A non-invasive stroke monitoring and early warning device based on oral chip

Abstract

The application discloses a non-invasive stroke monitoring and early warning device based on an oral cavity chip, and relates to the technical field of biomedical engineering.The device comprises a flexible oral cavity wearable base worn in the oral cavity, a microfluidic sampling module, a biological sensor module and a signal processing and control module integrated on the flexible oral cavity wearable base.The microfluidic sampling module collects saliva, the biological sensor module detects a plurality of biomarkers in the collected saliva to obtain the concentration of each biomarker, the biomarkers include nerve injury markers and inflammatory factors, the signal processing and control module preliminarily judges whether there is a stroke aura risk based on the concentration of each biomarker, when there is a stroke aura risk, the concentration of each biomarker is taken as input, an integrated model is used to determine the stroke aura risk grade, and an early warning instruction is issued based on the stroke aura risk grade.The application can realize stroke monitoring and early warning.

Description

Technical Field

[0001] This application relates to the field of biomedical engineering technology, and in particular to a non-invasive stroke monitoring and early warning device based on an oral chip. Background Technology

[0002] Stroke is a leading cause of death and disability, characterized by high incidence, high disability rate, and high recurrence rate. Currently, early warning of stroke mainly relies on medical imaging (such as CT and MRI) or blood tests, which are complex, costly, and dependent on hospital environments. Recent studies have found that some neurological damage markers (such as S100β, NSE, and GFAP) and inflammatory factors (such as IL-6, TNF-α, and CRP) are abnormally expressed in body fluids in the early stages of stroke. Saliva, as a convenient and non-invasive body fluid sample, is suitable for continuous collection and dynamic monitoring; however, wearable saliva detection devices for stroke monitoring and early warning are currently lacking. Summary of the Invention

[0003] The purpose of this application is to provide a non-invasive stroke monitoring and early warning device based on an oral chip, which can realize stroke monitoring and early warning through saliva detection.

[0004] To achieve the above objectives, this application provides the following solution.

[0005] This application provides a non-invasive stroke monitoring and early warning device based on an oral microchip, the non-invasive stroke monitoring and early warning device based on an oral microchip includes: A flexible, wearable oral substrate that is worn inside the oral cavity; A microfluidic sampling module, integrated on the flexible oral wearable substrate, is used to collect saliva; A biosensor module, integrated on the flexible oral wearable substrate, is used to detect multiple biomarkers in the collected saliva and obtain the concentration of each biomarker; the biomarkers include neurodegenerative markers and inflammatory factors; The signal processing and control module, integrated on the flexible oral wearable substrate, is communicatively connected to the biosensor module. It is used to preliminarily determine whether there is a risk of stroke precursors based on the concentration of each biomarker. When there is a risk of stroke precursors, the concentration of each biomarker is used as input to determine the risk level of stroke precursors using an integrated model, and an early warning command is issued based on the risk level of stroke precursors.

[0006] Optionally, the microfluidic sampling module includes: a main channel, multiple sub-channels, and multiple detection chambers integrated on the flexible oral wearable substrate. The number of sub-channels, the number of detection chambers, and the number of biomarkers are the same. The sub-channels, the detection chambers, and the biomarkers correspond one-to-one. No two sub-channels are connected to each other, and no two detection chambers are connected to each other. The main channel is used to collect saliva; The inlet of each of the sub-channels is connected to the outlet of the main channel, so that the saliva collected by the main channel enters each of the sub-channels respectively; each of the sub-channels is used to transport the saliva entering the sub-channel to the detection chamber corresponding to the sub-channel; The entrance to each of the detection chambers is connected to the exit of the corresponding sub-channel of the detection chamber; each of the detection chambers is used to store saliva.

[0007] Optionally, the flexible oral wearable substrate has an Mxene coating; A first filter membrane is provided at the entrance of the main channel, and a second filter membrane is provided at the entrance of each sub-channel. The pore size of the first filter membrane is larger than that of the second filter membrane. The main channel and each of the sub-channels are hydrophilic microchannels, and each of the detection chambers is equipped with a micropump. The micropump is used to generate suction to draw saliva from the sub-channel into the detection chamber.

[0008] Optionally, the biosensor module includes: multiple biosensors, the number of which is the same as the number of biomarkers, each biosensor and each biomarker corresponding one-to-one, and each biosensor being installed in a detection chamber corresponding to a biomarker; The biosensor is used to detect the biomarker in saliva within the detection chamber to obtain the concentration of the biomarker.

