M protein qualitative identification method and device based on convolutional neural network

Abstract

The invention provides an M protein qualitative identification method and device based on a convolutional neural network, and the method comprises the steps: obtaining to-be-analyzed M protein analysis data, and converting the to-be-analyzed M protein analysis data into a to-be-analyzed image; inputting the to-be-analyzed image into a pre-trained qualitative identification model to obtain a qualitative identification result output by the qualitative identification model; wherein the qualitative identification model is obtained by training an M protein analysis image and an antibody type label based on a pre-constructed convolutional neural network; the M protein analysis image is obtained by preprocessing M protein analysis data. The problems that in M protein qualitative identification, due to the fact that results are not easy to judge, wrong results are reported, the admission threshold of technicians participating in the work is high, and the training time is long are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to a method and device for qualitative identification of M protein based on a convolutional neural network. Background Art

[0002] M protein, namely monoclonal immunoglobulin, is an abnormal immunoglobulin produced by monoclonal malignant proliferation of plasma cells or B lymphocytes, and its essence is an immunoglobulin or a fragment of an immunoglobulin. When qualitatively identifying M protein by capillary immunotyping, the serum to be tested needs to be added into a blank cup and 5 antibody cups respectively. The 5 antibody cups contain IgA heavy chain (α chain) antibody, IgM heavy chain (μ chain) antibody, IgG heavy chain (γ chain) antibody, κ light chain antibody and λ light chain antibody respectively. After the antigen-antibody binding reaction, the samples in these 6 cups are taken for capillary electrophoresis simultaneously, and 6 capillary electrophoresis maps are obtained. Each capillary electrophoresis map consists of 300 points. Plotting these 300 points on a graph forms a curve (as Figure 1 shown). Technicians judge whether the sample contains M protein and the type of M protein by whether the curve has characteristic peak patterns (narrow-bottomed sharp peaks, irregular peak patterns) and the elimination of proteins with characteristic peak patterns in specific antibody cups (single-type elimination).

[0003] This manual interpretation method is highly subjective, and whether the result is correct depends on the experience accumulation and analysis ability of the interpreter.

[0004] In view of this, a method and device for qualitative identification of M protein based on a convolutional neural network are provided, aiming to solve the problems such as the reporting of incorrect results caused by difficult result judgment in the qualitative identification of M protein, the high entry threshold for technicians participating in this work, and the long training time. Summary of the Invention

[0005] The present invention provides a method and device for qualitative identification of M protein based on a convolutional neural network, aiming to solve the technical problems of high labor cost, low analysis efficiency and poor accuracy required in the qualitative identification of M protein.

[0006] The present invention provides a method for qualitative identification of M protein based on a convolutional neural network, and the method includes:

[0007] Obtaining M protein analysis data to be analyzed, and converting the M protein analysis data to be analyzed into an image to be analyzed;

[0008] Inputting the image to be analyzed into a pre-trained qualitative identification model, and obtaining a qualitative identification result output by the qualitative identification model;

[0009] Among them, the qualitative identification model is trained based on a pre-constructed convolutional neural network using M protein analysis images and antibody labels; the M protein analysis images are obtained by preprocessing M protein analysis data.

[0010] In some embodiments, the network structure of the pre-constructed convolutional neural network includes:

[0011] An input layer for receiving input images;

[0012] A convolutional layer with a convolutional kernel size of n×n and a stride greater than or equal to 1;

[0013] A normalization layer;

[0014] An activation layer using a ReLU activation function or a Sigmoid activation function;

[0015] An attention module;

[0016] A pooling layer, which is a max pooling layer or an average pooling layer for downsampling the feature map;

[0017] A fully connected layer. After the pooling layer, it is flattened using Flatten, and a softmax activation function is used in the fully connected layer to output the prediction probability of each antibody type.

[0018] In some embodiments, n takes a value of 1, 3, or 5.

