Concentration detection model training method, concentration detection method, device, electronic device and storage medium

Abstract

The embodiments of the present invention provide a training method for a concentration detection model, a concentration detection method, an apparatus, an electronic device and a storage medium, wherein the method comprises: collecting N groups of fluorescence data corresponding to the collection windows in different reagent sheets, wherein different reagent sheets correspond to different detection objects, and the same detection object corresponds to different concentrations. For each group of fluorescence data, a baseline value of each group of fluorescence data is determined, and based on the baseline value, the group of fluorescence data is preprocessed, and the preprocessed group of fluorescence data is converted into a word vector, wherein a group of word vectors corresponds to a concentration label, and the concentration detection model is trained based on N groups of word vectors carrying concentration labels, so that the input of the trained concentration detection model is the fluorescence data to be detected of the object to be detected in the reagent sheet, and the output is the concentration of the object to be detected. It is possible to improve the accuracy of determining the concentration of the object to be detected while reducing the process and experimental detection requirements.

Description

Technical Field

[0001] The present invention relates to the field of computer vision technology, and in particular to a concentration detection model training method, a concentration detection method, a device, an electronic device, and a storage medium. Background Art

[0002] Fluorescent immunoassay technology utilizes the specificity of the binding factors of the substance to be detected in immunology, combines fluorescent probes with the binding factors, and determines the content of the substance to be detected by measuring the intensity of the fluorescent signal after the reaction. It is an in vitro diagnostic technology that combines the principles of immunological reaction and fluorescence detection.

[0003] Fluorescent immunochromatography technology has high sensitivity and strong specificity, and has become a key development direction for the in vitro diagnostic industry in the future. The current fluorescent chromatography technology collects fluorescent signal data after the immune reaction through the detection system, and then converts it through an algorithm to obtain the concentration value of the corresponding detection object. A single detection system includes a detection instrument and a reagent sheet. Due to the superposition and combination of various factors, inconsistent differences will occur between the systems, resulting in different signal intensities and shapes under the same detection content, and then different detection results will be obtained, affecting the final detection accuracy. For example, when the reagent sheet is produced, the content and distribution of each group of fluorescent microspheres will not be completely consistent, and the volume of the liquid marked on the reagent sheet will also deviate due to human process factors, resulting in more or less differences in the TAP signal value calculated based on the fluorescence signal curve.

[0004] Clinical trials aim to minimize differences between groups at the same concentration while maintaining the ability to distinguish between test results at different concentrations. One approach is to reduce these differences by improving process quality and experimental precision, but this is difficult to achieve in practice, is costly, and results in inaccurate concentration measurements. Summary of the Invention

[0005] The present invention aims to provide a concentration detection model training method, a concentration detection method, a device, an electronic device and a storage medium, which can improve the accuracy of determining concentration detection.

[0006] In order to achieve the above objectives, the technical solutions adopted in the embodiments of the present application are as follows:

[0007] In a first aspect, an embodiment of the present application provides a method for training a concentration detection model, the method comprising:

[0008] Collecting N sets of fluorescence data corresponding to the collection windows in different reagent strips, wherein different reagent strips correspond to different test objects, and the same test object corresponds to different concentrations;

[0009] For each set of fluorescence data, determining a baseline value of each set of fluorescence data;

[0010] Preprocessing the group of fluorescence data based on the baseline value;

[0011] Converting the preprocessed set of fluorescence data into word vectors, wherein a set of word vectors corresponds to a concentration label;

[0012] The concentration detection model is trained based on N groups of word vectors carrying concentration labels, so that the input of the trained concentration detection model is the fluorescence data of the object to be detected in the reagent sheet, and the output is the concentration of the object to be detected.

[0013] In an optional embodiment, the step of preprocessing the set of fluorescence data based on the baseline value includes:

[0014] A first difference between each sub-fluorescence data value in the set of fluorescence data and a baseline value of the set of fluorescence data is calculated, wherein each set of fluorescence data is composed of a plurality of sub-fluorescence data values.

[0015] In an optional embodiment, the step of converting the preprocessed set of fluorescence data into a word vector includes:

[0016] determining a sequence value of each of the first difference values ​​in the set of fluorescence data;

[0017] Based on each of the sequence values ​​and each of the first difference values, forming a data pair of the set of fluorescence data;

[0018] Compile each data pair according to the preset compilation rules to obtain the word vector of each data pair.

