Feature extraction method, system and equipment for microbeads in high-density gene chip and medium

Abstract

The invention provides a method, a system and equipment for extracting features of microbeads in a high-density gene chip and a medium. The method comprises the following steps: acquiring a chip scanning image of the microbeads of which the features are to be extracted; inputting the chip scanning image into a constructed micro-bead brightness and darkness classification model for brightness and darkness degree classification; and according to the output weight of the trained bead brightness classification model, establishing a mapping relationship with the brightness degree classification result of the beads, generating a feature extraction kernel of the beads of which the features are to be extracted, and based on the feature extraction kernel, performing feature extraction on the beads of which the features are to be extracted. According to the method, the feature extraction kernel is constructed by constructing the brightness and darkness classification model of the microbeads and outputting the mapping relation between different weights and the brightness and darkness degrees of the microbeads, so that the microbeads are more adaptive to actual data during feature extraction, and the influence of subjective factors during manual setting of weighted mean weights is also avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of biochips, and particularly to a method, system, device and medium for extracting features of microbeads in a high-density gene chip. Background Art

[0002] The chip scan image (grayscale image) is the most original image of the chip quality. The grayscale value of the points on the image reflects the fluorescence intensity information of the probes at the corresponding chip positions. Through the grayscale image, preliminary quality control of the chip can be carried out. For each chip scan image, after successfully locating the central pixel coordinates of each microsphere in the image, if the grayscale representation of these microspheres is required, the most direct method is to take a rectangular area or a circular area with a suitable side length or diameter centered on the central pixel coordinates of each microbead, and calculate the average grayscale value of all pixels in this area, so as to achieve feature extraction.

[0003] However, in the prior art, the selection of the area is usually manually selected, and when calculating the average value of each pixel in the pre-fetch as the grayscale value, since each pixel value is not weighted, the pixel values of the pixels farther / closer to the center contribute equally to the final average value, which is obviously unreasonable. By the way of manually presetting a threshold that decreases from large to small according to the distance of the pixels from the center point, strong subjective factors of people will be introduced, resulting in inaccurate final results and lack of robustness. Therefore, it is necessary to design a more reasonable, accurate and robust feature extraction method that adapts to the actual data. Summary of the Invention

[0004] In order to solve the above technical problems, the present invention provides a method, system, device and medium for extracting features of microbeads in a high-density gene chip. By learning from a large amount of historical data, appropriate weights are obtained to weight each pixel value in the microbead area, making the feature extraction of the microbeads more adaptable to the actual data situation and avoiding the influence of subjective factors when setting the weighted average weight manually.

[0005] In the first aspect, the present invention provides a method for extracting features of microbeads in a high-density gene chip, including: Obtaining a chip scan image of the microbead whose features are to be extracted; Inputting the chip scan image into a constructed microbead bright-dark classification model for bright-dark degree classification; Establishing a mapping relationship between different weights output by the trained microbead bright-dark classification model and the bright-dark degree classification results, generating a feature extraction kernel for the microbead whose features are to be extracted, and performing feature extraction on the microbead whose features are to be extracted based on the feature extraction kernel; The weight is a parameter after the microbead bright-dark classification model is trained.

[0006] According to a specific embodiment of the present invention, the construction process of the bead bright-dark classification model includes the following steps: Select the beads with successful decoding from different high-density gene chips, obtain the bead regions with the same size as the desired feature extraction kernel and the corresponding label values on each bead with successful decoding, and use them as the data set; Construct a classification model architecture based on a weight extraction layer and an activation function; Input the data set into the classification model architecture for training on the bright-dark degree classification of beads, and then complete the construction of the bead bright-dark classification model; Among them, the weight extraction layer processes the data set based on a single-layer fully connected layer or a single-layer convolutional layer and outputs a feature vector or a feature map as the output result. After passing the output result through the activation function, the bright-dark classification of beads is performed, and the bead bright-dark classification model is trained with the classification result as the optimization target. After the training is completed, the parameters of the weight extraction layer are taken as weights to generate a feature extraction kernel with a mapping relationship between different weights and the bright-dark degree of beads.