[0019] In some embodiments, based on the pre-constructed convolutional neural network, using M protein analysis images and antibody type labels for training to obtain the qualitative identification model, specifically including:

[0020] Preprocess the obtained M protein analysis data to obtain M protein analysis images;

[0021] Based on the M protein analysis images and antibody type labels, construct a data set;

[0022] Divide the data set into a training set and a validation set;

[0023] Input the sample images and corresponding antibody type labels in the training set into the pre-constructed convolutional neural network for training to obtain a qualitative identification model;

[0024] Use the difference between the output result and the type label as feedback information to optimize the qualitative identification model;

[0025] Evaluate the accuracy of the output result of the qualitative identification model on the validation set.

[0026] In some embodiments, the obtained M protein analysis data is preprocessed to obtain an M protein analysis image, which specifically includes: based on a preset value range, scaling the six groups of capillary electrophoresis map data of the M protein analysis sample respectively, so that the values of the capillary electrophoresis map data fall within the value range;

[0027] Integrating the six groups of scaled capillary electrophoresis map data into arrays respectively;

[0028] Stitching the six capillary electrophoresis map arrays in a preset order to obtain the M protein analysis image.

[0029] In some embodiments, the types of M protein at least include: Aκ type immunoglobulin, Aλ type immunoglobulin, Gκ type immunoglobulin, Gλ type immunoglobulin, Mκ type immunoglobulin, Mλ type immunoglobulin, κ-type half molecule immunoglobulin, λ-type half molecule immunoglobulin, and M protein negative.

[0030] The present invention also provides an M protein qualitative identification device based on a convolutional neural network, and the device includes:

[0031] A data acquisition unit, configured to acquire M protein analysis data to be analyzed and convert the M protein analysis data to be analyzed into an image to be analyzed;

[0032] A result generation unit, configured to input the image to be analyzed into a pre-trained qualitative identification model to obtain a qualitative identification result output by the qualitative identification model;

[0033] Wherein, the qualitative identification model is trained based on a pre-constructed convolutional neural network by using M protein analysis images and antibody type labels; the M protein analysis image is obtained after preprocessing the M protein analysis data.

[0034] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the program, the above-mentioned method is implemented.

[0035] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned method is implemented.

[0036] The present invention also provides a computer program product, including a computer program, and when the computer program is executed by a processor, the above-mentioned method is implemented.

[0037] The M protein qualitative identification method based on a convolutional neural network provided by the present invention obtains M protein analysis data to be analyzed and converts the M protein data to be analyzed into an image to be analyzed; inputs the image to be analyzed into a pre-trained qualitative identification model, and the qualitative identification result output by the qualitative identification model can be obtained; wherein, the qualitative identification model is trained based on a pre-constructed convolutional neural network using M protein analysis images and antibody type labels; the M protein analysis images are obtained after preprocessing the M protein analysis data. This method converts existing data into images, and the images can use deep learning methods in the field of computer vision, making the analysis process more efficient and the analysis results more accurate, and solving the technical problems of high labor cost, low analysis efficiency, and poor accuracy in M protein qualitative identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0039] Figure 1 is a schematic diagram of a capillary electrophoresis map;

[0040] Figure 2 is one of the flowcharts of the M protein qualitative identification method based on a convolutional neural network provided by the present invention;

[0041] Figure 3 is another flowchart of the M protein qualitative identification method based on a convolutional neural network provided by the present invention;

[0042] Figure 4 is a schematic diagram of the network architecture of the convolutional neural network provided by the present invention;

[0043] Figure 5 is a third flowchart of the M protein qualitative identification method based on a convolutional neural network provided by the present invention;

[0044] Figure 6 is a fourth flowchart of the M protein qualitative identification method based on a convolutional neural network provided by the present invention;

[0045] Figure 7 is a schematic diagram of the structure of the M protein qualitative identification device based on a convolutional neural network provided by the present invention;

[0046] Figure 8 is a schematic diagram of the structure of the electronic device provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below with reference to the accompanying drawings in the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without creative efforts shall fall within the protection scope of the present invention.