[0019] In an optional embodiment, the step of determining the baseline value of each set of fluorescence data includes:

[0020] determining a test peak of a test line in the set of fluorescence data;

[0021] Selecting a first number of first target sub-fluorescence data values ​​within a first preset area of ​​the test peak;

[0022] Calculating a first average value based on each of the first target sub-fluorescence data values;

[0023] using the first average value as the baseline value of the fluorescence data of the group;

[0024] or,

[0025] Determining a test peak of a test line and a quality control peak of a quality control line in the set of fluorescence data;

[0026] selecting a second number of second target sub-fluorescence data values ​​between the test peak and the quality control peak;

[0027] Calculating a second average value based on each of the second target sub-fluorescence data values;

[0028] using the second average value as the baseline value of the fluorescence data of the group;

[0029] or,

[0030] Determining a test peak of a test line and a quality control peak of a quality control line in the set of fluorescence data;

[0031] Selecting a third number of third target sub-fluorescence data values ​​within a first preset region in the test peak and a third number of fourth target sub-fluorescence data values ​​within a second preset region in the quality control peak;

[0032] Calculating a third average value based on each of the third target sub-fluorescence data values ​​and each of the fourth target sub-fluorescence data values;

[0033] The third average value is used as the baseline value of the fluorescence data set.

[0034] In an optional embodiment, the step of compiling each of the data pairs according to a preset compilation rule to obtain a word vector for each of the data pairs includes:

[0035] Determining weight information of each first difference in each of the data pairs;

[0036] The first difference and the weight information corresponding to the first difference are used to form a word vector for the data pair.

[0037] In an optional embodiment, the step of determining weight information of each first difference in each data pair includes:

[0038] In a case where the baseline value is calculated based on each first target sub-fluorescence data value, determining a second sequence value of an intermediate fluorescence data value in each first target sub-fluorescence data value;

[0039] determining an interval threshold value of each of the first target sub-fluorescence data values;

[0040] determining first sequence values ​​of the first target sub-fluorescence data values;

[0041] For each first sequence value, calculating a difference between the first sequence value and the second sequence value;

[0042] When the difference is greater than zero, the weight information of the first difference is calculated based on the following formula:

[0043] Weight information = 1–0.1*(Int(first sequence value–second sequence value) / interval threshold), where Int is rounded;

[0044] When the difference is less than zero, the weight information of the first difference is calculated based on the following formula:

[0045] Weight information=1+0.1*(Int(second sequence value−first sequence value) / interval threshold).

[0046] In a second aspect, an embodiment of the present application provides a concentration detection method, the method comprising:

[0047] Acquire fluorescence data of the object to be detected;

[0048] Preprocessing and word vectorization of the fluorescence data to be detected;

[0049] The pre-processed and word-vectorized fluorescence data are input into the concentration detection model trained by the method described above, and the concentration of the object to be detected is output.

[0050] In a third aspect, an embodiment of the present application provides a training device for a concentration detection model, the device comprising:

[0051] An acquisition module is used to acquire N sets of fluorescence data corresponding to acquisition windows in different reagent strips, wherein different reagent strips correspond to different test substances, and the same test substance corresponds to different concentrations;

[0052] a determination module, configured to determine, for each set of fluorescence data, a baseline value of each set of fluorescence data;

[0053] a preprocessing module, configured to preprocess the group of fluorescence data based on the baseline value;

[0054] a conversion module, configured to convert the preprocessed set of fluorescence data into word vectors, wherein a set of word vectors corresponds to a concentration label;

[0055] The training module is used to train the concentration detection model based on N groups of word vectors carrying concentration labels, so that the input of the trained concentration detection model is the fluorescence data of the object to be detected in the reagent sheet, and the output is the concentration of the object to be detected.

[0056] In a fourth aspect, an embodiment of the present application provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the training method of the concentration detection model when executing the computer program.

[0057] In a fifth aspect, an embodiment of the present application provides a storage medium on which a computer program is stored, and when the computer program is executed by a processor, the steps of the training method of the concentration detection model are implemented.

