CN116577388A

CN116577388A - Cell pressure volume correction method, device, system and storage medium

Info

Publication number: CN116577388A
Application number: CN202310855082.7A
Authority: CN
Inventors: 翟敏; 杨语; 葛艳秋; 戈芷琪; 秦玉
Original assignee: Nanjing Jingjie Biotechnology Co ltd
Current assignee: Nanjing Jingjie Biotechnology Co ltd
Priority date: 2023-07-13
Filing date: 2023-07-13
Publication date: 2023-08-11
Anticipated expiration: 2043-07-13
Also published as: CN116577388B

Abstract

The invention provides a cell pressure volume correction method, device and system and a storage medium, and relates to the technical field of electrochemical detection. The cell pressure volume correction method comprises the following steps: establishing a data set by using an electrochemical test paper test sample; classifying the concentration segments by using a support vector machine; obtaining coefficient fitting of the classification hyperplane corresponding to each concentration section classification to obtain a compensation equation; and compensating the target sample data according to a compensation equation and calculating to obtain a corrected concentration value. The invention adopts a compensation mode of two variables of current and HCT, and utilizes multi-frequency point information to eliminate the influence of temperature on the HCT result. The measurement mode meets the requirement of rapid in-situ detection, can effectively avoid the temperature retardation effect caused by the built-in temperature sensor and the error caused by abnormal operation of a user, ensures that the applicable temperature range of blood detection is wider, improves the accuracy of HCT detection results and electrochemical blood detection correction results, and brings convenience to the electrochemical detection of blood.

Description

Cell pressure volume correction method, device, system and storage medium

Technical Field

The invention relates to the technical field of electrochemical detection, in particular to a cell pressure volume correction method, device and system and a storage medium.

Background

The electrochemical method is an effective method for measuring the content of specific substances in biological samples, and the measurement system is small and portable, so that the electrochemical detection application scene is wide, a creatinine detector and creatinine detection test paper are typical, collected blood or plasma samples are stored in a micro cavity through a hydrophilic layer with a siphon effect, chemical reactions are carried out in the cavity and in advance on an enzyme layer, an electron mediator and the like, a certain potential is applied between a working electrode and a reference electrode, so that creatinine micromolecules are subjected to oxidation-reduction reaction on the surface of the electrode, and current (Cottrell current) is generated.

The Cottrell current is affected by the temperature and viscosity of the sample to be tested, and the change in blood viscosity during actual blood testing is often affected by the hematocrit. Hematocrit (HCT), also known as hematocrit or hematocrit, refers to the volume percent of red blood cells in a volume of whole blood. HCT detection can provide information about the number and quality of red blood cells, as well as estimate hemoglobin values, often as an important indicator for assessing user health, particularly as an important means of assessing renal anemia.

The alternating current impedance method is a common method for measuring HCT in the electrochemical detection field, but the temperature of a sample to be measured and the frequency of an alternating current signal can influence the measurement accuracy of HCT, thereby influencing the compensation of Cottrell current.

It is therefore necessary to find a way to accurately measure HCT and make compensation corrections to the Cottrell current through HCT. The common scheme in the prior art is to test alternating current impedance at a certain fixed frequency and calibrate the influence of temperature on alternating current impedance by matching with a temperature sensor of an electrochemical detection instrument so as to improve the accuracy of HCT measurement and compensation correction.

Most instruments adopt a built-in temperature sensor to obtain the ambient temperature, the ambient temperature is approximately used as the reaction temperature on the test paper, and as the temperature sensor is built-in a machine shell, the time constant tau is increased, the response to the external temperature change becomes more 'dull', for example, a phenomenon of temperature mutation occurs in the test process (switching from outdoor to indoor), or the test paper is placed in a refrigerator for preservation by a user in error, the test paper is not completely rewarmed in use, or the user is erroneously pinched in a reaction area of the test paper when grabbing the test paper, so that the test paper is locally heated, and the like, the delay of the test is necessarily caused under similar test scenes, so that the temperature sensor cannot objectively evaluate the reaction temperature in a test paper channel to a certain extent, the accuracy of an HCT measurement result is influenced, and the accuracy of subsequent HCT compensation and the final concentration estimation value are finally influenced.

More importantly, to improve the accuracy of HCT compensation for current signals, most prior art techniques compensate for current bias by simple linear or piecewise functions, as shown in equations (1) and (2) below:

；(1)

；(2)

wherein, percentage _compensate To compensate for percentages, e.g. percentage _compensate By 30%, we mean that an additional 30% compensation is required for the current acquired, HCT is the measured HCT value, e.g., hct=30%, then percentage _compensate -20%, then the measured current is compensated to an additional-20%.

In the above compensation scheme, since only the influence of HCT on the current at a certain scale is considered, and when the target analyte concentration is different, influence factors have different degrees of influence. For example, at creatinine concentrations of less than 50 μmol/L, and greater than 500 μmol/L, respectively, the same hct=30% will exhibit a significantly different effect than the latter at two concentrations relative to hct=42%. It is therefore necessary to take into account the different analyte concentrations (over the entire linear range) when compensating for the current, which would otherwise lead to large deviations for samples of certain concentration segments.

In a word, in the existing compensation strategy for HCT, when the accuracy of HCT correction is improved through temperature sensor calibration temperature, the correction result is influenced due to temperature response delay, and a simple linear or piecewise compensation method is greatly influenced by a target analyte, so that the accurate target cannot be realized all the time by the HCT correction result, and great inconvenience is brought to electrochemical accurate detection of biological samples.

Disclosure of Invention

In view of the above, the present invention provides a method for correcting cell pressure accumulation, comprising:

testing a sample by using electrochemical test paper to obtain a current value, obtaining an HCT measured value of the sample, and constructing a data set based on the sample;

obtaining a corresponding HCT predicted value for each of the samples in the data set based on the HCT measured values;

according to the data set, classifying concentration segments of the detected HCT predicted value and current value by using a support vector machine;

obtaining coefficients of each classification hyperplane corresponding to each concentration section classification, and fitting compensation percentages of the HCT predicted value and the current value in each concentration section to obtain a corresponding compensation equation;

and compensating the target sample data according to the compensation equation of the corresponding concentration section, and calculating a corrected concentration value corresponding to the target sample data according to the compensated current.

Preferably, the constructing the data set includes:

collecting samples in different HCT ranges and different temperature ranges, performing alternating current test, and recording real part information and imaginary part information under each frequency;

acquiring an HCT measured value and a steady-state temperature of each sample, and combining data information of all the samples;

And establishing the data set according to the data information.

Preferably, the collecting samples of different HCT ranges and different temperature ranges and performing ac test, recording real part information and imaginary part information at each frequency, includes:

alternating current scanning is carried out on the sample based on the HCT working electrode and the reference electrode under different frequencies;

the real information and the imaginary information at each frequency are recorded.

Preferably, after recording the real part information and the imaginary part information at each frequency, the method further includes:

converting values of the real part information and the imaginary part information into impedance amplitude values and phase angles as converted real part information and imaginary part information;

the conversion relation among the values of the real part information and the imaginary part information converted into impedance amplitude values and phase angles is as follows:

；

) ；

wherein, z is amplitude, θ is phase angle, real part is real part information of alternating current test, and imaginary part is imaginary part information of alternating current test.

Preferably, the establishing the data set according to the data information includes:

dividing the data information into a training set and a testing set according to a preset proportion;

unified standardization processing is carried out on the training set and the testing set;

The unified normalization processing is to convert each data characteristic in the training set and the test set into the training set and the test set with the mean value of 0 and the standard deviation of 1;

the unified normalization processing formula is as follows:

；

wherein z is a unified normalized value, x is the HCT measured value or the current value in the data set,and s is the standard deviation of x and is the mean of x.

Preferably, said deriving a corresponding HCT predicted value for each of said samples in said data set based on said HCT measured values comprises:

and constructing a prediction model based on a nonlinear iterative partial least square method, and calculating the HCT predicted value of each sample according to the HCT measured value.

Preferably, the constructing a prediction model based on the nonlinear iterative partial least square method, and calculating the HCT predicted value of each sample according to the HCT measured value comprises:

performing factor decomposition on the HCT measured value in the data set through the prediction model by using the nonlinear iterative partial least square method to obtain a regression coefficient, and then determining the HCT predicted value;

wherein the obtaining regression coefficients comprises:

Determining the number of principal components according to the training set and the testing set of the data set;

and taking the number of principal components as an input parameter, and obtaining an HCT predicted value through the prediction model.

Preferably, the classifying the concentration segments of the measured HCT predicted value and the current value according to the data set by using a support vector machine includes:

establishing a classifier based on the support vector machine;

obtaining a classification equation coefficient through the classifier, and obtaining an equation slope and an intercept of the classifier, thereby constructing a classification model;

and inputting the HCT predicted value and the current value measured by each sample in the data set into the classification model to obtain a concentration segment classification result.

