CN116559143A - Method and system for analyzing composite Raman spectrum data of glucose component in blood - Google Patents
Method and system for analyzing composite Raman spectrum data of glucose component in blood Download PDFInfo
- Publication number
- CN116559143A CN116559143A CN202310543084.2A CN202310543084A CN116559143A CN 116559143 A CN116559143 A CN 116559143A CN 202310543084 A CN202310543084 A CN 202310543084A CN 116559143 A CN116559143 A CN 116559143A
- Authority
- CN
- China
- Prior art keywords
- raman spectrum
- glucose
- blood
- data
- blood glucose
- 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.)
- Pending
Links
- 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 188
- 239000008103 glucose Substances 0.000 title claims abstract description 188
- 238000001237 Raman spectrum Methods 0.000 title claims abstract description 187
- 210000004369 blood Anatomy 0.000 title claims abstract description 185
- 239000008280 blood Substances 0.000 title claims abstract description 185
- 238000000034 method Methods 0.000 title claims abstract description 72
- 239000002131 composite material Substances 0.000 title claims abstract description 23
- 238000001228 spectrum Methods 0.000 claims abstract description 33
- 238000004458 analytical method Methods 0.000 claims abstract description 17
- 238000005259 measurement Methods 0.000 claims abstract description 17
- 238000007781 pre-processing Methods 0.000 claims abstract description 13
- 238000000338 in vitro Methods 0.000 claims abstract description 12
- 238000001727 in vivo Methods 0.000 claims abstract description 12
- 238000013527 convolutional neural network Methods 0.000 claims description 22
- 238000012549 training Methods 0.000 claims description 22
- 238000004364 calculation method Methods 0.000 claims description 17
- 239000013598 vector Substances 0.000 claims description 17
- 238000013528 artificial neural network Methods 0.000 claims description 16
- 210000002569 neuron Anatomy 0.000 claims description 13
- 238000010606 normalization Methods 0.000 claims description 13
- 238000011176 pooling Methods 0.000 claims description 10
- 230000000694 effects Effects 0.000 claims description 8
- 238000012706 support-vector machine Methods 0.000 claims description 8
- 230000004913 activation Effects 0.000 claims description 7
- 238000009499 grossing Methods 0.000 claims description 7
- 238000007619 statistical method Methods 0.000 claims description 7
- 238000000605 extraction Methods 0.000 claims description 6
- JJWKPURADFRFRB-UHFFFAOYSA-N carbonyl sulfide Chemical compound O=C=S JJWKPURADFRFRB-UHFFFAOYSA-N 0.000 claims description 5
- 239000000284 extract Substances 0.000 claims description 5
- 238000005457 optimization Methods 0.000 claims description 5
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 238000010008 shearing Methods 0.000 claims description 4
- 238000012935 Averaging Methods 0.000 claims description 3
- 238000011002 quantification Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 238000001514 detection method Methods 0.000 abstract description 28
- 238000013136 deep learning model Methods 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 19
- 230000003595 spectral effect Effects 0.000 description 16
- 230000008569 process Effects 0.000 description 12
- 239000000523 sample Substances 0.000 description 12
- 238000010586 diagram Methods 0.000 description 9
- 238000007405 data analysis Methods 0.000 description 8
- 238000011156 evaluation Methods 0.000 description 8
- 238000004422 calculation algorithm Methods 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 6
- 238000012545 processing Methods 0.000 description 6
- 230000008859 change Effects 0.000 description 5
- 150000001875 compounds Chemical class 0.000 description 5
- 238000012360 testing method Methods 0.000 description 4
- 238000001069 Raman spectroscopy Methods 0.000 description 3
- 230000003044 adaptive effect Effects 0.000 description 3
- 238000012937 correction Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000004445 quantitative analysis Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000003068 static effect Effects 0.000 description 3
- 238000012795 verification Methods 0.000 description 3
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 2
- 238000010241 blood sampling Methods 0.000 description 2
- 206010012601 diabetes mellitus Diseases 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000007935 neutral effect Effects 0.000 description 2
- 230000000750 progressive effect Effects 0.000 description 2
- 230000001902 propagating effect Effects 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 230000002441 reversible effect Effects 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 208000017667 Chronic Disease Diseases 0.000 description 1
- 102000001554 Hemoglobins Human genes 0.000 description 1
- 108010054147 Hemoglobins Proteins 0.000 description 1
- 241000282412 Homo Species 0.000 description 1
- 241000102542 Kara Species 0.000 description 1
- 241001465754 Metazoa Species 0.000 description 1
- SVRXCFMQPQNIGW-UHFFFAOYSA-N [2-(dimethylcarbamothioylsulfanylamino)ethylamino] n,n-dimethylcarbamodithioate Chemical compound CN(C)C(=S)SNCCNSC(=S)N(C)C SVRXCFMQPQNIGW-UHFFFAOYSA-N 0.000 description 1
- 238000000862 absorption spectrum Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000012472 biological sample Substances 0.000 description 1
- 210000001124 body fluid Anatomy 0.000 description 1
- 239000010839 body fluid Substances 0.000 description 1
- 230000008602 contraction Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 208000018914 glucose metabolism disease Diseases 0.000 description 1
- 208000015181 infectious disease Diseases 0.000 description 1
- 150000002632 lipids Chemical class 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000004060 metabolic process Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012014 optical coherence tomography Methods 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 238000001834 photoacoustic spectrum Methods 0.000 description 1
- 230000010287 polarization Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 102000004169 proteins and genes Human genes 0.000 description 1
- 108090000623 proteins and genes Proteins 0.000 description 1
- 238000000611 regression analysis Methods 0.000 description 1
- 210000003296 saliva Anatomy 0.000 description 1
- 238000011895 specific detection Methods 0.000 description 1
- 238000010972 statistical evaluation Methods 0.000 description 1
- 208000011580 syndromic disease Diseases 0.000 description 1
- 210000003462 vein Anatomy 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/62—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light
- G01N21/63—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light optically excited
- G01N21/65—Raman scattering
-
- 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
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- 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
- G06N3/048—Activation functions
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Molecular Biology (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Chemical & Material Sciences (AREA)
- Biochemistry (AREA)
- Analytical Chemistry (AREA)
- Pathology (AREA)
- Immunology (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Investigating, Analyzing Materials By Fluorescence Or Luminescence (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention discloses a method and a system for analyzing composite Raman spectrum data of glucose in blood, which comprises the following steps: preprocessing blood glucose Raman spectrum data obtained by in-vivo or in-vitro measurement; quantitatively extracting glucose Raman spectrum characteristic information in blood from the pretreated blood glucose Raman spectrum data; and predicting the glucose content in blood according to the obtained glucose Raman spectrum characteristic information. According to the characteristic difference between the spectrum signal and the background noise, the physical parameters of the glucose characteristic peak are accurately extracted; and establishing a mixed deep learning model, effectively identifying common characteristics of different individual spectrum detection results, and accurately predicting the real content of blood sugar in the body. This is expected to form a more accurate and universal analysis method system for the Raman spectrum data of blood sugar.
