In-vivo component metabolism concentration identification method and system

Abstract

The invention discloses an in-vivo component metabolic concentration identification method. The method comprises the following steps: establishing an in-vivo component metabolic concentration identification and classification neural network model; obtaining a plurality of reagent reaction colors with different gradients as sample pictures, and processing the sample pictures to obtain a training data set corresponding to the in-vivo component metabolism concentration identification and classification neural network model; obtaining an in-vivo component metabolism concentration recognition numerical prediction neural network model based on the interaction between the in-vivo component metabolism concentration recognition classification neural network model and the training data set; and obtaining a real-time reaction color picture corresponding to the in-vivo component metabolism concentration, inputting the in-vivo component metabolism concentration identification value prediction neural network model, and taking the output of the in-vivo component metabolism concentration identification value prediction neural network model as the in-vivo component metabolism concentration. The invention further discloses a system, electronic equipment and a computer readable storage medium, classification detection of pictures is promoted to a numerical prediction task, and conversion from a classification problem to a regression problem is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of body composition detection, and in particular to a method and system for identifying metabolic concentrations of body components. Background Art

[0002] At present, the main methods for reagent reaction color recognition include color histogram, threshold segmentation, color moment, color space conversion and image color classification using neural network. Due to the presence of reflections in reagent images, errors in the device itself, shadows, uneven reagents, etc., each algorithm has certain limitations when applying traditional algorithms. For example, when using color histograms, there are many ways to set the weight when converting RGB colors to grayscale, and the results obtained by different weight setting methods will be slightly different; when threshold segmentation is used, pixel-level point segmentation cannot be segmented when the background and reaction colors are close; when RGB color space is converted to HSV (Hue, Saturation, Value) space, the threshold division is also unclear; neural network image color recognition, for example, Yolov8, deep neural network uses BCEloss when performing category detection, and the border regression loss uses DFL Loss + CIOU Loss. It can be seen that most neural network image color recognition currently belongs to category detection, and there are fewer models and applications for numerical prediction. Summary of the invention

[0003] The purpose of the present invention is to provide a method and system for identifying the metabolic concentration of in vivo components, which are used to identify the reaction color of a porous reagent test body fluid, and predict the concentration of the body fluid in each hole according to the depth of its color. The method performs corresponding numerical prediction after training with a classification model. Specifically, a deep neural network model is used to first fine-tune the training image, extract the last layer of the convolution layer as a feature, divide the convolution layer according to the different hole positions, and extract the features corresponding to each hole. The model weights are convolved with the corresponding features and reduced to one-dimensional data, wherein the model weights are obtained by linearly fitting one-dimensional data of different gradients of different samples and their corresponding concentration values, and this weight replaces the last layer of the original deep network model, thereby realizing the conversion of the deep neural network model from classification prediction to numerical prediction; so that this model can upgrade the classification detection of the image to the numerical prediction task of the image, and the model realizes the conversion from classification problem to regression problem.

[0004] The first aspect of the present invention is to provide a method for identifying the metabolic concentration of an in vivo component, which is used to identify the reaction color of a multi-well reagent test fluid and predict the concentration of the fluid in each well according to the color depth, comprising: S1, establish a neural network model for identifying and classifying metabolic concentrations of body components; S2, obtaining a plurality of reagent reaction colors with different gradients as sample images, and processing the sample images to obtain a training data set corresponding to the in vivo component metabolic concentration recognition and classification neural network model; S3, based on the interaction between the in vivo component metabolic concentration identification and classification neural network model and the training data set, a numerical prediction neural network model for in vivo component metabolic concentration identification is obtained; the last layer of the convolution layer of the in vivo component metabolic concentration identification and classification neural network model is extracted as a feature, the convolution layer is divided according to different hole positions to extract the features corresponding to each hole, the model weights are convolved with the corresponding features to reduce them to one-dimensional data, the one-dimensional data of different samples and different gradients are numerically trained, and the corresponding concentration values ​​are linearly fitted to obtain weight values; the weight values ​​are replaced with the last layer of the original deep network model, thereby realizing the conversion of the deep neural network model from classification prediction to numerical prediction; S4, obtaining a real-time reaction color image corresponding to the metabolic concentration of the in vivo component and inputting it into the in vivo component metabolic concentration identification numerical prediction neural network model, and the output of the in vivo component metabolic concentration identification numerical prediction neural network model is used as the in vivo component metabolic concentration.

