Intelligent quantitative detection method based on neural network algorithm and related equipment

Abstract

The invention discloses an intelligent quantitative detection method based on a neural network algorithm and related equipment. The method comprises the following steps: acquiring a tumor immunohistochemical staining image through a microscope imaging system; extracting key feature information in the tumor immunohistochemical staining image through a target neural network analysis model, wherein the key feature information comprises at least one of form, color and distribution information of positive cells; and the model automatically identifies and classifies the positive cells according to the marked training data, calculates the number and the distribution condition of the cells according to a preset rule, and generates an interpretation report. The method can solve the problem that a traditional interpretation method requires a pathologist to carefully observe a staining section under a microscope, evaluate staining quality, positive signal distribution and morphological characteristics, and count and classify the staining quality, the positive signal distribution and the morphological characteristics. And the process is long in time consumption and is easily influenced by factors such as experience of doctors, fatigue degree and visual field selection, so that the interpretation result is subjective and inconsistent.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of smart medical care. More specifically, the present invention relates to an intelligent quantitative detection method and related equipment based on a neural network algorithm. Background Art

[0002] In the field of tumor pathology diagnosis, immunohistochemical staining is an important detection method. It visualizes the target protein through the combination of specific antibodies and antigens in tissues, thereby assisting doctors in determining the type, stage, prognosis and treatment plan of the tumor. However, the traditional immunohistochemical staining interpretation process is highly dependent on the experience and subjective judgment of the pathologist, and is time-consuming, labor-intensive, and error-prone.

[0003] Traditional interpretation methods require pathologists to carefully observe stained sections under a microscope, evaluate staining quality, positive signal distribution and morphological characteristics, and count and classify them. This process is not only time-consuming, but also easily affected by factors such as doctor experience, fatigue, and field of view selection, resulting in subjectivity and inconsistency in interpretation results. In addition, with the increase in the number of tumor cases and the increase in the complexity of pathological diagnosis, the demand and workload for pathologists are also increasing, further exacerbating the challenges of interpretation efficiency and accuracy. Summary of the invention

[0004] A series of simplified concepts are introduced in the Summary of the Invention, which will be further described in detail in the Detailed Description of the Invention. The Summary of the Invention does not mean to attempt to define the key features and essential technical features of the claimed technical solution, nor does it mean to attempt to determine the scope of protection of the claimed technical solution.

[0005] In order to solve the problem that the traditional interpretation method requires pathologists to carefully observe the stained sections under a microscope, evaluate the staining quality, positive signal distribution and morphological characteristics, and count and classify. This process is not only time-consuming, but also easily affected by factors such as doctor experience, fatigue, and field of view selection, resulting in subjectivity and inconsistency in the interpretation results. In the first aspect, the present invention proposes an intelligent quantitative detection method based on a neural network algorithm, the method comprising:

[0006] Obtain tumor immunohistochemical staining images through a microscope imaging system;

[0007] Extracting key feature information from the tumor immunohistochemical staining image by using a target neural network analysis model, wherein the key feature information includes at least one of morphology, color, and distribution information of positive cells;

[0008] The model automatically identifies and classifies positive cells based on the labeled training data, calculates the number and distribution of cells according to preset rules, and generates a judgment report.

[0009] Optionally, extracting key feature information from the tumor immunohistochemical staining image by using a target neural network analysis model includes:

[0010] The key feature information in the tumor immunohistochemical staining image is extracted through the U-Net model.

[0011] Optionally, before the step of extracting key feature information from the tumor immunohistochemical staining image through the target neural network analysis model, the method further includes:

[0012] The tumor immunohistochemical staining image is preprocessed, and the preprocessing includes at least one of denoising, color correction and contrast enhancement.

[0013] Optionally, before the step of extracting key feature information from the tumor immunohistochemical staining image by using the target neural network analysis model, the method further includes:

[0014] Fine-tune pre-trained models for transfer learning;

[0015] The fine-tuned pre-trained model is trained to obtain the target neural network analysis model.

