by an embodiment of the present invention;

[0039] Figure 3 This is a structural block diagram of a personalized adoptive non-invasive blood glucose monitoring device provided by an embodiment of the present invention;

[0040] Figure 4 This is another structural block diagram of a personalized adoptive non-invasive blood glucose monitoring system provided by an embodiment of the present invention;

[0041] Figure 5 This is a schematic diagram of a personalized adoptive non-invasive blood glucose monitoring principle and measurement method provided by an embodiment of the present invention;

[0042] Figure 6 It is a flowchart of a personalized adoptive non-invasive blood glucose monitoring method provided by an embodiment of the present invention;

[0043] Figure 7 It is a schematic diagram of data collected by a microneedle blood glucose sensing module and a non-invasive blood glucose sensing module provided by an embodiment of the present invention;

[0044] Figure 8 It is a schematic diagram of the training of a preset artificial intelligence prediction model provided by an embodiment of the present invention;

[0045] Figure 9 It is a schematic diagram of the results of a correlation analysis of BG levels and MHC non-invasive parameters on the MI / NI-CGM dataset for all participants provided by an embodiment of the present invention;

[0046] Figure 10 It is a result difference graph between a personalized adoptive non-invasive CGM monitoring method and a common non-invasive CGM method provided by an embodiment of the present invention;

[0047] Figure 11 It is a Clark diagram of a training set using a personalized non-invasive dynamic blood glucose sensing method provided by an embodiment of the present invention;

[0048] Figure 12 It is a Clark diagram of a test set using a personalized non-invasive dynamic blood glucose sensing method and a common non-invasive blood glucose monitoring method provided by an embodiment of the present invention. Detailed implementation manners

[0049] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. For the step numbers in the following embodiments, they are only set for the convenience of explanation and illustration, and no limitation is imposed on the order between the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0050] It should be noted that although functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different module division from that in the device or a different order from that in the flowchart. Terms such as "first" and "second" in the description, claims, and the above-mentioned drawings are used to distinguish similar objects and do not necessarily describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units does not necessarily limit to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products, or devices.

[0051] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0052] As Figure 1 shown, an embodiment of the present invention provides a personalized adoptive non-invasive blood glucose monitoring method, including:

[0053] S100. Obtain a blood glucose sample and a physiological parameter sample of a target individual, and calibrate a trained universal artificial intelligence prediction model according to the blood glucose sample and the physiological parameter sample of the target individual to obtain a personalized artificial intelligence prediction model; the universal artificial intelligence prediction model is trained according to the blood glucose samples and physiological parameter samples of a group;

[0054] S200. Obtain the physiological parameters of the target individual, and monitor the blood glucose of the target individual according to the physiological parameters of the target individual and the personalized artificial intelligence prediction model; the physiological parameters are measured non-invasively.

[0055] It should be noted that the physiological parameters are determined according to actual applications and are not specifically limited in this embodiment. The physiological parameters include but are not limited to the temperature information and pulse information of the human body, etc., as well as the temperature information and humidity information of the detection environment, etc. The target individual refers to the individual to be monitored, and the target individual includes the human body. The blood glucose value is measured in a minimally invasive situation, and the physiological parameters are measured non-invasively.

[0056] Specifically, first, a general artificial intelligence prediction model is trained based on the blood glucose samples and physiological parameter samples of a group. Then, the trained general artificial intelligence prediction model is calibrated according to the blood glucose samples and physiological parameter samples of the target individual to obtain a personalized artificial intelligence prediction model, thereby improving the accuracy of monitoring the target individual. Finally, the blood glucose of the target individual is monitored according to the physiological parameters of the target individual and the personalized artificial intelligence prediction model.

[0057] Optionally, the general artificial intelligence prediction model is trained by the following method:

[0058] S010. Obtain the blood glucose samples and physiological parameter samples of a number of populations, where the number of populations includes hyperglycemic populations, hypoglycemic populations, and normoglycemic populations;

[0059] S020. Divide the blood glucose samples and physiological parameter samples of a number of populations into a training set and a test set. Train a preset artificial intelligence prediction model according to the training set, and test the trained preset artificial intelligence prediction model according to the test set until the preset requirements are met to obtain a trained general artificial intelligence prediction model.

[0060] The preset artificial intelligence prediction model is determined according to the actual application, and no specific limitation is made in this embodiment. The preset artificial intelligence prediction model includes, but is not limited to, AI models such as long short-term memory network (LSTM), gated recurrent unit (GRU), etc.