[0009] Optionally, the neurological injury markers include calcium-binding protein S100B, neuron-specific enolase, and glial fibrillary acidic protein, and the inflammatory factors include interleukin-6, tumor necrosis factor α, and C-reactive protein. The biosensors for detecting the calcium-binding protein S100B and the biosensors for detecting the glial fibrillary acidic protein are both based on the principle of electrochemiluminescence. The biosensors for detecting the neuron-specific enolase, the biosensors for detecting interleukin-6, and the biosensors for detecting tumor necrosis factor α are all based on the principle of electrochemistry. The biosensor for detecting C-reactive protein is based on the principle of optical fluorescence. The detection chambers corresponding to the calcium-binding protein S100B, the glial fibrillary acidic protein, and the C-reactive protein are not adjacent.

[0010] Optionally, in making a preliminary assessment of the risk of stroke precursors based on the concentration of each biomarker, the signal processing and control module is used to: The concentration of each biomarker was preprocessed to obtain the preprocessed concentration of each biomarker; the preprocessing included amplification, filtering, and analog-to-digital conversion; Determine whether the concentration of any pretreated biomarker is greater than a concentration threshold. If yes, there is a risk of stroke warning signs; if no, there is no risk of stroke warning signs. This is a preliminary assessment of whether there is a risk of stroke warning signs.

[0011] Optionally, in determining the risk level of stroke precursors using an ensemble model with the concentration of each biomarker as input, and issuing an early warning command based on the risk level of stroke precursors, the signal processing and control module is used to: Using the concentration of each biomarker as input, the risk value of stroke precursors was determined using logistic regression model, XGBoost model, random forest model, convolutional neural network model, deep neural network model and support vector machine, respectively. Calculate the average of all the stroke precursor risk values, and select the stroke precursor risk values ​​whose absolute value of the difference from the average value is less than a preset difference as the selected risk values; The selected risk values ​​are weighted and summed to obtain the merged risk value. Based on the fused risk value, the risk level of stroke precursors is determined; the risk level of stroke precursors includes multiple levels. Based on the level of the stroke precursor risk, a warning instruction corresponding to that level is issued.

[0012] Optionally, in obtaining the fused risk value by weighted summation of all selected risk values, the signal processing and control module is used to: Based on the prediction accuracy of the selected models corresponding to all the selected risk values, the weights corresponding to all the selected risk values ​​are calculated. Based on the weights corresponding to all the selected risk values, a weighted sum is performed on all the selected risk values ​​to obtain the fused risk value. The formula for calculating the weight is: ; in, For the first The weight corresponding to each selected risk value; For the first The prediction accuracy of the selected model corresponding to each selected risk value. For the first The root mean square error of the selected model corresponding to each selected risk value; The number of selected risk values; For the first The prediction accuracy of the selected model corresponding to each selected risk value. For the first The root mean square error of the selected model corresponding to each selected risk value; The formula for calculating the risk value after fusion is as follows: ; in, The risk value after fusion; For the first Selected risk values.

[0013] Optionally, in determining the risk level of stroke precursors using an ensemble model with the concentration of each biomarker as input, the signal processing and control module is used to: Obtain the dataset; the dataset includes multiple sample data and label data corresponding to each sample data, the sample data is the sample concentration of each biomarker, and the label data is the risk value of stroke precursors in the sample; Based on the dataset, the Gini coefficient, p-value, and removal coefficient for each biomarker were calculated. Target biomarkers are selected from all biomarkers based on the Gini coefficient, p-value, and removal coefficient of each biomarker. Using the concentration of each target biomarker as input, an integrated model is used to determine the risk level of stroke precursors. The formula for calculating the removal coefficient is: ; in, This is the removal coefficient; The AUC value of the trained model obtained from the dataset after removing biomarkers; The AUC value is the value of the trained model obtained from the dataset before removing biomarkers.

[0014] Optionally, the non-invasive stroke monitoring and early warning device based on an oral chip further includes: A wireless communication module, integrated on the flexible oral wearable substrate, is communicatively connected to the signal processing and control module and an external device, respectively, for transmitting the concentration of each biomarker, the judgment result of the presence of stroke precursor risk, and the stroke precursor risk level to the external device; the external device includes a mobile terminal and a remote server; An early warning module, integrated on the flexible oral wearable substrate, is communicatively connected to the signal processing and control module and is used to issue an early warning based on the early warning command; A power module, integrated on the flexible oral wearable substrate, is electrically connected to the biosensor module, the signal processing and control module, the wireless communication module, and the early warning module, respectively, and is used to supply power to the biosensor module, the signal processing and control module, the wireless communication module, and the early warning module.