[0016] Optionally, fine-tuning the pre-trained model for transfer learning includes:

[0017] Freeze the first few layers of the pre-trained model and replace the classification layers to fine-tune the pre-trained model.

[0018] Optionally, each patient includes multiple tumor immunohistochemical staining images corresponding to multiple slice samples. The model automatically identifies and classifies positive cells based on the labeled training data, calculates the number and distribution of cells according to preset rules, and generates an interpretation report, including:

[0019] Through the multi-instance learning framework, the analysis results of tumor immunohistochemistry staining images corresponding to all sections of the same patient are integrated through aggregation functions to generate a patient-level interpretation report.

[0020] Optionally, the model automatically identifies and classifies positive cells based on the labeled training data, calculates the number and distribution of cells according to preset rules, and generates a judgment report, including:

[0021] The model automatically identifies and classifies positive cells based on the labeled training data, and calculates the number and distribution of cells according to preset rules to generate a distribution map, quantity statistics table and morphological feature description of positive cells to form a judgment report.

[0022] In a second aspect, the present invention further proposes an intelligent quantitative detection device based on a neural network algorithm, comprising:

[0023] An acquisition unit, used for acquiring tumor immunohistochemical staining images through a microscope imaging system;

[0024] An extraction unit, configured to extract key feature information in the tumor immunohistochemical staining image through a target neural network analysis model, wherein the key feature information includes at least one of morphology, color and distribution information of positive cells;

[0025] The analysis unit is used for the model to automatically identify and classify positive cells based on the labeled training data, calculate the number and distribution of cells according to preset rules, and generate a judgment report.

[0026] In a third aspect, an electronic device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor is used to implement the steps of the intelligent quantitative detection method based on a neural network algorithm as described in any one of the first aspects above when executing the computer program stored in the memory.

[0027] In a fourth aspect, the present invention further proposes a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, it implements any one of the intelligent quantitative detection methods based on a neural network algorithm of the first aspect.

[0028] In summary, the intelligent quantitative detection method based on the neural network algorithm proposed in this application obtains tumor immunohistochemical staining images through a microscope imaging system; extracts key feature information in the tumor immunohistochemical staining image through a target neural network analysis model, and the key feature information includes at least one of the morphology, color and distribution information of positive cells; the model automatically identifies and classifies positive cells according to the labeled training data, calculates the number and distribution of cells according to preset rules, and generates a judgment report. Thus, the efficiency and accuracy of pathological diagnosis are greatly improved through automated image processing, feature extraction and classification. The implementation of each step is aimed at reducing human intervention and subjectivity, while enhancing the stability of judgment. Automatic learning and analysis of a large amount of image data by neural networks avoids subjective bias caused by manual interpretation. Compared with the traditional manual interpretation method, the system can quickly process a large amount of slice data, greatly shortening the diagnosis time. Automated image analysis and report generation functions can reduce the workload of pathologists, allowing them to focus on the analysis of complex cases.

[0029] The intelligent quantitative detection method based on the neural network algorithm of the present invention, and other advantages, objectives and features of the present invention will be reflected in part through the following description, and in part will also be understood by technicians in this field through research and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present specification. Also, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0031] Figure 1 A schematic diagram of a flow chart of an intelligent quantitative detection method based on a neural network algorithm provided in an embodiment of the present application;

[0032] Figure 2 A schematic diagram of the structure of an intelligent quantitative detection device based on a neural network algorithm provided in an embodiment of the present application;

[0033] Figure 3 A schematic diagram of the structure of an intelligent quantitative detection electronic device based on a neural network algorithm provided in an embodiment of the present application. DETAILED DESCRIPTION

[0034] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices. The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments.