[0061] The number of people in each population is determined according to the actual application, and no specific limitation is made in this embodiment. The ratio of the training set to the test set is determined according to the actual application, and no specific limitation is made in this embodiment. For example, the ratio of the training set to the test set is 9:1. The preset requirements for the completion of model training are determined according to the actual application, and no specific limitation is made in this embodiment. For example, the accuracy of model prediction reaches 95%, or the number of training times reaches 10,000 times.

[0062] The training process of the general artificial intelligence prediction model includes multiple links such as data collection, preprocessing, feature extraction, model selection, training, validation and tuning, and final deployment. First, a large amount of representative data needs to be collected, and data is collected from hyperglycemic, hypoglycemic, and normoglycemic populations to ensure that the model can cover various situations in the actual application scenario. In the training stage, the model parameters are learned using the training set, and the model hyperparameters are tuned with the help of the validation set to ensure that the model can generalize well to unseen data. During the training process, techniques such as regularization, cross-validation, and data augmentation can also be introduced to prevent the model from overfitting. After the model training and validation are completed, the test set is used to finally evaluate the model. After confirming that its performance indicators meet the expected requirements, the model is deployed to the actual application environment.

[0063] Optionally, the personalized adoptive non-invasive blood glucose monitoring method further includes:

[0064] S300. During the monitoring process, update the personalized artificial intelligence prediction model according to the feedback information.

[0065] After the personalized artificial intelligence prediction model is deployed, it is still necessary to monitor the running effect of the model in real time, and continuously optimize and update the model according to the feedback, so that it can always maintain a high prediction accuracy in the changing data environment.

[0066] When the target object is in use, the device adopts a two-stage blood glucose sensing mode of "constructing the association and individual data inheritance - non-invasive monitoring". The first stage is the association construction and individual inheritance stage (1 - 3 days). In this stage, the minimally invasive and non-invasive blood glucose sensing modules are combined. While using the microneedle electrode to detect the glucose concentration in tissue fluid as a reference value, the non-invasive module is used to collect various types of parameters of the metabolic heat integration method. Through the "minimally invasive + non-invasive" blood glucose data sets of a large number of patient groups, the internal relationship between the glucose concentration in tissue fluid and various types of parameters of the metabolic heat integration method is constructed. Furthermore, a universal "minimally invasive + non-invasive" blood glucose AI model can be built. At the same time, by collecting the minimally invasive + non-invasive blood glucose data sets of specific patient individuals, individualized migration calibration and minimally invasive - non-invasive transitional inheritance are realized. The original universal "minimally invasive + non-invasive" blood glucose model undergoes the transition and inheritance of "group - individual" and "minimally invasive - non-invasive". The second stage is the non-invasive monitoring stage, which is the actual application stage. After the system completes the association construction and individual inheritance, the minimally invasive blood glucose sensing module can be removed, and only the non-invasive blood glucose sensing module can be used to perform long-term and accurate personalized blood glucose monitoring on patient individuals. The core point of this method is to transition from the universal group minimally invasive blood glucose sensing combined with machine learning mode to the personalized individual non-invasive blood glucose sensing combined with machine learning. Therefore, the function of individualized non-invasive blood glucose monitoring is migrated and inherited from the basis of universal minimally invasive blood glucose sensing. Conventional universal sensing often has difficulty meeting the complex physiological differences between different individuals, such as complex factors like skin, tissue density, and blood components. This embodiment focuses on starting from the unique physiological characteristics of individuals, and through the way of migration and inheritance, reveals the unique correlation between minimally invasive physiological parameters and non-invasive physiological parameters, and is expected to overcome the similarities and differences between different patient individuals and adapt to the complex physiological specificity of patient individuals, that is, "adoptive" dynamic blood glucose monitoring. Different from the conventional universal sensing idea, although the adoptive individualized sensing requires patients to learn, transition, and inherit in the minimally invasive + non-invasive way during the individual inheritance stage, after this stage is completed, a non-invasive blood glucose sensor that fully conforms to the physiological state of the patient individual can be obtained. The idea provided by this method is an innovative sensing concept. The individualized adoptive non-invasive dynamic blood glucose monitoring method proposed in this application is expected to overcome the "bottleneck" in the field of non-invasive blood glucose monitoring and change the current situation of disease monitoring for diabetic patients. The individualized adoptive non-invasive sensing method can provide reference ideas for many methods that are difficult to accurately monitor non-invasively and expand the ideas of traditional universal sensors.

