CN113288131A - Non-invasive blood glucose detection method, processor and device based on graph convolution network - Google Patents
Non-invasive blood glucose detection method, processor and device based on graph convolution network Download PDFInfo
- Publication number
- CN113288131A CN113288131A CN202110489689.9A CN202110489689A CN113288131A CN 113288131 A CN113288131 A CN 113288131A CN 202110489689 A CN202110489689 A CN 202110489689A CN 113288131 A CN113288131 A CN 113288131A
- Authority
- CN
- China
- Prior art keywords
- graph
- blood glucose
- value
- ppg signal
- sample
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 239000008280 blood Substances 0.000 title claims abstract description 150
- 210000004369 blood Anatomy 0.000 title claims abstract description 150
- WQZGKKKJIJFFOK-GASJEMHNSA-N Glucose Natural products OC[C@H]1OC(O)[C@H](O)[C@@H](O)[C@@H]1O WQZGKKKJIJFFOK-GASJEMHNSA-N 0.000 title claims abstract description 108
- 239000008103 glucose Substances 0.000 title claims abstract description 108
- 238000001514 detection method Methods 0.000 title claims abstract description 37
- 239000011159 matrix material Substances 0.000 claims abstract description 95
- 238000000034 method Methods 0.000 claims abstract description 35
- 238000003860 storage Methods 0.000 claims abstract description 15
- 238000012545 processing Methods 0.000 claims abstract description 13
- 238000001914 filtration Methods 0.000 claims abstract description 12
- 238000012360 testing method Methods 0.000 claims description 60
- 238000012549 training Methods 0.000 claims description 23
- 238000012797 qualification Methods 0.000 claims description 4
- 238000013135 deep learning Methods 0.000 abstract description 4
- 238000013186 photoplethysmography Methods 0.000 description 116
- 238000010586 diagram Methods 0.000 description 11
- 238000004590 computer program Methods 0.000 description 10
- 230000006870 function Effects 0.000 description 9
- 230000008569 process Effects 0.000 description 8
- 238000004364 calculation method Methods 0.000 description 4
- 206010012601 diabetes mellitus Diseases 0.000 description 4
- 230000008859 change Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000003908 quality control method Methods 0.000 description 2
- 238000005096 rolling process Methods 0.000 description 2
- 208000017667 Chronic Disease Diseases 0.000 description 1
- 208000013016 Hypoglycemia Diseases 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 230000002641 glycemic effect Effects 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000002218 hypoglycaemic effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012886 linear function Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012806 monitoring device Methods 0.000 description 1
- 210000000056 organ Anatomy 0.000 description 1
- 238000010238 partial least squares regression Methods 0.000 description 1
- 229920001451 polypropylene glycol Polymers 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 210000000707 wrist Anatomy 0.000 description 1
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/145—Measuring characteristics of blood in vivo, e.g. gas concentration, pH value; Measuring characteristics of body fluids or tissues, e.g. interstitial fluid, cerebral tissue
- A61B5/1455—Measuring characteristics of blood in vivo, e.g. gas concentration, pH value; Measuring characteristics of body fluids or tissues, e.g. interstitial fluid, cerebral tissue using optical sensors, e.g. spectral photometrical oximeters
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
- A61B5/024—Detecting, measuring or recording pulse rate or heart rate
- A61B5/02416—Detecting, measuring or recording pulse rate or heart rate using photoplethysmograph signals, e.g. generated by infrared radiation
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/145—Measuring characteristics of blood in vivo, e.g. gas concentration, pH value; Measuring characteristics of body fluids or tissues, e.g. interstitial fluid, cerebral tissue
- A61B5/14532—Measuring characteristics of blood in vivo, e.g. gas concentration, pH value; Measuring characteristics of body fluids or tissues, e.g. interstitial fluid, cerebral tissue for measuring glucose, e.g. by tissue impedance measurement
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
Abstract
The invention relates to the field of signal processing, and discloses a non-invasive blood glucose detection method, a processor, a device and a storage medium based on a graph volume network. The method comprises the following steps: acquiring a PPG signal to be predicted; filtering a PPG signal to be predicted; converting the filtered PPG signal to be predicted into a node map; obtaining a corresponding adjacent matrix and a corresponding characteristic matrix according to the node map; and inputting the adjacency matrix and the characteristic matrix into a graph convolution network, and obtaining the corresponding blood sugar value through the graph convolution network. Through the technical scheme, the invention provides the deep learning method adopting the graph convolution network, so that the model can automatically find out the important characteristic information required for the blood sugar prediction problem, and meanwhile, through iterative updating of node information, the characteristic information is continuously optimized under the condition of keeping all the characteristic information, so that the accuracy of blood sugar prediction is greatly improved.
Description
Technical Field
The invention relates to the field of signal processing, in particular to a non-invasive blood glucose detection method based on a graph volume network, a processor, a device and a storage medium.
Background
As a chronic disease, on average, one person dies every seven seconds worldwide as a result of diabetes. In the early stage of diabetes, no obvious disease symptoms exist, and when complications are obvious, other organs are often damaged, so that complications are caused. In order to monitor the blood glucose concentration in real time and continuously, various non-invasive blood glucose detection techniques have been rapidly developed in recent years, especially by using photoplethysmography (PPG).
However, in the current prior art, more researches are conducted to build a model by using a traditional machine learning method, such as random forest, support vector machine, partial least squares regression, etc., to predict blood glucose concentration by using PPG signals. In the conventional method, characteristics need to be manually given, for example, operations such as characteristic dimension reduction, characteristic screening and the like are performed on the filtered PPG signal. Some useful data information may be screened out in the process, so that the accuracy of the prediction result is reduced.