[0035] In order to solve the problem that the traditional interpretation method requires pathologists to carefully observe the stained sections under a microscope, evaluate the staining quality, positive signal distribution and morphological characteristics, and count and classify them. This process is not only time-consuming, but also easily affected by factors such as doctor experience, fatigue, and field of view selection, resulting in subjective and inconsistent interpretation results. Please refer to Figure 1, is a schematic flow chart of an intelligent quantitative detection method based on a neural network algorithm provided in an embodiment of the present application, which may specifically include: steps S110 to S130.

[0036] S110, obtaining tumor immunohistochemical staining images through a microscope imaging system.

[0037] Exemplarily, high-quality digital images of tumor immunohistochemical staining sections are obtained through a microscope imaging system. Use a high-resolution microscope to capture images of stained sections to ensure that they have sufficient clarity for cell morphology, color and other characteristics. The imaging system should have functions such as autofocus, exposure control and multi-light source adjustment to ensure that the captured images are noise-free, bias-free, and can truly reflect the cell distribution and staining in the sections. Only through high-quality image acquisition can subsequent analysis and classification work be built on a reliable foundation. The quality of the image directly affects the accuracy and efficiency of the neural network model.

[0038] Exemplarily, the microscope imaging system can adopt advanced optical design, combined with high numerical aperture objectives and sophisticated focusing mechanisms, to ensure clear and detailed images at different magnifications. The microscope is equipped with a variety of light source options, such as LED, halogen lamp or laser confocal, to meet the needs of different immunohistochemical staining methods, ensuring accurate color reproduction and rich details. For the digital imaging part, the high-sensitivity CCD or CMOS camera that can be integrated can capture images under the microscope and convert them into high-quality digital signals. These cameras usually have high resolution and large dynamic range, and can record a wide range of changes from weak to strong signals, ensuring the integrity and accuracy of image information. In addition, the image quality is further improved through automatic exposure, white balance and color correction functions.

[0039] S120, extracting key feature information from the tumor immunohistochemical staining image through a target neural network analysis model, wherein the key feature information includes at least one of the morphology, color, and distribution information of positive cells.

[0040] Exemplarily, the key feature information in the image is extracted by the neural network model, including the morphology, color and distribution of positive cells. The multi-layer structure of the CNN model can be used to gradually extract features in the image. The low-level network is responsible for extracting basic image features, such as edges, textures, etc., while the high-level network is responsible for extracting more abstract features, such as cell morphology, color distribution, etc. The convolutional neural network can distinguish normal cells from positive cells based on the different cell morphologies. For example, positive cells may show obvious nuclear division or a specific morphology of nuclear staining depth. Immunohistochemical staining usually uses specific dyes to mark positive cells, and the neural network can identify these staining areas by analyzing RGB color values. The deep learning model can automatically learn and identify these color features. The neural network can capture the distribution pattern of positive cells through spatial information, such as whether positive cells are distributed in clusters or whether they are diffuse. Through feature extraction, the system can automatically learn and identify key cell features from a large amount of image data, instead of relying on manual feature selection and annotation, thereby greatly improving the efficiency and accuracy of detection.

[0041] Exemplarily, the model automatically identifies and classifies positive cells in an image based on labeled training data. The neural network is trained with a large amount of labeled training data so that it can recognize different types of positive cells. The training data includes positive cells of different morphologies, different staining degrees, and different distribution forms. During the training process, the neural network gradually learns to distinguish positive cells from other cells and classifies them according to different characteristics of the cells (such as morphology, size, and color). For example, positive cells can be further classified into different subtypes of tumor cells or immune cells. The automatic recognition and classification functions of the model greatly reduce the subjectivity and errors of manual interpretation and improve the standardization and consistency of the interpretation results.

[0042] S130, the model automatically identifies and classifies positive cells based on the labeled training data, calculates the number and distribution of cells according to preset rules, and generates a judgment report.