Description
Technical Field
The invention belongs to the technical field of Raman spectrum data analysis and glucose (blood glucose) Raman spectrum detection in blood, and particularly relates to a composite Raman spectrum data analysis method and system for glucose components in blood.
Background
Diabetes is a syndrome characterized by a disorder of glucose metabolism. Diabetes mellitus is a chronic disease progressive for life, and incurable. The blood sugar monitoring result is helpful for grasping the sugar metabolism disturbance degree of diabetics, knowing the blood sugar level and fluctuation rule in time, and is used for making a reasonable blood sugar reduction scheme, guiding and adjusting the treatment scheme, reducing the occurrence risk of complications and improving the life quality and treatment compliance of the diabetics. Traditional blood sampling and blood sugar detection methods by veins or fingertips are invasive operations, and can cause pain, infection, psychological burden and other negative factors, so that the life quality of patients and the enthusiasm of self-monitoring are seriously influenced.
The optical blood sugar detection technology has become a research hotspot of technology and industry at home and abroad because of the characteristics of convenient detection, rich information, capability of truly realizing noninvasive and painless and the like. Compared with optical technologies such as infrared polarization measurement, optical coherence tomography, near/middle infrared absorption spectrum, photoacoustic spectrum and the like, the Raman spectrum analysis technology can study the molecular structure and biochemical composition information of substances at a molecular level, has unique advantages in the noninvasive blood glucose detection technology and the application field thereof, namely has unique molecular fingerprint identification capability, and is particularly suitable for specific detection and quantitative analysis of biochemical components with larger water content. The existing research results show that the blood sugar content of animals and human bodies can be quantitatively identified from Raman spectrum characteristic information obtained in the living body noninvasive detection or in-vitro collection of blood samples.
However, quantitative analysis of the raman spectra of blood glucose obtained by detection is faced with a number of problems. Firstly, the availability of blood glucose Raman spectrum data can be greatly changed due to instrument state, equipment replacement and environmental noise, which brings a series of uncertainty factors for quantitatively analyzing the glucose content in blood; secondly, the raman spectrum information of blood sugar measured by living body/in vitro can be overlapped with other components (such as hemoglobin, lipid, protein and the like) in human tissues or blood, so that the difficulty is brought to judging the characteristic information of the raman spectrum of the blood sugar and further quantitatively analyzing the blood sugar content; finally, differences in individual biological tissue structures, biochemical structures thereof, and the like can lead to nonlinear differences in blood glucose concentrations and spectral detection results of different individuals and different measurement sites. In addition, the same individual measurement conditions change, which also causes a spectral change that is much greater than that caused by a change in blood glucose concentration.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a composite Raman spectrum data analysis method and system for glucose components in blood, which are used for preprocessing a blood Raman spectrum obtained by living body/in-vitro measurement, quantitatively extracting glucose Raman spectrum characteristic information in the blood from the preprocessed blood Raman spectrum, predicting the glucose content in the blood according to the obtained glucose Raman spectrum characteristic information, and quantitatively evaluating the accuracy of a prediction result by using a statistical analysis and Clark grid analysis method.
In order to achieve the above object, the present invention provides the following solutions:
a method for analyzing composite Raman spectrum data of glucose in blood comprises the following steps:
preprocessing blood glucose Raman spectrum data obtained by in-vivo or in-vitro measurement;
quantitatively extracting glucose Raman spectrum characteristic information in blood from the pretreated blood glucose Raman spectrum data;
and predicting the glucose content in blood according to the obtained glucose Raman spectrum characteristic information.
Preferably, the method for preprocessing blood glucose raman spectrum data obtained by in vivo or ex vivo measurement comprises:
selectively carrying out Raman spectrum wavelength interpolation on the blood glucose Raman spectrum data;
shearing the spectrum range of the blood glucose Raman spectrum data;
fluorescent background, cosmic rays, and other noise interference due to instrumentation and environment of the blood glucose raman spectrum data are eliminated.
Preferably, the method for quantitatively extracting the glucose raman spectrum characteristic information in the blood from the pretreated blood glucose raman spectrum data comprises the following steps:
and smoothing, normalizing and averaging the pretreated blood glucose Raman spectrum data, and identifying and extracting blood glucose Raman spectrum characteristic information.
Preferably, the method for predicting the glucose content in blood according to the obtained glucose Raman spectrum characteristic information comprises the following steps:
cutting the blood glucose Raman spectrum characteristic information into data sets for different targets, and selectively performing data expansion on the data sets;
constructing a blood glucose Raman spectrum model to identify Raman spectrum characteristic data;
training the blood glucose raman spectrum model by using the data set;
and extracting blood glucose Raman spectrum commonality quantification information by using the trained blood glucose Raman spectrum model, and predicting the in-vivo glucose content.
Preferably, the blood glucose Raman spectrum model is constructed by adopting a convolutional neural network, a feedback artificial neural network or a support vector regression machine.
Preferably, the method for constructing the blood glucose Raman spectrum model by adopting the convolutional neural network comprises the following steps:
performing tensor conversion on the blood glucose Raman spectrum data, converting two-dimensional blood glucose Raman spectrum data information into three-dimensional data identified by a convolutional neural network, and performing convolution after normalization;
the convolution layer extracts the blood glucose Raman spectrum characteristics from the normalized data, and creates new characteristic data input for the next layer;
the output of the convolution layer realizes nonlinear transformation through an activation function and transmits data to a pooling layer;
the positive correlation in the data is reserved by adopting a maximum pooling method, then the information of each feature map is connected in a full connection layer, and a regression value is obtained in a regression layer;
after the output of the last layer is obtained, the obtained error value is used as a criterion, the information of the error is fed back to the previous layer, the threshold value and the deviation value of the previous layer are changed until the output of the last layer is gradually adjusted to the previous layer, and the neuron at each position is subjected to optimization parameter adjustment;
and forward computing again by the network after forward error adjustment, and repeating the previous steps to obtain an output error effect until the obtained error meets the requirement.