[0007] According to a specific embodiment of the present invention, the step of the weight extraction layer generating output results with different weights based on a single-layer fully connected layer for processing the data set includes, during training, converting the result of processing the data set through the single-layer fully connected layer from a one-dimensional form to a two-dimensional form to generate output results based on different weights, where the number of nodes in the single-layer fully connected layer is set to 1.

[0008] According to a specific embodiment of the present invention, the step of the weight extraction layer generating output results with different weights based on a single-layer convolutional layer for processing the data set includes: during training, directly using the result of processing the data set through the single-layer convolutional layer as the output result based on different weights.

[0009] According to a specific embodiment of the present invention, the output bright-dark degree of beads includes the ratio r between the bright beads and the dark beads of the beads, where the bright bead error is 1 / r of the dark bead error, and the loss function is set to one of the BCE loss function, the Focal loss function, or the Contrastive loss function.

[0010] According to a specific embodiment of the present invention, the size of the feature extraction kernel generated during training is a preset size, and all the weights in the feature extraction kernel are multiplied by a fixed constant so that the sum of all the weights is 1.

[0011] According to a specific embodiment of the present invention, the activation function is one of the sigmoid function, the softsign function, or the swish function.

[0012] Second aspect, the present invention further provides a system for extracting features of microbeads in a high-density gene chip. The feature extraction system is used to implement the method for extracting features of microbeads in a high-density gene chip as described in the first aspect. The system includes: An acquisition module, configured to acquire a chip scan image of the microbeads whose features are to be extracted; A processing module, configured to pre-construct a microbead light-dark classification model, and input the chip scan image into the microbead light-dark classification model for light-dark degree classification; An extraction module, configured to establish a mapping relationship between different weights output by the trained microbead light-dark classification model and the light-dark degree classification result, generate a feature extraction kernel for the microbeads whose features are to be extracted, and perform feature extraction on the microbeads whose features are to be extracted based on the feature extraction kernel; Wherein, the weight is a parameter after the microbead light-dark classification model is trained.

[0013] Third aspect, the present invention further provides an electronic device, which includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for extracting features of microbeads in a high-density gene chip as described in the first aspect.

[0014] Fourth aspect, the present invention further provides a non-transitory computer-readable storage medium, which stores computer instructions for causing the computer to execute the method for extracting features of microbeads in a high-density gene chip as described in the first aspect.

[0015] In summary, the present invention provides a method, system, device and medium for extracting features of microbeads in a high-density gene chip, including: acquiring a chip scan image of the microbeads whose features are to be extracted; inputting the chip scan image into the constructed microbead light-dark classification model for light-dark degree classification; establishing a mapping relationship between different weights output by the trained microbead light-dark classification model and the light-dark degree classification result, generating a feature extraction kernel for the microbeads whose features are to be extracted, and performing feature extraction on the microbeads whose features are to be extracted based on the feature extraction kernel. By constructing a feature extraction kernel based on the mapping relationship between different weights output by the microbead light-dark classification model and the light-dark degree of the microbeads, the present invention makes the microbeads more adaptable to the actual data situation during feature extraction, and also avoids the influence of subjective factors when setting the weighted mean weight artificially. Description of the Drawings

[0016] To more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the accompanying drawings required for use in the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present disclosure. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0017] Figure 1 It is a schematic flowchart of the method for feature extraction of microbeads in the high-density gene chip provided by the present invention; Figure 2 It is a schematic application diagram of the feature extraction kernel in the method for feature extraction of microbeads in the high-density gene chip provided by the present invention; Figure 3 It is a schematic framework diagram of the system for feature extraction of microbeads in the high-density gene chip provided by the present invention. Specific Embodiments

[0018] The following uses specific specific examples to illustrate the embodiments of the present disclosure. Those skilled in the art can easily understand other advantages and effects of the present disclosure from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all of the embodiments. The present disclosure can also be implemented or applied through other different specific embodiments. The details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present disclosure. It should be noted that, without conflict, the following embodiments and the features in the embodiments can be combined with each other. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present disclosure without creative efforts belong to the scope of protection of the present disclosure.