[0043] Exemplarily, according to preset rules, the system calculates the number and distribution of positive cells and generates a final interpretation report. The system may include a counting function: based on the classification results, the system automatically counts the total number of positive cells and their spatial distribution in the slice image. Distribution pattern analysis function: the system can analyze the spatial distribution pattern of positive cells, such as whether it presents concentrated distribution, diffuse distribution and other characteristics. This is crucial for evaluating the invasiveness or reaction area of ​​the tumor. The ability to generate interpretation reports: the system automatically generates reports containing the number of positive cells, classification results, distribution maps and other contents for reference by pathologists. The report content not only lists the number of positive cells in each slice in detail, but also provides a distribution diagram of the cells and possible pathological diagnosis suggestions. The automatically generated interpretation report has a high degree of standardization, which can significantly improve the efficiency and accuracy of clinical diagnosis and reduce errors in manual interpretation. For example, suppose that when processing a set of tumor immunohistochemical staining slices, the traditional method requires the pathologist to observe and manually mark positive cells slice by slice. Through this neural network detection method, the system can process multiple slice images at the same time. After acquiring the image through the microscope imaging system, the neural network automatically extracts the positive cell features in each slice and classifies them as tumor cells or other immune cells. Finally, the system will automatically count the number of positive cells, analyze their distribution pattern, and generate a detailed diagnostic report, which not only reduces the manual workload but also improves the consistency and accuracy of the interpretation.

[0044] In summary, the intelligent quantitative detection method based on the neural network algorithm provided in the embodiment of the present application obtains tumor immunohistochemical staining images through a microscope imaging system; extracts key feature information in the tumor immunohistochemical staining image through a target neural network analysis model, and the key feature information includes at least one of the morphology, color and distribution information of positive cells; the model automatically identifies and classifies positive cells according to the labeled training data, calculates the number and distribution of cells according to preset rules, and generates a reading report. Thus, the efficiency and accuracy of pathological diagnosis are greatly improved through automated image processing, feature extraction and classification. The implementation of each step is intended to reduce human intervention and subjectivity, while enhancing the stability of interpretation. Automatic learning and analysis of a large amount of image data by neural networks avoids subjective bias caused by manual interpretation. Compared with traditional manual interpretation methods, the system can quickly process a large amount of slice data, greatly shortening the diagnosis time. Automated image analysis and report generation functions can reduce the workload of pathologists, allowing them to focus on the analysis of complex cases.

[0045] According to some embodiments, extracting key feature information from the tumor immunohistochemical staining image by using a target neural network analysis model includes:

[0046] The key feature information in the tumor immunohistochemical staining image is extracted through the U-Net model.

[0047] It can be understood that the U-Net model consists of two symmetrical structures: an encoder and a decoder. The encoder is responsible for extracting multi-level features in the image, while the decoder restores the resolution of the image layer by layer, ultimately achieving high-precision image segmentation. U-Net also introduces skip connections to connect the same-level features in the encoder and decoder to ensure that feature information is effectively transmitted between multiple levels. U-Net is widely used in the field of medical image processing because it can extract important contextual information at low resolution and combine local information with higher resolution for accurate segmentation. This is very helpful for identifying subtle differences in cell boundaries, size, and distribution.

[0048] Exemplarily, in the feature extraction stage, the input image is first processed by multiple convolutional layers and pooling layers to gradually extract low-level to high-level features. The convolutional layer is used to capture the edge, morphology and texture features in the image, while the pooling layer gradually reduces the resolution of the image but retains its contextual information. When processing tumor immunohistochemical staining images, the encoder part can extract the morphological features of positive cells (such as cell edges, nuclear shapes) and the color features of the stained area (such as the difference in color depth). The jump connection of U-Net directly transfers the feature map of each layer of the encoder to the corresponding decoder layer, retaining the fine features in the image. These jump connections can help the model recover detail information during the decoding process. The jump connection ensures the accurate segmentation of the edges of positive cells, avoids the loss of detail information caused by multi-layer feature extraction, and makes the segmented cell morphology more accurate. The decoder gradually restores the resolution of the image and combines the feature map of the jump connection to perform fine image segmentation. In each layer, the feature information is gradually restored from the coarser feature map to the original resolution, thereby generating accurate segmentation results. The decoder can separate positive cells from background areas and accurately mark the boundaries of each positive cell, ensuring that the morphology of each cell is fully extracted.