[0058] This application has the following beneficial effects:

[0059] The present application is by collecting N groups of fluorescence data corresponding to the collection window in different reagent sheets, wherein different reagent sheets correspond to different detection objects, and the same detection object corresponds to different concentrations. For each group of fluorescence data, the baseline value of each group of fluorescence data is determined, and based on the baseline value, the group of fluorescence data is pre-processed, and the pre-processed group of fluorescence data is converted into a word vector, wherein one group of the word vectors corresponds to a concentration label, and the concentration detection model is trained based on each word vector with a concentration label in N groups, so that the input of the trained concentration detection model is the fluorescence data to be detected in the reagent sheet of the object to be detected, and the output is the concentration of the object to be detected. It can improve the accuracy of determining the concentration of the object to be detected while reducing the process and experimental detection requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0061] Figure 1 A block diagram of an electronic device provided by an embodiment of the present invention;

[0062] Figure 2 One of the step flow charts of a concentration detection model training method provided in an embodiment of the present invention;

[0063] Figure 3 This is a second flow chart of a method for training a concentration detection model provided by an embodiment of the present invention;

[0064] Figure 4 This is a flowchart of the steps of a concentration detection model training method provided by an embodiment of the present invention;

[0065] Figure 5 A fourth step flow chart of a method for training a concentration detection model provided in an embodiment of the invention;

[0066] Figure 6 A schematic diagram of a reagent sheet provided in an embodiment of the present invention;

[0067] Figure 7 A flow chart of the steps of a concentration detection method provided by an embodiment of the present invention;

[0068] Figure 8 This is a structural block diagram of a concentration detection model training device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0069] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0070] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort shall fall within the scope of protection of the present invention.

[0071] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings.

[0072] In the description of the present invention, it should be noted that if the terms "upper", "lower", "inside", "outside", etc. appear, the orientation or position relationship indicated is based on the orientation or position relationship shown in the accompanying drawings, or is the orientation or position relationship in which the product of the invention is usually placed when in use. It is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be understood as a limitation on the present invention.

[0073] In addition, the terms "first", "second", etc., if used, are merely used to distinguish and describe, and should not be understood as indicating or implying relative importance.

[0074] It should also be noted that, in the description of this application, unless otherwise expressly specified or limited, the terms "disposed," "installed," "connected," and "connected" should be understood in a broad sense. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to internal connections between two components. Those skilled in the art will understand the specific meanings of the above terms in this application based on the specific circumstances.

[0075] After extensive research, the inventors found that the error can be reduced by improving the level of the process and the accuracy of the experiment, but this is difficult to achieve in actual operation, and the cost is high and the concentration detection is not accurate enough.

[0076] In view of the discovery of the above problems, the present embodiment provides a training method, a concentration detection method, a device, an electronic device and a storage medium for a concentration detection model, which can collect N groups of fluorescence data corresponding to the collection window in different reagent sheets, wherein different reagent sheets correspond to different detection objects, and the same detection object corresponds to different concentrations. For each group of fluorescence data, a baseline value of each group of fluorescence data is determined, and based on the baseline value, the group of fluorescence data is preprocessed, and the preprocessed group of fluorescence data is converted into a word vector, wherein a group of the word vectors corresponds to a concentration label, and the concentration detection model is trained based on N groups of word vectors carrying concentration labels, so that the input of the trained concentration detection model is the fluorescence data to be detected of the object to be detected in the reagent sheet, and the output is the concentration of the object to be detected. The accuracy of determining the concentration of the object to be detected can be improved while reducing the process and experimental detection requirements. The scheme provided by this embodiment is described in detail below.

[0077] This embodiment provides an electronic device that can train a concentration detection model. In one possible implementation, the electronic device can be a user terminal, such as, but not limited to, a server, a smartphone, a personal computer (PC), a tablet computer, a personal digital assistant (PDA), a mobile internet device (MID), etc.

[0078] Please refer to Figure 1 , Figure 1 The electronic device 100 provided in the embodiment of the present application is shown in FIG. Figure 1 More or fewer components than shown, or with Figure 1 Different configurations shown. Figure 1 Each component shown in the figure can be implemented by hardware, software or a combination thereof.

[0079] The electronic device 100 includes a concentration detection model training device 110 , a memory 120 , and a processor 130 .

[0080] The components of the memory 120 and the processor 130 are electrically connected to each other directly or indirectly to realize data transmission or interaction. For example, these components can be electrically connected to each other through one or more communication buses or signal lines. The training device 110 of the concentration detection model includes at least one software function module that can be stored in the memory 120 in the form of software or firmware or solidified in the operating system (OS) of the electronic device 100. The processor 130 is used to execute the executable modules stored in the memory 120, such as the software function modules and computer programs included in the training device 110 of the concentration detection model.

[0081] The memory 120 may be, but is not limited to, a random access memory (RAM), a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), etc. The memory 120 is used to store a program, and the processor 130 executes the program after receiving an execution instruction.

[0082] Please refer to Figure 2 , Figure 2 For application Figure 1 The flowchart of a method for training a concentration detection model of an electronic device 100 is shown, and the method including each step is described in detail below.

[0083] Step 201: Collect N groups of fluorescence data corresponding to the collection windows in different reagent strips.