[0019] It should be noted that the following describes various aspects of the embodiments within the scope of the appended claims. It should be obvious that the aspects described herein can be embodied in a wide variety of forms, and any specific structure and / or function described herein is illustrative only. Based on the present disclosure, those skilled in the art should understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects described herein can be used to implement the device and / or practice the method. Additionally, this device and / or this method can be implemented using other structures and / or functionality in addition to one or more of the aspects described herein.

[0020] In addition, in the following description, specific details are provided for the purpose of facilitating a thorough understanding of the examples. However, those skilled in the art will understand that the aspects can be practiced without these specific details.

[0021] Reference Figure 1, an embodiment of the present application provides a method for extracting features of microbeads in a high-density gene chip, including: S101. Obtain a chip scan image of the microbeads whose features are to be extracted; S102. Input the chip scan image into a constructed microbead light-dark classification model for light-dark degree classification; S103. Establish a mapping relationship between different weights output by the trained microbead light-dark classification model and the light-dark degree classification results, generate a feature extraction kernel for the microbeads whose features are to be extracted, and perform feature extraction on the microbeads whose features are to be extracted based on the feature extraction kernel; wherein, the weight is a parameter after the microbead light-dark classification model is trained.

[0022] In the embodiment of the present application, the light-dark degree classification of microbeads in a high-density gene chip is realized by constructing a microbead light-dark classification model. The implementation of the above steps is based on a specific application example after the microbead light-dark classification model is constructed. After the model is constructed, it can realize the light-dark degree classification of the obtained chip scan image. As the technical content for realizing this embodiment, the specific construction process of the microbead light-dark classification model will be elaborated below.

[0023] It should be noted that the construction process of the microbead light-dark classification model in this embodiment includes the production of a data set, the construction of a classification model architecture, and the model training process, which will be specifically elaborated below.

[0024] In this embodiment, microbeads with successful decoding are selected from different high-density gene chips, and a microbead region with the same size as the expected feature extraction kernel and the corresponding label value are obtained on each microbead with successful decoding, and used as a data set. Generally, the microbeads in a high-density gene chip are usually magnetic or fluorescent microspheres with a diameter of 3-5 and specific probes such as DNA and antibodies are fixed on their surfaces. Each microbead type corresponds to a unique probe and a unique code, where the unique probe is for a specific gene sequence / protein marker, and the unique code is realized by a combination of fluorescent dyes or physical markers for identity identification. Common decoding modes include fluorescence coding decoding, spatial positioning decoding, and sequence barcode decoding, etc. In this embodiment, the microbeads in a high-density gene chip can be decoded through these methods. After obtaining the microbeads with successful decoding, the microbead region with the same size as the expected feature extraction kernel is used to train the model. At the same time, the label value of the microbeads with successful decoding can also be obtained.

[0025] The data set constructed through the above processing has good quality, and the obtained labels used for training are also relatively reliable, so that different weights can be set according to the number ratio of different categories in the loss function below.

[0026] After the above dataset is constructed, the construction process of the classification model architecture begins. In this embodiment, the classification model architecture is implemented based on a weight extraction layer and an activation function. To achieve the purpose of this embodiment, the weight extraction layer can be set as a single-layer fully connected layer or a single-layer convolutional layer. The parameters after training the single-layer fully connected layer or the large-layer convolutional layer are used as weights, so as to generate output results with different weights.

[0027] In this embodiment, if a single-layer fully connected layer is used, the number of nodes in the single-layer fully connected layer is set to 1. In the input layer, the two-dimensional image data is flattened so that the number of nodes in the input layer is the number of pixels in the bead area. During training, the one-dimensional form output by the single-layer fully connected layer is converted into a two-dimensional form and then the output results with different weights are output.