[0048] In a specific embodiment, as Figure 2 shown, the M protein qualitative identification method based on a convolutional neural network provided by the present invention includes the following steps:

[0049] S210: Obtain M protein analysis data to be analyzed, and convert the M protein data to be analyzed into an image to be analyzed; specifically, preprocessing processes such as data scaling, reshape, and splicing can be used to complete the conversion of M protein analysis data into an image. For example, as Figure 3 shown, the original data can be six capillary electrophoresis maps. During the data preprocessing process, when converting the six original data containing 300 points into an image, some preprocessing of the data is required, mainly through data scaling, reshape, and splicing operations; among them, when scaling the data, since the value range of pixels in the picture is 0 to 255, but the values of these six maps fall outside this range, it is necessary to first constrain the value range of the six maps between 0 and 255; when reshaping (matrix reset), reshape 300 data into a 15x20 array; during the splicing process, splice the 6 curves in the order of 2 rows and 3 columns. The 3 curves in the first row are ELP, κ, and λ respectively, and the 3 curves in the second row are IgA, IgG, and IgM respectively. The size of the composed picture is 30x60.

[0050] S220: Input the image to be analyzed into a pre-trained qualitative identification model to obtain the qualitative identification result output by the qualitative identification model. In this embodiment, the qualitative identification result indicates that the capillary electrophoresis map is divided into 9 categories, namely IgAκ, IgAλ, IgGκ, IgGλ, IgMκ, IgMλ, κ, λ, and Negative (negative). The types of M protein samples include at least the following 9 types, namely Aκ-type immunoglobulin, Aλ-type immunoglobulin, Gκ-type immunoglobulin, Gλ-type immunoglobulin, Mκ-type immunoglobulin, Mλ-type immunoglobulin, κ-type half molecule immunoglobulin, λ-type half molecule immunoglobulin, and M protein negative. Specifically, when qualitatively identifying M protein by capillary immunotyping, the serum to be tested needs to be added to a blank cup and 5 antibody cups respectively. The 5 antibody cups contain IgA heavy chain (α chain) antibody, IgM heavy chain (μ chain) antibody, IgG heavy chain (γ chain) antibody, κ light chain antibody, and λ light chain antibody respectively. After the antigen-antibody binding reaction, the samples in these 6 cups are taken and capillary electrophoresis is performed simultaneously. The M protein is typed and identified by the elimination of the characteristic peak pattern (narrow-bottom sharp peak) in a specific antibody cup.

[0051] Among them, the qualitative identification model is trained based on a pre-constructed convolutional neural network using M protein analysis images and antibody type labels; the M protein analysis images are obtained after preprocessing M protein analysis data.

[0052] In this embodiment, the method provided by the present invention is based on the above convolutional algorithm. Input a picture with a size of 30×60, and output a vector with a size of 1×9, which represents the probabilities that the sample belongs to IgAκ, IgAλ, IgGκ, IgGλ, IgMκ, IgMλ, κ, λ, and negative. Select the type with the highest probability as the type of the sample.

[0053] In some embodiments, 1 image can be obtained through the above data preprocessing process. Next, the above-analyzed picture is classified using a deep learning algorithm. This patent uses a convolutional neural network to realize the typing and identification of M protein through a capillary electrophoresis map. As Figure 4 shown, the network structure of the pre-constructed convolutional neural network includes an input layer, a convolutional layer, a normalization layer, an activation layer, an attention module, a pooling layer, and a fully connected layer.

[0054] Among them, the input layer is used to receive the input image, that is, the image obtained after preprocessing; the convolution kernel size of the convolution layer is n×n, the stride of the convolution layer is greater than or equal to 1, and the value of n is 1, 3, or 5. The convolution layer extracts effective features by learning the parameters of the convolution kernel. The size of the convolution kernel can be selected as 3×3, 5×5, etc. Specifically, a 1×1 convolution kernel can also be selected, which can effectively fuse channel information without changing the spatial dimension. In addition, another important parameter in the convolution layer is the stride. When the stride is greater than 1, it can reduce the spatial dimension, increase the receptive field, and extract more features containing semantic information.