Preferably, the method for constructing the blood glucose Raman spectrum model by adopting the feedback artificial neural network comprises the following steps:
after the input layer inputs a data sample, performing feature capturing and voting after calculating the upper threshold value and the deviation of the neuron of the hidden layer, and gradually calculating from front to back to obtain each neuron parameter of the hidden layer and the predicted value of the output layer;
and calculating the error between the predicted value and the true value, and judging whether iteration calculation needs to be continued or not according to the set error allowable range.
Preferably, the method for constructing the blood glucose Raman spectrum model by adopting a support vector regression machine comprises the following steps:
based on the principle of a reference support vector machine part, regression model analysis is added, and the high-dimensional information such as Raman spectrum is used as an independent variable to be trained of the support vector machine.
Preferably, the accuracy of the predicted glucose content in blood is assessed by measuring glucose results using a combination of statistical analysis and Clark grid analysis methods, in conjunction with a blood glucose meter.
The invention also provides a composite Raman spectrum data analysis system of glucose component in blood, comprising: the device comprises a preprocessing module, a feature extraction module and a prediction module;
the pretreatment module is used for carrying out pretreatment on blood glucose Raman spectrum data obtained by in-vivo or in-vitro measurement;
the characteristic extraction module is used for quantitatively extracting glucose Raman spectrum characteristic information in blood from the preprocessed blood glucose Raman spectrum data;
the prediction module is used for predicting the glucose content in blood according to the obtained glucose Raman spectrum characteristic information.
Compared with the prior art, the invention has the beneficial effects that:
the invention provides a composite Raman spectrum data analysis method of glucose component (blood sugar) in blood, which adopts the characteristic difference between spectrum signals and background noise to accurately extract physical parameters of glucose characteristic peaks, and identifies and extracts the Raman spectrum characteristic information of glucose in living body and in-vitro blood; multiple regression analysis methods were fused: the method comprises the steps of building a composite deep learning model by a convolutional neural network (Convolutional Neural Network, CNN for short), a reverse artificial neural network (Back Propagating Artificial Neutral Network, BP-ANN for short) and a support vector regression machine (Support Vector Regression Machine, SVR for short), effectively identifying common characteristics of spectrum detection results of different individuals, and accurately predicting the real content of glucose in blood. And a blood glucose Raman spectrum analysis accuracy assessment method based on a statistical analysis and a Kara grid analysis method is established, so that blood glucose concentration information of living organisms is effectively and quantitatively analyzed, but the method is not limited to Raman spectrum characteristics of a certain organism and characteristics of a certain compound. The learned model can be flexibly used, so that the quantitative detection of the compound with different forms of original spectrum is performed, and the method is not limited to a specific spectrum data form. The method is expected to form a more accurate and universal noninvasive blood glucose Raman spectrum data analysis method system, and the application potential of the Raman spectrum detection technology in the noninvasive blood glucose detection field is further explored.
Drawings
In order to more clearly illustrate the technical solutions of the present invention, the drawings that are needed in the embodiments are briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a method for analyzing composite Raman spectrum data of glucose in blood according to an embodiment of the invention;
FIG. 2 is a schematic illustration of normalized spectra after pretreatment in an embodiment of the present invention;
FIG. 3 is a schematic diagram of a one-dimensional convolutional neural network model in an embodiment of the present invention;
FIG. 4 is a schematic diagram of a neural network calculation process in an embodiment of the invention;
FIG. 5 is a schematic diagram of a feedback artificial neural network model in an embodiment of the invention;
FIG. 6 is a schematic diagram of a support vector regression process according to an embodiment of the present invention;
FIG. 7 is a plot of error conditions between true and predicted values in an embodiment of the present invention;
FIG. 8 is a schematic view of a Clarke medical grid in an embodiment of the invention;
fig. 9 is a schematic diagram of evaluation parameters of a real value and a predicted value in an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to the appended drawings and appended detailed description.
The blood glucose Raman spectrum data which is analyzed in a targeted manner is derived from living body and in-vitro blood glucose Raman spectrums measured by various Raman spectrum detection devices. The Raman spectrum detection device comprises a large scientific research-level Raman spectrum detection instrument, a small portable Raman spectrum detection instrument and a Raman spectrum detection device which is independently researched and developed and is based on different optical structures. The living body blood glucose Raman spectrum refers to a blood glucose Raman spectrum obtained by percutaneous noninvasive detection by using the Raman spectrum detection device and blood glucose Raman spectra detected by other living body noninvasive detection methods. The isolated blood glucose Raman spectrum refers to blood glucose Raman spectrum obtained by measuring saliva, body fluid, blood samples, tissue slice samples and other various isolated biological samples by using the Raman spectrum detection device.
The reference blood sugar content related by the invention is derived from blood sugar content information obtained by measurement of various blood sugar detection instruments, such as various fingertip blood sampling blood sugar detection devices and blood sugar content obtained by clinical measurement.
Example 1
As shown in fig. 1, the present invention provides a method for analyzing composite raman spectrum data of a glucose component in blood, comprising the steps of:
preprocessing blood glucose Raman spectrum data obtained by in-vivo or in-vitro measurement;
quantitatively extracting glucose Raman spectrum characteristic information in blood from the pretreated blood glucose Raman spectrum data;
and predicting the glucose content in blood according to the obtained glucose Raman spectrum characteristic information.
In this embodiment, in the preprocessing section, the obtained blood glucose raman spectrum data is mainly subjected to processing operations of wavelength interpolation, preprocessing, data slicing, and data expansion, so as to obtain a high-quality and accurate raman spectrum. Firstly, selectively carrying out Raman spectrum wavelength interpolation, and eliminating prediction errors caused by the difference of spectrum data in a follow-up prediction data set and spectrum data in a training data set. Then eliminating fluorescent background, cosmic rays and other noise interference caused by instruments and environment of the original blood glucose Raman spectrum obtained by measurement; secondly, smoothing, normalizing and mean value centering are carried out on the spectrum data, and the characteristic information of the blood glucose Raman spectrum is identified and extracted; quantitatively comparing the pretreated Raman spectrum, and analyzing the characteristic information difference of the Raman spectrum of the blood sugar caused by the change of the blood sugar content; finally, the processed spectral data is cut into data sets for different targets for model training and model use. The specific flow is as follows:
wavelength interpolation: and carrying out interpolation processing on the spectrum to obtain spectrum characteristic information with more uniform wavelength interval.