[0028] In this embodiment, for the single-layer convolutional layer, the spatial structure is usually maintained and no adjustment is required. Therefore, the size of the single-layer convolutional layer can be set to be the same as the size of the bead area, and the two-dimensional image data is directly input in the input layer. During training, the result after processing the dataset by the single-layer convolutional layer is directly used as the output result with different weights. It should be noted that the weight extraction layer in this embodiment can be the single-layer fully connected layer, the single-layer convolutional layer as described above, or a logistic regression model. The purpose of the present invention is to establish a mapping relationship between different weights and the brightness and darkness of the beads by using the obtained model parameters as weights. Therefore, in practical applications, the single-layer fully connected layer, the single-layer convolutional layer, or the logistic regression model can all be optional choices for the weight extraction layer to achieve the purpose of the present invention.

[0029] After the above process, the weight extraction layer processes the dataset based on the single-layer fully connected layer or the single-layer convolutional layer and outputs a feature vector or a feature map, and uses the output feature vector or feature map as the output result. On the one hand, after the output result is binary classified through the activation function, the classification result of the brightness and darkness of the beads can be generated, and the entire model is trained with the classification result as the optimization target. Since the selected training data comes from the beads with successful decoding that meet the experimental expectations, the ratio r between the bright beads and the dark beads in the output brightness and darkness of the beads is included. Among them, the bright bead error is 1 / r of the dark bead error, and the loss function is set as the BCE loss function. It should be noted that the classification of the brightness and darkness of the beads is based on the gray value of each bead with successful decoding in the dataset as the input image and the result label, and is trained and extracted to achieve the classification of the brightness and darkness of the beads. This process can classify the brightness and darkness of the beads in any way, and this embodiment does not make specific limitations.

[0030] On the other hand, after the training is completed, the parameters of a single fully connected layer or a single convolutional layer are taken out as weights and converted into the desired feature extraction kernel size. At the same time, all weights are multiplied by a fixed constant so that the sum of all weights in the feature extraction kernel is 1. In this embodiment, a mapping relationship is established between the weights and the brightness of the microbeads, that is, each weight corresponds to the brightness value of a microbead. A large weight value can be set to correspond to the high brightness value of the microbead. An example will be given below. At this point, the feature extraction kernel completes the generation process of the training phase. In step S103, after the microbeads to be extracted are obtained, the brightness classification results are generated after the microbead brightness classification model is used. According to the parameters output by the trained model and the brightness values corresponding to the weights, a mapping relationship between the weights and the brightness classification results of the microbeads to be extracted is established to generate the corresponding feature extraction kernel, so as to perform feature extraction according to the feature extraction kernel.

[0031] It should be noted that the activation function in this embodiment can be a sigmoid function, the loss function can be a BCE loss function, the activation function can also be a softsign function or a swish function, the loss function can also be a Focal loss function or a Contrastive loss function, and in practice it can also be other activation functions or other loss functions to achieve the corresponding functions. This embodiment does not make any specific limitations.

[0032] In order to more clearly describe the contents of the embodiments of this application, Figure 2 shows a schematic diagram of the application of feature extraction kernel, such as Figure 2 As shown in the figure, the mapping relationship between the weight distribution in the feature extraction kernel and the brightness of the microbeads is that the closer the pixel is to the center, the higher the weight. Because the central area of the feature extraction kernel also corresponds to the center of each microbead, the pixel value in this area can better reflect the brightness of the microbead itself; the weight of the area slightly farther from the center is lower, because this area corresponds to the edge position of each microbead, and the grayscale value of the bright bead at these positions is attenuated compared to the center of the microbead, so it is not as representative of the grayscale value of the microbead as the center area of the microbead; the weight of the area farthest from the center is even negative, because this area corresponds to the edge position of the microbead or even the background area, which cannot reasonably reflect the brightness of the microbead, and the negative weight here also highlights the weight of the area close to the center.

[0033] Meanwhile, Table 1 shows the test results of comparative tests conducted between the existing common mean value method and the method of this embodiment by setting multiple groups of test data.