In order to solve the problems, the deep learning method of the graph convolutional network is adopted in the application, so that the model can automatically find out important characteristic information required for the blood sugar prediction problem, and meanwhile, the characteristic information is continuously optimized under the condition of keeping all the characteristic information through iterative updating of node information, so that the accuracy of blood sugar prediction is greatly improved.
Disclosure of Invention
The invention aims to solve the problem of low accuracy of a blood sugar prediction result in the prior art, and provides a non-invasive blood sugar detection method based on a graph convolution network.
In order to achieve the above object, an aspect of the present invention provides a non-invasive blood glucose detecting method based on a graph volume network, including:
acquiring a PPG signal to be predicted;
filtering a PPG signal to be predicted;
converting the filtered PPG signal to be predicted into a node map;
obtaining a corresponding adjacent matrix and a corresponding characteristic matrix according to the node map;
and inputting the adjacency matrix and the characteristic matrix into a graph convolution network, and obtaining the corresponding blood sugar value through the graph convolution network.
In the embodiment of the present invention, the PPG signal to be predicted is filtered by formula (1):
wherein, y (n) is the PPG signal at the nth moment output after filtering, akAnd bkIn order to be a filter coefficient, the filter coefficient,is the output value at the previous time instant,is an input value of a previous time, andk≠0。
in the embodiment of the present invention, converting the filtered PPG signal to be predicted into the node map includes: acquiring a node of a PPG signal to be predicted; determining the Euclidean distance between any two nodes in the nodes; determining connection between the two nodes under the condition that the Euclidean distance is smaller than a preset threshold value; determining that the two nodes are not connected under the condition that the Euclidean distance is greater than or equal to a preset threshold value; and determining a node graph according to the connection relation between the nodes.
In the embodiment of the present invention, the method further includes a step of training the graph convolution network, including: obtaining sample data, wherein the sample data comprises a plurality of sample PPG signals and a sample blood glucose value corresponding to each sample PPG signal; filtering the sample PPG signal; converting the filtered sample PPG signal into a sample node graph; extracting characteristics of the sample node graph to obtain a corresponding sample adjacency matrix and a sample characteristic matrix; inputting the sample adjacency matrix and the sample characteristic matrix into a convolution network of a graph to be trained to obtain a corresponding predicted blood glucose value; comparing the blood sugar value of the sample with the predicted blood sugar value; and adjusting the parameters of the convolutional network of the graph to be trained according to the comparison result so as to train the convolutional network of the graph to be trained.
In the embodiment of the present invention, adjusting the parameter of the convolutional network of the to-be-trained graph according to the comparison result, so as to train the convolutional network of the to-be-trained graph, includes: obtaining an absolute value of a difference value between a blood glucose value of the sample and a predicted blood glucose value; under the condition that the absolute value is higher than the first preset difference value, adjusting the parameters of the convolutional network of the graph to be trained; determining that the predicted blood glucose value is qualified under the condition that the absolute value is lower than the first preset difference value; and under the condition that the qualification rate of the predicted blood sugar value reaches a first preset proportion, determining that the convolutional network training of the graph to be trained is finished.
In an embodiment of the present invention, the method further comprises: obtaining test data, the test data comprising a plurality of test PPG signals and a test blood glucose value corresponding to each test PPG signal; processing the test PPG signal and inputting the processed test PPG signal to a convolution network of a graph to be trained to obtain a corresponding estimated blood glucose value; determining the absolute value of the difference between the estimated blood glucose value and the test blood glucose value of each test PPG signal; determining that the convolutional network test of the graph to be trained passes under the condition that the ratio of the absolute value of the difference value of the estimated blood glucose value and the test blood glucose value higher than the second preset difference value is larger than the second preset ratio
In an embodiment of the present invention, the method further comprises: and under the condition that the proportion is less than or equal to a second preset proportion, continuing training the convolutional network of the graph to be trained until the proportion is greater than the second preset proportion.
A second aspect of the present invention provides a processor configured to perform the method for non-invasive blood glucose detection based on a atlas network in any of the above embodiments.
The invention provides a non-invasive blood sugar detection device based on a graph volume network, which comprises a pulse detection device, a pulse detection unit and a pulse detection unit, wherein the pulse detection device is used for acquiring PPG signals; the blood glucose detection equipment is used for acquiring an actual blood glucose concentration value corresponding to the PPG signal; and the processor mentioned above.
A fourth aspect of the present invention provides a machine-readable storage medium having stored thereon instructions, which, when executed by a processor, cause the processor to execute the method for graph volume network-based noninvasive blood glucose detection according to any one of the above embodiments.
By the technical scheme, the deep learning method of the graph convolution network is adopted, so that the model can automatically find out important characteristic information required for the blood sugar prediction problem, and meanwhile, the characteristic information is continuously optimized under the condition of keeping all the characteristic information through iterative updating of the node information, so that the accuracy of blood sugar prediction is greatly improved. Meanwhile, the deep learning method of the graph convolution network is adopted, so that the method is suitable for the condition of large data volume, and can avoid the problems of dimension disaster and the like.
Drawings
FIG. 1 is a schematic flow chart illustrating a non-invasive blood glucose detection method based on a graph volume network according to an embodiment of the present invention;
FIG. 2 is a block diagram schematically illustrating the structure of a non-invasive blood glucose detecting apparatus based on a graph volume network according to an embodiment of the present invention;
fig. 3 schematically shows an internal structure diagram of a computer apparatus according to an embodiment of the present invention.
Detailed Description
The following detailed description of embodiments of the invention refers to the accompanying drawings. It should be understood that the detailed description and specific examples, while indicating embodiments of the invention, are given by way of illustration and explanation only, not limitation.