Preferably, the classifier is a linear kernel function based bi-classifier, and each bi-classifier is a linear function;

the equation for the linear function is as follows:

；

wherein ,is a coefficient; />Representing the input vector consisting of hematocrit HCT and current I; />Is the intercept; y is the output concentration section classification result;

the number of the classifier is N-1, and N is the classified number of the classification model.

Preferably, the obtaining the classification equation coefficient by the classifier and obtaining the equation slope and intercept of the classifier, thereby constructing a classification model includes:

Calling svm.SVC functions in the scikit-learn function library based on the Python learning library;

setting a kernel parameter of the svm.svc function type as a linear function;

and obtaining the classification equation coefficient through the classification model, and calculating the equation slope and the intercept.

Preferably, the classification equation coefficients include coef and interseptit;

the classifier comprises f1 and f2;

the equation for the equation slope and intercept is calculated as:

；

]；

where k and b are the slope and intercept of the function of the classifier f1 and/or f 2.

Preferably, the inputting the HCT predicted value and the current value measured by each sample in the data set into the classification model for prediction, to obtain a concentration segment classification result, includes:

bringing a training set and a testing set in a data set into classification equations of the classifiers f1 and f2 of the classification model to obtain corresponding classification discrimination values;

according to the classification discrimination value, a corresponding one-dimensional vector is obtained;

logically dividing the one-dimensional vector to obtain a corresponding concentration segment classification result;

and evaluating the concentration segment classification result by adopting a confusion matrix evaluation method to finish classification.

Preferably, the obtaining a corresponding one-dimensional vector according to the classification discrimination value includes:

and adopting 2-system coding, and converting the classification discrimination values according to the principle that the number is greater than 0 and is 1 and less than or equal to 0 to obtain the one-dimensional vector of 2-bit 2-system corresponding to each classification equation.

Preferably, the obtaining the coefficient of each classification hyperplane corresponding to each concentration segment classification, and fitting the compensation percentage of the HCT predicted value and the current value in each concentration segment to obtain a corresponding compensation equation includes:

recording coefficients of each classification hyperplane in the concentration segment classification;

based on the coefficients of the classification hyperplanes, using the measured HCT predicted value as an abscissa and using the current percentage of the current value to be compensated as an ordinate, and adopting a linear equation to fit the relation between the measured HCT predicted value and the current percentage of the current value to be compensated to obtain the compensation equation corresponding to each concentration section.

Preferably, the compensation equation corresponding to each concentration segment is:

；

wherein ,a compensation percentage which is the relative deviation of the measured current value and the reference current; / >Is a coefficient; />A predicted value for the HCT; />Is the intercept.

Preferably, the electrochemical test strip comprises:

hydrophilic film layer, barrier layer, single-sided adhesive layer, electrode layer and insulating layer;

the hydrophilic film layer is provided with air holes;

a siphon window is arranged on the barrier layer;

one end of the single-sided adhesive layer is provided with a reagent window;

the siphon window corresponds to the position of the reagent window and can be coincident when combined;

one end of the single-sided adhesive layer, which is provided with a reagent window, is overlapped with one end of the electrode layer;

the electrode layer is attached to the insulating layer;

the hydrophilic film layer, the barrier layer, the single-sided adhesive layer, the electrode layer and the insulating layer are sequentially arranged from top to bottom and are mutually attached.

Preferably, the number of the reagent windows on the single-sided adhesive layer is 4, and the reagent windows are respectively:

a first window, a second window, a third window, and a fourth window.

Preferably, the electrode layer is provided with a reaction area corresponding to the reagent window of the single-sided adhesive layer;

the reaction zone comprises: a first region, a second region, a third region, and a fourth region;

the first region carries a first reactive enzyme solution; the second zone carries a second reactive enzyme solution; the third zone carries a non-reactive substance; the fourth zone is free of application of a reactive enzyme solution and non-reactive materials;

The first region, the second region, the third region and the fourth region are sequentially arranged and are adjacent to each other.

Preferably, the electrode layer includes: a reference electrode assembly and an HCT working electrode, wherein the reference electrode assembly comprises a working reference electrode and an HCT reference electrode;

the reference electrode assembly and the HCT working electrode are electrically connected with a signal acquisition module of the cell pressure volume correction system when the reference electrode assembly and the HCT working electrode are connected with the cell pressure volume correction system.

Preferably, the electrode layer comprises at least one of the following features:

A. the electrode layer is made of a metal thin layer or printing carbon paste;

B. the thickness of the metal thin layer in the electrode layer is 1nm-50nm;

C. the metal in the metal thin layer in the electrode layer is an alloy formed by any one or a combination of more than one of gold, platinum, palladium, nickel or titanium.

In addition, in order to solve the above-mentioned problems, the present invention also provides a cell pressure volume correction device, comprising:

the data module is used for testing a sample by using electrochemical test paper to obtain a current value, obtaining an HCT measured value of the sample and constructing a data set based on the sample;

a prediction module for obtaining a corresponding HCT predicted value for each of the samples in the data set based on the HCT measured values;

The classification module is used for classifying the concentration section of the measured HCT predicted value and the current value by using a support vector machine according to the data set;

the calculation module is used for obtaining the coefficient of each classification hyperplane corresponding to each concentration section classification, fitting the compensation percentage of the HCT predicted value and the current value in each concentration section, and obtaining a corresponding compensation equation;

and the correction module is used for compensating the target sample data according to the compensation equation of the corresponding concentration section, and calculating a corrected concentration value corresponding to the target sample data according to the compensated current.

In addition, in order to solve the above problems, the present invention also provides a system for correcting cell pressure accumulation, which comprises a signal acquisition device, a memory and a processor; the signal acquisition device acquires signals returned by the reference electrode component and the HCT working electrode in the electrochemical test paper electrode layer; the memory stores a cell pressure accumulation correction program; the processor runs the cell pressure volume correction program to cause the cell pressure volume correction system to perform the cell pressure volume correction method as described above.

In addition, in order to solve the above-mentioned problems, the present invention also provides a computer-readable storage medium having stored thereon a cell pressure volume correction program which, when executed by a processor, implements the cell pressure volume correction method as described above.

The invention provides a cell pressure volume correction method, a device, a system and a storage medium, wherein the method comprises the following steps: testing a sample by using electrochemical test paper to obtain a current value, obtaining an HCT measured value of the sample, and constructing a data set based on the sample; obtaining a corresponding HCT predicted value for each of the samples in the data set based on the HCT measured values; according to the data set, classifying concentration segments of the detected HCT predicted value and current value by using a support vector machine; obtaining coefficients of each classification hyperplane corresponding to each concentration section classification, and fitting compensation percentages of the HCT predicted value and the current value in each concentration section to obtain a corresponding compensation equation; and compensating the target sample data according to the compensation equation of the corresponding concentration section, and calculating a corrected concentration value corresponding to the target sample data according to the compensated current. Based on the premise of fully considering the influence of temperature on HCT detection, the invention replaces the alternating current impedance of a single frequency point in the traditional scheme, gives up the single variable compensation method of the formula (1) or the formula (2), adopts a compensation mode of two variables of current and HCT, and utilizes the information of multiple frequency points to eliminate the influence of temperature on the HCT result. And the measurement mode meets the requirement of rapid in-situ detection, can effectively avoid the temperature retardation effect caused by the built-in temperature sensor and the error caused by abnormal operation of a user, ensures accurate HCT detection and correction compensation results, and brings convenience to electrochemical accurate detection of biological samples.