[0084] Among them, different reagent strips correspond to different detection objects, and the same detection object corresponds to different concentrations.

[0085] Step 202: For each set of fluorescence data, determine a baseline value of each set of fluorescence data.

[0086] Step 203: Preprocess the set of fluorescence data based on the baseline value.

[0087] Step 204: Convert the pre-processed set of fluorescence data into word vectors.

[0088] Among them, a set of word vectors corresponds to a concentration label.

[0089] Step 205: The concentration detection model is trained based on N groups of word vectors carrying concentration labels, so that the input of the trained concentration detection model is the fluorescence data of the object to be detected in the reagent sheet, and the output is the concentration of the object to be detected.

[0090] It should be noted that N groups of fluorescence data can be: on the reagent sheet, the capture antibody and the quality control antibody draw lines on the membrane, corresponding to the test line and the quality control line, and obtain the fluorescence signal values ​​of the reaction of the test substance with the test line and the quality control line on the reagent sheet under different concentrations of different test substances, as well as the fluorescence signal values ​​of the base area before the test line and the fluorescence signal values ​​of the base area between the test line and the quality control line, as the collected fluorescence data group.

[0091] Different detection objects may be different detection objects in the same liquid, and different detection objects are detected based on different reagent sheets, for example, different trace elements in blood are detected.

[0092] In another example, lines are drawn on the reagent sheet by capture antibodies, quality control antibodies and internal reference proteins, corresponding to the test line, quality control line and internal reference line, and the fluorescence signal values ​​of the reaction of the test substance with the test line, quality control line and internal reference line on the reagent sheet at different concentrations of different test substances, the fluorescence signal value of the base area before the test line, the fluorescence signal value of the base area between the test line and the quality control line, the fluorescence signal value of the base area between the quality control line and the internal reference line, and the fluorescence signal value of the base area after the internal reference line are obtained as the collected fluorescence data group.

[0093] The collection window can be the entire reagent sheet or a local area of ​​the reagent sheet. It should be noted that the collection window must include the above-mentioned test line, quality control line and base area. When the reagent sheet contains a test line, quality control line and internal reference line, the collection window must include the above-mentioned three lines and the base area.

[0094] In addition to the reaction fluorescence signal values ​​of the above-mentioned test lines, quality control lines or internal reference lines, the N groups of fluorescence data also include the fluorescence signal values ​​in the reaction substrate in the acquisition window. Since in actual reaction, the order of the test line, quality control line and internal reference line is fixed, the reaction reagent will diffuse from the first position to the last position at a certain flow rate. Therefore, there will also be fluorescence signal values ​​in the reaction substrate of the acquisition window. The fluorescent head will also feedback signal values ​​when it hits the area without fluorescent microspheres, but it is very low. The specific size is related to the color and material of the area. Therefore, it is necessary to collect the fluorescence data of the test line, quality control line and internal reference line on the reagent sheet, as well as the fluorescence data of the reaction substrate from the first position of the acquisition window, i.e., the initial position, to the test line, the fluorescence data of the reaction substrate between the test line and the quality control line, the fluorescence data of the reaction substrate between the quality control line and the internal reference line, and the fluorescence data of the reaction substrate between the internal reference line and the last position of the acquisition window, i.e., the termination position.

[0095] To eliminate the influence of background noise and extract more reliable absolute fluorescence data, it is necessary to preprocess the N sets of collected fluorescence data. That is, for each set of fluorescence data, a baseline value is determined for each set of fluorescence data, and the fluorescence data is preprocessed based on the baseline value. The baseline value is used to represent the value of the reaction base.

[0096] In one embodiment, the process of preprocessing the collected fluorescence data may be:

[0097] A first difference between each sub-fluorescence data value in the set of fluorescence data and a baseline value of the set of fluorescence data is calculated, wherein each set of fluorescence data is composed of a plurality of sub-fluorescence data values.

[0098] The N groups of preprocessed fluorescence data are converted into word vectors, wherein each word vector carries a concentration label, and the concentration label is used to indicate the concentration of the detection object corresponding to the word vector.

[0099] After preprocessing and word vector conversion of the collected N groups of fluorescence data, word vectors carrying concentration labels are obtained. Each word vector carrying the concentration label is used as a training set to train the concentration detection model, so that the input of the trained concentration detection model is the fluorescence data of the object to be detected in the reagent strip, and the output is the concentration of the object to be detected.