Shear spectral range: and selecting and shearing a proper range according to the Raman signal of the measured sample so as to obtain a spectrum with a more accurate range, and ensuring high quality and accuracy of the spectrum during subsequent spectrum processing.
Removing cosmic rays: the present patent provides two methods, including: the peak detection method is used for detecting and eliminating peaks which are obviously narrower than the spectrum; the abnormal peak detection method compares a series of repeated spectrums, and detects and eliminates the abnormal peak.
Baseline calibration: the present patent provides two methods, including: a polynomial fitting method (Polynomial Fitting) for adjusting the baseline position by setting polynomial order values and thresholds to eliminate the effect of the fluorescent background on the raman peak; an extended multiplication signal correction method (Extended Multiplicative Signal Correction, EMSC) regresses each input spectrum according to a reference (e.g., average spectrum) and the result is used to correct the input spectrum.
Spectral data smoothing: this patent uses SG (Savitzky-Golay) smoothing method to fit a continuous subset of adjacent data points to a low order polynomial by linear least squares. Noise is filtered out while the spectral shape and width are kept unchanged for baseline correction and smoothing/noise reduction.
The normalization process is as shown in fig. 2: the normalization processing method adopted in the pretreatment part comprises three types of mean normalization, spectral peak area normalization and spectral peak intensity normalization. In order to eliminate the influence of power disturbance and sample non-uniformity, the spectral intensity accurate prediction mean normalization can be selected; to discuss quantitative information of a substance, spectral peak area normalization may be selected; to further highlight certain material content variations in order to eliminate effects due to sample and instrument variations, spectral peak intensity normalization may be chosen.
Quantitative analysis: quantitatively analyzing spectral parameter differences caused by blood glucose content and changes thereof, comprising: peak position contrast, area contrast under peak and spectral peak intensity contrast. For example: analysis 1125cm -1 Peak intensity, peak width, area under peak as a function of blood glucose concentration: analysis 1125cm -1 Peak with other skin characteristics (e.g., 1450cm -1 ) Interval (C)Is proportional to the spectral parameters (e.g., intensity ratio: I) 1125 /I 1450 Peak width ratio W 1125 /W 1450 The method comprises the steps of carrying out a first treatment on the surface of the Area ratio under peak: s is S 1125 /S 1450 ) As a function of blood glucose concentration.
Cutting data: the patent provides two ways of choosing data sets, namely random data cutting and distance data cutting. Random data cutting, disturbing the data sequence, and selecting data according to the set target proportion of the data set; and (3) cutting the distance data, calculating Euclidean distance of a data vector space, and respectively attributing a pair of vectors with the farthest distance to different data sets each time until the set target proportion of the data sets is met. Through the two methods, the training set, the verification set and the data set of the test set are selected and cut, the verification set is used for the parameter adjusting process of the model trained by the training set, after the verification set obtains better parameters or interesting training results, fitting and testing are carried out on the test set, and the risk of data leakage is eliminated to a greater extent by not directly using the test set for parameter adjusting.
Data expansion: the data expansion processing is selectively performed according to whether or not convolution learning is required in the subsequent work. In a convolutional neural network algorithm, in order to meet the requirement of a model on the data volume of three-dimensional data and improve the robustness of the model, the data of a Raman spectrum is expanded by selectively carrying out expansion and contraction change according to the standard deviation of the data to generate more Raman spectrum data, and the data of the Raman spectrum is expanded. The offset variable is controlled to be randomly generated within +/-0.10 times of the standard deviation of the training set, and the adjusted data become new training set data, so that the problem of under fitting of the convolutional neural network model is solved to a greater extent; according to the data characteristics of the Raman spectrum data, setting a series of parameters such as a learning rate, iterative calculation times, a target loss value and the like.
In this embodiment, the method for predicting the glucose content in blood according to the obtained raman spectrum characteristic information of glucose includes:
cutting the blood glucose Raman spectrum characteristic information into data sets for different targets, and selectively performing data expansion on the data sets;
constructing a blood glucose Raman spectrum model;
training the blood glucose raman spectrum model by using the data set;
and extracting Raman spectrum characteristic data by using the trained blood glucose Raman spectrum model to identify blood glucose Raman spectrum commonality quantitative information, and predicting the glucose content in the body.
In this embodiment, in the model training portion, the processed raman spectrum data is firstly converted according to the form of tensor required by input data of the model, the converted data can be used as input of a machine learning model to be trained, normalization processing is performed on the input data after the input data is determined so as to optimize a data structure, then data training can be performed according to an algorithm to be used, finally the model output is unified into glucose concentration information in human blood and trained model information, complete evaluation is performed by adopting a statistical evaluation system of regression prediction according to output information of the model, and visual evaluation is also performed on output results by adopting visual graphs such as a clark grid.
In this embodiment, the blood glucose raman spectrum model adopts a convolutional neural network (Convolutional Neural Network, abbreviated as CNN), a reverse artificial neural network (Back Propagating Artificial Neutral Network, abbreviated as BP-ANN), and a support vector regression machine (Support Vector Regression Machine, abbreviated as SVR). And selecting any one of the methods to analyze the spectrum data set, and then, after model parameters are set on a machine learning interface, learning and evaluating the spectrum data set.
In this embodiment, the method for constructing the blood glucose raman spectrum model by adopting the convolutional neural network includes:
firstly, carrying out tensor conversion on Raman spectrum data, converting two-dimensional Raman spectrum data information into three-dimensional data which can be identified by a convolutional neural network, carrying out convolution after normalization, extracting blood glucose Raman spectrum characteristics from the normalized data by a convolution layer, and creating new input data (characteristic diagram) for the next layer. The output of the convolution layer realizes nonlinear transformation through an activation function, data is transmitted to a pooling layer, a maximal pooling method is adopted to keep positive correlation in the data, then information of each feature map is connected at a full connection layer, regression values are obtained at a regression layer, and the model structure and parameter details of the convolution neural network are shown in fig. 3. After the output of the last layer is obtained, the obtained error value is used as a criterion, the information of the error is fed back to the previous layer, the threshold value and the deviation value of the previous layer are changed until the output of the last layer is gradually adjusted to the previous layer, and the neuron at each position is subjected to optimization parameter adjustment; and finally, forward computing again by the network after forward error adjustment, and repeating the previous steps to obtain an output error effect until the obtained error meets the requirement. The convolutional neural network algorithm is an error gradient descent algorithm, the weight is gradually corrected along the opposite direction of the calculation direction, the calculation process follows the conceptual diagram of the neural network calculation process in fig. 4, and the calculation process can be briefly described as:
(1) The pooling layer performs downsampling on the local features extracted by the convolution layer, screens the features of the data, and outputs the features to the full connection layer again by using the Relu activation function.