Drawings

FIG. 1 is a schematic diagram of a hardware operating environment according to an embodiment of the present invention;

FIG. 2 is a flow chart of the method for correcting cell pressure accumulation according to the embodiment 1 of the invention;

FIG. 3 is a flowchart illustrating the refinement of step S100 in embodiment 2 of the cell pressure accumulation correction method according to the present invention;

FIG. 4 is a flowchart illustrating the refinement of step S110 in embodiment 2 of the cell pressure accumulation correction method according to the present invention;

FIG. 5 is a flowchart of an embodiment 3 of the method for correcting cell pressure accumulation according to the present invention;

FIG. 6 is a schematic diagram showing the ratio of principal component information according to the principal components in embodiment 3 of the cell volume correction method according to the present invention;

FIG. 7 is a diagram showing the effect of training and testing sets on HCT prediction when the principal component number is 6 in embodiment 3 of the method for correcting cell pressure accumulation according to the present invention;

FIG. 8 is a flowchart illustrating the refinement of step S300 in embodiment 4 of the cell pressure accumulation correction method according to the present invention;

FIG. 9 is a graph showing the response of measured currents to different HCT's when the concentration of creatinine (1-3) is different in example 4 of the method for correcting cell size deposition according to the present invention;

FIG. 10 is a schematic diagram of a third embodiment of the method for correcting cell volume according to the present invention;

FIG. 11 is a flowchart illustrating the refinement of step S320 in embodiment 4 of the cell pressure accumulation correction method according to the present invention;

FIG. 12 is a flowchart illustrating the refinement of step S330 in embodiment 4 of the cell pressure accumulation correction method according to the present invention;

FIG. 13 is a flowchart of the method for correcting cell size accumulation according to embodiment 4 of the present invention, wherein the step S332 is detailed and includes a step S3321;

FIG. 14 is a schematic diagram of classification interfaces (f 1 and f 2) for classifying different concentrations using a linear kernel in embodiment 4 of the method for correcting cell volume according to the present invention;

FIG. 15 is a flowchart showing the refinement of step S400 in embodiment 5 of the cell pressure accumulation correction method according to the present invention;

FIG. 16 is a schematic diagram showing compensation equations of the concentration C1 currents I and HCT according to the embodiment 5 of the hematocrit correction method of the present invention;

FIG. 17 is a schematic diagram showing compensation equations of the concentration C2 currents I and HCT according to the embodiment 5 of the hematocrit correction method of the present invention;

FIG. 18 is a schematic diagram showing compensation equations of the concentration C3 current I and HCT in the embodiment 5 of the cell pressure accumulation correction method according to the present invention;

FIG. 19 is a diagram showing the error before compensation in the embodiment 5 of the method for correcting the cell pressure accumulation according to the present invention;

FIG. 20 is a diagram showing the error after compensation in embodiment 5 of the method for correcting cell pressure accumulation according to the present invention;

FIG. 21 is a schematic diagram showing the structure of an electrochemical test strip used in embodiment 6 of the method for correcting cell pressure accumulation according to the present invention;

FIG. 22 is an exploded view of an electrochemical test strip used in example 6 of the method for correcting cell pressure accumulation according to the invention;

FIG. 23 is a schematic diagram showing the structure of an electrode layer of an electrochemical test strip used in example 6 of the cell size correction method according to the present invention;

FIG. 24 is a schematic diagram showing the connection of the modules of the cell pressure volume calibration device of the present invention.

Reference numerals:

1001, a processor; 1002, a communication bus; 1003, user interface; 1004, a network interface; 1005, a memory; 100, electrochemical test paper; 1, a hydrophilic film layer; 11, air holes; 2, a barrier layer; 21, a siphon window; 3, a single-sided adhesive layer; 31, a reagent window; 311, a first window; 312, a second window; 313, a third window; 314, fourth window; 4, an electrode layer; 41, a reference electrode assembly; 411, a working reference electrode; 412, hct reference electrode; 42, hct working electrode; 43, reaction zone; 431, a first zone; 432, a second region; 433, a third zone; 434, fourth zone; and 5, an insulating layer.

The achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.

Detailed Description

Embodiments of the present invention are described in detail below, wherein like or similar reference numerals refer to like or similar elements or elements having like or similar functions throughout.

Furthermore, the terms "first," "second," and the like, are used for descriptive purposes only and are not to be construed as indicating or implying a relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include one or more such feature. In the description of the present invention, the meaning of "a plurality" is two or more, unless explicitly defined otherwise.

In the present invention, unless explicitly specified and limited otherwise, the terms "mounted," "connected," "secured," and the like are to be construed broadly, and may be, for example, fixedly connected, detachably connected, or integrally formed; can be mechanically or electrically connected; can be directly connected or indirectly connected through an intermediate medium, and can be communicated with the inside of two elements or the interaction relationship of the two elements. The specific meaning of the above terms in the present invention can be understood by those of ordinary skill in the art according to the specific circumstances.

It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.

Fig. 1 is a schematic structural diagram of a hardware operating environment of a terminal according to an embodiment of the present invention.

The cell pressure volume correction system of the embodiment of the invention can be a PC, a mobile terminal device such as a smart phone, a tablet computer or a portable computer, and the like. The cell pressure volume correction system may include: a processor 1001, e.g. a CPU, a network interface 1004, a user interface 1003, a memory 1005 and a communication bus 1002. Wherein the communication bus 1002 is used to enable connected communication between these components. The user interface 1003 may include a display screen, an input unit such as a keyboard, a remote control, and the optional user interface 1003 may also include a standard wired interface, a wireless interface. The network interface 1004 may optionally include a standard wired interface, a wireless interface (e.g., WI-FI interface). The memory 1005 may be a high-speed RAM memory or a stable memory such as a disk memory. The memory 1005 may also optionally be a storage device separate from the processor 1001 described above. Optionally, the cell pressure volume correction system may further include an RF (Radio Frequency) circuit, an audio circuit, a WiFi module, and the like. In addition, the cell pressure volume correction system can be further provided with other sensors such as a gyroscope, a barometer, a hygrometer, an infrared sensor and the like, which are not described herein.

It will be appreciated by those skilled in the art that the cell volume correction system shown in FIG. 1 is not limiting and may include more or fewer components than shown, or certain components may be combined, or a different arrangement of components. As shown in fig. 1, an operating system, a data interface control program, a network connection program, and a cell pressure accumulation correction program may be included in a memory 1005 as one type of computer-readable storage medium.

In a word, on the premise of fully considering the influence of temperature on HCT detection, the invention replaces the alternating current impedance of a single frequency point in the traditional scheme, gives up the single variable compensation method of the formula (1) or the formula (2), adopts a compensation mode of two variables of current and HCT, and utilizes the information of multiple frequency points to eliminate the influence of temperature on the HCT result. And the measurement mode meets the requirement of rapid in-situ detection, can effectively avoid the temperature retardation effect caused by the built-in temperature sensor and the error caused by abnormal operation of a user, ensures accurate HCT detection and correction compensation results, and brings convenience to electrochemical accurate detection of biological samples.

Example 1

Referring to fig. 2, the present embodiment provides a method for correcting a cell pressure volume, including:

Step S100, testing a sample with the electrochemical test paper 100 to obtain a current value, obtaining an HCT measurement value of the sample, and constructing a data set based on the sample.

Step S200, obtaining a corresponding HCT predicted value for each of the samples in the data set based on the HCT measured value.

The electrochemical test paper 100 belongs to a current type sensor based on an electrochemical detection method, has the advantages of small blood sampling amount, high detection speed, high sensitivity, good accuracy, capability of avoiding endogenous substance interference and the like, and can obtain a test result only by 3-30 s, thereby achieving the purpose of instant detection.

As described above, the electrochemical test strip 100 can perform an ac test on a blood sample, and scan at different frequencies under the working electrode and the reference electrode, thereby obtaining a result at each frequency.

The electrochemical test paper 100 is used to test the sample to obtain the corresponding current value, and then the HCT measurement value and the current value are used to establish the data set. The samples of the data set are aimed at the construction and training of the model, and the data set of a large number of samples is established, so that the purpose of enabling the subsequent correction to be more accurate in the process of establishing the model training model can be achieved by collecting a large number of training data.

And step S300, classifying the concentration segments of the detected HCT predicted value and the current value by using a support vector machine according to the data set.

The support vector machine (Support Vector Machine, SVM) is a supervised learning algorithm for classification and regression problems. The basic idea is to find a hyperplane separating different categories of data. Compared with other classification algorithms, the SVM has better generalization performance and higher accuracy, and is good in processing complex data sets.

And aiming at the data set, adopting a support vector machine classifier to classify the concentration section of the measured HCT predicted value and the current value. The support vector machine is a classification algorithm, which can divide data into two or more categories, and has good classification effect.

Step S400, obtaining coefficients of each classification hyperplane corresponding to each concentration section classification, and fitting compensation percentages of the HCT predicted value and the current value in each concentration section to obtain a corresponding compensation equation;

and after the coefficients of the classification hyperplanes corresponding to the classification of each concentration section are obtained through the support vector machine, fitting the compensation percentages of the HCT predicted value and the current value in each concentration section to obtain a corresponding compensation equation. The method comprises the steps of processing training data according to a classification model, obtaining a compensation equation of each concentration section, and correcting a target sample.

And S500, compensating the target sample data according to the compensation equation of the corresponding concentration section, and calculating a corrected concentration value corresponding to the target sample data according to the compensated current.

And compensating the HCT predicted value of the target sample data according to a compensation equation of the corresponding concentration section, and calculating a corrected concentration value corresponding to the target sample according to the compensated current. This step is to apply the compensation equation obtained above to the actual sample data, resulting in a more accurate concentration value.

And (3) taking each sample into a compensation equation of the current concentration section, so as to calculate the compensation percentage required by the current point and compensate, finally taking the compensated current value into a code (the relation between the reference current and the reference concentration) and calculating the final correction concentration value, and completing the correction.