Fig. 1 schematically shows a flow chart of a non-invasive blood glucose detection method based on a graph volume network in an embodiment of the invention. As shown in fig. 1, in an embodiment of the present invention, a non-invasive blood glucose detecting method based on a graph volume network is provided, the method including:
The processor can acquire the PPG signal of presetting the crowd, and the PPG signal is the photoplethysmography that uses photoplethysmography, the photoplethysmography signal that obtains through detecting human blood volume change. The abscissa of the signal is the detection time, and the ordinate is the size of the human blood volume corresponding to the detection time. The predetermined population may be a population representative of foot measures, such as a population with diabetes, a population with normoglycemia, and a population with hypoglycemia. The wrist can be selected as the acquisition part of the PPG signal, and the PPG signal of the preset crowd is acquired through equipment. When the device is used for acquiring the PPG signals of the preset population, the PPG signals of the user at a preset time can be acquired for each person, for example, 6S can be acquired for each person to obtain the PPG signals of each person in 6S time as the signals to be predicted, and the acquired PPG signals to be predicted can be input to the processor.
And 102, filtering the PPG signal to be predicted.
After the processor acquires the PPG signal to be predicted, the PPG signal to be predicted may be filtered. Further, the processor may filter the PPG signal using a digital filter IIR.
In one implementation of the present invention, the PPG signal to be predicted is filtered by equation (1):
wherein, y (n) is the PPG signal at the nth moment output after filtering, akAnd bkIn order to be a filter coefficient, the filter coefficient,is the output value at the previous time instant,is an input value of a previous time, andk≠0。
and 103, converting the filtered PPG signal to be predicted into a node map.
After the PPG signal to be predicted is filtered by the digital filter, the processor can select nodes of the filtered PPG signal and calculate the selected nodes to determine a corresponding node map.
In an embodiment of the present invention, converting the filtered PPG signal to be predicted into a node map includes: acquiring a node of a PPG signal to be predicted; determining the Euclidean distance between any two nodes in the nodes; determining connection between the two nodes under the condition that the Euclidean distance is smaller than a preset threshold value; determining that the two nodes are not connected under the condition that the Euclidean distance is greater than or equal to a preset threshold value; and determining a node graph according to the connection relation between the nodes.
The processor may take one node at a preset time interval for the filtered PPG signal until a preset number of nodes are taken. For example, assuming that the user sets the preset time to 1S, the processor takes one node every 1S interval. Assuming that the user sets the preset number to 6, the processor takes a total of 6 nodes on the PPG signal. I.e. the processor may fetch every 1S one node on the PPG signal, then for a PPG signal of 6S, a total of 6 nodes may be fetched, and the corresponding node map may be further determined from the fetched nodes.
After the processor finishes selecting the nodes, the Euclidean distance between every two nodes can be calculated through an Euclidean distance formula (2):
wherein d is the Euclidean distance between two nodes, (x)1,y1) Is the coordinate of the first of any two nodes, (x)2,y2) The coordinates of the second of any two nodes. The processor may set a preset threshold, and determine that the two nodes are connected when the processor calculates that the euclidean distance between the two nodes is less than the preset threshold. And when the Euclidean distance between the two nodes calculated by the processor is greater than or equal to a preset threshold value, determining that no connection exists between the two nodes. And the processor determines the connection relation between the acquired nodes through calculation to obtain a node map.
And 104, obtaining a corresponding adjacency matrix and a corresponding feature matrix according to the node map.
The processor determines the connection relation between the nodes by calculating the Euclidean distance between the nodes, and then the corresponding node graph can be obtained. The processor may determine an adjacency matrix and a feature matrix corresponding to the node from the obtained node map.
In the node graph, a preset threshold value can be set through a processor, Euclidean distances between every two nodes are respectively calculated, and whether the nodes are connected or not can be determined by judging the size relationship between the Euclidean distances between the nodes and the preset threshold value. And the adjacency matrix is calculated according to whether the nodes are connected or not. Adjacency matrix A ∈ Rn×nElement A in (n is the number of nodes)ijRepresenting the connection relationship between the ith node and the jth node, AijThe values of (a) are as follows:
wherein a is a preset threshold value, d is the Euclidean distance between two nodes, AijDetermining the connection between two nodes when the Euclidean distance d is smaller than a preset threshold value, wherein the abscissa of the adjacent matrix is i, and the ordinate of the adjacent matrix is the corresponding element of j, and the value of the element A corresponding to the adjacent matrix is at the momentijIs 1, when the Euclidean distance d is greater than or equal to the preset threshold value, the two nodes are determined not to be connected, and at the moment, the element value A corresponding to the adjacency matrixijThe value of (d) is 0. Wherein d is represented as follows:
wherein (x)1,y1) Is the coordinate of the first of any two nodes, (x)2,y2) The coordinates of the second of any two nodes.
The feature matrix F ∈ Rn×m(m is the feature number of each node) represents the feature of each node, PPG signals within 0s to 1s are taken out every 10ms to form a 1 x 100 one-dimensional matrix as the feature of the node corresponding to 0s, and so on, the feature matrix corresponding to each node can be obtained, and the 6 feature matrices are spliced into a 6 x 100 feature matrix according to the node sequence, namely F.
And 105, inputting the adjacency matrix and the characteristic matrix into a graph convolution network, and obtaining the corresponding blood sugar value through the graph convolution network.
And the processor filters and converts the PPG signal to be predicted and obtains a corresponding adjacency matrix and a corresponding characteristic matrix. The adjacency matrix and the feature matrix can be input into a graph convolution network after training, analysis is carried out through the graph convolution network, and the blood sugar value output by the graph convolution network is obtained, so that the blood sugar value corresponding to the PPG signal to be predicted can be determined.