(2) The full connection layer is the mark space of the characteristic space mapping Raman spectrum sample obtained by calculation of the convolution layer and the pooling layer: the spectral characteristic representation is integrated into a value, the influence of the characteristic position on the classification result is reduced, the robustness of the whole network is improved, the regression layer is reached through the processing of a plurality of layers of full-connection layers, and finally a numerical value is output as a predicted value.
(3) Using a Mean Square Error (MSE) function in training a training set comprising a 1D-SCNN model of n Raman spectra, calculating a predicted valueAnd true value y i Error between:
(4) The adaptive learning rate is independently designed for different parameters using an ADAM (derived from Adaptive moment estimation, adaptive moment estimation) optimizer to calculate a first moment estimate and a second moment estimate of the gradient, iteratively calculating multiple times in updating the parameters into the various layers of the network structure as in fig. 4 in the direction of the decreasing mean square error gradient, obtaining a better value, and updating this better value into the layers of the network.
In this embodiment, the method for constructing the blood glucose raman spectrum model by using the feedback artificial neural network includes:
the calculation process of the back propagation neural network of the present invention can be roughly divided into the following two parts: firstly, after a data sample is input into an input layer, performing feature capturing and voting after calculation of upper threshold values and deviations of neurons of a hidden layer, and performing progressive calculation from front to back to obtain each neuron parameter of the hidden layer and a predicted value of an output layer; secondly, calculating the error of the predicted value and the true value, judging whether to continue iterative calculation according to the set error allowable range, wherein the calculation process is the same as the convolutional neural network and follows the conceptual diagram of the neural network calculation process in FIG. 4:
(1) Taking the length equal to the spectrum wave number as the length of the neural network input layer, and taking the intensity information omega of the ith wave number i Into the input layer, ω i The output value is obtained as the input of the hidden layer after the corresponding weight of the neuron is activated by the Relu activation function;
(2) After the hidden layer calculates the weight, the threshold and the activation function, the calculated output value is continuously input into the hidden layer with shorter length, so as to compress the characteristic information, the network learns more abstract spectrum characteristics, and finally the abstract spectrum characteristics are transmitted into the regression layer to obtain a value capable of describing the spectrum characteristics;
(3) Obtaining a true value and a predicted value error by taking a Mean Square Error (MSE) as a loss function, continuously optimizing through an ADAM algorithm, and calculating to obtain the learning rate of each parameter, wherein the learning rate when each parameter is updated is eta, and each parameter theta is calculated at the moment t i Updating:
(4) And (3) carrying out iterative computation for a plurality of times to obtain parameters with the optimization effect conforming to the loss target, and stopping computation to obtain the static model. The number of hidden layers depends on the volume of data, typically 2 to 3 layers in the task of the present invention, and the target loss value should be a value less than 0.001.
In this embodiment, the method for constructing the blood glucose raman spectrum model by using a support vector regression machine includes:
the support vector regression algorithm is based on the principle of a reference support vector machine part, regression model analysis is added, and high-dimensional information such as Raman spectrum is used as an independent variable to be trained of the support vector machine, namely, a given training sample D:
when to obtain f (x) =ω T +b can thus express a regression model of glucose concentration in humans, then it is necessary to make f (x i ) And y is i As close as possible, ω and b are the model parameters to be determined. For sample D, the conventional regression model is typically directly based on the model output f (x i ) And true output y i The difference between them to calculate the loss if and only if f (x i ) And y is i The loss is zero when it is exactly the same. The support vector regression algorithm assumes that we can tolerate f (x i ) And y is i With deviation of at most epsilon between, i.eOnly when f (x i ) And y is i The loss is calculated only when the absolute value of the difference is larger than epsilon. As shown in fig. 6, this corresponds to a value of f (x i ) For the center, a interval band with the width of 2 epsilon is constructed, if the training sample falls into the interval band, the training sample is considered to be predicted correctly, and the process can be briefly described as:
(1) The processed spectrum data and the glucose concentration information in human body are x and y respectively, and the appointed hyperplane omega is agreed T x+b=y; wherein ω is a normal vector, b is a displacement, and the distance d from the point to the hyperplane y is calculated;
(2) Maximizing the classification interval;
(3) Introducing a relaxation variable for the classification interval, allowing some data to be misclassified, preventing overfitting;
(4) Optionally using a linear kernel function or a gaussian kernel function K (x i X) after entering the kernel function, the SVM will define a hyperplane interface and maximize the hyperplane to closest data point spacing, constructing a classification function.
(x i ，y i ) For the ith spectrum of the training set, alpha is Lagrangian multiplier, beta 0 To bias, a solution function of the nonlinear problem is constructed:
in which:
linear kernel function: k (x) i ，x j )＝x i ×x j
Gaussian kernel function:
(5) Introducing SVM related functions to regression problem to obtain weight omega i And threshold b i It is noted that in the task of this patent, a linear kernel function is generally employed to achieve the desired effect. In this example, the accuracy of the predicted glucose content in blood was evaluated by using a statistical analysis and a clark grid analysis method in combination with a glucose meter to measure the glucose results.
In this example, the accuracy of the predicted glucose content in blood was evaluated by using a statistical analysis and a clark grid analysis method in combination with a glucose meter to measure the glucose results.
In this embodiment, after model training is completed, model evaluation is performed to obtain the performance of different data sets on the model: the performance of the raman spectrum quantitative detection compound model is evaluated by drawing a line graph of error conditions of a true value and a predicted value and Clarke medical grid evaluation (the true value and the predicted value are used as an abscissa and an ordinate and are drawn in grids of a standard partition, wherein a region A represents a plurality of better predictions, a region B represents an acceptable error condition, a region C, D possibly brings medical accidents on a certain length, and a region E possibly brings serious medical accidents), and obtaining sum variance, mean square error, root mean square error, average absolute error, R-square value and correlation coefficient of the true value and the predicted value.
