Method and system for realizing head and neck tumor immunohistochemical classification by using hyperspectrum

Abstract

The invention discloses a method and a system for realizing head and neck tumor immunohistochemical classification by using hyperspectrum, and relates to the field of image processing, and the method comprises the following steps: labeling a hyperspectral image corresponding to a head and neck tumor pathological section sample to obtain pixel-level immunohistochemical indexes including epithelium, interstitial, negative, weak positive and strong positive; inputting the hyperspectral image corresponding to the head and neck tumor pathological section sample into a preset deep learning classification framework, and performing training by taking a corresponding immunohistochemical index as a label to obtain an immunohistochemical index prediction model; the preset deep learning classification framework is a lightweight encoder-decoder network architecture integrating a waveband selection strategy, a self-adaptive local and global feature fusion strategy and a self-adaptive channel fusion strategy; and inputting a hyperspectral image corresponding to a to-be-detected head and neck tumor pathological section into the immunohistochemical index prediction model to obtain a corresponding pixel-level immunohistochemical index. According to the invention, rapid and accurate immunohistochemical index prediction can be realized.

Description

Technical Field

[0001] The present application relates to the field of image processing, and in particular to a method and system for realizing immunohistochemical classification of head and neck tumors using hyperspectral. Background Art

[0002] Head and neck tumors are malignant tumors that occur on the mucosal surfaces of the oral cavity, nasopharynx, pharynx (nasopharynx, oropharynx, and hypopharynx), and larynx. In the past few decades, the five-year survival rate of patients with advanced head and neck tumors has remained low, at approximately 40%-50%. In order to improve the prognosis of these patients, early diagnosis and effective treatment are both essential. Pathological histology is the "gold standard" for clinical diagnosis of cancer. In head and neck tumors, the expression of some immunohistochemical marker proteins has been shown to be a marker of prognosis. However, immunohistochemical markers are easily affected by the depth of background staining, and the presence, expression, and distribution of target molecules are also easily affected by the pathologist's reading experience. Therefore, there is an urgent need for an auxiliary tool to help pathologists improve the quality and efficiency of their work. Summary of the invention

[0003] The purpose of this application is to provide a method and system for realizing immunohistochemical classification of head and neck tumors using hyperspectral, which can realize rapid and accurate prediction of immunohistochemical indexes.

[0004] To achieve the above objectives, this application provides the following solutions: In a first aspect, the present application provides a method for realizing immunohistochemical classification of head and neck tumors using hyperspectral, comprising the following steps.

[0005] Acquire multiple head and neck tumor pathology section samples and corresponding hyperspectral images.

[0006] The hyperspectral image corresponding to each head and neck tumor pathological section sample is annotated to obtain corresponding pixel-level immunohistochemical indicators; the pixel-level immunohistochemical indicators include epithelial, stroma, negative, weakly positive and strongly positive.

[0007] The hyperspectral image corresponding to each of the head and neck tumor pathology section samples is input into a preset deep learning classification framework, and the immunohistochemical indexes corresponding to the head and neck tumor pathology section samples are used as labels for training to obtain an immunohistochemical index prediction model; wherein the preset deep learning classification framework is a lightweight encoder-decoder network architecture that integrates a band selection strategy, an adaptive local and global feature fusion strategy, and an adaptive channel fusion strategy.

[0008] The hyperspectral image corresponding to the head and neck tumor pathological section to be detected is input into the immunohistochemical index prediction model for prediction to obtain the corresponding pixel-level immunohistochemical index.

[0009] In a second aspect, the present application provides a system for realizing immunohistochemical classification of head and neck tumors using hyperspectral spectrum, including the following multiple modules.

[0010] The acquisition module is used to obtain multiple head and neck tumor pathological section samples and corresponding hyperspectral images.

[0011] The labeling module is used to label the hyperspectral image corresponding to each head and neck tumor pathological section sample to obtain the corresponding pixel-level immunohistochemical index; the pixel-level immunohistochemical index includes epithelial, stroma, negative, weak positive and strong positive.