[0067] Implementing the embodiments of the present invention includes the following beneficial effects: First, a general artificial intelligence prediction model is trained through the blood glucose samples and physiological parameter samples of a group. The general artificial intelligence prediction model constructs the internal relationship between blood glucose and physiological parameters. Then, the trained general artificial intelligence prediction model is calibrated through the blood glucose samples and physiological parameter samples of the target individual to obtain a personalized artificial intelligence prediction model. The personalized artificial intelligence prediction model takes into account the individual differences of the target individual, thus realizing individualized transfer calibration and minimally invasive-noninvasive transitional inheritance. Finally, based on the physiological parameters measured non-invasively by the target individual and the personalized artificial intelligence prediction model, the blood glucose of the target individual is monitored, thereby realizing non-invasive, accurate, and continuous monitoring of the blood glucose of the target individual.

[0068] Referring to Figure 2 , the embodiments of the present invention provide a personalized adoptive non-invasive blood glucose monitoring system, including:

[0069] A first module for obtaining the blood glucose samples and physiological parameter samples of the target individual, and calibrating the trained general artificial intelligence prediction model according to the blood glucose samples and physiological parameter samples of the target individual to obtain a personalized artificial intelligence prediction model; the general artificial intelligence prediction model is trained according to the blood glucose samples and physiological parameter samples of a group;

[0070] A second module for obtaining the physiological parameters of the target individual, and monitoring the blood glucose of the target individual according to the physiological parameters of the target individual and the personalized artificial intelligence prediction model; the physiological parameters are measured non-invasively.

[0071] It can be seen that the content in the above method embodiments is applicable to the system embodiments of the present invention. The functions specifically implemented by the system embodiments are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those in the above method embodiments.

[0072] Referring to Figure 3 , the embodiments of the present invention provide a personalized adoptive non-invasive blood glucose monitoring device, including:

[0073] At least one processor;

[0074] At least one memory for storing at least one program;

[0075] When at least one program is executed by at least one processor, at least one processor implements the above method.

[0076] Among them, the memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. The memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory optionally includes a remote memory remotely disposed relative to the processor, and these remote memories can be connected to the processor through a network. Examples of the above networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0077] It can be seen that the content in the above method embodiments is applicable to the device embodiments of the present invention. The functions specifically implemented by the device embodiments of the present invention are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.

[0078] In addition, the embodiments of the present application also disclose a computer program product or a computer program. The computer program product or the computer program is stored in a computer-readable storage medium. The processor of the computer device can read the computer program from the computer-readable storage medium, and the processor executes the computer program to enable the computer device to execute the above method. Similarly, the content in the above method embodiments is applicable to the storage medium embodiments of the present invention. The functions specifically implemented by the storage medium embodiments of the present invention are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.

[0079] The embodiments of the present invention also provide a computer-readable storage medium. The computer-readable storage medium stores a program executable by a processor, and the program executable by the processor is used to implement the above method when executed by the processor.

[0080] It can be understood that all or some of the steps and systems disclosed in the above methods can be implemented as software, firmware, hardware, and their appropriate combinations. Some physical components or all physical components can be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include a computer storage medium (or non-transitory medium) and a communication medium (or transitory medium). As is well known to those of ordinary skill in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, as is well known to those of ordinary skill in the art, a communication medium typically contains computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transmission mechanism, and can include any information delivery medium.

[0081] Refer to Figure 4 , an embodiment of the present invention provides a personalized adoptive non-invasive blood glucose monitoring system, including a microneedle blood glucose sensing module (MI-CGM), a non-invasive blood glucose sensing module (NI-CGM), and a computer device, where the computer device is connected to the microneedle blood glucose sensing module and the non-invasive blood glucose sensing module; wherein,

[0082] The microneedle blood glucose sensing module is used to measure the blood glucose value and send the blood glucose value to the computer device;

[0083] The non-invasive blood glucose sensing module is used to measure physiological parameters and send the physiological parameters to the computer device;

[0084] The computer device includes:

[0085] At least one processor; [[ID=IS]]

[0086] At least one memory for storing at least one program;

[0087] [[ID=IS]]When at least one program is executed by at least one processor, at least one processor implements the above method.