After the compensated current value is obtained, the final concentration value of the target sample data needs to be obtained by carrying into the code calculation.

In a biochemical detection device, for example, a glucometer, a code represents a calculation equation of a concentration value obtained by generating a current to a test paper by the glucometer, and represents a relationship between the current value and a corresponding concentration value, so as to represent a concentration of glucose in a detected blood sample. The code is converted from a current signal measured by an optical sensor or an electrochemical sensor used in the device.

The code is calculated by reacting electrochemical test paper with blood sample to generate electric signal, and then converting the electric signal into blood sugar concentration value by biochemical detection equipment according to a certain algorithm. Different algorithms and calculation modes can be adopted for obtaining codes for different brands and models of blood glucose meters.

In addition, each batch of blood glucose strips has its own corresponding code, as there may be minor differences between batches of strips, resulting in different responses and readings to current, i.e., the labels or numbers of the batches of test electrochemical strips or the test electrochemical strips themselves correspond to the codes. Therefore, in order to obtain accurate and reliable test results, the same batch of test strips needs to be used in successive tests.

In this embodiment, the obtained compensated current value is brought into a code relational expression corresponding to the electrochemical test paper, and a corrected concentration value corresponding to the target sample data can be obtained through calculation.

In this embodiment, on the premise of fully considering the influence of temperature on HCT detection, the influence of temperature on HCT results is eliminated by using information of multiple frequency points. And the measurement mode meets the requirement of rapid in-situ detection, can effectively avoid the temperature retardation effect caused by the built-in temperature sensor and the error caused by abnormal operation of a user, ensures accurate HCT correction result, and brings convenience for electrochemical accurate detection of biological samples.

Example 2

Referring to fig. 3, embodiment 2 of the present invention provides a method for correcting a cell volume, which is based on embodiment 1. In the step S100, constructing a data set includes:

step S110, collecting samples in different HCT ranges and different temperature ranges, performing alternating current test, and recording real part information and imaginary part information under each frequency;

as mentioned above, samples of different HCT ranges and different temperature ranges are used, the purpose of which is to achieve a guaranteed representation of the data set.

The HCT ranges are different, for example, hct=42% (about 10% to about 65%);

the ac test was performed on samples having different temperature ranges, for example, 23 degrees (about 10 degrees to about 40 degrees).

The ac test is performed as described above, and the electrochemical test paper 100 is used to detect the blood sample, thereby obtaining data.

In the above-mentioned alternating current test, both the real part information and the imaginary part information refer to a numerical attribute of the measurement result. The impedance response of the substance to be measured to the current is reflected for describing various parameters of the electrical properties.

Wherein the real part information, usually expressed as resistance (in ohms), reflects the degree of obstruction of the current by the material. Whereas the imaginary information is usually expressed as inductance or capacitance (in henries or farads) reflecting the phase effect of the material on the current.

In ac testing, the purpose of obtaining real and imaginary information at each frequency is to determine the electrical characteristics of the sample at different temperatures and HCT ranges from these data, so as to better understand the electrical properties of the sample and optimize its design and production process.

Further, referring to fig. 4, in step S110, samples of different HCT ranges and different temperature ranges are collected and ac test is performed, and real part information and imaginary part information of each frequency are recorded, including:

step S111, alternating current scanning is carried out on the sample under different frequencies based on a working electrode and a reference electrode;

step S112, recording the real part information and the imaginary part information at each frequency.

The above-mentioned scanning of different frequencies for the blood sample is performed by using the working electrode and the reference electrode of the electrochemical test strip 100, wherein the test conditions may be: the scanning range is 10Hz-30KHz, and the amplitude of the applied alternating current is about 20mv-50 mv.

The real and imaginary information at each frequency is then recorded, wherein the amount of data for the frequency can be specifically selected depending on the performance of the ac module of the machine, e.g. the minimum number required in this embodiment is 2.

Further, in the step S112, after recording the real part information and the imaginary part information of each frequency, the method further includes:

step S113 of converting values of the real part information and the imaginary part information into impedance magnitude and phase angle as the converted real part information and imaginary part information;

；（3）

) ；（4）

In this embodiment, the values of the real part and the imaginary part are converted into the impedance magnitude and the phase angle for further processing of the subsequent data, and the specific conversion relationships are shown in the following formulas (3) and (4).

For example, the real and imaginary parts of the measurement at 200Hz are [6108.97, 15815.3], translated to [16954.1, 68.8799]; the real and imaginary parts at 30KHz are [1194.19, 264.263], converted to [1223.08, 12.4779]; combining features at multiple frequency bins, e.g., combining 200Hz and 30KHz, can yield vector features of [16954.1, 68.8799, 1223.08, 12.4779] size 1 x 4. Similarly, if the frequency points tested are m, each sample obtains a vector with a feature size of 1×2×m.

Step S120, obtaining HCT measured values and steady-state temperatures of each sample, and combining data information of all the samples;

and step S130, establishing the data set according to the data information.

The HCT measurement value refers to the ratio of blood cells in a sample, and is generally used for diagnosing diseases such as anemia, leukocytosis, and infection. The blood sample may be obtained for measurement by finger prick or blood drawing.

For example, the measurement method thereof may be: a certain amount of whole blood is sucked by a glass capillary tube for centrifugation, the centrifugation parameter is 4000rpm, the centrifugation is carried out for 3min, and then the total volume is the HCT value on a comparison card by using the volume ratio of red blood cells.

Steady state temperature refers to the temperature at which the sample reaches a constant state after heating or cooling. In this embodiment, an accurate thermometer (qualified measurement) is used to ensure the accuracy of the temperature.

For example, a sample is at a temperature of 10 ℃, hct=53%.

The data information is the sum steady-state temperature of each sample. Combining the data information of all samples, a certain amount of data sets can be obtained, which can be used to construct a machine learning model to predict the electrical performance characteristics at unknown HCT and steady-state temperatures.

According to the method, a series of samples are acquired, so that the data volume of the samples in the data set can be kept at about 100, and the information of all the samples is combined.

For example, if the number of samples n=100, the size of the data set X is 100×2×m, the size of Y is 100×2, and the two rows of samples are in one-to-one correspondence.

Further, in the step S100, the establishing the data set according to the data information includes:

step S140, dividing the data information into a training set and a testing set according to a preset proportion;

to ensure the representativeness of the model, the data sets X and Y are divided into training sets and test sets in a certain proportion before data processing. The training set is used for finding out the optimal model parameters, and the testing set is used for evaluating whether the model parameters of the training set are reasonable.

For example, the training set and the test set are randomly divided in a ratio of 7:3, specifically, 70 samples are randomly extracted as the training set, 30 samples are randomly extracted as the test set, and in order to ensure the randomness and the representativeness of the samples, hierarchical sampling is adopted, that is, the ratio of each class of HCT in the training set to the ratio of each class of temperature sample to the ratio in the test set is equivalent.

For example, if the duty cycle of hct=55% in the training set is 30%, the duty cycle of hct=55% in the test set should be kept around 30%. Thus, by this step, X and Y are divided into four data sets, X_train, Y_train, X_test, Y_test, respectively. Wherein the X_train and Y_train lines are in one-to-one correspondence, the sizes are (70,2 ×m), (70,2), and the X_test and Y_test lines are in one-to-one correspondence, and the sizes are (30, 2×m), (30, 2).

And step S150, carrying out unified standardization processing on the training set and the test set.

the unified normalization processing formula is as follows:

；（5）

The above-mentioned formula conversion according to formula (5) ensures that the mean value of each feature in the dataset is 0 and the standard deviation is 1.

The above formula (5) is used to perform unified normalization on the HCT measurement value and the current value of each sample in the data set, that is, the HCT measurement value is carried into formula (5) to obtain a HCT measurement value corresponding to the unified normalization, and the current value is carried into formula (5) to obtain a current value corresponding to the unified normalization.

For example, the real part of X_train at 30KHz is distributed at 1200-17000, the phase angle is distributed at 12-68, the normalized mean value of the two is 0, and the standard deviation is 1.

The operation of unified normalization processing is to ensure that the variables are on the same number level, and features with larger orders of magnitude do not affect features with smaller orders of magnitude.

After unified standardization processing, the data in the data set can realize that the variables are in the same quantity level, and the characteristics with larger orders of magnitude can not influence the characteristics with smaller orders of magnitude.

Example 3

Referring to fig. 5, embodiment 3 of the present invention provides a method for correcting a cell volume, based on embodiment 1 above, the step S200 of obtaining a corresponding HCT predicted value for each of the samples in the data set based on the HCT measured value includes:

step S210, constructing a prediction model based on a nonlinear iterative partial least square method, and calculating the HCT predicted value of each sample according to the HCT measured value.