[0015] According to the specific embodiments provided in this application, this application has the following technical effects: This application provides a non-invasive stroke monitoring and early warning device based on an oral chip, including a flexible oral wearable substrate worn inside the oral cavity and a microfluidic sampling module, a biosensor module, and a signal processing and control module integrated on the flexible oral wearable substrate. The microfluidic sampling module collects saliva, and the biosensor module detects multiple biomarkers in the collected saliva to obtain the concentration of each biomarker. The biomarkers include neurodegenerative markers and inflammatory factors. Based on the concentration of each biomarker, the signal processing and control module makes a preliminary judgment on whether there is a risk of stroke precursors. When there is a risk of stroke precursors, the concentration of each biomarker is used as input to determine the risk level of stroke precursors using an integrated model, and an early warning command is issued based on the risk level of stroke precursors. This application designs a flexible oral wearable substrate, a microfluidic sampling module, a biosensor module, and a signal processing and control module, which can be worn inside the oral cavity as an oral chip to detect biomarkers in saliva, further determining the presence and level of stroke precursor risk. It is a wearable saliva detection device that can conveniently and non-invasively collect and detect saliva, enabling stroke monitoring and early warning. Users can decide whether to seek medical attention based on the stroke precursor risk level, and after seeking medical attention, doctors will diagnose whether the user has a stroke. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a structural block diagram of a non-invasive stroke monitoring and early warning device based on an oral chip, provided in Embodiment 1 of this application.

[0018] Figure label: 1 - Flexible oral wearable substrate. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] Example 1 This embodiment provides a non-invasive stroke monitoring and early warning device based on an oral microchip, such as... Figure 1 As shown, the non-invasive stroke monitoring and early warning device based on an oral chip includes: Flexible oral wearable substrate 1, worn inside the oral cavity; A microfluidic sampling module, integrated on a flexible oral wearable substrate 1, is used to collect saliva; A biosensor module, integrated on a flexible oral wearable substrate 1, is used to detect multiple biomarkers in collected saliva and obtain the concentration of each biomarker, including neurodegenerative markers and inflammatory factors. The signal processing and control module, integrated on the flexible oral wearable substrate 1, communicates with the biosensor module. It is used to preliminarily determine whether there is a risk of stroke precursors based on the concentration of each biomarker. When there is a risk of stroke precursors, the concentration of each biomarker is used as input to determine the risk level of stroke precursors using the integrated model, and a warning command is issued based on the risk level of stroke precursors.

[0021] An oral chip refers to an integrated analysis system with microfluidic control, which is constructed by assembling micro-functional components using microfabrication technology. By designing a flexible oral wearable substrate, a microfluidic sampling module, a biosensor module, and a signal processing and control module, it can be worn inside the oral cavity as an oral chip to detect biomarkers in saliva, and further determine the presence and level of stroke precursor risk. It is suitable for early risk warning of cerebrovascular diseases such as stroke, and aims to monitor in real time biomarkers such as neurodegenerative markers and inflammatory factors related to stroke in saliva, so as to achieve automatic identification and early warning of stroke precursor risk.

[0022] The non-invasive stroke monitoring and early warning device based on an oral microchip in this embodiment also includes: The wireless communication module, integrated on the flexible oral wearable substrate 1, communicates with the signal processing and control module and external devices respectively. It is used to transmit the concentration of each biomarker, the judgment result of the presence of stroke precursor risk, and the stroke precursor risk level to the external devices, including mobile terminals and remote servers. The early warning module, integrated on the flexible oral wearable substrate 1, communicates with the signal processing and control module and is used to issue early warnings based on early warning commands. The power supply module, integrated on the flexible oral wearable substrate 1, is electrically connected to the biosensor module, signal processing and control module, wireless communication module and early warning module, respectively, and is used to supply power to the biosensor module, signal processing and control module, wireless communication module and early warning module.

[0023] The following, combined with Figure 1 This embodiment provides a detailed description of the non-invasive stroke monitoring and early warning device based on an oral chip.

[0024] (a) Flexible oral wearable substrate The flexible oral wearable substrate is worn inside the oral cavity, specifically conforming to the inner wall of the user's mouth (such as the palate) or the surface of the teeth, and is suitable for long-term wear.

[0025] (ii) Microfluidic sampling module The microfluidic sampling module is integrated on a flexible oral wearable substrate for collecting saliva, specifically through capillary action and / or micropump technology.

[0026] The microfluidic sampling module includes a main channel integrated on a flexible oral wearable substrate, multiple sub-channels, and multiple detection chambers. The number of sub-channels, detection chambers, and biomarkers are the same, and each sub-channel, detection chamber, and biomarker corresponds one-to-one. No two sub-channels or detection chambers are connected to each other. In other words, each sub-channel and each detection chamber is physically isolated to prevent cross-contamination.

[0027] The main channel is used to collect saliva.

[0028] Each subchannel's inlet is connected to the main channel's outlet, allowing saliva collected from the main channel to enter each subchannel. Each subchannel is used to transport the saliva entering the subchannel to the corresponding detection chamber.

[0029] The entrance to each detection chamber is connected to the exit of the corresponding sub-channel, and each detection chamber is used to store saliva.