Before inputting the adjacency matrix and the feature matrix corresponding to the PPG signal to be predicted into the graph convolution network model to obtain the corresponding blood glucose value, the processor needs to train the constructed graph convolution network model to obtain the trained graph convolution network. Only inputting the adjacency matrix and the feature matrix into the trained graph convolution network model can obtain the accurate blood sugar value corresponding to the PPG signal.
Firstly, a 2-layer graph convolution network is required to be built, an adjacent matrix and a characteristic matrix are obtained from individual PPG signals in a set for training and are used as input of the graph convolution network, and the individual real-time blood glucose concentration measured by a micro-invasive glucometer is used as a label for network learning correspondingly to learn and output the blood glucose concentration of the individual.
In particular, a 2-layer graph convolution network includes two graph convolution layers and two pooling layers, wherein a graph convolution layer can be written as such a non-linear function:
Hl+1=f(Hl,A)
wherein Hl+1The feature matrix obtained for layer l +1, when l is 0, H0F is the input of the first layer, i.e. the feature matrix F obtained directly from the graph structure, a is the adjacency matrix obtained directly from the graph, F is a function, and the difference achieved by the function F is the difference point of different models.
The commonly used function f is as follows:
Hl+1=σ(LsymHlWl)
where σ is the activation function, Lsym=D-1/2AD-1/2For Laplace matrix, use Laplace matrix LsymThe role of replacing the original adjacency matrix a is mainly two-fold. Firstly, introducing a self-degree matrix D of an adjacent matrix A, namely A + I (I is an identity matrix), and solving the self-transmission problem; next, the adjacency matrix a is normalized by multiplying both sides of the adjacency matrix a by the degree-opening of the node and then taking the inverse. WlThe weight of the l-th layer, namely a transfer coefficient of the neighbor node in the l-th layer to the target node.
The loss function is the most common mean square error (also called quadratic loss, L) in the regression problem2Loss), the sum of squared distances between predicted and true values is minimized.
Where MSE is the mean square error, yiIn order to be the true value of the value,is a predicted value, and n is the number of samples.
In one embodiment of the invention, the training of the graph convolution network comprises: obtaining sample data, wherein the sample data comprises a plurality of sample PPG signals and a sample blood glucose value corresponding to each sample PPG signal; filtering the sample PPG signal; converting the filtered sample PPG signal into a sample node graph; extracting characteristics of the sample node graph to obtain a corresponding sample adjacency matrix and a sample characteristic matrix; inputting the sample adjacency matrix and the sample characteristic matrix into a convolution network of a graph to be trained to obtain a corresponding predicted blood glucose value; comparing the blood sugar value of the sample with the predicted blood sugar value; and adjusting the parameters of the convolutional network of the graph to be trained according to the comparison result so as to train the convolutional network of the graph to be trained.
The processor may acquire sample data comprising a plurality of sample PPG signals and a sample blood glucose value corresponding to the sample PPG signals. The sample PPG signal may be a PPG signal of a predetermined population, which may be representative of a population, including, for example, a diabetic population and a glycemic health population. Through carrying out PPG signal acquisition to predetermineeing the crowd, can gather the PPG signal of predetermineeing the crowd everybody preset time through test equipment. After PPG signals of preset crowds are collected, real-time blood glucose concentration of corresponding individuals can be measured through minimally invasive blood glucose detection equipment and used as blood glucose values of samples. For example, the preset time is set to 6S, the PPG signal of each person 6S of the preset population is acquired by the test equipment, and the real-time blood glucose value of the individual is acquired by the minimally invasive blood glucose detection equipment for the person acquiring the PPG signal. Ideally, the subject's blood glucose level is collected with a minimally invasive blood glucose monitoring device while the PPG signal is collected. The PPG signal and the blood glucose value are obtained over the same time period. The collected PPG signal and blood sugar value are used as sample data to be input into the processor.
After receiving the sample PPG signal, the processor filters the acquired PPG signal by a digital filter IIR according to equation (1):
wherein, y (n) is the PPG signal at the nth moment output after filtering, akAnd bkIn order to be a filter coefficient, the filter coefficient,is the output value at the previous time instant,is an input value of a previous time, andk≠0。
the processor may select a node on the filtered sample PPG signal. And taking one node every preset time until a preset number of nodes are taken. Assuming that the preset time is set to 1S and the preset number is set to 6, the processor takes one node every 1S on the filtered sample PPG signal, and takes a total of 6 nodes.
After the processor finishes selecting the nodes, the Euclidean distance between every two nodes can be calculated through an Euclidean distance formula (2):
wherein d is the Euclidean distance between two nodes, (x)1,y1) Is the coordinate of the first node, (x)2,y2) Is the coordinates of the second node. The processor may set a preset threshold, and determine that the two nodes are connected when the processor calculates that the euclidean distance between the two nodes is less than the preset threshold. And when the Euclidean distance between the two nodes calculated by the processor is greater than or equal to a preset threshold value, determining that no connection exists between the two nodes. And the processor determines the connection relation between the acquired nodes through calculation to obtain a sample node graph. Corresponding sample adjacency matrices and sample feature matrices may then be determined from the sample node map.
Wherein for a sample adjacency matrix A ∈ Rn×nElement A in (n is the number of nodes)ijThe connection relationship between the ith node and the jth node is represented, a threshold value a can be set, the Euclidean distance d between every two nodes is calculated respectively, whether the nodes are connected or not is determined by judging the magnitude relationship between d and a, and then
Wherein A isijThe element corresponding to the abscissa of i and the ordinate of j in the adjacent matrix;
wherein (x)1,y1) Is the coordinate of the first node, (x)2,y2) Is the coordinates of the second node.