[0088] In a specific embodiment, the manufacturing process of the microneedle blood glucose sensing module is as follows: The core microelectrodes were fabricated on a flexible PET substrate (50 μm) using screen printing, where the microelectrode patterns were prepared by laser cutting. The patterned PET substrate was subjected to via-hole treatment to establish the electrical connection electrode paths on the front and back sides of the electrodes. Carbon ink was screen printed layer by layer on the front and back sides of the PET substrate to form the working electrode, counter electrode, and reference electrode, respectively. The working electrode substrate was insulated by a printed resin layer, while the sensing area was exposed. Ag / AgCl ink was further patterned on the surface of the PET substrate to form the reference electrode substrate. In addition, carbon ink was screen printed onto the back of the PET substrate to form the counter electrode, which was connected to the electrode interface through a via-hole. The counter electrode and reference electrode were also insulated with resin to ensure a stable electrical connection. The glucose oxidase solution was mixed with a redox polymer mediator and a crosslinking agent evenly. The polymer enzyme mixture solution was coated on the working electrode by a high-precision enzyme dispenser. A thin polymer outer membrane was further coated on the surface of the microelectrodes by dip coating to reduce the absorption of non-specific proteins in vivo. The MN electrode may also include a rigid hollow metal MN shell for penetrating the cortex and an electrode core for subcutaneous glucose detection. To prepare the MN shell, a 23g syringe needle was cut into a hollow microneedle approximately 10 mm long by laser cutting. Subsequently, the MN shell was further ultrasonically treated in 75% alcohol for 1 hour and then insulated with parylene. The flexible core microelectrodes were integrated inside the metal NM shell and combined with the electrochemistry signal acquisition circuit to form the MI-CGM module. The core-shell microelectrodes were embedded in a via-hole substrate to facilitate the adjustment of the exposed microneedle length. In another design, the flexible microelectrodes can be inserted into the skin with the help of an external needle injection device, just as used in traditional CGM devices.

[0089] Optionally, the non-invasive blood glucose sensing module includes a metabolic heat sensor and a cardiovascular sensor; wherein;

[0090] The metabolic heat sensor is used to measure metabolic heat-related parameters;

[0091] The cardiovascular sensor is used to measure the pulse wave signal of the cardiovascular system.

[0092] The non-invasive blood glucose sensing module further includes a multi-parameter processing microsystem, which processes the parameters of the metabolic heat sensor and the cardiovascular sensor. The metabolic heat-related parameters include, but are not limited to, the surface temperature, thermal radiation temperature of the target individual, and the temperature and humidity of the measurement detection environment.

[0093] Optionally, the metabolic heat sensor includes a surface temperature sensor (STS), an ambient temperature sensor (ATS), an ambient humidity sensor (AHS), and an infrared radiation sensor (RTS).

[0094] In a specific embodiment, the STS and ATS are based on the SI7021-A20 chip. In the temperature sensor, two identical semiconductor triodes are used as the core components for temperature sensing. The sensor has a medical accuracy of ±0.3°C in the range of -10°C to 85°C. When the STS is attached to the wrist, the surface temperature of the participant can be detected through contact sensing. When the ATS is exposed to the test environment, the corresponding ambient temperature can be directly obtained. The RTS is based on the MLX90614-DAA chip. The RTS uses a thermopile as the core component for measuring thermal radiation. In the temperature range of 16°C to 40°C, the accuracy of the sensor is less than ±0.2°C. When testing human skin, the emissivity is set to 0.985 without recalibrating with a black body. By aligning the RTS with the skin surface of the human wrist (about 3 mm) and ensuring that the human skin is completely covered and focused, the thermal radiation temperature of the human skin can be detected. The AHS is also based on the SI7021-A20 chip. The AHS uses a metal-insulator-metal (MIM) capacitor as its main sensing element. The sensor has an accuracy of ±2%RH in the range of 20%-80%RH. The ambient temperature also affects the humidity, and the relative humidity usually decreases by about 5% for every 1°C. Therefore, the humidity sensor is equipped with a built-in temperature compensation mechanism, effectively improving the accuracy of humidity detection. By exposing the ambient humidity sensor to the test environment, the corresponding humidity value of the environment can be directly detected.

[0095] Optionally, the cardiovascular sensor includes a plurality of photoplethysmogram sensors (PPGs) for measuring the pulse wave signal at a preset site.