[0030] The flexible oral wearable substrate has an Mxene coating. The Mxene coating is formed by spin-coating a Ti3C2Tx (titanium carbide) solution onto the flexible oral wearable substrate and then drying it. The surface is enriched with hydroxyl and oxygen functional groups, which can reduce non-specific protein adsorption and reduce background signals in the sensor signals collected by the biosensor module.

[0031] A diverter is installed between the main channel and the sub-channels. This diverter ensures that the saliva collected by the main channel is evenly distributed into each sub-channel, resulting in equal saliva content in each sub-channel.

[0032] A first filter membrane is installed at the entrance of the main channel, and a second filter membrane is installed at the entrance of each sub-channel. The pore size of the first filter membrane is larger than that of the second filter membrane. As an example, the first filter membrane can be a 5μm filter membrane to remove large particles, and the second filter membrane can be a 0.22μm filter membrane to remove bacteria.

[0033] Capillary action utilizes the Laplace pressure difference generated by the surface tension of the liquid within the microchannel to drive the flow of saliva without the need for external energy. As a high-viscosity non-Newtonian fluid, saliva can overcome viscous resistance and achieve self-drive in hydrophilic microchannels (contact angle <90°). Therefore, in this embodiment, the main channel and each sub-channel are designed as hydrophilic microchannels to drive saliva into the detection chamber based on capillary action.

[0034] Optionally, each detection chamber is equipped with a micropump, which generates suction to draw saliva from the subchannel into the detection chamber, thereby accelerating the delivery speed of saliva and allowing the saliva entering the subchannel to quickly enter the detection chamber.

[0035] In this embodiment, the width of both the main channel and the sub-channel of the microfluidic sampling module can be controlled within the range of 100-300μm, supporting automatic saliva sampling.

[0036] (III) Biosensor Module The biosensor module is integrated on a flexible oral wearable substrate and connected to a microfluidic sampling module. It is used to detect multiple biomarkers in saliva collected by the microfluidic sampling module and obtain the concentration of each biomarker. The biomarkers include neurodegenerative markers and inflammatory factors. Neurodegenerative markers reflect the destruction of brain cell structure, while inflammatory factors drive secondary brain injury.

[0037] Specifically, the biosensor module is used to detect neurological damage markers and inflammatory factors in saliva. Neurological damage markers include, but are not limited to, at least one of the following: calcium-binding protein S100B (S100β, released from damaged astrocytes, enters body fluids through the disrupted blood-brain barrier, reflecting glial cell damage and blood-brain barrier permeability), neuron-specific enolase (NSE, a glycolytic enzyme in neuronal cytoplasm, released into blood / cerebrospinal fluid during neuronal necrosis, specifically reflecting the degree of neuronal damage), and glial fibrillary acidic protein (GFAP, intermediate filament protein of the astrocyte cytoskeleton, leaked during brain structural damage, a marker of astrocyte activation and mechanical damage). Inflammatory factors include, but are not limited to, those mentioned above. The biosensor module is used to simultaneously detect biomarkers such as S100β, NSE, GFAP, IL-6, TNF-α, and CRP, obtaining the concentration of each biomarker, including at least one of the following: interleukin-6 (IL-6, produced by endothelial cells, microglia, astrocytes and leukocytes after ischemia, mediating leukocyte adhesion and infiltration, and aggravating inflammatory response), tumor necrosis factor-α (TNF-α, secreted by M1 type microglia, an acute phase reaction product, inducing vascular endothelial necrosis and disrupting the blood-brain barrier), and C-reactive protein (CRP, an acute phase reaction protein synthesized by the liver, a systemic inflammatory marker, exacerbating atherosclerosis, and predicting stroke risk).

[0038] The biosensor module includes: multiple biosensors, the number of which is the same as the number of biomarkers, with a one-to-one correspondence between the biosensors and biomarkers, and the biosensors are installed in the detection chamber corresponding to the biomarkers.

[0039] Biosensors are used to detect biomarkers in saliva within a detection chamber, determining their concentration. These biosensors employ immunoimmobilization-based detection technology, a highly sensitive bioanalytical technique that combines antibody immobilization with sensing. By directionally immobilizing antibodies on the electrode surface of the biosensor, changes in electron transport resistance or redox reactions occur on the electrode surface when antigens such as S100β, NSE, GFAP, IL-6, TNF-α, and CRP bind to the antibodies. This quantifies the detection signal, enabling rapid and sensitive detection of biomarkers. In short, by immobilizing antibodies on the electrode surface of the biosensor, highly sensitive detection of biomarkers is achieved.

[0040] In this embodiment, the neurological injury markers include calcium-binding protein S100B, neuron-specific enolase, and glial fibrillary acidic protein. The inflammatory factors include interleukin-6, tumor necrosis factor-α, and C-reactive protein. In this embodiment, there are 6 sub-channels, 6 detection chambers, and 6 biosensors. One biomarker corresponds to one sub-channel, one detection chamber, and one biosensor. For a certain biomarker, saliva enters the detection chamber corresponding to that biomarker through the sub-channel. The biosensor corresponding to that biomarker, located in the detection chamber, detects the biomarker in the saliva to obtain the concentration of the biomarker.