[0100] During the training of the concentration detection model, a cross-entropy loss function is used to calculate the predicted concentration of the analyte in the reagent strip based on the fluorescence data. The distance between the predicted concentration and the concentration represented by the sample's corresponding concentration label is then calculated. The network parameters of the concentration detection model are updated based on this calculated distance until the loss function is minimized, completing the training.

[0101] Specifically, AdamOptimizer is used to update the network parameters of the concentration detection model. The Adam algorithm uses the first-order moment estimation and second-order moment estimation of the gradient to dynamically adjust the learning rate of each parameter. The tf.train.AdamOptimizer provided by TensorFlow can control the learning speed. After bias correction, the learning rate of the parameters for each iteration has a certain range. The learning rate (step size) will not become very large due to a large gradient, and the parameter value is relatively stable. This function can be called directly to update the network parameters of the concentration detection model so that the concentration detection model moves closer to the direction of minimizing the loss function. Modify the hyperparameters such as the learning rate, dropout ratio, number of training times, etc. according to the results of the predicted concentration. After obtaining a high accuracy rate in the test set, save the concentration detection model for judging and identifying new fluorescence data of the same type.

[0102] It should be noted that for the training of the concentration detection model, it is possible to train concentration detection models corresponding to different detection objects based on different detection objects in the same liquid, such as concentration detection model A, concentration detection model B and concentration detection model C, to detect different detection objects respectively.

[0103] How to convert the pre-processed fluorescence data into word vectors? Regarding the above step 204, in another embodiment of the present application, Figure 3 As shown, a training method for a concentration detection model is provided, which specifically includes the following steps:

[0104] Step 204 - 1 : Determine the sequence value of each first difference value in the set of fluorescence data.

[0105] Step 204 - 2 : constructing data pairs of the set of fluorescence data based on each sequence value and each first difference value.

[0106] Step 204 - 3 : Compile each data pair according to the preset compilation rules to obtain the word vector of each data pair.

[0107] For example, if a group of fluorescence data is {1, 2, 3, 1, 4, 2, 5, 2}, and the baseline value of the group of fluorescence data is A, then the first difference values ​​calculated in the group are {1-A, 2-A, 3-A, 1-A, 4-A, 2-A, 5-A, 2-A}.

[0108] Determine the sequence values ​​corresponding to each first difference value: the sequence value corresponding to 1-A is 1, the sequence value corresponding to 2-A is 2, the sequence value corresponding to 3-A is 3, the sequence value corresponding to 1-A is 4, the sequence value corresponding to 4-A is 5, the sequence value corresponding to 2-A is 6... The sequence value corresponding to the last element 2-A in the set is 8.

[0109] Based on each sequence value and the first difference value corresponding to the sequence value, a data pair of the set of fluorescence data is constructed. For example, based on the above set, the obtained data pairs are (1, 1-A), (2, 2-A) ... (8, 2-A).

[0110] Based on the preset compilation rules, the word vectors of each data pair are obtained, that is, the word vectors of (1,1-A), (2,2-A)...(8,2-A) are converted to obtain the word vectors of each data pair.

[0111] How to determine the baseline value of each set of fluorescence data? Regarding the above step 202, in another embodiment of the present application, Figure 4 As shown, a training method for a concentration detection model is provided, which specifically includes the following steps:

[0112] Step 202 - 1 : Determine a test peak of a test line in the set of fluorescence data.

[0113] Step 202 - 2 : Select a first number of first target sub-fluorescence data values ​​within a first preset region of a test peak.

[0114] Step 202 - 3 : Calculate a first average value based on each first target sub-fluorescence data value.

[0115] Step 202 - 4 : Using the first average value as the baseline value of the set of fluorescence data.

[0116] It should be noted that the test peak of the test line is the test peak obtained from the fluorescence data after the test object reacts with the capture antibody of the test line.

[0117] The average value of the first number of first target fluorescence data values ​​at the front end of the test peak is selected as the baseline value; wherein the front end refers to the position of the front part of the test peak.

[0118] It should be noted that the first number can be 10, 15, 20, etc., and the embodiment of the present invention does not impose any specific limitation on this.

[0119] In addition, in one example, the baseline value may be determined by:

[0120] Determine the test peak of the test line and the quality control peak of the quality control line in the group of fluorescence data, select a second number of second target sub-fluorescence data values ​​between the test peak and the quality control peak, calculate a second average value based on each second target sub-fluorescence data value, and use the second average value as the baseline value of the group of fluorescence data.