[0096] In a specific embodiment, the cardiovascular sensor includes two PPG sensors. The PPG sensors are based on two optical analog front-end (AFE) chips (MAX30101 and MAX30102). The light-emitting diode and the photodetector are the core detection elements of the PPG sensor. MAX30101 has a built-in LED that can emit green light (530 nm), red light (660 nm), and infrared light (880 nm), while MAX30102 has red light and infrared light. As the wavelength increases, the depth of light penetration deepens. Green light, red light, and infrared light can penetrate the epidermis, dermis, and subcutaneous layer respectively. Physiological differences between individuals, such as skin color and fat layer thickness, may affect the penetration distance and reflection intensity. The detector (PD) receives and detects the transmitted or reflected light signal from the tissue. When light of different intensities irradiates the PN junction of the photodiode, electron-hole pairs in the photodiode form a photocurrent. The photocurrent is converted into a voltage signal by a transimpedance amplifier, enabling the detection of the reflected light signal. The reflected light signal received by the PD consists of an alternating current (AC) signal and a direct current (DC) signal. The AC signal is the periodic change of the reflected light signal caused by the change in blood volume during arterial blood vessel pulsation. Therefore, the AC signal is the effective signal component of the PPG signal and accounts for 10% of the total signal intensity. The DC signal is the reflected light signal generated by venous blood, non-pulsating arterial blood, and other tissues (such as skin, bone, and muscle, etc.). Therefore, the DC signal is the invalid signal component of the PPG signal. Therefore, when the pulse wave propagates in the blood vessel, the contraction and relaxation of the arterial blood vessel system cause periodic changes in blood volume. The periodic change in blood volume affects the absorption and scattering of the emitted light signal, resulting in the AC signal detected by the PPG changing with the fluctuation of the pulse wave signal. By wearing two PPG sensors at the wrist and finger positions, the double pulse wave signals at the wrist and finger positions can be effectively detected. The PPG sensor can synchronize one or more light-emitting diode (LED) channels for optical signal acquisition and store the digital signal in the internal FIFO memory. Then the digital data is transmitted through the IIC protocol. The sampling rate of the PPG sensor is set to 200 Hz. The MAX30101 chip is configured in a mixed LED mode, where two LED channels are alternately mixed to achieve the pulse wave signal. The MAX30102 chip is configured in a blood oxygen mode, with the LED channels mixed, alternately completing the pulse wave signal acquisition and SaO2 calculation.

[0097] In the embodiments of the present invention, the wearable circuit of the non-invasive CGM device mainly consists of a non-invasive metabolic heat module, a minimally invasive microneedle module, and a multi-parameter processing microsystem. The circuit of the metabolic heat module is composed of a photoelectric sensor and a metabolic heat sensor. The photoelectric sensor can convert the collected analog signal into a digital signal and perform preprocessing operations, including amplification, filtering, sampling, and quantization of the electrical signal. The peripheral circuit outside the photoelectric ensures power supply and data communication. The metabolic heat sensor consists of an infrared radiation sensor, a surface temperature sensor, and an environmental temperature and humidity sensor. The metabolic heat sensor can collect human physiological-related parameters and transmit the signals to the MCU for further processing. The minimally invasive module is constructed by the AD5941 analog front end (AFE). The AD5941 is a high-precision, low-power AFE designed for portable applications that require high-precision electrochemical measurement technology. The AD5941 has an adjustable three-electrode detection system that converts the current signal of the microelectrode into a voltage signal by configuring the internal programmable gain amplifier (PGA) or adjustable resistor. Then, the processed signal is transmitted to the MCU through the SPI protocol. By connecting the RE pin, CE pin, and WE pin to the interface of the MI-CGM module, the microelectrode can be integrated into the microneedle module. The multi-parameter processing microsystem consists of a microcontroller unit, a Bluetooth module, and a power module. Based on the ARM Cortex-M4 core of STM32, the MCU provides sufficient external interfaces to coordinate signal acquisition, data processing, and real-time transmission between different components. The Bluetooth module consists of a BLE2U-A chip and peripheral circuits, and realizes wireless data transmission and reception between the wearable device and the host through the UART protocol. The power module converts the lithium-ion battery power supply (3.7V) into a stable 3.3V power supply through a low-dropout linear regulator (LDO, ME6209A33).

[0098] In a specific embodiment, refer to Figure 5 , the principle of adoptive non-invasive CGM and the dynamic blood glucose monitoring process are as follows. The monitoring device includes an MI-CGM module for minimally invasive and continuous BG recording, and an NI-CGM module for collecting MHC-related parameters. Minimally invasive and continuous blood glucose detection is performed through the MI-CGM module. The sensor is based on the second-generation glucose detection principle, in which the redox polymer promotes the connection between GOx and the electrode surface, enabling electrons to quickly transfer from the enzyme to the electrode for glucose detection. The NI-CGM module is used to collect human metabolic heat-related parameters for non-invasive blood glucose derivation.