Sample feature matrix F ∈ Rn×m(m is the feature number of each node) represents the feature of each node, PPG signals within 0s to 1s are taken out every 10ms to form a 1 x 100 one-dimensional matrix as the feature of the node corresponding to 0s, and so on, the feature matrix corresponding to each node can be obtained, and the 6 feature matrices are spliced into a 6 x 100 feature matrix according to the node sequence, namely F; and inputting the sample adjacency matrix and the sample characteristic matrix which are obtained by the calculation of the processor into the convolutional network of the graph to be trained.
And (4) building the graph convolution network, wherein the built graph convolution network is used as the graph convolution network to be trained. And taking the calculated sample adjacency matrix and sample characteristic matrix as the input of the graph convolution network. And after the graph convolution network calculates the input sample adjacency matrix and the sample characteristic matrix, outputting the predicted blood glucose value corresponding to the sample PPG signal.
The processor compares the predicted blood glucose value corresponding to each sample PPG signal calculated by the graph convolution network with the sample blood glucose value corresponding to each sample PPG signal. The processor can adjust the parameters of the convolution network model of the graph to be trained according to the comparison result. And training the convolution network of the graph to be trained until the training is finished.
In an embodiment of the present invention, adjusting parameters of the convolutional network of the graph to be trained according to the comparison result to train the convolutional network of the graph to be trained includes: obtaining an absolute value of a difference value between a blood glucose value of the sample and a predicted blood glucose value; under the condition that the absolute value is higher than the first preset difference value, adjusting the parameters of the convolutional network of the graph to be trained; determining that the predicted blood glucose value is qualified under the condition that the absolute value is lower than the first preset difference value; and under the condition that the qualification rate of the predicted blood sugar value reaches a first preset proportion, determining that the convolutional network training of the graph to be trained is finished.
And the processor inputs a sample adjacency matrix and a sample characteristic matrix which are obtained by processing the obtained sample PPG signal into the built training graph convolution model. The graphical volume model may computationally output a predicted blood glucose value corresponding to the sample PPG signal. The processor may compare a sample blood glucose value corresponding to the sample PPG signal acquired when the sample PPG signal was acquired with a predicted blood glucose value computed and output by the training graph convolutional network. The absolute difference MAE between the two is obtained. The processor may set the first preset difference value to 0.83 according to national standards for blood glucose meters issued by the national quality control Bureau. In case the absolute difference between the predicted blood glucose value and the sample blood glucose value is higher than a first preset difference set by the processor. The processor can adjust parameters of the graph convolution network model to be trained, after adjustment is completed, the output predicted blood glucose value is compared with the sample blood glucose value, and the absolute value difference MAE between the output predicted blood glucose value and the sample blood glucose value is judged to be compared with a first preset difference set by the processor. When the processor judges that the absolute difference value MAE between the two is lower than a first preset difference value. The processor determines that the predicted blood glucose value obtained at this time is acceptable.
For example, assume that the first preset difference set by the processor is 0.83 mmol/L. When the PPG signal No. 1 is collected, the real-time blood sugar value No. 1 corresponding to the signal is 4.1 mmol/L. And the processor obtains an adjacency matrix and a feature matrix corresponding to the PPG signal No. 1 by processing the PPG signal No. 1. And after the matrix is input into the graph rolling model as an input value, the graph rolling model predicts that the blood sugar value is 5.1mmol/L through the No. 1 output by calculation, and the absolute value difference value MAE between the No. 1 predicted blood sugar value and the No. 1 real-time blood sugar value is 1 mmol/L. The absolute difference MAE at this time is higher than the first predetermined difference by 0.83 mmol/L. The processor determines that the predicted blood glucose level No. 1 is not qualified at this time. And adjusting parameters of the graph convolution network model. If the processor collects the PPG signal No. 2, the real-time blood sugar value No. 2 corresponding to the signal is 4.3 mmol/L. The processor obtains an adjacent matrix and a characteristic matrix corresponding to the PPG signal No. 2 by processing the PPG signal No. 2, and inputs the adjacent matrix and the characteristic matrix into the graph convolution model, the predicted blood sugar value No. 2 output by the graph convolution model is 4.9mmol/L, and the absolute value difference value MAE between the predicted blood sugar value No. 2 and the real-time blood sugar value No. 2 is 0.6 mmol/L. The absolute difference MAE at this time is lower than the first predetermined difference of 0.83 mmol/L. At this point, the processor may determine that the predicted blood glucose value is acceptable.
The processor may determine that training of the graph convolution network model is complete when a ratio of the number of qualified predicted blood glucose values output by the graph convolution network to the total number reaches a first preset ratio. For example, assume that the first preset proportion of the processor setting is 99%. And under the condition that the qualification rate of the predicted blood sugar value output by the graph convolution reaches 99%, the processor determines that the training of the graph convolution network model to be trained is finished.
In an embodiment of the invention, test data is acquired, the test data comprising a plurality of test PPG signals and a test blood glucose value corresponding to each test PPG signal; processing the test PPG signal and inputting the processed test PPG signal to a convolution network of a graph to be trained to obtain a corresponding estimated blood glucose value; determining the absolute value difference of the estimated blood glucose value and the test blood glucose value of each test PPG signal; and determining that the convolutional network test of the graph to be trained passes under the condition that the ratio of the difference value between the estimated blood sugar value and the test blood sugar value is lower than a second preset difference value and is larger than a second preset ratio.
In an embodiment of the present invention, in a case that the ratio is smaller than or equal to a second preset ratio, the convolutional network to be trained continues to be trained until the ratio is larger than the second preset ratio.