[0012] The training module is used to: input the hyperspectral image corresponding to each of the head and neck tumor pathological section samples into a preset deep learning classification framework, and train with the immunohistochemical indicators corresponding to the head and neck tumor pathological section samples as labels to obtain an immunohistochemical indicator prediction model; wherein the preset deep learning classification framework is a lightweight encoder-decoder network architecture that integrates a band selection strategy, an adaptive local and global feature fusion strategy, and an adaptive channel fusion strategy.

[0013] The prediction module is used to: input the hyperspectral image corresponding to the head and neck tumor pathological section to be detected into the immunohistochemical index prediction model for prediction to obtain the corresponding pixel-level immunohistochemical index.

[0014] According to the specific embodiments provided by the present application, the present application has the following technical effects: the present application provides a method and system for realizing immunohistochemical classification of head and neck tumors using hyperspectral, inputting the hyperspectral image corresponding to the head and neck tumor pathological section sample into the preset deep learning classification framework, and training with immunohistochemical indicators as labels. Among them, the preset deep learning classification framework set in the present application is a lightweight encoder-decoder network architecture that integrates band selection strategy, adaptive local and global feature fusion strategy, and adaptive channel fusion strategy, which can realize the training of sparse pixel annotations. The band selection strategy adopted in the framework can identify a subset of bands with significant discriminatory power; then, through the setting of the adaptive local and global feature fusion strategy, the classification performance is effectively improved by enhancing the connection between pixels; at the same time, the adaptive channel fusion strategy is adopted, combining local and global features, so as to achieve high-precision classification. So far, the immunohistochemical indicator prediction model trained by the present application can quickly and accurately predict the immunohistochemical indicators of the hyperspectral image corresponding to the head and neck tumor pathological section to be detected. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 This is a diagram of the application environment of a method for realizing immunohistochemical classification of head and neck tumors using hyperspectral in one embodiment of the present application.

[0017] Figure 2 A flowchart of a method for realizing immunohistochemical classification of head and neck tumors using hyperspectral imaging is provided in accordance with an embodiment of the present application.

[0018] Figure 3 A schematic diagram of a preset deep learning classification framework provided for an embodiment of the present application.

[0019] Figure 4 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0021] The spatial resolution spectral imaging obtained by hyperspectral imaging can provide diagnostic information about tissue physiology, morphology and composition. Hyperspectral imaging systems can accurately capture color information using narrowband filters and are not affected by super-fragmentation phenomena similar to RGB images. Therefore, the application of hyperspectral imaging technology to evaluate immunohistochemically stained tissue sections to improve the quantification of markers that may be relevant in diagnosis or prognosis can be a new detection method.

[0022] Based on this, the present application provides a method for predicting immunohistochemical indices of head and neck tumor pathological sections using hyperspectral imaging technology to solve the problem of insufficient accuracy of current immunohistochemical diagnosis.

[0023] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0024] The method for realizing immunohistochemical classification of head and neck tumors using hyperspectral provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set up separately, or it can be integrated on the server 104, or it can be placed on the cloud or other servers. The terminal 102 can send the hyperspectral image corresponding to the head and neck tumor pathological section to be detected to the server 104. After the server 104 receives it, it inputs it into the immunohistochemical index prediction model for prediction to obtain the corresponding pixel-level immunohistochemical index. The server 104 can feedback the obtained pixel-level immunohistochemical index to the terminal 102.

[0025] In an exemplary embodiment, Figure 2 As shown, a method for realizing immunohistochemical classification of head and neck tumors using hyperspectral is provided. The method is executed by a computer device, and specifically can be executed by a computer device such as a terminal or a server alone, or can be executed by a terminal and a server together. In the embodiment of the present application, the method is applied to Figure 1 The server 104 in the example is used for explanation, and the steps include the following steps 201 to 204.

[0026] Step 201 : Acquire multiple head and neck tumor pathological slice samples and corresponding hyperspectral images.

[0027] In an application example, the process of obtaining multiple head and neck tumor pathology section samples includes: obtaining multiple head and neck tumor pathology sections of different immunohistochemicals and screening them to obtain multiple head and neck tumor pathology section samples.