[0099] In a specific embodiment, refer to Figure 6, the dynamic blood glucose sensing method provided by the embodiments of the present invention is as follows. The non-invasive CGM based on personalized sensors adopts a two-stage strategy to convert the general MI-CGM into personalized NI-CGM. During the sensor development process, a general dataset of MHC-related parameters and BG data was collected from different populations for the pre-training of the AI model. After being applied to an individual, in the initial stage, individual-specific MHC-related parameters and BG datasets are continuously collected for a short period (about 2 days). Among them, the MI-CGM module is used to collect accurate BG data, and the NI-MHC module is used to collect MHC-related multi-parameter data. At the same time, based on the personalized training dataset, a personalized artificial intelligence model tailored for this individual is generated. Then the MI-CGM module can be removed from the device (detachable), and the personalized AI model enables the NI-MHC module to accurately monitor BG from the MI-MHC module, especially for the BG of a specific individual.

[0100] The dynamic blood glucose sensing device includes two modules, MI-CGM and NI-CGM, which can be worn on the wrist. After collecting MI / NI-CGM data and developing a personalized AI model (2 days), the MI-CGM module can be detached from the device, leaving only the NI-CGM module for further detection of MHC-related parameters to obtain long-term non-invasive BG.

[0101] In a specific embodiment, refer to Figure 7 , the schematic diagram of the human body data collection by the two modules, MI-CGM and NI-CGM, provided by the embodiments of the present invention. Diabetic patients and the general population are sought to collect data continuously for several days at two time periods every day, and the two modules, MI-CGM and NI-CGM, collect data simultaneously as the dataset.

[0102] In a specific embodiment, refer to Figure 8 , the schematic diagram of the training of the personalized non-invasive CGM model provided by the present invention. After collecting the individual-specific MI / NI-CGM dataset through MI-CGM + NI-CGM, the customized AI model is retrained or migrated from the pre-trained AI model. Then the personalized adoptive NI-CGM can be used for accurate BG monitoring.

[0103] In a specific embodiment, refer to Figure 9, in the embodiments of the present invention, a correlation analysis was performed on the BG levels and MHC non-invasive parameters in the MI / NI-CGM dataset for all participants. The results showed that there were significant differences in the MI / NI-CGM datasets among different participants. When directly applied to individual patients after training using a general BG dataset, the general BG model is often affected by biases caused by inter-individual physiological variations. Therefore, the development of precise BG sensing technology needs to shift towards continuous tracking and personalized calibration of individual physiological monitoring.

[0104] In a specific embodiment, refer to Figure 10 , the embodiments of the present invention compared the result differences between the personalized adoptive non-invasive CGM monitoring method and the ordinary non-invasive CGM method. Figure 10 (a) in [reference] compares the BG values and trend errors among individuals using a radar chart. When using the AI model of personalized non-invasive CGM, the total area of the BG values and trend errors of 6 subjects is smaller, and the average error of each individual has good uniformity. It can effectively capture the correlation between MHC-related parameters and BG values of different individuals, and obtain a personalized AI model for different patients. Therefore, the BG values and trends of different subjects show better effects, and the overall distribution shows a uniform symmetric distribution. In contrast, for the AI model of traditional non-invasive CGM, the BG values and trend errors of all participants are larger, and the average errors of different participants show a high degree of asymmetry. Figure 10 (b) in [reference] performs a statistical analysis on the average error and trend of the BG values of all subjects. Radar charts of the average BG values and trend errors of each individual are plotted by two detection methods: the adoptive non-invasive CGM model and the ordinary non-invasive GCM. The smaller the area of the chart, the lower the BG value error or trend error, and the symmetric feature of the chart indicates that the result error of the AI model built by the adoptive personalized non-invasive CGM is significantly reduced. Figure 10 (c) in [reference] analyzes and compares the BG values of diabetic patients and the general population. As can be seen from the figure, the BG results using the adoptive personalized non-invasive CGM monitoring method are closer to the results of MI-CGM.

[0105] In a specific embodiment, 6 participants were randomly recruited for the adoptive non-invasive CGM test, including 3 healthy participants and 3 participants with a tendency to diabetes. All participants were clearly informed of the details involved in the experiment and gave informed consent before being included in this study. The experiment was conducted in a comfortable and quiet environment to ensure that the participants were not disturbed by the surrounding environment. The NI-CGM module and the MI-CGM module were applied to the human body simultaneously, but to different body parts. The MI-CGM module was worn on the participant's arm to record the reference minimally invasive BG value. The NI-CGM module was designed as a wearable device worn on the wrist to collect MHC-related parameters, which had been approved for ethical certification of human trials. During the 5-day test, the MI / NI-CGM dataset was recorded for about 5 to 6 hours per day, about 2 to 3 hours in the morning and about 2 to 4 hours in the afternoon. Throughout the test, the MI-CGM remained on the body. To accommodate the subject's daily activities, the NI-MHC could be removed at the end of the day and the location of the NI-MHC marked for subsequent testing in the following days. The test time, meal time, and relevant personal information of the participants, such as age, gender, and diabetes history, were also recorded separately. The 5-day MI / NI-CGM data from 6 participants were used to train and test the personalized AI model.