It should be noted that the nonlinear iterative partial least squares method (Nonlinear Iterative Partial Least Squares, abbreviated as nipars) is a data analysis method, and is particularly suitable for multivariate statistical analysis in the field of bioinformatics. It is a specific implementation of partial least squares regression (Partial Least Squares Regression, PLS for short).

The NIPALS is used for continuously and iteratively updating the estimated value of the regression vector on the basis of the partial least square algorithm until convergence. This approach is typically applied to dimension reduction of high-dimensional data in order to better understand and interpret the correlation in the data.

The nonlinear iterative partial least square method is used as a predictive modeling method, and the main purpose and function of the nonlinear iterative partial least square method are to build a mathematical model capable of predicting a target variable by analyzing and learning data. In the method of correcting the cell pressure product in the present embodiment, the purpose of using the nonlinear iterative partial least squares method is to predict the HCT predicted value of each sample.

In a word, the nonlinear iterative partial least square method is adopted to fully utilize the information characteristics of each frequency point in the data in the cell pressure product prediction, and a more accurate and effective prediction model is established.

Further, the step S210 is configured to construct a prediction model based on a nonlinear iterative partial least square method, and calculate the HCT predicted value of each sample based on the HCT measured value, and includes:

step S211, performing factor decomposition on the HCT measured value in the data set through the prediction model by using the nonlinear iterative partial least square method to obtain regression coefficients of the independent variable matrix and the dependent variable matrix after decomposition, and then determining the HCT predicted value;

In the step S211, the obtaining the regression coefficient includes:

step S2111, determining the number of principal components according to the training set and the testing set of the data set;

and step S2112, taking the number of principal components as an input parameter, and obtaining an HCT predicted value through the prediction model.

In step S211, the median value of the data set is first factorized by the prediction model by using the nonlinear iterative partial least square method to obtain regression coefficients of the independent variable matrix and the dependent variable matrix after decomposition, and then the HCT predicted value is determined;

specifically, obtaining regression coefficients includes two steps:

step S2111: the number of principal components is determined from the training set and the test set. This step is mainly to determine how many principal components need to be preserved in order to be able to interpret most of the information in the original data.

Step S2112: based on the prediction model, taking the number of principal components as a reference, it is required to explain that the iterative decomposition of the X matrix and the Y matrix takes into consideration that all X feature dimensions are 2×m (where m is the frequency of the alternating current test), and the regression coefficient matrix which is the best superior to the Y space is obtained through continuous iterative decomposition by adopting the NIPALS algorithm, namely the nonlinear iterative partial least square method.

In a specific implementation manner, the plsregressions function in the sklearn function library can be directly called to realize the function, and the main reference of the function is n_components, namely, a reasonable number of principal components are required to be input to interpret the information of the X space.

The selection of the main component can be obtained by combining the information duty ratio graph and the score of the test set, and when the duty ratio reaches more than 90%, the variable obtained by the representative decomposition can approximate to the whole system.

Referring to fig. 6, as the main component is continuously increased, the information duty ratio is gradually increased, and when the main component is around 10, the steady state is substantially reached, so that the main component number can be controlled within substantially 10, and when the main component is between 1 and 7, the main component is substantially in an increased state, so that the main component number can be determined to be around 5 to 8. More accurate principal component scores may then be available through cross-validation or direct evaluation of the scores of the test set.

Referring to fig. 7, when different numbers of principal components are selected, the scores of the training set and the test set change, when the number of the components is 5, the model has certain under fitting, and when the number of the principal components is 8-10, the model has over fitting, so in order to ensure the generalization capability of the model, the number of the principal components can be 6. When the principal component number of the final model is set to 6, and the predicted effects of the training set and the test set on HCT are respectively: r2_score= 0.9932, r2_score= 0.9911.

The effect of specific predictions can be seen with reference to fig. 7 above, from which it can be seen that the current model has good predictive power for HCT. Even if the temperature difference is large, the absolute error of the HCT prediction is controlled within 5 percent, thereby completely meeting the existing use requirements.

Example 4

Referring to fig. 8, embodiment 4 of the present invention provides a method for correcting a cell pressure volume, based on embodiment 1 above, the step S300 of classifying the concentration segments of the measured HCT predicted value and the measured current value by using a support vector machine according to the data set includes:

step S310, a classifier is established based on the support vector machine;

step S320, obtaining a classification equation coefficient through the classifier, and obtaining an equation slope and an intercept of the classifier, thereby constructing a classification model;

and step S330, inputting the HCT predicted value and the current value measured by each sample in the data set into the classification model to obtain a concentration segment classification result.

In this embodiment, the single variable compensation method of the formula (1) or the formula (2) is abandoned, and the compensation method of two variables of the current and HCT is adopted. In particular, as shown in fig. 9, the x-axis in fig. 9 is a different HCT distribution, the y-axis is that the Cottrell current measured under the HCT is equivalent to the standard HCT (hct=43+++ -2): percent change in current. From this figure, it is possible to obtain: when the creatinine concentration is around (1) a=50 μmol/L (concentration=1, low concentration), (2) a=400 μmol/L (concentration=2, medium concentration), (3) a=800 μmol/L (concentration=3, high concentration), the effect of HCT on the current at 8% -65% can be seen: the compensation percentage gradually becomes smaller with the increase of HCT, and the change trend of each concentration is not consistent, and the high HCT interval of the low concentration section and the change percentage of the low HCT interval of the high concentration section are obviously different from other concentration sections. This shows that significant deviations can be introduced if a uniform compensation is used for all concentrations.

Therefore, in the operation of HCT for current compensation, it is necessary to perform a process of segmenting the concentration segment, that is, to first perform accurate concentration classification, and then input a corresponding compensation coefficient for each type through steps S310 to S330. This will increase the accuracy of the overall compensation model.

In step S310, a classifier is built based on the support vector machine, which is to create a model for classifying tasks by using the support vector machine algorithm. The main purpose of this is to train a classifier with a known sample data set so that the classifier can accurately classify new unknown data. Specifically, in this step, the support vector machine classifier is trained using the training data set that has been labeled for categories, thereby learning how to assign the unknown data into different categories. During training, the support vector machine algorithm finds an optimal hyperplane to distinguish between the classes and maximize the distance of the support vector from it as much as possible. By adjusting model parameters, the support vector machine can adapt to different data sets and classification tasks, thereby providing accurate and reliable classification results.

In step S320, the classification equation coefficient is obtained by the classifier, and the equation slope and intercept of the classifier are obtained, so as to construct a classification model. In this step, the classification equation coefficient, equation slope and intercept are obtained based on the support vector machine classifier trained in the previous step, and are used for constructing the classification model. In particular, a support vector machine classifier typically outputs a decision function that predicts its class based on the eigenvalues of the input data. From the form of the decision function, the coefficients, slopes and intercepts of the classification equation can be calculated and then used to construct a classification model. For example, if a support vector machine classifier has been trained, the coefficients, slope and intercept of the classification equation can be solved from the decision function of its output, and these parameters can then be used to construct a classification model.

In step S330, the HCT predicted value and the current value measured for each sample in the data set are input into the classification model to obtain a concentration segment classification result. In this step, the unknown data in the data set will be classified using the already constructed classification model. Specifically, HCT values and current values measured by each sample in the data set are input into a classification model, so that a concentration segment classification result to which the sample belongs is obtained.

The steps have the advantages that the characteristics and the rules of the model can be learned through the data set, different data sets and classification tasks can be automatically adapted, and a highly accurate and reliable classification result is provided. In addition, the support vector machine algorithm has strong generalization capability, and can effectively solve the problems of high-dimensional data and complex classification.

Further, the classifier in the step S310 is a two-classifier based on a linear kernel function, and each of the two classifiers is a linear function; the equation for the linear function is as follows:

；（6）

；（7）

It should be noted that, since the same creatinine concentration generates different current signals I at different HCTs, the concentration segments are divided only by the difference of the currents I, which may generate erroneous classification results.

Specifically, the concentration was 894.8 μmol/L, hct=60% generated current i1= -160nA and 458.3 μmol/L, hct=20% generated current i2= -153.9nA were almost identical, but the two were in different concentration ranges (but the former was high concentration and the latter was medium concentration).

Therefore, considering the effect of HCT on current, in this embodiment, the two dimensional changes of HCT and current I are used to achieve more accurate concentration classification.

Specifically, the principle of the classification model is given in the following formula (8):

classification result = f (HCT, dc); (8)

The specific form of f in equation (8) should be a piecewise function, such as:

(1) When HCT is <30% and dc > -20nA, the classification result is low concentration;

(2) When HCT >50% and dc < -200nA, the classification result is medium concentration.