[0005] Preferably, the in vivo component metabolic concentration identification and classification neural network model in S1 is a deep neural network model.

[0006] Preferably, S2 includes: S21, obtaining a plurality of reagent reaction colors with different gradients as an initial sample image; S22, performing a preprocessing operation on the initial sample image, wherein the preprocessing operation includes: performing one or more operations of image shadow removal, image noise removal, and invalid area removal on the initial sample image to obtain a sample image to be labeled; S23, annotating the different colors of different holes in the sample image to be annotated with multi-hole indicators and concentrations, and finally forming the training data set.

[0007] Preferably, the S3, obtaining a numerical prediction neural network model for identifying metabolic concentrations of in vivo components based on the interaction between the in vivo component metabolic concentration identification classification neural network model and the training data set, comprises: S31, performing multiple fine-tuning on the training data set based on the in vivo component metabolic concentration recognition and classification neural network model and the preprocessing operation to obtain multiple fine-tuning results; S32, verifying the recognition and classification effects of the fine-tuning results, extracting the first key model parameters in the in vivo component metabolic concentration recognition and classification neural network model corresponding to the preset recognition and classification effects and the second key model parameters of the in vivo component metabolic concentration recognition and classification neural network model in S1; the first key model parameters include the last convolution layer of the model; the second key model parameters include the activation function layer; S33, using the last convolutional layer of the model as a feature layer, extracting features of different pore positions based on the feature layer, and obtaining a multidimensional data formula after weighting the features of different pore positions and the in vivo component metabolic concentration recognition and classification neural network model; reducing the multidimensional data in the multidimensional data formula to one-dimensional data; S34, performing numerical training on the one-dimensional data of different samples and different gradients, and performing linear fitting on the corresponding concentration values ​​to obtain weight values; S35, updating the network based on the weight values ​​and the corresponding relationship, replacing the activation function layer with the weight values ​​to form the neural network model for numerical prediction of the metabolic concentration of the in vivo components.

[0008] Preferably, the features of the different hole positions in S33 are the relative position features of the holes obtained after the last convolutional layer division.

[0009] Preferably, the numerical training and linear fitting in S34 are linear regression fitting.

[0010] Preferably, S4 includes: S41, obtaining a real-time reaction color image corresponding to the metabolic concentration of a component in the body and inputting the image into the neural network model for identifying the numerical value prediction of the metabolic concentration of the component in the body; S42, identifying the reaction color of the multi-well reagent test body fluid based on the reaction color image; predicting the metabolic concentration of the in vivo component in each well based on the color depth in the identification result and the in vivo component metabolic concentration identification numerical prediction neural network model; S43, outputting and displaying the metabolic concentration of the body component in each well.

[0011] The second aspect of the present invention is to provide a system for identifying metabolic concentrations of body components, which is used to implement the method for identifying metabolic concentrations of body components of the first aspect, comprising: The first model building module (101) is used to build a neural network model for identifying and classifying metabolic concentrations of body components; A training data set acquisition module (102) is used to acquire a plurality of reagent reaction colors with different gradients as sample images, and to obtain a training data set corresponding to the in vivo component metabolic concentration recognition and classification neural network model after processing the sample images; The second model building module (103) is used to obtain a neural network model for numerical prediction of metabolic concentration identification of in vivo components based on the interaction between the in vivo component metabolic concentration identification and classification neural network model and the training data set; extract the last layer of the convolution layer of the in vivo component metabolic concentration identification and classification neural network model as a feature, divide the convolution layer according to different hole positions to extract the features corresponding to each hole, convolve the model weights with the corresponding features and reduce them to one-dimensional data, perform numerical training on the one-dimensional data of different samples with different gradients, and perform linear fitting with the corresponding concentration values ​​to obtain weight values; replace the last layer of the original deep network model with the weight values, thereby realizing the conversion of the deep neural network model from classification prediction to numerical prediction; The metabolic concentration identification module (104) is used to obtain a real-time reaction color image corresponding to the metabolic concentration of the body component and input it into the body component metabolic concentration identification numerical prediction neural network model, and the output of the body component metabolic concentration identification numerical prediction neural network model is used as the body component metabolic concentration.