[0106] Referring to Figure 11 , the embodiment of the present invention provides a Clark diagram using a training set of a personalized non-invasive dynamic blood glucose sensing method. The BG value obtained by minimally invasive CGM is used as a reference. The results show that for each participant, most of the data points using the proposed personalized AI model fall in area A, meeting the error < 15%.

[0107] Referring to Figure 12 , the embodiment of the present invention provides a Clark diagram using a test set of a personalized non-invasive dynamic blood glucose sensing method and a common non-invasive blood glucose monitoring method. The BG value obtained by minimally invasive CGM is used as a reference. The results show that for each participant, most of the data points using the proposed personalized AI model fall in area A, and the error is significantly smaller than that of the common non-invasive blood glucose monitoring method.

[0108] The personalized minimally invasive and non-invasive combined adoptive dynamic blood glucose sensing method provided by the embodiments of the present invention has a BG error of 12.5±10.3%. Clinically, the standard for FDA, MDR or NMPA to approve BG measurement is MARD<15%. The calculated value of MARD of the non-invasive CGM model developed in the embodiments of the present invention is <15.3%, meeting the clinical accuracy standard. This is the first time to prove that non-invasive CGM meets the clinical accuracy rate standard (MARD≈15%) in multi-day continuous monitoring. In contrast, the MARD of ordinary non-invasive CGM is ~49.1%, far from reaching the standard for clinical application. Recently, the performance of non-invasive blood glucose measurement instruments by various μ-space offset Raman (MARD>20%) or mid-infrared photoacoustic signal methods (MARD<14.6%) has been reported, but due to the huge requirements of the instruments and settings, it is difficult to develop these technologies into wearable devices and continuous blood glucose monitoring cannot be carried out. The performance of recently commercialized MI-CGM, such as the products of Abbott, Dexcom, Medtronic and Microtech, where MARD is mostly distributed around 10%. The BG measurement accuracy of the adoptive non-invasive CGM in this embodiment is almost close to these commercialized MI-CGM products. This is the first time that non-invasive CGM can reach the clinical accuracy rate standard (MARD≈15%) in multi-day continuous monitoring. By using long-term learning of MI-CGM data or calibrating the sensor with finger BG values before the test day, the accuracy of adoptive non-invasive CGM may be improved in the future.

[0109] Converting minimally invasive BG sensing into a personalized device of non-invasive CGM based on an AI model is called "adoptive non-invasive biosensing". The personalized non-invasive CGM (NI-CGM) inherits the precise sensing ability of the minimally invasive CGM (MI-CGM) of the AI model. The device includes a short-term used MI-CGM module and a long-term used NI-CGM module based on metabolic thermal conformation. After collecting MI-CGM and NI-CGM data sets from an individual for a short term (about 2 days) for artificial intelligence training, a personalized artificial intelligence model suitable for a specific individual is developed. Subsequently, the MI-CGM module is removed, and only the NI-CGM module can be used for a long time. In operation, the adoptive non-invasive CGM focuses on capturing the unique characteristics and physiological states of an individual, understanding the specific correlation of the short-term MI / NI-CGM data sets, and then transferring it to long-term non-invasive sensing.

[0110] This patent is significant in several aspects. First, it provides an important methodological breakthrough, enabling NI-CGM to adopt the ability of MI-CGM to accurately record BG levels and trends, thus extending traditional minimally invasive sensors to customized non-invasive sensors. Second, personalized non-invasive CGM addresses the issue of individual diversity that still plagues general technologies and general sensors, providing a new paradigm for personalized non-invasive CGM. Finally, it demonstrates how to utilize advanced AI models to solve a challenging problem in precision biomedical applications. The concept of personalized sensors proposed in this work may change the traditional strategy for developing general sensors by suggesting that the development of customized sensors and personalized AI models can be adapted to different individuals. It is expected to solve the "bottleneck" in the field of non-invasive CGM and provide unique opportunities for personalized diagnosis and treatment of diabetes patients.

[0111] It can be seen that the content in the above method embodiments is applicable to the system embodiments herein. The functions specifically implemented by the system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.