Thus (30%, -20 nA) and (50%, -200 nA) are both some points of the classification parameter, e.g. the classification equation is a linear equation, and the above information is the point on this linear equation.

In this embodiment, a support vector machine is used to obtain the classification equation, because the SVM has a good generalization capability on a small sample data set, and an optimal classification plane is found by maximizing the distance from the classification hyperplane to the support vector. In the kernel function selection of the support vector machine, a linear kernel function is selected due to the fact that specific feature distribution and later generalization capability are considered, namely each two-classifier is a linear equation, and specific forms are shown in the formula (6) and the formula (7). Wherein, in the formula (6), HCT and current I of an input sample are shown, and y is output to obtain a classification result; can be simply positive with more than or equal to 0 and negative with less than 0. The above procedure will obtain the functional form of one of the two classifiers and a specific decision flow.

Considering multi-classification (i.e., the piecewise function form of f), there are a plurality of equations similar to the equations of the formulas (9) and (10), in this embodiment, 3 concentration segments are considered, a plurality of two classifiers are built to realize the multi-classification splitting task, the splitting strategy adopts a similar idea of a pair of redundant (OVR), and N-1 classifiers can be used to realize the N-classification operation. Specifically, n=3 (classification of 3 concentrations) in this embodiment can be implemented by only 2 classifiers, and the implementation principle is as shown in fig. 10:

as can be seen from fig. 10, by constructing four data sets { C1, C2} and { C2, C3}, two classifiers f1 and f2 are obtained, and then the sample is brought into each classifier to obtain a corresponding classification result; for example, (1) if the result of each classifier is [ +, + ], the corresponding class is the first class; (2) [ -, + ] is of the second type and (3) [ -, - ] is of the third type. The method has the advantage of being very convenient for the compatibility of the later embedded machine.

Further, referring to fig. 11, the step S320 of obtaining the classification equation coefficient by the classifier and obtaining the equation slope and intercept of the classifier, thereby constructing a classification model includes:

Step S321, calling svm.SVC functions in the scikit-learn function library based on the Python learning library;

step S322, setting the kernel parameter of the svm.SVC function type as a linear function;

step S323, obtaining the classification equation coefficient through the classification model, and calculating the equation slope and the intercept.

When it is desired to classify a set of data or each data in a data set, step S320 refers to training a model using a classifier algorithm and obtaining equation coefficients of the model. This step is an important step in machine learning in the present example for generating a model that can correctly classify the input data.

In step S321, the svm.svc function of the scikit-learn function library is called by the learning library of Python to construct the classifier model. SVM (support vector machine) as a support vector machine algorithm, the SVC function of which is one implementation in a scikit-learn library.

In step S322, the kernel parameter of the svm.svc function is set as a linear function. The Kernel function describes how the SVM algorithm classifies data in a high-dimensional space. A linear function is one of the Kernel functions that can be used to classify linearly separable data.

In step S323, the classification equation coefficient is acquired using the constructed classification model, and the slope and intercept of the equation are calculated. This equation can be used to classify new data points into the correct categories.

Through the above steps, training a Support Vector Machine (SVM) model may be achieved to facilitate use in classifying given data. This model will automatically learn how to assign new data points into the corresponding categories based on the learning set data provided.

The SVM model using a linear Kernel function has the following advantages:

(1) For linearly separable data, the SVM model can quickly converge and find the optimal solution.

(2) The SVM model can handle high-dimensional data and in most cases exhibits a better generalization ability, i.e. it is able to accurately classify new data.

(3) The SVM model also has some adaptability to nonlinear problems, and core skills can be used to map the original features into a high-dimensional space, so that the nonlinear problems become linearly separable problems.

In summary, by using an SVM model and a linear Kernel function, a high accuracy classification result can be obtained, and the model has a good generalization capability.

Further, in the step S323, the classification equation coefficients include coef and interscept;

the classifier comprises f1 and f2;

the equation for the equation slope and intercept is calculated as:

；（9）

]；（10）

The above-mentioned specific data, named x_sub and y_sub, may be selected from the data set by y=1 or y=2, and the data set x_sub is normalized by using the standard coefficient, so as to obtain x_sub_std. X_sub_std is made up of hct_std and i_std, then call svm.svc functions in sklearn function library, and kernel= 'linear' (with linear kernel function).

Then, the direct fit (x_sub_std, y) obtains the classification equation coefficient coef_ (vector magnitude is 2) and the intercept (vector magnitude is 1), and further obtains the equation slope k and the intercept b by the following equation (9) and equation (10), that is, "w×in equation (6)".

In the above formula, k and b may be coefficients of the function f1 and/or the function f2, and thus, the form of the obtained f1 is: i_std= -0.1774 hct_std+ -1.0004, the form of f2 obtained is: i_std= -0.6148 hct_std+ 0.9086.

Further, referring to fig. 12, the step S330 of inputting the HCT predicted value and the current value measured by each sample in the data set into the classification model for prediction, to obtain a concentration segment classification result includes:

Step S331, bringing a training set and a testing set in a data set into classification equations of the classifiers f1 and f2 of the classification model to obtain corresponding classification discrimination values;

step S332, obtaining a corresponding one-dimensional vector according to the classification discrimination value;

step S333, logically dividing the one-dimensional vector to obtain a corresponding concentration segment classification result;

and step 334, evaluating the concentration segment classification result by adopting a confusion matrix evaluation method to finish classification.

The data in the training set amount test set in the data set is predicted by using the two classifiers f1 and f 2.

The Confusion matrix evaluation method (fusion matrix) is a classification model performance evaluation method. The model prediction result is compared with the real label in a form of a table, so that the classification accuracy and the misjudgment condition of the model are measured. In the confusion matrix, the rows represent the true labels, the columns represent the model predictions, and each element of the matrix represents the number of samples of the classification.

The confusion matrix typically includes four indices: true (1 Positive, TP), false (0 Negative, FN), false (0 Positive, FP) and true (1 Negative, TN). Where TP represents the number of positive samples that the model correctly classifies as positive samples; FN represents the number of positive samples that the model incorrectly classifies as negative samples; FP represents the number of negative samples that the model incorrectly classifies as positive samples; TN represents the number of negative samples that the model correctly classifies as negative. According to the indexes, evaluation indexes such as accuracy, recall rate, F1 value and the like can be calculated, and classification performance of the model can be comprehensively known.

In the step S331, the HCT predicted value and the current value measured by each sample in the training set are input into the classification equations of the classifiers f1 and f2 of the classification model to obtain the corresponding classification discrimination value. This step is to input the features in the training dataset into the classification model, and calculate the classification discrimination value of each sample by the classification equations of the classifiers f1 and f2, which can be used for the subsequent classification operation. For example, if a Support Vector Machine (SVM) classifier is used, the classification equation may be expressed as a linear or nonlinear function and find an optimal hyperplane in the feature space to segment samples of different classes.

In step S332, a corresponding one-dimensional vector is obtained from the classification discrimination value. And processing according to the classification discrimination value obtained in the previous step to obtain a one-dimensional vector, wherein the one-dimensional vector contains classification information of all samples. For example, in an SVM classifier, the classification decision value may be expressed as a real number, if the value is greater than zero, indicating that the sample belongs to a positive class, otherwise belongs to a negative class.

In step S333, the one-dimensional vector is logically divided to obtain a corresponding concentration segment classification result. Each sample in the training dataset is divided into different concentration segments according to the value of the one-dimensional vector. For example, the samples with zero or more are above the hyperplane (positive samples) and the samples with zero or less are below the hyperplane (negative samples) by performing logical division based on the real values obtained by the SVM classifier.

In step S334, the concentration segment classification result is evaluated by using the confusion matrix evaluation method, and classification is completed. And (3) evaluating the classification result by using a confusion matrix evaluation method, and calculating indexes such as accuracy, recall rate, F1-score and the like of the classification model. For example, the number of positive class prediction errors (true examples TP), the number of positive class prediction errors (false negative examples FN), the number of negative class prediction errors (false positive examples FP) and the number of negative class prediction errors (true negative examples TN) can be calculated by the confusion matrix, so that indexes such as model accuracy and recall rate are calculated to evaluate the performance of the model.

Through the method, based on the classification equation of the classifiers f1 and f2, the concentration segment classification result is evaluated by adopting the confusion matrix evaluation method, so that the performance of the model can be comprehensively evaluated. Finally, the steps can obtain a classification model established based on the training data set, and classify the new sample by using the model, thereby realizing automatic classification of the sample concentration section.

Further, referring to fig. 13, in step S332, a corresponding one-dimensional vector is obtained according to the classification discrimination value, which includes:

and S3321, adopting 2-system coding, and converting the classification discrimination values according to the principle that the classification discrimination values are greater than 0 and equal to or less than 0 and 0 to obtain the 2-bit 2-system one-dimensional vector corresponding to each classification equation.