[0121] The average value of the second target sub-fluorescence data values ​​of the second number between the test peak and the quality control peak is selected as the baseline value; wherein, for two-line data, that is, when the reagent sheet only includes the test line and the quality control line, the second target sub-fluorescence data between the test peak and the quality control peak is taken; for three-line data, that is, when the reagent sheet includes the test line, the quality control line and the internal reference line, the second target sub-fluorescence data between the test peak and the internal reference peak is taken.

[0122] The quality control peak of the quality control line is the quality control peak obtained by the fluorescence data after the test object reacts with the quality control antibody of the quality control line, and the internal reference peak of the internal reference line is the internal reference peak obtained by the fluorescence data after the test object reacts with the internal reference protein of the internal reference line.

[0123] In another example, the baseline value may be determined by:

[0124] Determine the test peak of the test line and the quality control peak of the quality control line in the group of fluorescence data, select a third number of third target sub-fluorescence data values ​​within the first preset area in the test peak, and a third number of fourth target sub-fluorescence data values ​​within the second preset area in the quality control peak, calculate a third average value based on the third target sub-fluorescence data values ​​and the fourth target sub-fluorescence data values, and use the third average value as the baseline value of the group of fluorescence data.

[0125] Exemplarily, a third number of third target sub-fluorescence data values ​​and fourth target sub-fluorescence data values ​​are selected at the front end of the test peak and the back end of the quality control peak; wherein the front end refers to the position of the front end of the test peak, and the back end refers to the position of the back end of the quality control peak.

[0126] How to compile each data pair according to the preset compilation rules to obtain the word vector of each data pair, for the above step 204-3, in another embodiment of the present application, as Figure 5 As shown, a training method for a concentration detection model is provided, which specifically includes the following steps:

[0127] Step 204-3-1: Determine the weight information of each first difference in each data pair.

[0128] Step 204-3-2: The first difference and the weight information corresponding to the first difference are used to form a word vector for the data pair.

[0129] The step of determining weight information of each first difference in each data pair includes:

[0130] When the baseline value is calculated based on each first target sub-fluorescence data value, the second sequence value of the intermediate fluorescence data value in each first target sub-fluorescence data value is determined, the interval threshold of each first target sub-fluorescence data value is determined, and each first sequence value of each first target sub-fluorescence data value is determined. For each first sequence value, the difference between the first sequence value and the second sequence value of the intermediate fluorescence data value is calculated. When the difference is greater than zero, the weight information of the first difference is calculated based on the following formula: weight information = 1-0.1*(Int(first sequence value-second sequence value) / interval threshold), where Int is rounded. When the difference is less than zero, the weight information of the first difference is calculated based on the following formula: weight information = 1+0.1*(Int(second sequence value-first sequence value) / interval threshold).

[0131] The baseline values ​​of the reaction samples, i.e., the N sets of fluorescence data collected, are not uniform everywhere. Due to the different stacking effects of the markers, i.e., the capture antibody, quality control antibody, and internal reference protein, as they flow along with the reacted test material, the accumulation of markers will be more obvious at the position passed first. Therefore, generally speaking, the baseline value at the front end is greater than the baseline value at the back end. When subtracting the baseline value to calculate the absolute signal value, the baseline value used is a fixed value calculated from a certain interval and cannot fully reflect the actual corresponding base at each coordinate. Based on the trend that the front end baseline value is generally higher than the back end baseline value, the weight information is assigned according to the distance between the actual position and the position where the calculated baseline value is intercepted. The weight of the interval near the baseline value interception is 1. When the actual position is in front of the baseline value interception position, the weight information increases with increasing distance. When the actual position is behind the baseline value interception position, the weight information decreases with increasing distance. This increases the amount of information contained in the input fluorescence data, thereby making the trained concentration detection model more accurate.

[0132] Based on the weight information of each calculated first difference, the word vectors for each data pair are obtained, where a set of word vectors corresponds to a concentration label. Different concentration labels are one-hot encoded. Here, taking the five concentration label categories of a certain test object as an example, assuming that the fluorescence data to be classified is divided into five different concentration labels, namely 5500pg / ml, 3200pg / ml, 1600pg / ml, 100pg / ml, and 0pg / ml, the one-hot encoding of different concentration labels can be set as follows:

[0133] Concentration 5500pg / ml: [1, 0, 0, 0, 0];

[0134] Concentration 3200 pg / ml: [0, 1, 0, 0, 0];

[0135] Concentration 1600 pg / ml: [0, 0, 1, 0, 0];

[0136] Concentration 100 pg / ml: [0, 0, 0, 1, 0];

[0137] Concentration 0pg / ml: [0, 0, 0, 0, 1];

[0138] The word vector and concentration label corresponding to each set of fluorescence data correspond one to one.