[0112] It should be understood that in this application, "at least one (item)" means one or more, and "a plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships can exist. For example, "A and / or B" can represent: only A exists, only B exists, and both A and B exist simultaneously. Here, A and B can be singular or plural. The character " / " generally indicates that the associated objects before and after are in an "or" relationship. "At least one (one) of the following" or its similar expression refers to any combination of these items, including any combination of single item (one) or plural items (ones). For example, at least one (one) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0113] In several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection to each other can be through some interfaces, and the indirect coupling or communication connection of devices or units can be in electrical, mechanical, or other forms.

[0114] The unit described as a separation component may or may not be physically separated. The component displayed as a unit may or may not be a physical unit, that is, it may be located in one place or may be distributed across multiple network units. One or all of these units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0115] The above is a specific description of the preferred embodiment of the present invention. However, the present invention is not limited to the described embodiment. Those skilled in the art can make various equivalent deformations or substitutions without departing from the spirit of the present invention, and these equivalent deformations or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A personalized adoptive non-invasive blood glucose monitoring method, characterized in that, Including: Obtaining a blood glucose sample and a physiological parameter sample of a target individual, and calibrating a trained general artificial intelligence prediction model according to the blood glucose sample and the physiological parameter sample of the target individual to obtain a personalized artificial intelligence prediction model; the general artificial intelligence prediction model is trained according to the blood glucose samples and physiological parameter samples of a group; Obtaining the physiological parameters of the target individual, and monitoring the blood glucose of the target individual according to the physiological parameters of the target individual and the personalized artificial intelligence prediction model; The physiological parameters are measured non-invasively.

2. The method according to claim 1, characterized in that, The general artificial intelligence prediction model is trained by the following method: Obtaining blood glucose samples and physiological parameter samples of a number of populations, where the number of populations includes hyperglycemic populations, hypoglycemic populations, and normoglycemic populations; Dividing the blood glucose samples and physiological parameter samples of a number of the populations into a training set and a test set, training a preset artificial intelligence prediction model according to the training set, and testing the trained preset artificial intelligence prediction model according to the test set until a preset requirement is met to obtain a trained general artificial intelligence prediction model.

3. The method according to claim 1, wherein The method further includes: During the monitoring process, updating the personalized artificial intelligence prediction model according to feedback information.

4. A personalized adoptive non-invasive blood glucose monitoring system, characterized in that, Including: A first module for obtaining a blood glucose sample and a physiological parameter sample of a target individual, and calibrating a trained general artificial intelligence prediction model according to the blood glucose sample and the physiological parameter sample of the target individual to obtain a personalized artificial intelligence prediction model; the general artificial intelligence prediction model is trained according to the blood glucose samples and physiological parameter samples of a group; A second module for obtaining the physiological parameters of the target individual, and monitoring the blood glucose of the target individual according to the physiological parameters of the target individual and the personalized artificial intelligence prediction model; The physiological parameters are measured non-invasively.

5. A personalized adoptive non-invasive blood glucose monitoring device, characterized in that, Including: 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 method according to any one of claims 1-3.

6. A computer-readable storage medium storing a program executable by a processor, characterized in that, The program executable by the processor, when executed by the processor, is used to execute the method according to any one of claims 1-3.

7. A personalized adoptive non-invasive blood glucose monitoring system, characterized in that, Including a microneedle blood glucose sensing module, a non-invasive blood glucose sensing module, and a computer device, the computer device being connected to the microneedle blood glucose sensing module and the non-invasive blood glucose sensing module; wherein, The microneedle blood glucose sensing module is used to measure the blood glucose value and send the blood glucose value to the computer device; The non-invasive blood glucose sensing module is used to measure physiological parameters and send the physiological parameters to the computer device; The computer device includes: 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 method according to any one of claims 1-3.

8. The system according to claim 7, wherein The non-invasive blood glucose sensing module includes a metabolic heat sensor and a cardiovascular sensor; wherein; The metabolic heat sensor is used to measure metabolic heat-related parameters; The cardiovascular sensor is used to measure the pulse wave signal of the cardiovascular system.

9. The system according to claim 8, characterized in that, The metabolic heat sensor includes a surface temperature sensor, an environmental temperature and humidity sensor, and an infrared radiation sensor; wherein, The surface temperature sensor is used to measure the surface temperature of the target individual; The environmental temperature and humidity sensor is used to measure the temperature and humidity of the detection environment; The infrared radiation sensor is used to measure the thermal radiation temperature of the target individual.

10. The system according to claim 8, wherein The cardiovascular sensor includes a plurality of photoplethysmogram sensors, and the photoplethysmogram sensors are used to measure the pulse wave signal of a preset part.

Images