[0055] The activation layer adopts the ReLU activation function or the Sigmoid activation function; since the input image is composed of 6 pieces of data containing 300 points, among the 300 points, the points in some ranges are very important for judging the type, while the points in some ranges can provide relatively less information. Therefore, in addition to the above modules, a channel attention module and a spatial attention module are additionally added, as Figure 4 shown. The attention module means that among N computing units, this module only exists in some computing units. The channel attention module explores the relationship between channels and extracts features according to the importance of each channel, while the spatial attention module aims to explore the importance of features at different spatial positions. These two modules make the extracted features more effective and can further improve the classification accuracy.

[0056] The pooling layer is a maximum pooling layer or an average pooling layer, and the pooling layer is used to downsample the feature map; the pooling layer can choose maximum pooling or average pooling. This module can extract significant features, reduce the spatial dimension, and increase the receptive field. The residual connection layer directly transfers the features of the previous layer to the next layer, which can make the algorithm more stable and achieve higher accuracy. The fully connected layer is usually used to map the features to each category to achieve the classification task. After the pooling layer, Flatten is used for flattening, and the softmax activation function is used in the fully connected layer to output the prediction probability of each antibody type.

[0057] In some embodiments, based on a pre-constructed convolutional neural network, the qualitative identification model is obtained by training with the M protein analysis image and the antibody type label, as Figure 5 shown, and specifically includes the following steps:

[0058] S510: Preprocess the obtained M protein analysis data to obtain the M protein analysis image; as Figure 6 shown, the preprocessing process specifically includes the following steps:

[0059] S610: Based on a preset value range, scale the capillary electrophoresis map data of 6 groups of M protein analysis samples respectively, so that the values of the capillary electrophoresis map data fall within the value range;

[0060] S620: Integrate the 6 groups of scaled capillary electrophoresis map data into arrays respectively;

[0061] S630: Splice the 6 capillary electrophoresis map arrays in a preset order to obtain the M protein analysis image.

[0062] It should be understood that the preprocessing process of the sample data is the same as the preprocessing process of the data to be classified in step S210.

[0063] S520: Based on the M protein analysis image and the antibody type label, construct a data set;

[0064] S530: Divide the data set into a training set and a validation set;

[0065] S540: Input the sample images and the corresponding antibody type labels in the training set into a pre-constructed convolutional neural network for training to obtain a qualitative identification model;

[0066] S550: Optimize the qualitative identification model according to the difference between the output result and the type label as feedback information;

[0067] S560: Evaluate the accuracy of the output result of the qualitative identification model on the validation set.

[0068] That is to say, this algorithm inputs an image, passes through N computing units, then passes through an average pooling layer, and finally passes through a fully connected layer. Specifically, the deep learning algorithm based on a convolutional neural network mainly consists of the following parts, namely the convolutional layer, the pooling layer, the residual connection layer, the fully connected layer, etc. The activation functions involved in the algorithm are mainly Sigmoid, ReLU, etc. The loss functions used in the algorithm are mainly the cross-entropy loss function and the improvement based on the cross-entropy loss function. For example, using the Focal loss function can effectively solve the classification problems of positive and negative samples and easy and difficult samples. Based on the cross-entropy loss function, by introducing additional parameters, the algorithm pays more attention to the samples that are easy to be misclassified, thereby improving the accuracy of the algorithm. The calculation method is shown in Equation (1).

[0069] FL(p t )=-α t (1 - p t ) γ log(p t ) (1)

[0070] Among them, FL(p t ) represents the Focal loss function, α t is the imbalance parameter of positive and negative sample quantities, and γ is the sample difficulty parameter.

[0071] In the above specific implementation manner, the method for qualitative identification of M protein provided by the present invention obtains the M protein analysis data to be analyzed and converts the M protein analysis data to be analyzed into an image to be analyzed; inputs the image to be analyzed into a pre-trained qualitative identification model, and the qualitative identification result output by the qualitative identification model can be obtained; wherein, the qualitative identification model is trained based on a pre-constructed convolutional neural network using M protein analysis images and antibody type labels; the M protein analysis images are obtained after preprocessing the M protein analysis data. This method converts existing data into images, and the images can use deep learning methods in the field of computer vision, making the analysis process more efficient and the analysis results more accurate, solving the technical problems of high labor cost, low analysis efficiency, and poor accuracy in the qualitative identification of M protein.