In the model using part, the interested model (the static model containing the trained model parameters) output in the model training process is stored, the model using part is used for carrying out the model using link, at the moment, only the data to be predicted is input, at the moment, the Raman spectrum data to be predicted activates the parameters of each layer of the model to obtain the prediction result under the static model, and finally, the regression evaluation consistent with the model training step is carried out. And evaluating the conditions of the predicted value and the true value to obtain: the error condition line diagram of the true value and the predicted value shown in fig. 7 and the Clarke medical grid shown in fig. 8 are evaluated, and evaluation parameters of the true value and the predicted value shown in fig. 9 are obtained, wherein the evaluation parameters comprise sum variance, mean square error, root mean square error, mean absolute error, R-square value and correlation coefficient, so that the prediction accuracy of the compound quantitatively detected by the raman spectrum is evaluated from different angles, and the prediction accuracy of the compound quantitatively detected by the raman spectrum is obtained, wherein:
(1) And variance, the sum of squares due to error, SSE.
(2) Mean square error, mean squared error, MSE, mean of the sum of squares of the point errors corresponding to the predicted and original data.
(3) The root mean square error, root mean squared error, RMSE, also known as the fitting standard deviation of the regression system, is the square root of the MSE.
(4) Mean absolute error, mean Absolute Error, MAE, is the mean of absolute errors.
(5) R-square, is the ratio of the (1-model does not capture the amount of information) to the total amount of information, whereIs the average value of the true values
(6) And evaluating the result of the regression model by using the correlation coefficient, r, and quantitatively evaluating the data correlation by using the pearson correlation coefficient in the program. The pearson correlation coefficient between two variables is defined as the quotient of the covariance and standard deviation between the two variables:
the above equation defines the overall correlation coefficient, with the greek lower case ρ being used as a representative symbol. The covariance and standard deviation of the sample are estimated to obtain the pearson correlation coefficient, and the common english lowercase letter r represents:
after calculating the correlation coefficient, the correlation strength of the variable can be judged by the value range of table 1:
TABLE 1
| |r| | Correlation strength |
| 0.8-1.0 | Extremely strong correlation |
| 0.6-0.8 | Strong correlation |
| 0.4-0.6 | Moderate correlation |
| 0.2-0.4 | Weak correlation |
| 0.0-0.2 | Very weak correlation or no correlation |
The invention initiates a composite Raman spectrum data analysis method of glucose component in blood, namely, according to the characteristic difference between spectrum signal and background noise, accurately extracting physical parameters of glucose characteristic peak; and establishing a mixed deep learning model, effectively identifying common characteristics of different individual spectrum detection results, and accurately predicting the real content of blood sugar in the body. This is expected to form a more accurate and universal analysis method system for the Raman spectrum data of blood sugar.
Example two
The invention also provides a composite Raman spectrum data analysis system of glucose component in blood, comprising: the device comprises a preprocessing module, a feature extraction module and a prediction module;
the pretreatment module is used for carrying out pretreatment on blood glucose Raman spectrum data obtained by in-vivo or in-vitro measurement;
the characteristic extraction module is used for quantitatively extracting glucose Raman spectrum characteristic information in blood from the preprocessed blood glucose Raman spectrum data;
the prediction module is used for predicting the glucose content in blood according to the obtained glucose Raman spectrum characteristic information.
In this embodiment, the method for preprocessing blood glucose raman spectrum data obtained by in vivo or ex vivo measurement includes:
selectively carrying out Raman spectrum wavelength interpolation on the blood glucose Raman spectrum data;
shearing the spectrum range of the blood glucose Raman spectrum data;
fluorescent background, cosmic rays, and other noise interference due to instrumentation and environment are eliminated from the blood glucose raman spectral data.
In this embodiment, the method for quantitatively extracting the raman spectrum characteristic information of glucose in blood from the pretreated raman spectrum data of blood glucose includes:
and smoothing, normalizing and averaging the pretreated blood glucose Raman spectrum data, and identifying and extracting blood glucose Raman spectrum characteristic information.
In this embodiment, the method for predicting the glucose content in blood according to the obtained raman spectrum characteristic information of glucose includes:
cutting the blood glucose Raman spectrum characteristic information into data sets for different targets, and selectively performing data expansion on the data sets;
constructing a blood glucose Raman spectrum model to identify Raman spectrum characteristic data;
training the blood glucose raman spectrum model by using the data set;
and identifying and extracting blood glucose Raman spectrum commonality quantification information by utilizing the trained blood glucose Raman spectrum model to extract Raman spectrum characteristic data, and predicting the glucose content in the body.
In this embodiment, the blood glucose raman spectrum model is constructed by adopting a convolutional neural network, a feedback artificial neural network or a support vector regression machine.
In this embodiment, the method for constructing the blood glucose raman spectrum model by adopting the convolutional neural network includes:
tensor conversion is carried out on the blood glucose Raman spectrum data, two-dimensional blood glucose Raman spectrum data information is converted into three-dimensional data recognized by a convolutional neural network, and convolution is carried out after normalization;
the convolution layer extracts features from the normalized data and creates a new output for the next layer;
the output of the convolution layer will be transformed by an activation function to achieve nonlinearity and the data will be passed to the pooling layer;
the positive correlation in the data is reserved by adopting a maximum pooling method, then the information of each feature map is connected in a full connection layer, and a regression value is obtained in a regression layer;
after the output of the last layer is obtained, the obtained error value is used as a criterion, the information of the error is fed back to the previous layer, the threshold value and the deviation value of the previous layer are changed until the output of the last layer is gradually adjusted to the previous layer, and the neuron at each position is subjected to optimization parameter adjustment;
and forward computing again by the network after forward error adjustment, and repeating the previous steps to obtain an output error effect until the obtained error meets the requirement.
In this embodiment, the method for constructing the blood glucose raman spectrum model by using the feedback artificial neural network includes:
after the input layer inputs a data sample, performing feature capturing and identification after calculating the upper threshold value and the deviation of the neuron of the hidden layer, and performing step-by-step calculation from front to back to obtain each neuron parameter of the hidden layer and the predicted value of the output layer;
and calculating the error between the predicted value and the true value, and judging whether iteration calculation needs to be continued or not according to the set error allowable range.