[0041] Optionally, the biosensor for detecting calcium-binding protein S100B and the biosensor for detecting glial fibrillary acidic protein are both based on the principle of electrochemiluminescence; the biosensor for detecting neuron-specific enolase, the biosensor for detecting interleukin-6, and the biosensor for detecting tumor necrosis factor α are all based on the principle of electrochemistry; and the biosensor for detecting C-reactive protein is based on the principle of optical fluorescence.

[0042] Considering that both electrochemiluminescence and optical fluorescence principles emit light, they may interfere with each other if they are adjacent. Therefore, the detection chambers corresponding to calcium-binding protein S100B, glial fibrillary acidic protein, and C-reactive protein are set to be non-adjacent to avoid light signal interference.

[0043] (iv) Signal Processing and Control Module The signal processing and control module is integrated on a flexible oral wearable substrate and communicates with the biosensor module. It is used to collect and process the sensor signals of the biosensor module and has an artificial intelligence risk assessment function. It is used to comprehensively analyze the concentration of each biomarker through an integrated machine learning model (i.e., ensemble model), and output the assessment result of the stroke precursor risk level (i.e., the classification result) and the warning command. Specifically, it is used to preliminarily determine whether there is a risk of stroke precursor based on the concentration of each biomarker. When there is a risk of stroke precursor, the concentration of each biomarker is used as input to determine the stroke precursor risk level using the ensemble model and issue a warning command based on the stroke precursor risk level.

[0044] When the signal processing and control module communicates with the biosensor module, wireless or wired communication methods can be used.

[0045] When collecting and processing sensor signals from the biosensor module, the sensor signals can be amplified, filtered, converted from analog to digital, and preliminarily analyzed. Specifically, moving average filtering can be used to remove low-frequency drift and improve the signal-to-noise ratio. In this embodiment, in order to preliminarily determine whether there is a risk of stroke precursors based on the concentration of each biomarker, the signal processing and control module is used to: preprocess the concentration of each biomarker to obtain the preprocessed concentration of each biomarker. The preprocessing includes amplification, filtering, and analog-to-digital conversion; determine whether the preprocessed concentration of any biomarker is greater than a concentration threshold; if so, there is a risk of stroke precursors; if not, there is no risk of stroke precursors, thus preliminarily determining whether there is a risk of stroke precursors and completing the preliminary analysis.

[0046] Considering that abnormalities in biomarkers do not begin after a stroke, but rather show a chronic elevation some time before the stroke, this embodiment uses the concentration of each biomarker to initially determine the presence of a risk of stroke precursors. Stroke precursor risk refers to the possibility of acute stroke indicated by transient, reversible neurological deficits that appear some time before a stroke. After identifying a risk of stroke precursors, a stroke precursor risk level assessment is then conducted. It should be noted that determining the stroke precursor risk level does not determine whether the user has a stroke; a doctor's diagnosis is still required. Therefore, after obtaining the stroke precursor risk level, the user can decide whether to seek medical attention. After seeking medical attention, a doctor will make a diagnosis to determine whether the user has a stroke.

[0047] In this embodiment, the signal processing and control module is responsible for determining the risk level of stroke precursors using an ensemble model with the concentration of each biomarker as input, and issuing warning commands based on the risk level of stroke precursors. (1) Using the concentration of each biomarker as input, the risk value of stroke precursors was determined by logistic regression model, XGBoost model, random forest model, convolutional neural network model, deep neural network model and support vector machine respectively.

[0048] Logistic Regression models offer linear interpretability; XGBoost models capture nonlinear interactions; Random Forest models are highly accurate, automatically handle feature interactions, and are robust to outliers; Convolutional Neural Network models are highly accurate, automatically extract multi-scale spatiotemporal features without requiring manual feature engineering, and possess high-dimensional nonlinear learning capabilities; Deep Neural Network models are highly accurate, possess deep learning capabilities, and can automatically capture extremely complex nonlinear relationships between biomarkers; Support Vector Machines (SVMs) offer high-dimensional pattern recognition. Therefore, this embodiment selects the above six models. For each model, the concentration of each biomarker is used as input to determine the risk value of stroke precursors.

[0049] (2) Calculate the average value of all stroke precursor risk values, and select the stroke precursor risk value whose absolute value of the difference from the average value is less than the preset difference value as the selected risk value.