[0139] like Figure 6 As shown, the coordinates in the reagent sheet are regarded as a time sequence, and the intensity of the fluorescence data corresponding to each position is regarded as a sub-fluorescence data value.

[0140] During the training process of the concentration detection model, dropout parameters were set between each layer to prevent data overfitting. The entire concentration detection model follows the classic LSTM model. The sub-fluorescence data values ​​at each coordinate point are preprocessed and converted into corresponding word vectors. These are then sequentially input into the RNN model. The RNN at each moment can be considered a separate fully connected network, with the input being the current word vector and the hidden layer output at the previous moment. This enables the correlation between neuron parameters at different moments, ensuring that the previous output influences the subsequent output.

[0141] In order to detect the concentration of the substance to be detected, in another embodiment of the present application, Figure 7 As shown, a concentration detection method is provided, which specifically includes the following steps:

[0142] Step 301: Obtain fluorescence data of an object to be detected.

[0143] Step 302: Preprocess and vectorize the fluorescence data to be detected.

[0144] Step 303: Input the pre-processed and word-vectorized fluorescence data into the concentration detection model obtained through training, and output the concentration of the object to be detected.

[0145] Preprocessing and word-vectorizing the acquired fluorescence data of the target object can eliminate background noise and extract a more reliable absolute signal. The preprocessing method for the fluorescence data is consistent with the training method for the concentration detection model. The fluorescence data is then input into the trained concentration detection model to determine the concentration of the target object.

[0146] Please refer to Figure 8 The present application also provides an embodiment of a method for Figure 1 The training device 110 for the concentration detection model of the electronic device 100 includes:

[0147] The acquisition module 111 is used to acquire N sets of fluorescence data corresponding to acquisition windows in different reagent strips, wherein different reagent strips correspond to different detection objects, and the same detection object corresponds to different concentrations;

[0148] a determination module 112, configured to determine a baseline value of each set of fluorescence data for each set of fluorescence data;

[0149] A preprocessing module 113, configured to preprocess the set of fluorescence data based on the baseline value;

[0150] a conversion module 114 for converting the preprocessed set of fluorescence data into word vectors, wherein a set of word vectors corresponds to a concentration label;

[0151] The training module 115 is used to train the concentration detection model based on N groups of word vectors carrying concentration labels, so that the input of the trained concentration detection model is the fluorescence data of the object to be detected in the reagent sheet, and the output is the concentration of the object to be detected.

[0152] The present application is by collecting N groups of fluorescence data corresponding to the collection window in different reagent sheets, wherein different reagent sheets correspond to different detection objects, and the same detection object corresponds to different concentrations. For each group of fluorescence data, the baseline value of each group of fluorescence data is determined, and based on the baseline value, the group of fluorescence data is pre-processed, and the pre-processed group of fluorescence data is converted into a word vector, wherein a group of word vectors corresponds to a concentration label, and the concentration detection model is trained based on each word vector with a concentration label in N groups, so that the input of the trained concentration detection model is the fluorescence data to be detected in the reagent sheet of the object to be detected, and the output is the concentration of the object to be detected. It can improve the accuracy of determining the concentration of the object to be detected while reducing the process and experimental detection requirements.

[0153] The present application also provides an electronic device 100, which includes a processor 130 and a memory 120. The memory 120 stores computer-executable instructions, which, when executed by the processor 130, implement the training method of the concentration detection model.

[0154] An embodiment of the present application further provides a storage medium storing a computer program. When the computer program is executed by the processor 130, the training method of the concentration detection model is implemented.

[0155] In the embodiments provided in this application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of the code, and the module, program segment or a part of the code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or can be implemented using a combination of dedicated hardware and computer instructions.

[0156] In addition, the functional modules in each embodiment of the present application can be integrated together to form an independent part, or each module can exist separately, or two or more modules can be integrated to form an independent part. If the function is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0157] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