In this embodiment, the method for constructing the blood glucose raman spectrum model by using a support vector regression machine includes:
based on the principle of a reference support vector machine part, regression model analysis is added, and the high-dimensional information such as Raman spectrum is used as an independent variable to be trained of the support vector machine.
In this example, the accuracy of the predicted glucose content in blood was evaluated by using a statistical analysis and a clark grid analysis method in combination with a glucose meter to measure the glucose results.
The above embodiments are merely illustrative of the preferred embodiments of the present invention, and the scope of the present invention is not limited thereto, but various modifications and improvements made by those skilled in the art to which the present invention pertains are made without departing from the spirit of the present invention, and all modifications and improvements fall within the scope of the present invention as defined in the appended claims.
Claims (10)
1. The method for analyzing the composite Raman spectrum data of the glucose component in the blood is characterized by comprising the following steps:
preprocessing blood glucose Raman spectrum data obtained by in-vivo or in-vitro measurement;
quantitatively extracting glucose Raman spectrum characteristic information in blood from the pretreated blood glucose Raman spectrum data;
and predicting the glucose content in blood according to the obtained glucose Raman spectrum characteristic information.
2. The method for analyzing composite raman spectrum data of a glucose component in blood according to claim 1, wherein the method for preprocessing the blood glucose raman spectrum data obtained by in vivo or ex vivo measurement comprises:
selectively carrying out Raman spectrum wavelength interpolation on the blood glucose Raman spectrum data;
shearing the spectrum range of the blood glucose Raman spectrum data;
fluorescent background, cosmic rays, and other noise interference due to instrumentation and environment of the blood glucose raman spectrum data are eliminated.
3. The method for analyzing composite raman spectrum data of a glucose component in blood according to claim 1, wherein the method for quantitatively extracting the raman spectrum characteristic information of the glucose in blood from the pretreated raman spectrum data of the blood glucose comprises:
and smoothing, normalizing and averaging the pretreated blood glucose Raman spectrum data, and identifying and extracting blood glucose Raman spectrum characteristic information.
4. The method for analyzing composite raman spectrum data of a glucose component in blood according to claim 1, wherein the method for predicting the glucose content in blood based on the obtained glucose raman spectrum characteristic information comprises:
cutting the blood glucose Raman spectrum characteristic information into data sets for different targets, and selectively performing data expansion on the data sets;
constructing a blood glucose Raman spectrum model to identify Raman spectrum characteristic data;
training the blood glucose raman spectrum model by using the data set;
and extracting blood glucose Raman spectrum commonality quantification information by using the trained blood glucose Raman spectrum model, and predicting the in-vivo glucose content.
5. The method for analyzing composite raman spectrum data of glucose in blood according to claim 4, wherein the blood glucose raman spectrum model is constructed by using a convolutional neural network, a feedback artificial neural network or a support vector regression machine.
6. The method for analyzing composite raman spectrum data of a glucose component in blood according to claim 5, wherein the method for constructing the blood glucose raman spectrum model by using a convolutional neural network comprises:
performing tensor conversion on the blood glucose Raman spectrum data, converting two-dimensional blood glucose Raman spectrum data information into three-dimensional data identified by a convolutional neural network, and performing convolution after normalization;
the convolution layer extracts the blood glucose Raman spectrum characteristics from the normalized data, and creates new characteristic data input/output for the next layer;
the output of the convolution layer realizes nonlinear transformation through an activation function and transmits data to a pooling layer;
the positive correlation in the data is reserved by adopting a maximum pooling method, then the information of each feature map is connected in a full connection layer, and a regression value is obtained in a regression layer;
after the output of the last layer is obtained, the obtained error value is used as a criterion, the information of the error is fed back to the previous layer, the threshold value and the deviation value of the previous layer are changed until the output of the last layer is gradually adjusted to the previous layer, and the neuron at each position is subjected to optimization parameter adjustment;
and forward computing again by the network after forward error adjustment, and repeating the previous steps to obtain an output error effect until the obtained error meets the requirement.
7. The method for analyzing composite raman spectrum data of glucose components in blood according to claim 5, wherein the method for constructing the blood glucose raman spectrum model by using a feedback artificial neural network comprises:
after the input layer inputs a data sample, performing feature capturing and voting after calculating the upper threshold value and the deviation of the neuron of the hidden layer, and gradually calculating from front to back to obtain each neuron parameter of the hidden layer and the predicted value of the output layer;
and calculating the error between the predicted value and the true value, and judging whether iteration calculation needs to be continued or not according to the set error allowable range.
8. The method for analyzing composite raman spectrum data of a glucose component in blood according to claim 5, wherein the method for constructing the blood glucose raman spectrum model by using a support vector regression machine comprises:
based on the principle of a reference support vector machine part, regression model analysis is added, and the high-dimensional information such as Raman spectrum is used as an independent variable to be trained of the support vector machine.
9. The method for analyzing composite raman spectrum data of a glucose component in blood according to claim 1, wherein the accuracy of the predicted glucose content in blood is evaluated by measuring the glucose result by using a combination of a statistical analysis and a clark lattice analysis method and a glucometer.