[0012] Application examples:

[0013] This embodiment mainly applies multiple-well reagents in the six-item urine test and gives the concentration value of each well according to the depth of the reagent reaction color in each well.

[0014] The detection steps are as follows: (1) Training sample preparation: Prepare reagent samples with different gradients in each well and obtain training sample images through the optical detection system, such as Figure 4 As shown: (2) Perform training preprocessing on the training sample images and label the image data with labelme; (3) Select but not limited to the Yolov8 deep network model, fine-tune the labeled training data, and save the model; Figure 5 shown.

[0015] (4) Extract the last convolution layer of the model and divide it to take the relative position characteristics of each well as the characteristics of this sample and this well. After the model weight is applied, it is reduced to one-dimensional data. This data is regressed and fitted with the concentration corresponding to this sample and this well to obtain its weight; (5) Replace the last activation function layer in the model of step (3) with the weights obtained in step (4) and save the model. The model can be converted from a classification model to a numerical prediction model.

[0016] A third aspect of the present invention is to provide an electronic device, comprising a processor and a memory, wherein the memory stores a plurality of instructions, and the processor is used to read the instructions and execute the method described in the first aspect.

[0017] A fourth aspect of the present invention is to provide a computer-readable storage medium, wherein the computer-readable storage medium stores a plurality of instructions, and the plurality of instructions can be read by a processor to execute the method described in the first aspect.

[0018] Beneficial effects of the method and system of the present invention: (1) The last convolutional layer of the trained model is used as the feature layer, which reduces the impact of algorithm rule mining and different weights during color extraction.

[0019] (2) Perform regression fitting on the feature layer and the known concentration, thus realizing the transformation from classification prediction to numerical prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the related technologies, the drawings required for use in the specific embodiments or the related technical descriptions will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0021] Figure 1 A flow chart of a method for identifying metabolic concentrations of components in vivo provided according to an embodiment of the present invention; Figure 2 A flow chart of converting a classification model to a numerical prediction model according to an embodiment of the present invention; Figure 3 This is a diagram showing the architecture of a system for identifying metabolic concentrations of components in the body according to an embodiment of the present invention; Figure 4 A schematic diagram of the principle of preparing training samples according to an embodiment of the present invention, preparing reagent samples with different gradients in each well and obtaining training sample images through an optical detection system; Figure 5 A diagram of the Yolov8 model architecture provided according to an embodiment of the present invention; Figure 6 A schematic diagram of an electronic device provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0022] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0024] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances. Embodiment 1

[0025] like Figure 1-2 As shown, this embodiment provides a method for identifying the metabolic concentration of a body component, which is used to identify the reaction color of a multi-well reagent test fluid and predict the concentration of the fluid in each well according to the color depth, including: S1, establish a neural network model for identifying and classifying metabolic concentrations of body components; S2, obtaining a plurality of reagent reaction colors with different gradients as sample images, and processing the sample images to obtain a training data set corresponding to the in vivo component metabolic concentration recognition and classification neural network model; S3, based on the interaction between the in vivo component metabolic concentration identification and classification neural network model and the training data set, a numerical prediction neural network model for in vivo component metabolic concentration identification is obtained; the last layer of the convolution layer of the in vivo component metabolic concentration identification and classification neural network model is extracted as a feature, the convolution layer is divided according to different hole positions to extract the features corresponding to each hole, the model weights are convolved with the corresponding features to reduce them to one-dimensional data, the one-dimensional data of different samples and different gradients are numerically trained, and the corresponding concentration values ​​are linearly fitted to obtain weight values; the weight values ​​are replaced with the last layer of the original deep network model, thereby realizing the conversion of the deep neural network model from classification prediction to numerical prediction; S4, obtaining a real-time reaction color image corresponding to the metabolic concentration of the in vivo component and inputting it into the in vivo component metabolic concentration identification numerical prediction neural network model, and the output of the in vivo component metabolic concentration identification numerical prediction neural network model is used as the in vivo component metabolic concentration.