[0034] Table 1 In Table 1, the "decoding rate" is defined as the ratio of the number of beads whose decoding results meet the experimental expectations to the total number of all beads on the chip; the "proportion of bead types decoded" is defined as the ratio of the number of bead types successfully decoded (when a certain number of beads of each bead type are decoded successfully, it is considered that the bead type is decoded successfully) to the total number of all bead types in the experiment. After being processed by the feature extraction method described in this embodiment, it has been significantly improved compared with the existing common mean value calculation methods.

[0035] Based on the same inventive concept, the embodiments of the present application also provide a feature extraction system for implementing the feature extraction method of beads in the high-density gene chip involved above. The implementation solutions provided by this system to solve problems are similar to the implementation solutions described in the above method. Therefore, the specific limitations in one or more embodiments of the feature extraction system of beads in the high-density gene chip provided below can refer to the limitations on the feature extraction method in the above text, and will not be repeated here.

[0036] In one embodiment, referring to Figure 3 , this embodiment also provides a feature extraction system 300 for beads in a high-density gene chip, including: An acquisition module 301, configured to acquire a chip scan image of the beads whose features are to be extracted; A processing module 302, configured to pre-construct a bead light-dark classification model, and input the chip scan image into the bead light-dark classification model for light-dark degree classification; An extraction module 303, configured to establish a mapping relationship between different weights output by the trained bead light-dark classification model and the light-dark degree classification result, generate a feature extraction kernel for the beads whose features are to be extracted, and perform feature extraction on the beads whose features are to be extracted based on the feature extraction kernel; Wherein, the weight is a parameter after the bead light-dark classification model is trained.

[0037] In one of the embodiments, the processing module 302 is specifically configured to pre-construct a bead light-dark classification model, including acquiring a data set for model training, constructing a classification model architecture based on a weight extraction layer and an activation function, and inputting the data set into the classification model architecture for light-dark degree classification training of the beads, and then completing the construction of the bead light-dark classification model.

[0038] Figure 3 The specific implementation manners of the system shown can refer to the specific implementation manners in the foregoing method embodiments, and will not be repeated here.

[0039] In addition, the embodiments of the present application also provide an electronic device, including: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method for extracting features of microbeads in a high-density gene chip in the foregoing method embodiments.

[0040] An embodiment of the present application further provides a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the method for extracting features of microbeads in a high-density gene chip in the foregoing method embodiments.

[0041] It should be noted that the computer-readable medium in the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the two. The computer-readable storage medium may be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, semiconductor system, apparatus, or device, or any combination of the above. More specific examples of the computer-readable storage medium may include, but are not limited to, an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0042] In the present disclosure, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In the present disclosure, the computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer-readable signal medium may also be any computer-readable medium other than the computer-readable storage medium, and the computer-readable signal medium may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium may be transmitted by any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0043] The above computer-readable medium may be included in the above electronic device; or it may exist separately without being assembled into the electronic device.

[0044] The above computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to: obtain a chip scan image of the beads from which features are to be extracted; input the chip scan image into a constructed bead brightness and darkness classification model for brightness and darkness classification; establish a mapping relationship between different weights output by the trained bead brightness and darkness classification model and the brightness and darkness classification results, generate a feature extraction kernel for the beads from which features are to be extracted, and perform feature extraction on the beads from which features are to be extracted based on the feature extraction kernel; wherein the weights are parameters after the bead brightness and darkness classification model is trained.

[0045] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages or combinations thereof. The programming languages include object-oriented programming languages such as Java, Smalltalk, C++, and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, executed as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., by connecting through the Internet using an Internet service provider).

[0046] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagram may represent a module, a segment of code, or a part of the code that contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than marked in the accompanying drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and combinations of blocks in the block diagram and / or flowchart, may be implemented by a dedicated hardware-based system for performing the specified functions or operations, or may be implemented by a combination of dedicated hardware and computer instructions.

[0047] It should be understood that the various parts of the present disclosure may be implemented in hardware, software, firmware, or a combination thereof.

[0048] As described above, it is only the specific implementation manner of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present disclosure should be covered by the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure shall be subject to the protection scope of the claims.