[0028] Specifically, pathological tissue sections of patients with multiple immunohistochemical labels (P16, EGFR, P63, P53, EGFR) of head and neck tumors who have been treated in a hospital within a preset period of time (such as half a year, three years, five years, etc.) and have undergone radical surgery can be obtained, totaling 53 cases. 43 of them can be used as training sample sets for extracting hyperspectral information and building artificial intelligence classification models in subsequent steps, and the remaining 10 cases can be used as test sample sets for evaluation. So far, multiple head and neck tumor pathological sections with different immunohistochemistry have been obtained. Then, experienced pathologists evaluate the reaction results of the head and neck tumor pathological sections in the training sample set without knowing the basic information of the patients and clinical related data, and screen the head and neck tumor pathological sections according to the obtained reaction results, and obtain multiple head and neck tumor pathological section samples in the training sample set.

[0029] In another application example, the process of acquiring the hyperspectral image corresponding to the head and neck tumor pathological section sample includes: (11) Assembling a hyperspectral microscopic imaging system; the hyperspectral microscopic imaging system comprises a microscope, a scanning acquisition component and a halogen illumination source; wherein the microscope is an upright imaging microscope, the halogen illumination source is a 12V / 100W halogen illumination source, the scanning acquisition component is an integration of a liquid crystal tunable filter and a CCD (charge-coupled device) camera, and the CCD camera is a 12-bit CCD camera. In addition, the hyperspectral microscopic imaging system may also include a hyperspectral data processing terminal.

[0030] (12) Placing the head and neck tumor pathology slice sample on the stage of the microscope, irradiating the head and neck tumor pathology slice sample with the halogen illumination source, and collecting multi-band spectral information through the scanning and acquisition component while the microscope automatically focuses to obtain an initial hyperspectral image. Specifically, after the microscope automatically focuses, the entire head and neck tumor pathology slice sample is scanned. During the scanning process, the CCD camera collects spectral information of 260 uniform bands in the spectral range of 400nm-1000nm in the head and neck tumor pathology slice sample through the CCD sensor with a resolution of 2.3nm.

[0031] (13) Performing black-and-white calibration and normalization processing on the initial hyperspectral image to obtain a hyperspectral image corresponding to the head and neck tumor pathological section sample. In order to avoid interference from ambient light, the following formula is used when performing black-and-white calibration and normalization processing on the initial hyperspectral image: .

[0032] in, I ref is the reflection spectrum intensity value obtained after black and white calibration normalization processing, I Raw is the reflected spectrum intensity value in the initial hyperspectral image, I dark is the dark current reference intensity, I White is the total reflection reference intensity.

[0033] Step 202 , annotating the hyperspectral image corresponding to each of the head and neck tumor pathological section samples to obtain corresponding pixel-level immunohistochemical indicators; the pixel-level immunohistochemical indicators include epithelial, stroma, negative, weakly positive and strongly positive.

[0034] In an application example, the step of annotating the hyperspectral image corresponding to each of the head and neck tumor pathological slice samples includes: 1) using ENVI5.6 to annotate each pixel in the hyperspectral image corresponding to the head and neck tumor pathological slice sample to obtain the pixel-level immunohistochemical index corresponding to each pixel; wherein the annotation process is: manually annotated by two head and neck surgeons with more than five years of work experience, and then evaluated and corrected by a pathologist. 2) Counting the number of pixels corresponding to different categories of pixel-level immunohistochemical indicators, such as the number of pixels in the epithelial category, the number of pixels in the stroma category, the number of pixels in the negative category, the number of pixels in the weak positive category, and the number of pixels in the strong positive category. In order to facilitate statistics, different colors are used to represent different categories. For example, for the hyperspectral images of the five immunohistochemical indicators of p16, p53, p63, egfr, and Eber, cyan represents epithelial EPI, blue represents stroma MEM, yellow represents negative, green represents weak positive, and red represents strong positive, as shown in Table 1 below, which is a schematic table of partial pixel number statistics.

[0035] Table 1

[0036] During the labeling process, the negative refers to the staining intensity of any pixel in the image being within the first preset range; the weak positive refers to the staining intensity of any pixel in the image being within the second preset range; the strong positive refers to the staining intensity of any pixel in the image being within the third preset range; the pixels in the stroma and the epithelium refer to the tissue category to which the pixels stained by immunohistochemistry are different.