[0158] The above descriptions are merely examples of various embodiments of the present application, but the scope of protection of the present application is not limited thereto. Any modifications or substitutions that can be readily conceived by a person skilled in the art within the technical scope disclosed in the present application should be included within the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A method for training a concentration detection model, characterized in that: The method comprises: Collecting N sets of fluorescence data corresponding to the collection windows in different reagent strips, wherein different reagent strips correspond to different test objects, and the same test object corresponds to different concentrations; For each set of fluorescence data, determining a baseline value of each set of fluorescence data; Preprocessing the group of fluorescence data based on the baseline value; Converting the preprocessed set of fluorescence data into word vectors, wherein a set of word vectors corresponds to a concentration label; The concentration detection model is trained based on N groups of word vectors carrying concentration labels, so that the input of the trained concentration detection model is the fluorescence data of the object to be detected in the reagent strip, and the output is the concentration of the object to be detected; the step of preprocessing the group of fluorescence data based on the baseline value includes: Calculating a first difference between each sub-fluorescence data value in the set of fluorescence data and a baseline value of the set of fluorescence data, wherein each set of fluorescence data is composed of a plurality of sub-fluorescence data values; and converting the preprocessed set of fluorescence data into a word vector comprises: determining a sequence value of each of the first difference values ​​in the set of fluorescence data; Based on each of the sequence values ​​and each of the first difference values, forming a data pair of the set of fluorescence data; Compile each data pair according to the preset compilation rules to obtain the word vector of each data pair.

2. The method according to claim 1, characterized in that The step of determining the baseline value of each set of fluorescence data comprises: determining a test peak of a test line in the set of fluorescence data; Selecting a first number of first target sub-fluorescence data values ​​within a first preset area of ​​the test peak; Calculating a first average value based on each of the first target sub-fluorescence data values; using the first average value as the baseline value of the fluorescence data of the group; or, Determining a test peak of a test line and a quality control peak of a quality control line in the set of fluorescence data; selecting a second number of second target sub-fluorescence data values ​​between the test peak and the quality control peak; Calculating a second average value based on each of the second target sub-fluorescence data values; using the second average value as the baseline value of the fluorescence data of the group; or, Determining a test peak of a test line and a quality control peak of a quality control line in the set of fluorescence data; Selecting a third number of third target sub-fluorescence data values ​​within a first preset region in the test peak and a third number of fourth target sub-fluorescence data values ​​within a second preset region in the quality control peak; Calculating a third average value based on each of the third target sub-fluorescence data values ​​and each of the fourth target sub-fluorescence data values; The third average value is used as the baseline value of the fluorescence data set.

3. The method according to claim 1, characterized in that The step of obtaining a word vector for each data pair according to a preset compilation rule includes: Determining weight information of each first difference in each of the data pairs; The first difference and the weight information corresponding to the first difference are used to form a word vector for the data pair.

4. The method according to claim 3, characterized in that The step of determining the weight information of each first difference in each data pair includes: In a case where the baseline value is calculated based on each first target sub-fluorescence data value, determining a second sequence value of an intermediate fluorescence data value in each first target sub-fluorescence data value; determining an interval threshold value of each of the first target sub-fluorescence data values; determining first sequence values ​​of the first target sub-fluorescence data values; For each first sequence value, calculating a difference between the first sequence value and the second sequence value; When the difference is greater than zero, the weight information of the first difference is calculated based on the following formula: Weight information = 1 – 0.1*(Int(first sequence value – second sequence value) / interval threshold), where Int is rounded; When the difference is less than zero, the weight information of the first difference is calculated based on the following formula: Weight information = 1 + 0.1*(Int(second sequence value – first sequence value) / interval threshold).

5. A concentration detection method, characterized in that: The method comprises: Acquire fluorescence data of the object to be detected; Preprocessing and word vectorization of the fluorescence data to be detected; The preprocessed and word-vectorized fluorescence data is input into a concentration detection model trained by the method described in any one of claims 1 to 4, and the concentration of the object to be detected is output.

6. A training device for a concentration detection model, characterized in that: The device comprises: An acquisition module is used to acquire N sets of fluorescence data corresponding to acquisition windows in different reagent strips, wherein different reagent strips correspond to different test substances, and the same test substance corresponds to different concentrations; a determination module, configured to determine, for each set of fluorescence data, a baseline value of each set of fluorescence data; a preprocessing module, configured to calculate a first difference between each sub-fluorescence data value in the set of fluorescence data and a baseline value of the set of fluorescence data, wherein each set of fluorescence data is composed of a plurality of sub-fluorescence data values; a conversion module, configured to determine a sequence value of each of the first difference values ​​in the set of fluorescence data; construct data pairs of the set of fluorescence data based on each of the sequence values ​​and each of the first difference values; and obtain a word vector for each of the data pairs according to a preset compilation rule, wherein a set of the word vectors corresponds to one concentration label; The training module is used to train the concentration detection model based on N groups of word vectors carrying concentration labels, so that the input of the trained concentration detection model is the fluorescence data of the object to be detected in the reagent sheet, and the output is the concentration of the object to be detected.

7. An electronic device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method according to any one of claims 1 to 5 when executing the computer program.

8. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images