Claims

1. A method for feature extraction of microbeads in a high-density gene chip, characterized in that It includes the following steps: Obtain a chip scan image of the beads from which features are to be extracted; Input the chip scan image into the constructed bead bright-dark classification model for bright-dark degree classification; Establish a mapping relationship between different weights output by the trained bead bright-dark classification model and the bright-dark degree classification results, generate a feature extraction kernel for the beads from which features are to be extracted, and perform feature extraction on the beads from which features are to be extracted based on the feature extraction kernel; The weight is a parameter after the bead bright-dark classification model is trained.

2. The method for extracting features of microbeads in a high-density gene chip according to claim 1, wherein The construction process of the bead bright-dark classification model includes the following steps: Select decoded beads from different high-density gene chips, obtain bead regions with the same size as the expected feature extraction kernel and corresponding label values on each decoded bead, and use them as a data set; Construct a classification model architecture based on a weight extraction layer and an activation function; Input the data set into the classification model architecture for bright-dark degree classification training of the beads, and complete the construction of the bead bright-dark classification model; Among them, the weight extraction layer processes the data set based on a single-layer fully connected layer or a single-layer convolutional layer and outputs a feature vector or a feature map as the output result. After passing the output result through the activation function, it performs bright-dark classification of the beads, and trains the bead bright-dark classification model with the classification result as the optimization target. After training is completed, take out the parameters of the weight extraction layer as weights, and generate a feature extraction kernel for the mapping relationship between different weights and the bright-dark degree of the beads.

3. The method for extracting the characteristics of microbeads in the high-density gene chip according to claim 2, wherein The step in which the weight extraction layer processes the data set based on a single-layer fully connected layer to generate output results with different weights includes: during training, convert the result after processing the data set by the single-layer fully connected layer from a one-dimensional form to a two-dimensional form to generate output results based on different weights, where the number of nodes in the single-layer fully connected layer is set to 1.

4. The method for extracting features of microbeads in the high-density gene chip according to claim 2, wherein, The step in which the weight extraction layer processes the data set based on a single-layer convolutional layer to generate output results with different weights includes: during training, directly use the result after processing the data set by the single-layer convolutional layer as the output result based on different weights.

5. The method for feature extraction of microbeads in the high-density gene chip according to claim 3 or 4, characterized in that, The output bright-dark degree of the beads includes the ratio r between the bright beads and the dark beads of the beads. Among them, the bright bead error is 1 / r of the dark bead error, and the loss function is set to one of the BCE loss function, the Focal loss function, or the Contrastive loss function.

6. The method for extracting features of microbeads in the high-density gene chip according to claim 3 or 4, characterized in that, The size of the feature extraction kernel generated during training is a preset size, and all weights in the feature extraction kernel are multiplied by a fixed constant so that the sum of all weights is 1.

7. The method for extracting the characteristics of microbeads in the high-density gene chip according to claim 2, wherein The activation function is one of the sigmoid function, the softsign function, or the swish function.

8. A feature extraction system for beads in a high-density gene chip, the feature extraction system is used to implement the method for extracting features of beads in a high-density gene chip according to any one of claims 1-7, characterized in that The system includes: An acquisition module for acquiring a chip scan image of the beads from which features are to be extracted; A processing module, configured to pre-build a microbead bright-dark classification model, and input the chip scan image into the microbead bright-dark classification model for bright-dark degree classification; An extraction module, configured to establish a mapping relationship between different weights output by the trained microbead bright-dark classification model and the bright-dark degree classification result, generate a feature extraction kernel for the microbead whose features are to be extracted, and perform feature extraction on the microbead whose features are to be extracted based on the feature extraction kernel; wherein the weight is a parameter after the microbead bright-dark classification model is trained.

9. An electronic device, characterized in that, Comprising: At least one processor; And, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor, so that the at least one processor can execute the method for extracting features of microbeads in the high-density gene chip described in any one of the preceding claims 1-7.

10. A non-transitory computer-readable storage medium, characterized in that, This non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to cause the computer to execute the method for extracting features of microbeads in the high-density gene chip described in any one of the preceding claims 1-7.

Images