[0049] Exemplarily, the output of the U-Net model can be used to segment and identify the positive cell area in the image. The positive cell area can be segmented from the background and negative cells through the final output image of U-Net. Positive cells may show color differences due to different staining depths, and U-Net can classify them according to the color features and morphological differences in the image. For immunohistochemical staining images, positive cells may be stained dark brown, while the background is light or transparent. The U-Net model can distinguish these dark cells and mark their locations. Through the features extracted by U-Net, the model can automatically classify the positive cell areas and mark their locations and morphological characteristics. For different types of tumors or immune cells, U-Net can further classify and mark them according to their morphology and color. In the analysis and report generation stage, for example, if the U-Net model segments a large number of clustered positive cells, it may indicate the rapid proliferation of tumor cells, while scattered cells may suggest that the tumor is in an early or milder state.

[0050] Therefore, U-Net, through its unique network structure, can perform accurate cell segmentation at high resolution and retain important image details. Compared with traditional image processing methods, U-Net not only performs well in processing large-scale data, but also has a high sensitivity to cell features in complex medical images. Utilizing U-Net's multi-level feature extraction capabilities, it can effectively distinguish tumor cells from normal cells, especially in the extraction of cell edges and morphological features. This precision helps to improve the accuracy of diagnosis. U-Net's structure is suitable for large-scale automated processing, and can process a large number of immunohistochemical staining sections, significantly reducing the time and errors of manual operations. In addition, the U-Net model can optimize its performance through continuous training to adapt to different types of medical images.

[0051] In some examples, before the step of extracting key feature information from the tumor immunohistochemical staining image through the target neural network analysis model, the method further includes:

[0052] The tumor immunohistochemical staining image is preprocessed, and the preprocessing includes at least one of denoising, color correction and contrast enhancement.

[0053] In some examples, before the step of extracting key feature information from the tumor immunohistochemical staining image through the target neural network analysis model, the method further includes:

[0054] Fine-tune pre-trained models for transfer learning;

[0055] The fine-tuned pre-trained model is trained to obtain the target neural network analysis model.

[0056] In some examples, fine-tuning the pre-trained model for transfer learning includes:

[0057] Freeze the first few layers of the pre-trained model and replace the classification layers to fine-tune the pre-trained model.

[0058] Exemplarily, model fine-tuning may include: freezing some layers. The early layers in the pre-trained model are usually responsible for extracting more general features (such as edges, textures, etc.). These layers can be directly used for new tasks without retraining. Therefore, these layers are usually "frozen", that is, their weights are kept unchanged; modifying the last layer. The latter layers (usually classification layers) of the pre-trained model need to be adjusted according to the new task (such as the interpretation of tumor immunohistochemistry images). These layers will be replaced with layers suitable for the current task, and the weights of these layers will be retrained according to the new data; and fine-tuning. After introducing new task data (i.e., tumor immunohistochemistry staining image data), the entire model is fine-tuned. In this process, although the first few layers of the model remain frozen, the parameters of the latter layers will be updated and optimized according to the data of the new task to adapt to the new task. After loading the pre-processed image data, these images will be input into the fine-tuned model. Since the model already has general feature extraction capabilities, only light training for specific tasks is required, which will significantly speed up the speed. The system will adjust the model weights according to the characteristics of the new data, especially the morphology and distribution of positive cells, to improve the accuracy of classification and interpretation. Through multiple rounds of training and optimization, the model's recognition and classification capabilities on tumor immunohistochemical staining images are gradually improved. Since transfer learning utilizes the feature extraction capabilities of existing models, it can greatly reduce the time to train a new model from scratch.