The specific steps of the method can be as follows:

(1) The original input values were [ hct=8%, i= 25.2024uA ], normalized to: [ -1.4962, 2.4976];

(2) Then respectively taking into the equations corresponding to f1 and f2 to obtain [ -3.14579, -2.451980]2 classification discrimination values;

(3) 2 discrimination values are coded by a 2-system code (more than 0 is 1 and less than 0 is 1), so that 2-bit 2-system one-dimensional vectors [0,0] corresponding to 2 classification equations are obtained;

(4) Based on the logical division of the three-classification implementation principle in fig. 10, it can be determined that the current sample is C3.

And by analogy, all classification results of the current test set samples can be obtained. The performance decisions for the above results were evaluated using a confusion matrix, as shown in tables 1 and 2 below. Also the graphically classifying effect of f1 and f2 is shown in fig. 14.

TABLE 1 training set confusion matrix table

TABLE 2 test set confusion matrix table

From fig. 14, and the data and related contents in tables 1 and 2, it is shown that macro-average (macro-avg) of both data sets can be maintained at 0.95 or more, indicating excellent classification effect of the model. Can meet the current use requirement.

Example 5

Referring to fig. 15, embodiment 5 of the present invention provides a method for correcting a cell pressure volume, based on embodiment 1 above, the step S400 of obtaining coefficients of each classification hyperplane corresponding to each concentration segment classification, and fitting the compensation percentages of the HCT predicted value and the current value in each concentration segment to obtain a corresponding compensation equation, including:

Step S410, recording coefficients of each classification hyperplane in the concentration segment classification;

and step S420, based on the coefficients of the classification hyperplanes, using the measured HCT predicted value as an abscissa and using the current percentage of the current value to be compensated as an ordinate, and adopting a linear equation to fit the relation between the measured HCT predicted value and the current percentage of the current value to be compensated to obtain the compensation equation corresponding to each concentration section.

In step S410, coefficients of the respective classification hyperplane in each concentration segment classification are recorded, which coefficients are to be used for calculating the relevant variables in the compensation equation.

For example, assume that there are 3 different concentration segments: a low concentration section, a medium concentration section, and a high concentration section. For each concentration segment, its corresponding classification hyperplane coefficient is recorded for use in subsequent calculations.

Step S420 is to fit the relationship between the measured HCT value and the percentage of current to be compensated based on the respective classified hyperplane coefficients. In this step, a linear equation is used to fit the relationship between the measured HCT value and the percentage of current to be compensated and to derive a compensation equation for each concentration segment, i.e. the compensation equation.

For example, assume that a set of data is obtained in a first concentration segment, where HCT has a value of A, the measured current is B mA, and the reference current is C mA. In the compensation process, the interference influence in the concentration segment can be calculated by using the corresponding classified hyperplane coefficient stored in the step S310, and the current compensation percentage required to be performed can be calculated by using the linear equation in the step S320, so as to eliminate the influence of the interference.

By recording the classified hyperplane coefficient of each concentration section and establishing a compensation equation, the influence of the interferents on the measurement result can be effectively eliminated, and the measurement accuracy and precision are improved. Finally, through the steps, a complete compensation equation is obtained and can be used for correcting the concentration measurement result of the to-be-detected object after the influence of the interfering object.

Further, in the step S420, the compensation equation corresponding to each concentration segment is:

；（11）

wherein ,a compensation percentage which is the relative deviation of the measured current value and the reference current; />Is a coefficient; />A predicted value for the HCT; />Is the intercept.

In the above, in each divided concentration section, the measured HCT is taken as the abscissa, the current percentage to be compensated is taken as the ordinate, and the relationship between the two is fitted by adopting the linear equation, so that the HCT compensation equation corresponding to each concentration section can be obtained and recorded.

Referring to fig. 16, 17 and 18, linear equations and corresponding fitting degrees in the concentrations 1, 2 and 3 are shown, and each concentration segment compensation equation takes the form in the formula (11).

Further, each sample is brought into a compensation equation of the current concentration section, the compensation percentage required by the current point is calculated and compensated, and finally, the compensated current is brought into a code (a relation between the reference current and the reference concentration) and the final concentration value is calculated.

The effects before and after compensation are shown in fig. 19 and 20. As can be seen from fig. 19 and 20, the effect of different HCTs on the current can be effectively controlled within ±20% and most samples can be controlled within ±15% by the above-mentioned compensation method. This way of compensation is explained to be effective. Compared with a uniform concentration compensation mode, the method reduces the MARD of the current from 7.25% to 5.85%, which shows that the concentration section compensation can effectively improve the compensation accuracy.

Example 6

Referring to fig. 21, embodiment 6 of the present invention provides a method for correcting a cell volume, based on embodiment 1 above, in the step S100, the electrochemical test strip 100 is used comprising:

Hydrophilic film layer 1, barrier layer 2, single-sided adhesive layer 3, electrode layer 4 and insulating layer 5;

the hydrophilic membrane layer 1 is provided with air holes 11;

a siphon window 21 is arranged on the barrier layer 2;

one end of the single-sided adhesive layer 3 is provided with a reagent window 31;

the siphon window 21 corresponds to the position of the reagent window 31 and can coincide when combined;

one end of the single-sided adhesive layer 3, provided with a reagent window 31, coincides with one end of the electrode layer 4;

the electrode layer 4 is attached to the insulating layer 5;

the hydrophilic film layer 1, the barrier layer 2, the single-sided adhesive layer 3, the electrode layer 4 and the insulating layer 5 are sequentially arranged from top to bottom and are mutually attached.

Referring to the explosion diagram of the electrochemical test strip 100 in fig. 22, the structure of the electrochemical test strip 100 includes five parts, namely a hydrophilic film layer 1, a barrier layer 2, a single-sided adhesive layer 3, an electrode layer 4 and an insulating layer 5. And, five parts of the hydrophilic film layer 1, the barrier layer 2, the single-sided adhesive layer 3, the electrode layer 4 and the insulating layer 5 are sequentially arranged from top to bottom, and are mutually attached to each other two by two to form the electrochemical test paper 100.

The reagent window 31 may be a single window or a plurality of windows for setting different solvents and detection substances, so as to achieve more detection functions.

The size of the siphon window 21 may be matched to the size of the reagent window 31 at the lower end, so as to enable the siphon window 21 to correspond to the position of the reagent window 31 and to coincide when combined.

For example, when the reagent window 31 is formed in a plurality of lines, the siphon window 21 may have a long shape.

The electrochemical test paper 100 provided in the embodiment belongs to a current sensor based on an electrochemical detection method, has the advantages of small blood sampling amount, high detection speed, high sensitivity, good accuracy, capability of avoiding endogenous substance interference and the like, and can obtain a test result only by 3s-30s, thereby achieving the purpose of instant detection. In addition, the electrochemical test paper 100 provided by the embodiment has simple reaction and low reagent cost, and is beneficial to industrialized popularization and application.

Further, the number of the reagent windows 31 on the single-sided adhesive layer 3 is 4, which are respectively:

a first window 311, a second window 312, a third window 313, and a fourth window 314.

Further, the electrode layer 4 is provided with a reaction zone 43 corresponding to the reagent window 31 of the single-sided adhesive layer 3;

the reaction zone 43 comprises: a first region 431, a second region 432, a third region 433 and a fourth region 434;

The first region 431 carries a first reaction enzyme solution; the second zone 432 carries a second reactant enzyme solution; the third zone 433 carries a non-reactive substance; the fourth zone 434 is free of the application of the reactive enzyme solution and the non-reactive species.

For example, the first reactive enzyme solution includes: creatine, sarcosine oxidase, horseradish peroxidase and ascorbate oxidase;

the second reaction enzyme solution comprises: creatinine amide hydrolase, creatinase, sarcosine oxidase, horseradish peroxidase, and ascorbate oxidase;

the non-reactive material includes: a thickener, a protectant, and a surfactant;

the first region 431, the second region 432 and the third region 433 are sequentially arranged, and are adjacent to each other.

Further, referring to fig. 23, the electrode layer 4 includes: the reference electrode assembly 41 and the HCT working electrode 42, and may further include a test paper type judgment electrode and an insertion judgment electrode;

wherein the reference electrode assembly 41 comprises a working reference electrode 411 and an HCT reference electrode 412;

the reference electrode assembly 41 and the HCT working electrode 42 are electrically connected to a signal acquisition module of the cell pressure and volume correction system when they are connected to the cell pressure and volume correction system.