The processor may acquire test data including a plurality of test PPG signals and a test blood glucose value corresponding to the test PPG signals. Namely, when the PPG signal is acquired, the minimally invasive blood glucose detection equipment is used for acquiring the blood glucose value to be tested in the same time period. And after the processor processes the test PPG signal to obtain a corresponding test adjacency matrix and a test characteristic matrix, the test adjacency matrix and the test characteristic matrix are used as input values and input into a graph convolution network, and the graph convolution network calculates to obtain an estimated blood glucose value corresponding to the test PPG signal.
The processor may compare the estimated blood glucose value corresponding to each test PPG signal with the test blood glucose value corresponding to each test PPG signal. And obtaining the absolute value error of the estimated blood sugar value and the test blood sugar value. The processor may set a second predetermined difference value and a second predetermined ratio. For example, the processor may set a glucose meter national standard of 0.83 issued by the national Bureau of quality control to a second preset difference. The processor may set the second preset ratio to 98%. And when the processor compares the estimated blood sugar value with the test blood sugar value, and the obtained absolute value error between the two parties is smaller than a second preset difference value and is larger than a second preset ratio. And determining that the test of the convolutional network model of the graph to be trained passes, namely completing the training of the convolutional network model of the graph to be trained.
When the processor tests the graph convolution network model through the test data, if the estimated blood sugar value corresponding to the test PPG signal output by the graph convolution model is qualified, the proportion of all the PPG signal data to be tested is smaller than a second preset proportion. The graph convolution network model continues to be trained until the proportion of the quantity of the output of the graph convolution network model reaching the qualified condition is larger than or equal to a second preset proportion.
For example, assume that the processor sets the second preset difference to 0.83 and the second preset proportion to 98%. And inputting a test adjacency matrix and a test characteristic matrix corresponding to the test PPG signal into the graph convolution network model, and comparing the absolute value difference of the output estimated blood glucose value and the test blood glucose value with a second preset difference. And when the absolute value difference is smaller than a second preset difference, the processor determines that the estimated blood sugar value is qualified. And when the absolute value difference is larger than a second preset difference, the processor determines that the estimated blood sugar value is unqualified. When all the tested PPGs were calculated and aligned. The percentage of qualified data to the total data is less than or equal to 98%. At this time, the processor judges the graph convolution network model as passing the test. Training and adjusting the graph convolution network model are required to be carried out continuously. If the data is qualified, the processor judges that the proportion of the qualified data to all the data is more than 98 percent. The processor determines that the training of the graph convolution network model is complete.
At this point, the graph convolution network model training is completed. The processor can process the PPG signal to be predicted to obtain an adjacency matrix to be predicted and a characteristic matrix to be predicted, and the adjacency matrix to be predicted and the characteristic matrix to be predicted are input into the graph convolution network model which is trained. And outputting the blood sugar value corresponding to the PPG signal to be predicted by the trained graph convolution network model.
Fig. 2 is a schematic diagram showing a structure of a non-invasive blood glucose detecting apparatus based on a graph volume network according to an embodiment of the present invention. As shown in fig. 2, in the embodiment of the present invention, a non-invasive blood glucose detecting apparatus 200 based on a graph volume network is provided, which includes a pulse detecting device 201, a minimally invasive blood glucose detecting device 202 and a processor 203. Wherein:
a pulse detection device 201 for acquiring a PPG signal.
A minimally invasive blood glucose detection device 202 for obtaining an actual blood glucose concentration value corresponding to the PPG signal.
The processor 203 may process the PPG signals acquired by the pulse detection device 201, and the processing may include filtering the PPG signals, selecting nodes and converting the nodes into a node map, and determining an adjacency matrix and a feature matrix corresponding to the PPG signals according to the node map. And inputting the matrix as an input value to the graph convolution network model. To obtain a predicted blood glucose level.
The processor 203 may also train the established atlas network model by obtaining a set of training PPG signals by the pulse detection device 201. Until the graph convolution network model training is obtained. The processor 203 may further obtain a test PPG signal set through the pulse detection device 201 to test the trained convolution network model, and determine whether the training of the convolution network model is completed.
The processor 203 includes a kernel, which calls the corresponding program unit from the memory. The kernel can be set to be one or more than one, and the noninvasive blood glucose detection method based on the graph volume network is realized by adjusting the kernel parameters.
In one embodiment, a computer device is provided, which may be a server, the internal structure of which may be as shown in fig. 3. The computer device includes a processor a01, a network interface a02, a memory (not shown), and a database (not shown) connected by a system bus. Wherein processor a01 of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises an internal memory a03 and a non-volatile storage medium a 04. The non-volatile storage medium a04 stores an operating system B01, a computer program B02, and a database (not shown in the figure). The internal memory a03 provides an environment for the operation of the operating system B01 and the computer program B02 in the nonvolatile storage medium a 04. The database of the computer device may be used to store the acquired PPG signal data and the acquired real-time blood glucose data corresponding to the PPG signal. The network interface a02 of the computer device is used for communication with an external terminal through a network connection. The computer program B02 is executed by the processor a01 to implement a graph volume network based non-invasive blood glucose detection method.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
In a typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.
The memory may include forms of volatile memory in a computer readable medium, Random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). The memory is an example of a computer-readable medium.
Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, a computer readable medium does not include a transitory computer readable medium such as a modulated data signal and a carrier wave.
It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in the process, method, article, or apparatus that comprises the element.
The above are merely examples of the present application and are not intended to limit the present application. Various modifications and changes may occur to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.