[0037] Among them, the first preset range, the second preset range and the third preset range are all expressed in the form of a ratio relative to the maximum staining intensity. For example, for a head and neck tumor pathology section sample, the area indicated by the staining index color is the positive control area, the area lacking the staining index is the negative control area, and the tissue-containing area of ​​the entire section is the overall control area. According to the average percentage of immunostained cells and the immunostaining intensity, the global 0-20% (first preset range) staining intensity is defined as negative immunohistochemical indicators (marked as -), 20-50% (second preset range) staining intensity is defined as weakly positive immunohistochemical indicators (marked as +), and greater than 50% staining intensity is defined (third preset range) as strongly positive immunohistochemical indicators (marked as ++).

[0038] Step 203, input the hyperspectral image corresponding to each of the head and neck tumor pathology section samples into a preset deep learning classification framework, and train with the immunohistochemical indexes corresponding to the head and neck tumor pathology section samples as labels to obtain an immunohistochemical index prediction model; wherein the preset deep learning classification framework is a lightweight encoder-decoder network architecture that integrates a band selection strategy, an adaptive local and global feature fusion strategy, and an adaptive channel fusion strategy.

[0039] like Figure 3 As shown, the preset deep learning classification framework includes an encoder network architecture, an adaptive local and global feature fusion layer, and a decoder network architecture, which are arranged in sequence. This architecture can realize the effective classification of microscopic hyperspectral images.

[0040] Specifically, the encoder network architecture includes a band selection layer, a first encoding layer, a second encoding layer and a third encoding layer arranged in sequence; the decoder network architecture includes a third decoding layer, a second decoding layer, a first decoding layer and an output layer arranged in sequence; the first encoding layer is jump-connected with the first decoding layer, the second encoding layer is jump-connected with the second decoding layer; the third encoding layer is jump-connected with the third decoding layer.

[0041] The first coding layer, the second coding layer and the third coding layer all include a spatial spectrum attention layer and a first convolution layer arranged in sequence; the first decoding layer, the second decoding layer and the third decoding layer all include a second convolution layer and a spatial semantic fusion layer arranged in sequence. The first convolution layer is an integration of a convolution layer, a group normalization layer and a ReLu activation layer, and the second convolution layer is a 3 The output layer includes a second convolutional layer and a sub-output layer arranged in sequence; the sub-output layer is 1 1 is an integration of the convolutional layer and the softmax classification layer.

[0042] The band selection layer is used to: use a band selection algorithm based on maximum quality to perform band screening on the received hyperspectral image to obtain a significant band subset, and then perform convolution processing to obtain spectral features. In addition, the layer used in the convolution processing can use the same structure as the first convolution layer.

[0043] According to the settings of the above preset deep learning classification framework, it can be seen that this framework is a new weakly supervised learning framework trained based on sparse pixel annotations. It first identifies the subset of bands with significant discriminative power based on the maximum quality band selection strategy; then, the classification process is completed in a lightweight encoder-decoder network architecture, which integrates an adaptive local and global feature fusion layer to effectively improve the classification performance by enhancing the connection between pixels. At the same time, an adaptive channel fusion strategy is adopted to cleverly combine local and global features, thereby achieving high-precision classification. After the above framework is constructed, the spatial and spectral features of the samples in the training set are trained, and then the trained immunohistochemical index prediction model is tested using the test sample set.

[0044] In the actual test, to ensure the comprehensiveness of the evaluation, 1-3 instances were selected from the microscopic hyperspectral images corresponding to different immunohistochemical methods in the test sample set, and a total of 10 immunohistochemical hyperspectral images were selected for testing. Table 2 lists the immunohistochemical selection types selected for the test dataset.

[0045] Table 2

[0046] During the test, the confusion matrix was used to calculate the three indicators of overall classification accuracy (OA), average classification accuracy (AA) and Kappa coefficient to evaluate the actual effect of immunohistochemical index classification. The prediction results obtained after the above tests are summarized in Table 3, as shown below.