[0059] Understandably, the fine-tuning process only requires adjustments to the high-level features of the model, which are closely related to the specific task. For example, in tumor immunohistochemical staining images, the model needs to learn to identify and distinguish positive cells from background cells. Through fine-tuning, the model can quickly adapt to these specific tasks without having to relearn the underlying general features. During the fine-tuning process, the first few layers of the model remain frozen, so there is no need to retrain. Only the last few layers (layers responsible for classification and high-level feature representation) need to be retrained, which greatly reduces the need for labeled data. Because most of the features have been learned from large-scale datasets through transfer learning, even when data is scarce, only less labeled data is needed to complete the fine-tuning of the task. For example, only a small amount of labeled data for tumor immunohistochemical staining is needed for the model to be fine-tuned and achieve high-accuracy interpretation. And when labeled data is scarce, the model is prone to overfitting, that is, it performs well on the training data but poorly on new data. Transfer learning can help the model better generalize to new tasks by using features learned on large datasets, and can effectively avoid overfitting even when labeled data is limited.

[0060] In some examples, each patient includes multiple tumor immunohistochemical staining images corresponding to multiple slice samples. The model automatically identifies and classifies positive cells based on the labeled training data, calculates the number and distribution of cells according to preset rules, and generates a judgment report, including:

[0061] Through the multi-instance learning framework, the analysis results of tumor immunohistochemistry staining images corresponding to all sections of the same patient are integrated through aggregation functions to generate a patient-level interpretation report.

[0062] Exemplarily, after the system acquires multiple tumor immunohistochemical staining slice images of the same patient, each image will be preprocessed and feature extracted. These slice images may contain heterogeneous information such as different cell densities and morphological differences, so each slice needs to be processed separately. In the framework of multi-instance learning, each slice sample of the same patient is regarded as an "instance", and the collection of all slices is regarded as a "bag". The system takes the interpretation result of each slice as the output result of an instance. The system uses a multi-instance learning model, takes the output result of each slice as input for training, and tries to extract global information from the common features of multiple slices. For example, the system learns the distribution pattern and heterogeneity of positive cells in each slice to improve the judgment ability of the model. For each individual slice, the neural network model will first perform feature extraction and intelligent interpretation, and calculate the number of positive cells and related features of the slice. This is an instance-level interpretation, and the model generates a separate result for each slice. The key step of multi-instance learning is to integrate the results of multiple instances (slices). The system integrates the interpretation results of each slice through an aggregation function (such as average, maximum or other more complex functions). This aggregation method ensures that the differences in the distribution of positive cells in different slices can be considered uniformly, thereby more comprehensively reflecting the patient's overall condition. After integrating the interpretation results of each slice, the system generates a final interpretation result at the package level. This result is more accurate than the interpretation of a single slice because it takes into account the heterogeneous information of all slices. Through this integration method, the system can eliminate the bias or accidental errors caused by a single slice and use the information of multiple slices to further improve the accuracy of diagnosis. For example, if the cell density in a slice is abnormally high, it may be only a local phenomenon, and the normal cell distribution of other slices can help the model make a comprehensive judgment, thereby avoiding over-reliance on the characteristics of a single slice.

[0063] In some examples, the model automatically identifies and classifies positive cells based on labeled training data, calculates the number and distribution of cells based on preset rules, and generates a judgment report, including:

[0064] The model automatically identifies and classifies positive cells based on the labeled training data, and calculates the number and distribution of cells according to preset rules to generate a distribution map, quantity statistics table and morphological feature description of positive cells to form a judgment report.

[0065] Exemplarily, based on the extracted feature information and the preset interpretation rules and standards, the automatic calculation of the number of positive cells is realized. The model can quickly traverse the entire image area, identify and count each positive cell, and generate corresponding statistical data. In addition to the calculation of the number of positive cells, the present invention can also generate a detailed interpretation report according to the interpretation rules and standards. The report content includes but is not limited to the distribution map of positive cells, the number statistics table, the morphological feature description and the possible pathological diagnosis suggestions. These results not only provide important reference information for pathologists, but also can help doctors to formulate treatment plans and evaluate prognosis more accurately.