Further, the electrode layer 4 includes at least one of the following features:

A. the electrode layer 4 is made of a metal thin layer or printing carbon paste;

B. the thickness of the metal thin layer in the electrode layer 4 is 1nm-50nm; for example, it may be 1nm, 5nm, 10nm, 15nm, 20nm, 25nm, 30nm, 35nm, 40nm, 45nm, 50nm or the like;

C. the metal in the metal thin layer in the electrode layer 4 is an alloy formed of any one or a combination of several of gold, platinum, palladium, nickel or titanium.

In addition, in the present embodiment, regarding the electrochemical test strip 100, the detection principle related to the detection may be specifically as follows:

(1) In a sample to be tested, endogenous creatine generates sarcosine under the action of creatine enzyme, and then the sarcosine generates H under the action of sarcosine oxidase ₂ O ₂ At the same time, endogenous ascorbic acid further produces H under the action of novel ascorbic acid oxidase ₂ O ₂ H generated in two parts ₂ O ₂ Under the action of horseradish peroxidase (HRP), an oxidation-reduction type electron mediator is subjected to negative excitation potential to obtain a current signal W1;

(2) Creatinine of an object to be detected in the sample to be detected sequentially generates H under the action of creatinine amide hydrolase, creatinase and sarcosine oxidase ₂ O ₂ At the same time, endogenous Ascorbic Acid (AA) further generates H under the action of novel ascorbic acid oxidase (ASO-3) ₂ O ₂ H generated in two parts ₂ O ₂ Under the action of horseradish peroxidase (HRP), an oxidation-reduction type electron mediator is subjected to negative excitation potential to obtain a current signal W2;

(3) The accurate creatinine concentration in the sample can be obtained through the signal value difference W obtained by the W2-W1. And then, the current signal of creatinine is combined with the HCT prediction result to compensate, so that the influence of the hematocrit on the current signal can be eliminated, and the more accurate creatinine concentration can be obtained.

In this embodiment, the detection principle of the electrochemical test strip 100 includes the following chemical reactions:

（1）；

（2）；

（3）；

（4）；

（5）；

（6）。

in a preferred embodiment, the mass concentration of creatinine amide hydrolase in the second reaction enzyme solution is 5% -10%, and the mass ratio of creatinine amide hydrolase, creatinase, sarcosine oxidase, horseradish peroxidase (HRP), and ascorbate oxidase (ASO-3) in the second reaction enzyme solution may be: 5 (1-4): 2-3): 2-5): 0.1-0.6; the first reaction enzyme solution reduces only the creatinine amide hydrolase component compared to the second reaction enzyme solution.

In addition, referring to fig. 24, the present invention also provides a cell pressure volume correction device, including:

A data module 10 for testing a sample with electrochemical test paper to obtain a current value, obtaining an HCT measurement value of the sample, and constructing a data set based on the sample;

a prediction module 20 for deriving a corresponding HCT predicted value for each of the samples in the data set based on the HCT measured values;

the classification module 30 is configured to classify the concentration segments of the measured HCT predicted value and the measured current value by using a support vector machine according to the data set;

the calculation module 40 is configured to obtain coefficients of each classification hyperplane corresponding to each concentration segment classification, and fit the compensation percentages of the HCT predicted value and the current value in each concentration segment to obtain a corresponding compensation equation;

the correction module 50 is configured to compensate the target sample data according to the compensation equation corresponding to the concentration segment, and calculate a corrected concentration value corresponding to the target sample data according to the compensated current.

In addition, the invention also provides a cell pressure accumulation correction system, which comprises a signal acquisition device, a memory and a processor; the signal acquisition device acquires signals returned by the reference electrode component and the HCT working electrode in the electrochemical test paper electrode layer; the memory stores a cell pressure accumulation correction program; the processor runs the cell pressure volume correction program to cause the cell pressure volume correction system to perform the cell pressure volume correction method as described above.

Furthermore, the present invention provides a computer-readable storage medium having stored thereon a cell pressure volume correction program which, when executed by a processor, implements a cell pressure volume correction method as described above.

The foregoing embodiment numbers of the present invention are merely for the purpose of description, and do not represent the advantages or disadvantages of the embodiments.

From the above description of the embodiments, it will be clear to those skilled in the art that the above-described embodiment method may be implemented by means of software plus a necessary general hardware platform, but of course may also be implemented by means of hardware, but in many cases the former is a preferred embodiment. Based on such understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art in the form of a software product stored in a storage medium (e.g. ROM/RAM, magnetic disk, optical disk) as described above, comprising instructions for causing a terminal device (which may be a mobile phone, a computer, a server, or a handheld electrochemical detection terminal, etc.) to perform the method according to the embodiments of the present invention. The foregoing description is only of the preferred embodiments of the present invention, and is not intended to limit the scope of the invention, but rather is intended to cover any equivalents of the structures or equivalent processes disclosed herein or in the alternative, which may be employed directly or indirectly in other related arts.

Claims

1. A method of correcting a cell volume, comprising:

2. The method of cell pressure volume correction of claim 1, wherein the constructing a data set comprises:

And establishing the data set according to the data information.

3. The method of claim 2, wherein the steps of collecting samples of different HCT ranges and different temperature ranges and performing ac tests, recording real part information and imaginary part information at each frequency, comprises:

4. The method of claim 3, wherein after recording the real and imaginary information at each frequency, further comprising:

；

) ；

5. The method of claim 2, wherein said establishing said data set from said data information comprises:

the unified normalization processing formula is as follows:

；

6. The method of claim 1, wherein said deriving a corresponding HCT predicted value for each of said samples in said data set based on said HCT measurements comprises:

7. The method of claim 6, wherein constructing a predictive model based on a nonlinear iterative partial least squares method and calculating the HCT predictive value for each of the samples based on the HCT measured values comprises:

wherein the obtaining regression coefficients comprises:

8. The method of claim 1, wherein classifying the measured HCT predicted value and the current value according to the data set using a support vector machine comprises:

establishing a classifier based on the support vector machine;

9. The method of claim 8, wherein the classifier is a linear kernel-based bi-classifier and each of the bi-classifiers is a linear function;

The equation for the linear function is as follows:

；

10. The method of claim 8, wherein said obtaining, by said classifier, classification equation coefficients and obtaining equation slopes and intercepts of said classifier, thereby constructing a classification model, comprises:

setting a kernel parameter of the svm.svc function type as a linear function;

11. The method of claim 10, wherein the classification equation coefficients include coef and interseptit;

the classifier comprises f1 and f2;

the equation for the equation slope and intercept is calculated as:

；

]；

12. The method of claim 8, wherein inputting the HCT predicted value and the current value measured for each sample in the data set into the classification model for prediction, and obtaining a concentration segment classification result, comprises:

13. The method of claim 12, wherein said deriving a corresponding one-dimensional vector from said classification discrimination values comprises:

14. The method of claim 1, wherein said obtaining coefficients of the hyperplane of each of the concentration segments for each of the concentration segments and fitting the percentage of compensation of the HCT predicted value and the current value in each of the concentration segments to obtain a corresponding compensation equation comprises:

15. The method of claim 14, wherein the compensation equation for each concentration segment is:

；

wherein ,a compensation percentage which is the relative deviation of the measured current value and the reference current; />Is a coefficient;a predicted value for the HCT; />Is the intercept.

16. The method of correcting for cell pressure accumulation of claim 1 wherein the electrochemical test strip comprises:

the hydrophilic film layer is provided with air holes;

a siphon window is arranged on the barrier layer;

one end of the single-sided adhesive layer is provided with a reagent window;

the electrode layer is attached to the insulating layer;

17. The method of claim 16, wherein the number of reagent windows on the single facer layer is 4, each:

a first window, a second window, a third window, and a fourth window.

18. The method of claim 17, wherein the electrode layer is provided with a reaction zone corresponding to the reagent window of the single sided adhesive layer;

19. The method of correcting for cell volume of claim 16, wherein the electrode layer comprises: a reference electrode assembly and an HCT working electrode, wherein the reference electrode assembly comprises a working reference electrode and an HCT reference electrode;

20. The method of claim 19, wherein the electrode layer comprises at least one of the following features:

A. the electrode layer is made of a metal thin layer or printing carbon paste;

B. the thickness of the metal thin layer in the electrode layer is 1nm-50nm;

21. A cell pressure accumulation correction apparatus, comprising:

22. The cell pressure accumulation correction system is characterized by comprising a signal acquisition device, a memory and a processor; the signal acquisition device acquires signals returned by the reference electrode component and the HCT working electrode in the electrochemical test paper electrode layer; the memory stores a cell pressure accumulation correction program; the processor runs the cell pressure volume correction program to cause the cell pressure volume correction system to perform the cell pressure volume correction method of any one of claims 1-20.

23. A computer readable storage medium, characterized in that the computer readable storage medium has stored thereon a cell pressure volume correction program, which when executed by a processor, implements the cell pressure volume correction method according to any of claims 1-20.