[0072] In addition to the above method, the present invention also provides a device for qualitative identification of M protein based on a convolutional neural network, as Figure 7 shown, the device includes:

[0073] A data acquisition unit 710, configured to acquire the M protein analysis data to be analyzed and convert the M protein analysis data to be analyzed into an image to be analyzed;

[0074] A result generation unit 720, configured to input the image to be analyzed into a pre-trained qualitative identification model to obtain the qualitative identification result output by the qualitative identification model;

[0075] wherein, the qualitative identification model is trained based on a pre-constructed convolutional neural network using M protein analysis images and antibody type labels; the M protein analysis images are obtained after preprocessing the M protein analysis data.

[0076] In some embodiments, the network structure of the pre-constructed convolutional neural network includes:

[0077] An input layer, which is configured to receive the input image;

[0078] A convolutional layer, the convolutional kernel size of the convolutional layer is n×n, and the stride of the convolutional layer is greater than or equal to 1;

[0079] A normalization layer;

[0080] An activation layer, and the activation layer adopts a ReLU activation function or a Sigmoid activation function;

[0081] Attention module;

[0082] Pooling layer, where the pooling layer is a max - pooling layer or an average - pooling layer, and the pooling layer is used to downsample the feature map;

[0083] Fully - connected layer. After the pooling layer, it is flattened using Flatten, and the softmax activation function is used in the fully - connected layer to output the predicted probability of each antibody category.

[0084] In some embodiments, n takes values of 1, 3, or 5.

[0085] In some embodiments, based on a pre - constructed convolutional neural network, training is performed using M - protein analysis images and antibody type labels to obtain the qualitative identification model, which specifically includes:

[0086] Pre - process the obtained M - protein analysis data to obtain M - protein analysis images;

[0087] Based on the M - protein analysis images and antibody type labels, construct a data set;

[0088] Divide the data set into a training set and a validation set;

[0089] Input the sample images and corresponding antibody type labels in the training set into the pre - constructed convolutional neural network for training to obtain a qualitative identification model;

[0090] Optimize the qualitative identification model according to the difference between the output result and the type label as feedback information;

[0091] Evaluate the accuracy of the output result of the qualitative identification model on the validation set.

[0092] In some embodiments, pre - process the obtained M - protein analysis data to obtain M - protein analysis images, which specifically includes:

[0093] Based on a preset value range, scale the 6 - group capillary electrophoresis map data of the M - protein sample respectively so that the values of the capillary electrophoresis map data fall within the value range;

[0094] Integrate the 6 - group scaled capillary electrophoresis map data into arrays respectively;

[0095] Stitch the 6 capillary electrophoresis map arrays in a preset order to obtain the M - protein analysis image.

[0096] In some embodiments, the types of M protein samples include at least the following nine types, namely immunoglobulin of type Aκ, immunoglobulin of type Aλ, immunoglobulin of type Gκ, immunoglobulin of type Gλ, immunoglobulin of type Mκ, immunoglobulin of type Mλ, half-molecule immunoglobulin of type κ, half-molecule immunoglobulin of type λ, and M protein negative.

[0097] In the above specific embodiments, the M protein qualitative identification device provided by the present invention obtains the M protein analysis data to be analyzed and converts the M protein analysis data to be analyzed into an image to be analyzed; inputs the image to be analyzed into a pre-trained qualitative identification model, and the qualitative identification result output by the qualitative identification model can be obtained; wherein, the qualitative identification model is trained based on a pre-constructed convolutional neural network using M protein analysis images and antibody type labels; the M protein analysis images are obtained after preprocessing the M protein analysis data. This device converts the acquired data into an image, and the image can utilize deep learning methods in the field of computer vision, making the analysis process more efficient and the analysis result more accurate, and solving the technical problems of high labor cost, low analysis efficiency, and poor accuracy in M protein qualitative identification.

[0098] Figure 8 Illustrates a schematic physical structure diagram of an electronic device, as Figure 8 shown, the electronic device may include: a processor 810, a communication interface 820, a memory 830, and a communication bus 840. Among them, the processor 810, the communication interface 820, and the memory 830 complete mutual communication through the communication bus 840. The processor 810 can call the logical instructions in the memory 830 to execute the above method.