[0026] As a preferred embodiment, the in vivo component metabolic concentration identification and classification neural network model in S1 is a deep neural network model.

[0027] See also Figure 5 As a preferred implementation, the deep network model includes but is not limited to Yolov 8.

[0028] As a preferred embodiment, the S2 includes: S21, obtaining a plurality of reagent reaction colors with different gradients as an initial sample image; S22, performing a preprocessing operation on the initial sample image, wherein the preprocessing operation includes: performing one or more operations of image shadow removal, image noise removal, and invalid area removal on the initial sample image to obtain a sample image to be labeled; S23, annotating the different colors of different holes in the sample image to be annotated with multi-hole indicators and concentrations, and finally forming the training data set.

[0029] As a preferred embodiment, the S3, obtaining a neural network model for numerical prediction of metabolic concentration identification of in vivo components based on the interaction between the in vivo component metabolic concentration identification and classification neural network model and the training data set, comprises: S31, performing multiple fine-tuning on the training data set based on the in vivo component metabolic concentration recognition and classification neural network model and the preprocessing operation to obtain multiple fine-tuning results; S32, verifying the recognition and classification effects of the fine-tuning results, extracting the first key model parameters in the in vivo component metabolic concentration recognition and classification neural network model corresponding to the preset recognition and classification effects and the second key model parameters of the in vivo component metabolic concentration recognition and classification neural network model in S1; the first key model parameters include the last convolution layer of the model; the second key model parameters include the activation function layer; S33, using the last convolutional layer of the model as a feature layer, extracting features of different pore positions based on the feature layer, and obtaining a multidimensional data formula after weighting the features of different pore positions and the in vivo component metabolic concentration recognition and classification neural network model; reducing the multidimensional data in the multidimensional data formula to one-dimensional data; S34, performing numerical training on the one-dimensional data of different samples and different gradients, and performing linear fitting on the corresponding concentration values ​​to obtain weight values; S35, updating the network based on the weight values ​​and the corresponding relationship, replacing the activation function layer with the weight values ​​to form the neural network model for numerical prediction of the metabolic concentration of the in vivo components.

[0030] As a preferred implementation, the features of the different hole positions in S33 are the relative position features of the holes obtained after the last convolutional layer division.

[0031] As a preferred embodiment, the numerical training and linear fitting of S34 are linear regression fitting. Of course, those skilled in the art should know that other regression fitting methods or higher-level fitting methods can also be used, all of which are within the protection scope of the present invention.

[0032] As a preferred implementation, S4 includes: S41, obtaining a real-time reaction color image corresponding to the metabolic concentration of a component in the body and inputting the image into the neural network model for identifying the numerical value prediction of the metabolic concentration of the component in the body; S42, identifying the reaction color of the multi-well reagent test body fluid based on the reaction color image; predicting the metabolic concentration of the in vivo component in each well based on the color depth in the identification result and the in vivo component metabolic concentration identification numerical prediction neural network model; S43, outputting and displaying the metabolic concentration of the body component in each well. Embodiment 2

[0033] like Figure 3 As shown, this embodiment provides a system for identifying metabolic concentrations of components in the body, which is used to implement the method for identifying metabolic concentrations of components in the body of embodiment 1, including: The first model building module 101 is used to build a neural network model for identifying and classifying metabolic concentrations of body components; A training data set acquisition module 102 is used to acquire multiple reagent reaction colors with different gradients as sample images, and obtain a training data set corresponding to the in vivo component metabolic concentration recognition and classification neural network model after processing the sample images; The second model building module 103 is used to obtain a neural network model for numerical prediction of metabolic concentration identification of in vivo components based on the interaction between the in vivo component metabolic concentration identification and classification neural network model and the training data set; extract the last layer of the convolution layer of the in vivo component metabolic concentration identification and classification neural network model as a feature, divide the convolution layer according to different hole positions to extract the features corresponding to each hole, convolve the model weights with the corresponding features and reduce them to one-dimensional data, perform numerical training on the one-dimensional data of different samples with different gradients, and perform linear fitting with the corresponding concentration values ​​to obtain weight values; replace the last layer of the original deep network model with the weight values, thereby realizing the conversion of the deep neural network model from classification prediction to numerical prediction; The metabolic concentration identification module 104 is used to obtain a real-time reaction color image corresponding to the metabolic concentration of the body component and input it into the body component metabolic concentration identification numerical prediction neural network model, and the output of the body component metabolic concentration identification numerical prediction neural network model is used as the body component metabolic concentration.