10. A system for analyzing composite raman spectrum data of a glucose component in blood, comprising: the device comprises a preprocessing module, a feature extraction module and a prediction module;
the pretreatment module is used for carrying out pretreatment on blood glucose Raman spectrum data obtained by in-vivo or in-vitro measurement;
the characteristic extraction module is used for quantitatively extracting glucose Raman spectrum characteristic information in blood from the preprocessed blood glucose Raman spectrum data;
the prediction module is used for predicting the glucose content in blood according to the obtained glucose Raman spectrum characteristic information.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310543084.2A CN116559143A (en) | 2023-05-15 | 2023-05-15 | Method and system for analyzing composite Raman spectrum data of glucose component in blood |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310543084.2A CN116559143A (en) | 2023-05-15 | 2023-05-15 | Method and system for analyzing composite Raman spectrum data of glucose component in blood |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116559143A true CN116559143A (en) | 2023-08-08 |
Family
ID=87501483
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310543084.2A Pending CN116559143A (en) | 2023-05-15 | 2023-05-15 | Method and system for analyzing composite Raman spectrum data of glucose component in blood |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116559143A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117617962A (en) * | 2023-12-25 | 2024-03-01 | 暨南大学 | Subcutaneous blood sugar measurement method based on space shift Raman |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110575181A (en) * | 2019-09-10 | 2019-12-17 | 重庆大学 | Near infrared spectrum noninvasive blood glucose detection network model training method |
| CN111050645A (en) * | 2017-08-23 | 2020-04-21 | 株式会社理光 | Measuring device and measuring method |
| CN112712108A (en) * | 2020-12-16 | 2021-04-27 | 西北大学 | Raman spectrum multivariate data analysis method |
| CN112754479A (en) * | 2020-12-21 | 2021-05-07 | 四川大学华西医院 | Method and equipment for detecting blood glucose concentration |
| CN113662538A (en) * | 2021-08-24 | 2021-11-19 | 西安理工大学 | Non-invasive blood glucose detection method based on time-frequency domain comprehensive analysis |
| CN114544592A (en) * | 2022-02-28 | 2022-05-27 | 上海应用技术大学 | Intelligent Raman spectrum classification method based on convolutional neural network |
-
2023
- 2023-05-15 CN CN202310543084.2A patent/CN116559143A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111050645A (en) * | 2017-08-23 | 2020-04-21 | 株式会社理光 | Measuring device and measuring method |
| CN110575181A (en) * | 2019-09-10 | 2019-12-17 | 重庆大学 | Near infrared spectrum noninvasive blood glucose detection network model training method |
| CN112712108A (en) * | 2020-12-16 | 2021-04-27 | 西北大学 | Raman spectrum multivariate data analysis method |
| CN112754479A (en) * | 2020-12-21 | 2021-05-07 | 四川大学华西医院 | Method and equipment for detecting blood glucose concentration |
| CN113662538A (en) * | 2021-08-24 | 2021-11-19 | 西安理工大学 | Non-invasive blood glucose detection method based on time-frequency domain comprehensive analysis |
| CN114544592A (en) * | 2022-02-28 | 2022-05-27 | 上海应用技术大学 | Intelligent Raman spectrum classification method based on convolutional neural network |
Non-Patent Citations (5)
| Title |
|---|
| FEIFEI PIAN等: "《A shallow convolutional neural network with elastic nets for blood glucose quantitative analysis using Raman spectroscopy》", 《SPECTROCHIMICA ACTA PART A: MOLECULAR AND BIOMOLECULAR SPECTROSCOPY》, pages 1 - 6 * |
| NARAHARA CHARI DINGARI等: "《Wavelength selection-based nonlinear calibration for transcutaneous blood glucose sensing using Raman spectroscopy》", 《JOURNAL OF BIOMEDICAL OPTICS》 * |
| 苏永健;解世涛;李雪冰;李闯;李鹏竹;谭祥帅;李昭;辛志波;赵如宇;王林;何川;张宏元: "基于ADAM-BPNN优化模型的磨煤机智能启动技术研究", 工业加热, vol. 52, no. 001, 31 December 2023 (2023-12-31), pages 31 - 32 * |
| 谭祥帅;白世雄;李昭;高景辉;陈景勇;冯培峰;刘文仓;辛志波;高奎;牛利涛;刘冲: "基于ADAM-BPNN模型单汽泵故障联动控制技术研究应用", 中国电机工程学报, vol. 43, no. 003, 31 December 2023 (2023-12-31), pages 1098 - 1099 * |
| 陈红岩;盛伟铭;刘嘉豪;黄翰;朱俊江;赵永佳;: "基于TensorFlow架构的Adam-BPNN在红外甲烷传感器系统误差修正的应用", 传感技术学报, no. 04, 15 April 2020 (2020-04-15), pages 531 - 533 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117617962A (en) * | 2023-12-25 | 2024-03-01 | 暨南大学 | Subcutaneous blood sugar measurement method based on space shift Raman |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP1250082B1 (en) | Classification and characterization of tissue through features related to adipose tissue | |
| US20070179367A1 (en) | Method and Apparatus for Noninvasively Estimating a Property of an Animal Body Analyte from Spectral Data | |
| KR102095959B1 (en) | Artificial Neural Network Model-Based Methods for Analyte Analysis | |
| JP2004528083A (en) | A multi-step method to build a localized calibration model for non-invasive blood sample prediction | |
| Ashok et al. | Determination of blood glucose concentration by using wavelet transform and neural networks | |
| CN111599470B (en) | Method for improving near-infrared noninvasive blood glucose detection precision | |
| CN108542402B (en) | Blood glucose detection method based on self-organizing competitive neural network model and infrared spectrum | |
| CN112806977B (en) | Physiological parameter measuring method based on multi-scale fusion network | |
| CN112200770A (en) | Tumor detection method based on Raman spectrum and convolutional neural network | |
| CN116559143A (en) | Method and system for analyzing composite Raman spectrum data of glucose component in blood | |
| CN116754511B (en) | Indocyanine green detection method and system based on spectrum technology | |
| CN116840214A (en) | Method for diagnosing brain tumor and cerebral infarction | |
| CN116451110A (en) | Blood glucose prediction model construction method based on signal energy characteristics and pulse period | |
| Suryakala et al. | Investigation of goodness of model data fit using PLSR and PCR regression models to determine informative wavelength band in NIR region for non-invasive blood glucose prediction | |
| Philip et al. | Continous monitoring of blood glucose using photophlythesmograph signal | |
| CN115089139A (en) | Personalized physiological parameter measuring method combining biological characteristic identification | |
| CN108680748B (en) | Blood glucose detector based on acetone sensor and detection method thereof | |
| US20030087456A1 (en) | Within-sample variance classification of samples | |
| Li et al. | Noninvasive detection and analysis of human globulin based on dynamic spectrum | |
| CN111134689B (en) | Blood glucose concentration detection method based on time-frequency Teager-Kaiser energy | |
| Ham et al. | Multivariate determination of glucose concentrations from optimally filtered frequency-warped NIR spectra of human blood serum | |
| CN113974618B (en) | Noninvasive blood glucose testing method based on water peak blood glucose correction | |
| CN114145725B (en) | PPG sampling rate estimation method based on noninvasive continuous blood pressure measurement | |
| Zhao et al. | An automatic glucose monitoring signal denoising method with noise level estimation and responsive filter updating | |
| Ming et al. | A comparison analysis between partial least squares and neural network in non-invasive blood glucose concentration monitoring system |
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 |