Claims (10)
1. A non-invasive blood glucose detection method based on a graph convolution network is characterized by comprising the following steps:
acquiring a PPG signal to be predicted;
filtering the PPG signal to be predicted;
converting the filtered PPG signal to be predicted into a node map;
obtaining a corresponding adjacency matrix and a corresponding feature matrix according to the node map;
and inputting the adjacency matrix and the feature matrix into a graph convolution network, and obtaining the corresponding blood glucose value through the graph convolution network.
2. The method of non-invasive blood glucose detection based on graph volume network according to claim 1, wherein the PPG signal to be predicted is filtered by formula (1):
3. the non-invasive blood glucose detection method based on graph volume network according to claim 1, wherein the converting the filtered PPG signal to be predicted into the node graph comprises:
acquiring a node of the PPG signal to be predicted;
determining the Euclidean distance between any two nodes in the nodes;
determining the connection between the two nodes under the condition that the Euclidean distance is smaller than a preset threshold value;
determining that the two nodes are not connected under the condition that the Euclidean distance is greater than or equal to the preset threshold value;
and determining a node graph according to the connection relation between the nodes.
4. The method of non-invasive blood glucose detection based on a graph volume network of claim 1, further comprising a training step of the graph volume network, comprising:
obtaining sample data, the sample data comprising a plurality of sample PPG signals and a sample blood glucose value corresponding to each sample PPG signal;
filtering the sample PPG signal;
converting the filtered sample PPG signal into a sample node graph;
extracting characteristics of the sample node graph to obtain a corresponding sample adjacency matrix and a sample characteristic matrix;
inputting the sample adjacency matrix and the sample characteristic matrix into a convolution network of a graph to be trained to obtain a corresponding predicted blood glucose value;
comparing the sample blood glucose value to the predicted blood glucose value;
and adjusting the parameters of the convolutional network of the graph to be trained according to the comparison result so as to train the convolutional network of the graph to be trained.
5. The method of claim 4, wherein the adjusting parameters of the convolutional network of the graph to be trained according to the comparison result to train the convolutional network of the graph to be trained comprises:
obtaining an absolute value of a difference between the sample blood glucose value and the predicted blood glucose value;
under the condition that the absolute value is higher than a first preset difference value, adjusting parameters of the convolutional network of the graph to be trained;
determining that the predicted blood glucose value is qualified if the absolute value is lower than the first preset difference value;
and under the condition that the qualification rate of the predicted blood sugar value reaches a first preset proportion, determining that the convolutional network training of the graph to be trained is finished.
6. The method of non-invasive blood glucose detection based on graph volume network of claim 4, wherein the method comprises:
obtaining test data comprising a plurality of test PPG signals and a test blood glucose value corresponding to each of the test PPG signals;
processing the test PPG signal and inputting the processed test PPG signal to the convolution network of the graph to be trained to obtain a corresponding estimated blood glucose value;
determining the absolute value of the difference between the estimated blood glucose value and the test blood glucose value of each test PPG signal;
and determining that the convolution network test of the graph to be trained passes under the condition that the ratio of the absolute value of the difference value between the estimated blood glucose value and the test blood glucose value lower than a second preset difference value is larger than a second preset ratio.
7. The method of non-invasive glucose detection based on graph and volume network according to claim 6, further comprising:
and under the condition that the proportion is less than or equal to the second preset proportion, continuing training the convolutional network of the graph to be trained until the proportion is greater than the second preset proportion.
8. A processor, characterized in that the processor is configured to execute the graph volume network based non-invasive blood glucose detecting method according to any one of claims 1 to 7.
9. A non-invasive blood glucose detecting device based on a graph volume network is characterized by comprising:
a pulse detection device for acquiring a PPG signal;
the minimally invasive blood glucose detection equipment is used for acquiring an actual blood glucose concentration value corresponding to the PPG signal; and
the processor of claim 8.
10. A machine-readable storage medium having instructions stored thereon, wherein the instructions, when executed by a processor, cause the processor to perform the graph volume network-based noninvasive blood glucose detection method according to any one of claims 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110489689.9A CN113288131B (en) | 2021-05-06 | 2021-05-06 | Non-invasive blood glucose detection method, processor and device based on graph convolution network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110489689.9A CN113288131B (en) | 2021-05-06 | 2021-05-06 | Non-invasive blood glucose detection method, processor and device based on graph convolution network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113288131A true CN113288131A (en) | 2021-08-24 |
CN113288131B CN113288131B (en) | 2022-07-12 |
Family
ID=77320857
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110489689.9A Expired - Fee Related CN113288131B (en) | 2021-05-06 | 2021-05-06 | Non-invasive blood glucose detection method, processor and device based on graph convolution network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113288131B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114420301A (en) * | 2022-01-28 | 2022-04-29 | 广东工业大学 | Method, system and storage medium for predicting blood glucose based on segmented domain RF modeling |