[0034] Application examples:

[0035] This embodiment mainly applies multiple-well reagents in the six-item urine test and gives the concentration value of each well according to the depth of the reagent reaction color in each well.

[0036] The detection steps are as follows: (1) Training sample preparation: Prepare reagent samples with different gradients in each well and obtain training sample images through the optical detection system, such as Figure 4 As shown: (2) Perform training preprocessing on the training sample images and label the image data with labelme; (3) Select but not limited to the Yolov8 deep network model, fine-tune the labeled training data, and save the model; Figure 5 shown.

[0037] (4) Extract the last convolution layer of the model and divide it to take the relative position characteristics of each well as the characteristics of this sample and this well. After the model weight is applied, it is reduced to one-dimensional data. This data is regressed and fitted with the concentration corresponding to this sample and this well to obtain its weight; (5) Replace the last activation function layer in the model of step (3) with the weights obtained in step (4) and save the model. The model can be converted from a classification model to a numerical prediction model.

[0038] The present invention also provides a memory storing a plurality of instructions, wherein the instructions are used to implement the method as in the first embodiment.

[0039] like Figure 6 As shown, the present invention also provides an electronic device, including a processor 301 and a memory 302 connected to the processor 301, the memory 302 stores a plurality of instructions, and the instructions can be loaded and executed by the processor so that the processor can execute the method as in the first embodiment.

[0040] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above embodiments can be implemented by software, or by means of software plus necessary general hardware platforms. Based on such understanding, the technical solutions of the above embodiments can be embodied in the form of software products, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), including several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0041] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying metabolic concentration of body components, characterized in that: include: S1, establish a neural network model for identifying and classifying metabolic concentrations of body components; S2, obtaining a plurality of reagent reaction colors with different gradients as sample images, and processing the sample images to obtain a training data set corresponding to the in vivo component metabolic concentration recognition and classification neural network model; S3, based on the interaction between the in vivo component metabolic concentration identification and classification neural network model and the training data set, a numerical prediction neural network model for in vivo component metabolic concentration identification is obtained; the last layer of the convolution layer of the in vivo component metabolic concentration identification and classification neural network model is extracted as a feature, the convolution layer is divided according to different hole positions to extract the features corresponding to each hole, the model weights are convolved with the corresponding features to reduce them to one-dimensional data, the one-dimensional data of different samples and different gradients are numerically trained, and the corresponding concentration values ​​are linearly fitted to obtain weight values; the weight values ​​are replaced with the last layer of the original deep network model, thereby realizing the conversion of the deep neural network model from classification prediction to numerical prediction; S4, obtaining a real-time reaction color image corresponding to the metabolic concentration of the in vivo component and inputting it into the in vivo component metabolic concentration identification numerical prediction neural network model, and the output of the in vivo component metabolic concentration identification numerical prediction neural network model is used as the in vivo component metabolic concentration.

2. A method for identifying metabolic concentrations of body components according to claim 1, characterized in that: The in vivo component metabolic concentration recognition and classification neural network model described in S1 is a deep neural network model.

3. A method for identifying metabolic concentrations of body components according to claim 2, characterized in that: The S2 includes: S21, obtaining a plurality of reagent reaction colors with different gradients as an initial sample image; S22, performing a preprocessing operation on the initial sample image, wherein the preprocessing operation includes: performing one or more operations of image shadow removal, image noise removal, and invalid area removal on the initial sample image to obtain a sample image to be labeled; S23, annotating the different colors of different holes in the sample image to be annotated with multi-hole indicators and concentrations, and finally forming the training data set.