[0047] Table 3

[0048] Step 204 , input the hyperspectral image corresponding to the head and neck tumor pathological section to be detected into the immunohistochemical index prediction model for prediction to obtain the corresponding pixel-level immunohistochemical index.

[0049] In summary, this application combines hyperspectral imaging technology with artificial intelligence classification algorithms to predict the immunohistochemical indices of head and neck tumor pathological sections, and predicts the presence and intensity of five immunohistochemical indices related to the prognosis of head and neck tumors, so as to improve the diagnostic accuracy of immunohistochemical indices of head and neck tumors, which is of great significance for the accurate classification and individualized, decision-making diagnosis and treatment of head and neck tumors.

[0050] Based on the same inventive concept, the embodiment of the present application also provides a system for realizing immunohistochemical classification of head and neck tumors using hyperspectral. The implementation scheme for solving the problem provided by the system is similar to the implementation scheme recorded in the above method, so the specific limitations of one or more system embodiments provided below can refer to the limitations of the method above, and will not be repeated here. The system for realizing immunohistochemical classification of head and neck tumors using hyperspectral includes an acquisition module, a labeling module, a training module and a prediction module.

[0051] The acquisition module is used to obtain multiple head and neck tumor pathological section samples and corresponding hyperspectral images.

[0052] The labeling module is used to label the hyperspectral image corresponding to each head and neck tumor pathological section sample to obtain the corresponding pixel-level immunohistochemical index; the pixel-level immunohistochemical index includes epithelial, stroma, negative, weak positive and strong positive.

[0053] The training module is used to: input the hyperspectral image corresponding to each of the head and neck tumor pathology section samples into a preset deep learning classification framework, and train with the immunohistochemical indicators corresponding to the head and neck tumor pathology section samples as labels to obtain an immunohistochemical indicator prediction model; wherein the preset deep learning classification framework is a lightweight encoder-decoder network architecture that integrates a band selection strategy, an adaptive local and global feature fusion strategy, and an adaptive channel fusion strategy.

[0054] The prediction module is used to: input the hyperspectral image corresponding to the head and neck tumor pathological section to be detected into the immunohistochemical index prediction model for prediction to obtain the corresponding pixel-level immunohistochemical index.

[0055] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 4 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for realizing immunohistochemical classification of head and neck tumors using hyperspectral is implemented.

[0056] Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components. In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0057] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0058] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0059] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0060] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0061] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0062] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0063] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for realizing immunohistochemical classification of head and neck tumors using hyperspectral, characterized in that: The method for realizing immunohistochemical classification of head and neck tumors using hyperspectral spectrum comprises: Acquire multiple head and neck tumor pathology slice samples and corresponding hyperspectral images; Annotating the hyperspectral image corresponding to each of the head and neck tumor pathological section samples to obtain corresponding pixel-level immunohistochemical indicators; the pixel-level immunohistochemical indicators include epithelial, stroma, negative, weakly positive and strongly positive; The hyperspectral image corresponding to each of the head and neck tumor pathological section samples is input into a preset deep learning classification framework, and the pixel-level immunohistochemical index corresponding to the head and neck tumor pathological section sample is used as a label for training to obtain an immunohistochemical index prediction model; wherein the preset deep learning classification framework is a lightweight encoder-decoder network architecture that integrates a band selection strategy, an adaptive local and global feature fusion strategy, and an adaptive channel fusion strategy; The hyperspectral image corresponding to the head and neck tumor pathological section to be detected is input into the immunohistochemical index prediction model for prediction to obtain the corresponding pixel-level immunohistochemical index.

2. The method for realizing immunohistochemical classification of head and neck tumors using hyperspectral according to claim 1, characterized in that: The preset deep learning classification framework includes an encoder network architecture, an adaptive local and global feature fusion layer, and a decoder network architecture arranged in sequence; The encoder network architecture includes a band selection layer, a first encoding layer, a second encoding layer, and a third encoding layer arranged in sequence; the decoder network architecture includes a third decoding layer, a second decoding layer, a first decoding layer, and an output layer arranged in sequence; the first encoding layer is jump-connected with the first decoding layer, the second encoding layer is jump-connected with the second decoding layer; the third encoding layer is jump-connected with the third decoding layer; The first encoding layer, the second encoding layer and the third encoding layer each include a spatial spectrum attention layer and a first convolution layer arranged in sequence; the first decoding layer, the second decoding layer and the third decoding layer each include a second convolution layer and a spatial semantic fusion layer arranged in sequence; The band selection layer is used to: for the received hyperspectral image, use a band selection algorithm based on maximum quality to perform band screening to obtain a significant band subset, and then perform convolution processing to obtain spectral features.