[0050] When integrating the stroke precursor risk values ​​determined by the six models, inaccurate stroke precursor risk values ​​are first removed. In this embodiment, the average value of the six stroke precursor risk values ​​is first calculated. For each stroke precursor risk value, the absolute value of the difference between the stroke precursor risk value and the average value is calculated, and it is determined whether the absolute value is less than a preset difference. If so, the stroke precursor risk value is selected as the risk value for subsequent fusion. If not, the stroke precursor risk value is removed.

[0051] (3) The selected risk values ​​are weighted and summed to obtain the merged risk value.

[0052] In terms of weighted summation of all selected risk values ​​to obtain the fused risk value, the signal processing and control module is used to: calculate the weights corresponding to all selected risk values ​​based on the prediction accuracy of the selected models corresponding to all selected risk values; and perform weighted summation of all selected risk values ​​based on the weights corresponding to all selected risk values ​​to obtain the fused risk value.

[0053] The formula for calculating the weight is as follows: ; in, For the first The weight corresponding to each selected risk value; For the first The prediction accuracy of the selected model corresponding to each selected risk value (i.e., the model that yields that selected risk value; for example, if the risk value of stroke precursors determined by a logistic regression model is used as the selected risk value, then the selected model corresponding to that risk value is the logistic regression model). For the first The root mean square error of the selected model corresponding to each selected risk value; The number of selected risk values; For the first The prediction accuracy of the selected model corresponding to each selected risk value. For the first The root mean square error of the selected model corresponding to each selected risk value.

[0054] The formula for calculating the post-fusion risk value is as follows: ; in, The risk value after fusion; For the first Selected risk values.

[0055] (4) Based on the fused risk value, the risk level of stroke precursor is determined. The risk level of stroke precursor includes multiple levels.

[0056] The higher the risk value after fusion, the higher the level of stroke precursor risk. In this embodiment, the numerical range corresponding to each level is predetermined. The level of stroke precursor risk is determined according to the numerical range in which the risk value after fusion falls.

[0057] (5) Based on the risk level of stroke precursors, issue the corresponding warning instruction.

[0058] The risk level of stroke precursors can be low, medium or high. When the risk level of stroke precursors is low, a low-risk warning instruction is issued; when the risk level of stroke precursors is medium, a medium-risk warning instruction is issued; and when the risk level of stroke precursors is high, a high-risk warning instruction is issued.

[0059] To improve efficiency, the signal processing and control module is used to perform the following steps in determining the risk level of stroke precursors using an ensemble model with the concentration of each biomarker as input: (1) Obtain the dataset, which includes multiple sample data and label data corresponding to each sample data. The sample data is the sample concentration of each biomarker, and the label data is the risk value of stroke precursors in the sample.

[0060] (2) Based on the dataset, the Gini coefficient, p-value and removal coefficient of each biomarker were calculated.

[0061] A random forest model is trained based on the dataset, and the Gini coefficient for each biomarker is further calculated. Based on the dataset, the p-value between the sample concentration of each biomarker and the risk value of stroke precursors in the sample can be calculated, and the p-value for each biomarker is obtained. An initial model is trained based on the dataset, and the AUC (Area Under ROC Curve) value of the trained model is obtained. A certain biomarker is removed from the dataset, and the dataset after removal is obtained. The initial model is trained using the dataset after removal, and the AUC value of the trained model is obtained. The removal coefficient of the certain biomarker is calculated based on the AUC values ​​of the trained model before and after removing the certain biomarker, and the removal coefficient of each biomarker is obtained. The initial model can be any machine learning model.

[0062] The formula for calculating the removal coefficient is as follows: ; in, This is the removal coefficient; The AUC value of the trained model obtained after removing the biomarker, based on a dataset (the dataset does not contain the biomarker); The AUC value is the value of the trained model obtained from training on the dataset (which contains the biomarker) before the biomarker is removed.

[0063] (3) Based on the Gini coefficient, p-value and removal coefficient of each biomarker, the target biomarker is screened from all biomarkers.

[0064] Based on the Gini coefficient of each biomarker, the importance of each biomarker is determined, and biomarkers with Gini coefficients greater than a preset Gini coefficient are selected as the first biomarker. Based on the p-value of each biomarker, the importance of each biomarker is determined, and biomarkers with p-values ​​less than a preset p-value are selected as the second biomarker. Based on the removal coefficient of each biomarker, the importance of each biomarker is determined, and biomarkers with removal coefficients less than a preset removal coefficient are selected as the third biomarker. The intersection of the first, second, and third biomarkers is taken, and the biomarkers that are the same among the first, second, and third biomarkers are selected as the target biomarkers.

[0065] (4) Using the concentration of each target biomarker as input, the risk level of stroke precursors is determined by using an integrated model.

[0066] At this point, it is not necessary to use the concentration of each biomarker as input to predict the risk level of stroke precursors, which can reduce the number of input variables and improve efficiency.

[0067] The signal processing and control module can be designed based on the STM32 low-power microcontroller and has a built-in BLE communication module.