[0066] Exemplarily, in order to verify the accuracy and reliability of the interpretation results, a control test method can also be used. By comparing and analyzing the intelligent interpretation results with the manual interpretation results, the interpretation performance of the model in different situations can be evaluated. On the basis of the control test, experienced pathologists can also be invited to review the interpretation results. Expert review can not only further verify the accuracy of the model, but also find out the deficiencies and errors of the model in specific situations, and provide direction for subsequent optimization. According to the feedback from the results of the control test and expert review, the algorithm model can be continuously optimized. This includes adjusting the parameter settings of the model, improving the network structure of the model, adding new training data, etc. Through continuous iteration and optimization process, the interpretation accuracy and robustness of the model are improved to ensure its stability and reliability in clinical applications.

[0067] See also Figure 2 In the embodiments of the present application, an embodiment of an intelligent quantitative detection device based on a neural network algorithm may include:

[0068] An acquisition unit 21 is used to acquire tumor immunohistochemical staining images through a microscope imaging system;

[0069] An extraction unit 22, configured to extract key feature information in the tumor immunohistochemical staining image through a target neural network analysis model, wherein the key feature information includes at least one of the morphology, color and distribution information of positive cells;

[0070] The analysis unit 23 is used for the model to automatically identify and classify positive cells according to the labeled training data, calculate the number and distribution of cells according to preset rules, and generate a judgment report.

[0071] In summary, the intelligent quantitative detection device based on the neural network algorithm provided in the embodiment of the present application obtains tumor immunohistochemical staining images through a microscope imaging system; extracts key feature information in the tumor immunohistochemical staining image through a target neural network analysis model, and the key feature information includes at least one of the morphology, color and distribution information of positive cells; the model automatically identifies and classifies positive cells according to the labeled training data, calculates the number and distribution of cells according to preset rules, and generates a reading report. Thus, the efficiency and accuracy of pathological diagnosis are greatly improved through automated image processing, feature extraction and classification. The implementation of each step is aimed at reducing human intervention and subjectivity, while enhancing the stability of interpretation. Automatic learning and analysis of a large amount of image data by neural networks avoids subjective bias caused by manual interpretation. Compared with traditional manual interpretation methods, the system can quickly process a large amount of slice data, greatly shortening the diagnosis time. Automated image analysis and report generation functions can reduce the workload of pathologists, allowing them to focus on the analysis of complex cases.

[0072] like Figure 3 As shown, the embodiment of the present application further provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored in the memory 320 and executable on the processor. When the processor 320 executes the computer program 311, the steps of any method of intelligent quantitative detection based on a neural network algorithm are implemented:

[0073] Obtain tumor immunohistochemical staining images through a microscope imaging system;

[0074] Extracting key feature information from the tumor immunohistochemical staining image by using a target neural network analysis model, wherein the key feature information includes at least one of morphology, color, and distribution information of positive cells;

[0075] The model automatically identifies and classifies positive cells based on the labeled training data, calculates the number and distribution of cells according to preset rules, and generates a judgment report.

[0076] Since the electronic device introduced in this embodiment is a device used to implement an intelligent quantitative detection device based on a neural network algorithm in the embodiment of the present application, based on the method introduced in the embodiment of the present application, the technical personnel in this field can understand the specific implementation mode of the electronic device of the present embodiment and its various variations. Therefore, how the electronic device implements the method in the embodiment of the present application is not introduced in detail here. As long as the equipment used by the technical personnel in this field to implement the method in the embodiment of the present application is within the scope of protection of this application.

[0077] In the specific implementation process, when the computer program 311 is executed by the processor, it can achieve Figure 1 Any implementation manner in the corresponding embodiment:

[0078] Obtain tumor immunohistochemical staining images through a microscope imaging system;

[0079] Extracting key feature information from the tumor immunohistochemical staining image by using a target neural network analysis model, wherein the key feature information includes at least one of morphology, color, and distribution information of positive cells;

[0080] The model automatically identifies and classifies positive cells based on the labeled training data, calculates the number and distribution of cells according to preset rules, and generates a judgment report.