3. The method for realizing immunohistochemical classification of head and neck tumors using hyperspectral according to claim 1, characterized in that: The negative refers to the staining intensity of any pixel in the image being within the first preset range; the weak positive refers to the staining intensity of any pixel in the image being within the second preset range; the strong positive refers to the staining intensity of any pixel in the image being within the third preset range; the pixels in the stroma and the epithelium refer to the tissue categories that are different from the pixels stained by immunohistochemistry.

4. The method for realizing immunohistochemical classification of head and neck tumors using hyperspectral spectrum according to claim 1, characterized in that: The step of annotating the hyperspectral image corresponding to each head and neck tumor pathological section sample comprises: ENVI5.6 was used to annotate each pixel in the hyperspectral image corresponding to the head and neck tumor pathological section sample to obtain the pixel-level immunohistochemical index corresponding to each pixel; the number of pixels corresponding to different categories of pixel-level immunohistochemical indexes was counted.

5. The method for realizing immunohistochemical classification of head and neck tumors using hyperspectral according to claim 1, characterized in that: The process of obtaining multiple head and neck tumor pathology section samples includes: obtaining multiple head and neck tumor pathology sections of different immunohistochemicals and screening them to obtain multiple head and neck tumor pathology section samples.

6. The method for realizing immunohistochemical classification of head and neck tumors using hyperspectral spectrum according to claim 1, characterized in that: The process of acquiring the hyperspectral image corresponding to the head and neck tumor pathological section sample includes: Assembling a hyperspectral microscopic imaging system; the hyperspectral microscopic imaging system comprises a microscope, a scanning acquisition component and a halogen illumination source; Placing the head and neck tumor pathological slice sample on the stage of the microscope, irradiating the head and neck tumor pathological slice sample with the halogen illumination source, and collecting multi-band spectral information through the scanning and acquisition component while the microscope automatically focuses to obtain an initial hyperspectral image; The initial hyperspectral image is subjected to black-and-white calibration and normalization processing to obtain a hyperspectral image corresponding to the head and neck tumor pathological section sample.

7. The method for realizing immunohistochemical classification of head and neck tumors using hyperspectral spectrum according to claim 6, characterized in that: When performing black and white calibration normalization processing on the initial hyperspectral image, the following formula is used: ； in, I ref is the reflection spectrum intensity value obtained after black and white calibration normalization processing, I Raw is the reflected spectrum intensity value in the initial hyperspectral image, I dark is the dark current reference intensity, I White is the total reflection reference intensity.

8. A system for realizing immunohistochemical classification of head and neck tumors using hyperspectral, characterized in that: The system for realizing immunohistochemical classification of head and neck tumors using hyperspectral spectrum comprises: An acquisition module is used to: acquire multiple head and neck tumor pathological section samples and corresponding hyperspectral images; The labeling module is used to label the hyperspectral image corresponding to each head and neck tumor pathological section sample to obtain the corresponding pixel-level immunohistochemical index; the pixel-level immunohistochemical index includes epithelial, stroma, negative, weak positive and strong positive; A training module is used to: input the hyperspectral image corresponding to each of the head and neck tumor pathological section samples into a preset deep learning classification framework, and use the immunohistochemical indexes corresponding to the head and neck tumor pathological section samples as labels for training to obtain an immunohistochemical index prediction model; wherein the preset deep learning classification framework is a lightweight encoder-decoder network architecture that integrates a band selection strategy, an adaptive local and global feature fusion strategy, and an adaptive channel fusion strategy; The prediction module is used to: input the hyperspectral image corresponding to the head and neck tumor pathological section to be detected into the immunohistochemical index prediction model for prediction to obtain the corresponding pixel-level immunohistochemical index.

Images