4. A method for identifying metabolic concentrations of body components according to claim 3, characterized in that: The step S3, obtaining a neural network model for numerical prediction of metabolic concentration identification of in vivo components based on the interaction between the neural network model for identifying and classifying metabolic concentration of in vivo components and the training data set, comprises: S31, performing multiple fine-tuning on the training data set based on the in vivo component metabolic concentration recognition and classification neural network model and the preprocessing operation to obtain multiple fine-tuning results; S32, verifying the recognition and classification effects of the fine-tuning results, extracting the first key model parameters in the in vivo component metabolic concentration recognition and classification neural network model corresponding to the preset recognition and classification effects and the second key model parameters of the in vivo component metabolic concentration recognition and classification neural network model in S1; the first key model parameters include the last convolution layer of the model; the second key model parameters include the activation function layer; S33, using the last convolutional layer of the model as a feature layer, extracting features of different pore positions based on the feature layer, and obtaining a multidimensional data formula after weighting the features of different pore positions and the in vivo component metabolic concentration recognition and classification neural network model; reducing the multidimensional data in the multidimensional data formula to one-dimensional data; S34, performing numerical training and linear fitting on the one-dimensional data to obtain a weight value; S35, updating the network based on the weight values ​​and the corresponding relationship, replacing the activation function layer with the weight values ​​to form the neural network model for numerical prediction of the metabolic concentration of the in vivo components.

5. A method for identifying metabolic concentrations of body components according to claim 4, characterized in that: The features of the different hole positions in S33 are the relative position features of the holes obtained after the last convolutional layer division.

6. A method for identifying metabolic concentrations of body components according to claim 5, characterized in that: The numerical training and linear fitting of S34 are linear regression fitting.

7. A method for identifying metabolic concentrations of body components according to claim 6, characterized in that: The S4 includes: S41, obtaining a real-time reaction color image corresponding to the metabolic concentration of the in vivo component and inputting it into the in vivo component metabolic concentration recognition numerical prediction neural network model; S42, identifying the reaction color of the multi-well reagent test body fluid based on the reaction color image; predicting the metabolic concentration of the in vivo component in each well based on the color depth in the identification result and the in vivo component metabolic concentration identification numerical prediction neural network model; S43, outputting and displaying the metabolic concentration of the body component in each well.

8. A system for identifying metabolic concentration of body components, used for implementing the method for identifying metabolic concentration of body components according to any one of claims 1 to 7, characterized in that: include: The first model building module (101) is used to build a neural network model for identifying and classifying metabolic concentrations of body components; A training data set acquisition module (102) is used to acquire a plurality of reagent reaction colors with different gradients as sample images, and to obtain a training data set corresponding to the in vivo component metabolic concentration recognition and classification neural network model after processing the sample images; The second model building module (103) is used to obtain a neural network model for numerical prediction of metabolic concentration identification of in vivo components based on the interaction between the in vivo component metabolic concentration identification and classification neural network model and the training data set; extract the last layer of the convolution layer of the in vivo component metabolic concentration identification and classification neural network model as a feature, divide the convolution layer according to different hole positions to extract the features corresponding to each hole, convolve the model weights with the corresponding features and reduce them to one-dimensional data, perform numerical training on the one-dimensional data of different samples with different gradients, and perform linear fitting with the corresponding concentration values ​​to obtain weight values; replace the last layer of the original deep network model with the weight values, thereby realizing the conversion of the deep neural network model from classification prediction to numerical prediction; The metabolic concentration identification module (104) is used to obtain a real-time reaction color image corresponding to the metabolic concentration of the body component and input it into the body component metabolic concentration identification numerical prediction neural network model, and the output of the body component metabolic concentration identification numerical prediction neural network model is used as the body component metabolic concentration.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores a plurality of instructions, and the processor is used to read the instructions and execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a plurality of instructions, and the plurality of instructions can be read by a processor to execute the method according to any one of claims 1 to 7.

Images