CN114403866A (en) * | 2022-01-28 | 2022-04-29 | 广东工业大学 | Noninvasive blood glucose prediction device based on near-infrared light wavelength conversion |
CN116942151A (en) * | 2023-07-25 | 2023-10-27 | 迈德医疗科技(深圳)有限公司 | Blood glucose detection method based on wavelet scattering fusion Markov angle field |
CN116942151B (en) * | 2023-07-25 | 2024-05-14 | 迈德医疗科技(深圳)有限公司 | Blood glucose detection method based on wavelet scattering fusion Markov angle field |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106446777A (en) * | 2016-08-29 | 2017-02-22 | 广东工业大学 | Noninvasive blood sugar data processing method and noninvasive blood sugar data processing system based on convolutional neural network |
CN107361776A (en) * | 2017-05-20 | 2017-11-21 | 深圳市前海安测信息技术有限公司 | Noninvasive system for detecting blood sugar and method |
CN108020565A (en) * | 2017-11-24 | 2018-05-11 | 天津大学 | Blood sugar concentration detection method based on neural network algorithm |
CN108542402A (en) * | 2018-05-17 | 2018-09-18 | 吉林求是光谱数据科技有限公司 | Blood sugar detecting method based on Self-organizing Competitive Neutral Net model and infrared spectrum |
CN109743286A (en) * | 2018-11-29 | 2019-05-10 | 武汉极意网络科技有限公司 | A kind of IP type mark method and apparatus based on figure convolutional neural networks |
CN110111325A (en) * | 2019-05-14 | 2019-08-09 | 深圳大学 | Neuroimaging classification method, terminal and computer readable storage medium |
CN110522448A (en) * | 2019-07-12 | 2019-12-03 | 东南大学 | A kind of brain network class method based on figure convolutional neural networks |
CN110934587A (en) * | 2019-12-13 | 2020-03-31 | 深圳龙岗智能视听研究院 | Alzheimer disease auxiliary diagnosis method based on atlas neural network |
-
2021
- 2021-05-06 CN CN202110489689.9A patent/CN113288131B/en not_active Expired - Fee Related
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106446777A (en) * | 2016-08-29 | 2017-02-22 | 广东工业大学 | Noninvasive blood sugar data processing method and noninvasive blood sugar data processing system based on convolutional neural network |
CN107361776A (en) * | 2017-05-20 | 2017-11-21 | 深圳市前海安测信息技术有限公司 | Noninvasive system for detecting blood sugar and method |
CN108020565A (en) * | 2017-11-24 | 2018-05-11 | 天津大学 | Blood sugar concentration detection method based on neural network algorithm |
CN108542402A (en) * | 2018-05-17 | 2018-09-18 | 吉林求是光谱数据科技有限公司 | Blood sugar detecting method based on Self-organizing Competitive Neutral Net model and infrared spectrum |
CN109743286A (en) * | 2018-11-29 | 2019-05-10 | 武汉极意网络科技有限公司 | A kind of IP type mark method and apparatus based on figure convolutional neural networks |
CN110111325A (en) * | 2019-05-14 | 2019-08-09 | 深圳大学 | Neuroimaging classification method, terminal and computer readable storage medium |
CN110522448A (en) * | 2019-07-12 | 2019-12-03 | 东南大学 | A kind of brain network class method based on figure convolutional neural networks |
CN110934587A (en) * | 2019-12-13 | 2020-03-31 | 深圳龙岗智能视听研究院 | Alzheimer disease auxiliary diagnosis method based on atlas neural network |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114420301A (en) * | 2022-01-28 | 2022-04-29 | 广东工业大学 | Method, system and storage medium for predicting blood glucose based on segmented domain RF modeling |
CN114403866A (en) * | 2022-01-28 | 2022-04-29 | 广东工业大学 | Noninvasive blood glucose prediction device based on near-infrared light wavelength conversion |
CN114420301B (en) * | 2022-01-28 | 2022-08-05 | 广东工业大学 | Method, system and storage medium for predicting blood glucose based on segmented domain RF modeling |
CN116942151A (en) * | 2023-07-25 | 2023-10-27 | 迈德医疗科技(深圳)有限公司 | Blood glucose detection method based on wavelet scattering fusion Markov angle field |
CN116942151B (en) * | 2023-07-25 | 2024-05-14 | 迈德医疗科技(深圳)有限公司 | Blood glucose detection method based on wavelet scattering fusion Markov angle field |
Also Published As
Publication number | Publication date |
---|---|
CN113288131B (en) | 2022-07-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113288131B (en) | Non-invasive blood glucose detection method, processor and device based on graph convolution network | |
CN110071913B (en) | Unsupervised learning-based time series anomaly detection method | |
CN104970789B (en) | Electrocardiogram sorting technique and system | |
CN108229667A (en) | Trimming based on artificial neural network classification | |
CN110686633B (en) | Landslide displacement prediction method and device and electronic equipment | |
CN112587153B (en) | End-to-end non-contact atrial fibrillation automatic detection system and method based on vPPG signal | |
CN109192305B (en) | Heart function automatic analysis method based on deep circulation neural network | |
CN106355195B (en) | System and method for measuring image definition value | |
WO2017041651A1 (en) | User data classification method and device | |
EP3617953A1 (en) | An adaptable neural network | |
CN108681697B (en) | Feature selection method and device | |
CN110738235A (en) | Pulmonary tuberculosis determination method, pulmonary tuberculosis determination device, computer device, and storage medium | |
CN112132801A (en) | Lung bullae focus detection method and system based on deep learning | |
CN111179235A (en) | Image detection model generation method and device, and application method and device | |
CN114330647A (en) | Model training method and device and silicon rod weight prediction method | |
CN109065176B (en) | Blood glucose prediction method, device, terminal and storage medium | |
CN111370120A (en) | Method for detecting diastolic dysfunction based on heart sound signals | |
US10437944B2 (en) | System and method of modeling irregularly sampled temporal data using Kalman filters | |
CN113313682A (en) | No-reference video quality evaluation method based on space-time multi-scale analysis | |
CN116327133A (en) | Multi-physiological index detection method, device and related equipment | |
CN112164468B (en) | Method for processing missing data of pregnancy examination data | |
CN111785379B (en) | Brain function connection prediction method, device, computer equipment and storage medium | |
CN113435321A (en) | Method, system and equipment for evaluating state of main shaft bearing and readable storage medium | |
CN113096815A (en) | Chronic nephropathy prediction method based on logistic regression | |
CN110786839A (en) | Method, device, equipment and medium for generating instantaneous waveform-free ratio |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20220712 |