[0068] (v) Wireless communication module The wireless communication module is integrated on the flexible oral wearable substrate and communicates with the signal processing and control module and external devices respectively. It is used to upload the data processed by the signal processing and control module to the external devices. Specifically, it is used to transmit the concentration of each biomarker, the judgment result of the presence of stroke precursor risk, and the stroke precursor risk level to the external devices, including mobile terminals and remote servers.

[0069] The wireless communication module can connect to the mobile terminal via BLE or NFC to enable real-time data upload.

[0070] (vi) Early warning module The early warning module is integrated on a flexible oral wearable substrate and communicates with the signal processing and control module. It is used to issue early warnings based on the early warning commands issued by the signal processing and control module. It can issue early warnings in different ways at different levels of stroke precursor risk, or it can issue early warnings only when the stroke precursor risk level is high.

[0071] (vii) Power supply module The power module is integrated on the flexible oral wearable substrate and is electrically connected to the biosensor module, signal processing and control module, wireless communication module and early warning module respectively. It is used to provide working power to each module, specifically providing stable low-power power, that is, to power the biosensor module, signal processing and control module, wireless communication module and early warning module.

[0072] The non-invasive stroke monitoring and early warning device based on an oral chip in this embodiment can be calibrated, data viewed, and early warning commands received via a mobile APP.

[0073] Developing an intelligent oral chip that integrates multiple biomarker detection functions and is suitable for daily wear and data uploading has significant clinical and social implications. This embodiment achieves multi-parameter non-invasive monitoring and intelligent analysis through an oral chip worn by the user daily, overcoming the shortcomings of existing technologies such as strong dependence on equipment and lack of continuous monitoring capabilities, and has significant practical and promotional value.

[0074] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0075] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A non-invasive stroke monitoring and early warning device based on an oral microchip, characterized in that, The non-invasive stroke monitoring and early warning device based on an oral microchip includes: A flexible, wearable oral substrate that is worn inside the oral cavity; A microfluidic sampling module, integrated on the flexible oral wearable substrate, is used to collect saliva; A biosensor module, integrated on the flexible oral wearable substrate, is used to detect multiple biomarkers in the collected saliva and obtain the concentration of each biomarker; the biomarkers include neurodegenerative markers and inflammatory factors; The signal processing and control module, integrated on the flexible oral wearable substrate, is communicatively connected to the biosensor module. It is used to preliminarily determine whether there is a risk of stroke precursors based on the concentration of each biomarker. When there is a risk of stroke precursors, the concentration of each biomarker is used as input to determine the risk level of stroke precursors using an integrated model, and an early warning command is issued based on the risk level of stroke precursors.

2. The non-invasive stroke monitoring and early warning device based on an oral chip according to claim 1, characterized in that, The microfluidic sampling module includes: a main channel, multiple sub-channels, and multiple detection chambers integrated on the flexible oral wearable substrate. The number of sub-channels, the number of detection chambers, and the number of biomarkers are the same. The sub-channels, detection chambers, and biomarkers correspond one-to-one. No two sub-channels are connected to each other, and no two detection chambers are connected to each other. The main channel is used to collect saliva; The inlet of each of the sub-channels is connected to the outlet of the main channel, so that the saliva collected by the main channel enters each of the sub-channels respectively; each of the sub-channels is used to transport the saliva entering the sub-channel to the detection chamber corresponding to the sub-channel; The entrance to each of the detection chambers is connected to the exit of the corresponding sub-channel of the detection chamber; each of the detection chambers is used to store saliva.

3. The non-invasive stroke monitoring and early warning device based on an oral chip according to claim 2, characterized in that, The flexible oral wearable substrate has an Mxene coating; A first filter membrane is provided at the entrance of the main channel, and a second filter membrane is provided at the entrance of each sub-channel. The pore size of the first filter membrane is larger than that of the second filter membrane. The main channel and each of the sub-channels are hydrophilic microchannels, and each of the detection chambers is equipped with a micropump. The micropump is used to generate suction to draw saliva from the sub-channel into the detection chamber.

4. The non-invasive stroke monitoring and early warning device based on an oral chip according to claim 2, characterized in that, The biosensor module includes: multiple biosensors, the number of which is the same as the number of biomarkers, and there is a one-to-one correspondence between the biosensors and the biomarkers. The biosensors are installed in the detection chambers corresponding to the biomarkers. The biosensor is used to detect the biomarker in saliva within the detection chamber to obtain the concentration of the biomarker.