[0081] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0082] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0083] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0084] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0085] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0086] The present application also provides a computer program product, which includes computer software instructions. When the computer software instructions are executed on a processing device, the processing device is caused to execute the following Figure 1 The process of intelligent quantitative detection based on neural network algorithm in the corresponding embodiment.

[0087] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on the computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website site, a computer, a server or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (digital subscriber line, DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium can be any available medium that a computer can store or a data storage device such as a server or a data center that includes one or more available media integration. The available medium can be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid state drive (SSD)), etc.

[0088] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0089] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0090] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0091] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0092] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), disk or optical disk and other media that can store program codes.

[0093] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An intelligent quantitative detection method based on a neural network algorithm, characterized in that: include: Obtain tumor immunohistochemical staining images through a microscope imaging system; Extracting key feature information from the tumor immunohistochemical staining image by using a target neural network analysis model, wherein the key feature information includes at least one of morphology, color, and distribution information of positive cells; The model automatically identifies and classifies positive cells based on the labeled training data, calculates the number and distribution of cells according to preset rules, and generates a judgment report.

2. The method according to claim 1, characterized in that The extracting key feature information in the tumor immunohistochemical staining image by using a target neural network analysis model includes: The key feature information in the tumor immunohistochemical staining image is extracted through the U-Net model.

3. The method according to claim 1, characterized in that Before the step of extracting key feature information from the tumor immunohistochemical staining image through the target neural network analysis model, the method further includes: The tumor immunohistochemical staining image is preprocessed, and the preprocessing includes at least one of denoising, color correction and contrast enhancement.

4. The method according to claim 1, characterized in that Before the step of extracting key feature information from the tumor immunohistochemical staining image through the target neural network analysis model, the method further includes: Fine-tune pre-trained models for transfer learning; The fine-tuned pre-trained model is trained to obtain the target neural network analysis model.

5. The method according to claim 4, characterized in that The method of fine-tuning the pre-trained model for transfer learning includes: Freeze the first few layers of the pre-trained model and replace the classification layers to fine-tune the pre-trained model.

6. The method according to claim 1, characterized in that Each patient includes multiple tumor immunohistochemical staining images corresponding to multiple slice samples. The model automatically identifies and classifies positive cells based on the labeled training data, calculates the number and distribution of cells according to preset rules, and generates a judgment report, including: Through the multi-instance learning framework, the analysis results of tumor immunohistochemistry staining images corresponding to all sections of the same patient are integrated through aggregation functions to generate a patient-level interpretation report.

7. The method according to any one of claims 1 to 6, characterized in that The model automatically identifies and classifies positive cells based on the labeled training data, calculates the number and distribution of cells according to preset rules, and generates a judgment report, including: The model automatically identifies and classifies positive cells based on the labeled training data, and calculates the number and distribution of cells according to preset rules to generate a distribution map, quantity statistics table and morphological feature description of positive cells to form a judgment report.

8. An intelligent quantitative detection device based on a neural network algorithm, characterized in that: include: An acquisition unit, used for acquiring tumor immunohistochemical staining images through a microscope imaging system; An extraction unit, configured to extract key feature information in the tumor immunohistochemical staining image through a target neural network analysis model, wherein the key feature information includes at least one of morphology, color and distribution information of positive cells; The analysis unit is used for the model to automatically identify and classify positive cells based on the labeled training data, calculate the number and distribution of cells according to preset rules, and generate a judgment report.

9. An electronic device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor is used to implement the steps of the intelligent quantitative detection method based on a neural network algorithm as described in any one of claims 1 to 7 when executing the computer program stored in the memory.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the intelligent quantitative detection method based on a neural network algorithm as described in any one of claims 1 to 7 is implemented.

Images