[0099] In addition, when the logical instructions in the above-mentioned memory 830 are implemented in the form of a software functional unit and sold or used as an independent product, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical disks, and other various media that can store program codes.

[0100] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the above method.

[0101] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0102] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0103] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present invention.

Claims

1. A method for qualitative identification of M protein based on convolutional neural network, characterized in that: The method comprises: Acquiring M protein analysis data to be analyzed, and converting the M protein analysis data to be analyzed into an image to be analyzed; Inputting the image to be analyzed into a pre-trained qualitative identification model to obtain a qualitative identification result output by the qualitative identification model; Wherein, the qualitative identification model is based on a pre-constructed convolutional neural network, which is trained using M protein analysis images and antibody type labels; the M protein analysis images are obtained after preprocessing using M protein analysis data.

2. The method for qualitative identification of M protein based on convolutional neural network according to claim 1, characterized in that: The network structure of the pre-built convolutional neural network includes: An input layer, wherein the input layer is used to receive an input image; A convolution layer, wherein the convolution kernel size of the convolution layer is n×n, and the step length of the convolution layer is greater than or equal to 1; Normalization layer; An activation layer, wherein the activation layer adopts a ReLU activation function or a Sigmoid activation function; Attention module; A pooling layer, wherein the pooling layer is a maximum pooling layer or a mean pooling layer, and the pooling layer is used to downsample the feature map; The fully connected layer, after the pooling layer, is flattened using Flatten, and a softmax activation function is used in the fully connected layer to output the predicted probability of each antibody type.

3. The method for qualitative identification of M protein based on convolutional neural network according to claim 2, characterized in that: The value of n is 1, 3 or 5.

4. The method for qualitative identification of M protein based on convolutional neural network according to claim 2, characterized in that: Based on the pre-built convolutional neural network, the qualitative identification model is obtained by training using the M protein analysis image and the antibody type label, which specifically includes: Preprocessing the acquired M protein analysis data to obtain an M protein analysis image; Constructing a data set based on the M protein analysis images and antibody type labels; Dividing the data set into a training set and a validation set; Inputting the sample images and corresponding antibody type labels in the training set into a pre-built convolutional neural network for training to obtain a qualitative identification model; The qualitative identification model is optimized based on the difference between the output results and the type label as feedback information; The accuracy of the output results of the qualitative identification model is evaluated on the validation set.

5. The method for qualitative identification of M protein based on convolutional neural network according to claim 4, characterized in that: The acquired M protein analysis data is preprocessed to obtain an M protein analysis image, specifically including: Based on the preset value range, the six groups of capillary electrophoresis pattern data of the M protein analysis sample are scaled respectively so that the values ​​of the capillary electrophoresis pattern data fall within the value range; The 6 sets of scaled capillary electrophoresis pattern data were respectively integrated into arrays; The six capillary electrophoresis pattern arrays are spliced ​​in a preset order to obtain the M protein analysis image.

6. The method for qualitative identification of M protein based on convolutional neural network according to claim 4, characterized in that: The types of M protein samples include at least the following 9 types, namely Aκ type immunoglobulin, Aλ type immunoglobulin, Gκ type immunoglobulin, Gλ type immunoglobulin, Mκ type immunoglobulin, Mλ type immunoglobulin, κ type half molecule immunoglobulin, λ type half molecule immunoglobulin and M protein negative.

7. A device for qualitative identification of M protein based on convolutional neural network, characterized in that: The device comprises: A data acquisition unit, used for acquiring the M protein analysis data to be analyzed, and converting the M protein analysis data to be analyzed into an image to be analyzed; A result generating unit, used for inputting the image to be analyzed into a pre-trained qualitative identification model to obtain a qualitative identification result output by the qualitative identification model; Wherein, the qualitative identification model is based on a pre-constructed convolutional neural network, which is trained using M protein analysis images and antibody type labels; the M protein analysis images are obtained after preprocessing using M protein analysis data.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images