5. The non-invasive stroke monitoring and early warning device based on an oral chip according to claim 4, characterized in that, The neurological injury markers include calcium-binding protein S100B, neuron-specific enolase, and glial fibrillary acidic protein; the inflammatory factors include interleukin-6, tumor necrosis factor-α, and C-reactive protein. The biosensors for detecting the calcium-binding protein S100B and the biosensors for detecting the glial fibrillary acidic protein are both based on the principle of electrochemiluminescence. The biosensors for detecting the neuron-specific enolase, the biosensors for detecting interleukin-6, and the biosensors for detecting tumor necrosis factor α are all based on the principle of electrochemistry. The biosensor for detecting C-reactive protein is based on the principle of optical fluorescence. The detection chambers corresponding to the calcium-binding protein S100B, the glial fibrillary acidic protein, and the C-reactive protein are not adjacent.

6. The non-invasive stroke monitoring and early warning device based on an oral chip according to claim 1, characterized in that, In the preliminary assessment of the risk of stroke precursors based on the concentration of each biomarker, the signal processing and control module is used for: The concentration of each biomarker was preprocessed to obtain the preprocessed concentration of each biomarker; the preprocessing included amplification, filtering, and analog-to-digital conversion; Determine whether the concentration of any pretreated biomarker is greater than a concentration threshold. If yes, there is a risk of stroke warning signs; if no, there is no risk of stroke warning signs. This is a preliminary assessment of whether there is a risk of stroke warning signs.

7. The non-invasive stroke monitoring and early warning device based on an oral chip according to claim 1, characterized in that, In determining the risk level of stroke precursors using an ensemble model with the concentration of each biomarker as input, and issuing early warning commands based on the stroke precursor risk level, the signal processing and control module is used for: Using the concentration of each biomarker as input, the risk value of stroke precursors was determined using logistic regression model, XGBoost model, random forest model, convolutional neural network model, deep neural network model and support vector machine, respectively. Calculate the average of all the stroke precursor risk values, and select the stroke precursor risk values ​​whose absolute value of the difference from the average value is less than a preset difference as the selected risk values; The selected risk values ​​are weighted and summed to obtain the merged risk value. Based on the fused risk value, the risk level of stroke precursors is determined; The risk levels for stroke warning signs include multiple levels; Based on the level of the stroke precursor risk, a warning instruction corresponding to that level is issued.

8. The non-invasive stroke monitoring and early warning device based on an oral chip according to claim 7, characterized in that, In terms of weighted summation of all selected risk values ​​to obtain the fused risk value, the signal processing and control module is used for: Based on the prediction accuracy of the selected models corresponding to all the selected risk values, the weights corresponding to all the selected risk values ​​are calculated. Based on the weights corresponding to all the selected risk values, a weighted sum is performed on all the selected risk values ​​to obtain the fused risk value. The formula for calculating the weight is: ； in, For the first The weight corresponding to each selected risk value; For the first The prediction accuracy of the selected model corresponding to each selected risk value. For the first The root mean square error of the selected model corresponding to each selected risk value; The number of selected risk values; For the first The prediction accuracy of the selected model corresponding to each selected risk value. For the first The root mean square error of the selected model corresponding to each selected risk value; The formula for calculating the risk value after fusion is as follows: ； in, The risk value after fusion; For the first Selected risk values.

9. The non-invasive stroke monitoring and early warning device based on an oral chip according to claim 7, characterized in that, In determining the risk level of stroke precursors using an integrated model with the concentration of each biomarker as input, the signal processing and control module is used for: Obtain the dataset; the dataset includes multiple sample data and label data corresponding to each sample data, the sample data is the sample concentration of each biomarker, and the label data is the risk value of stroke precursors in the sample; Based on the dataset, the Gini coefficient, p-value, and removal coefficient for each biomarker were calculated. Target biomarkers are selected from all biomarkers based on the Gini coefficient, p-value, and removal coefficient of each biomarker. Using the concentration of each target biomarker as input, an integrated model is used to determine the risk level of stroke precursors. The formula for calculating the removal coefficient is: ； in, This is the removal coefficient; The AUC value of the trained model obtained from the dataset after removing biomarkers; The AUC value is the value of the trained model obtained from the dataset before removing biomarkers.

10. The non-invasive stroke monitoring and early warning device based on an oral chip according to claim 1, characterized in that, The non-invasive stroke monitoring and early warning device based on an oral chip also includes: A wireless communication module, integrated on the flexible oral wearable substrate, is communicatively connected to the signal processing and control module and an external device, respectively, for transmitting the concentration of each biomarker, the judgment result of the presence of stroke precursor risk, and the stroke precursor risk level to the external device; the external device includes a mobile terminal and a remote server; An early warning module, integrated on the flexible oral wearable substrate, is communicatively connected to the signal processing and control module and is used to issue an early warning based on the early warning command; A power module, integrated on the flexible oral wearable substrate, is electrically connected to the biosensor module, the signal processing and control module, the wireless communication module, and the early warning module, respectively, and is used to supply power to the biosensor module, the signal processing and control module, the wireless communication module, and the early warning module.

Images