HE image ovarian cancer polyploid tumor giant cell detection method based on deep learning

Abstract

The invention relates to a deep learning-based Hamp; the invention discloses an E image ovarian cancer polyploid tumor giant cell (PGCCs) detection method. The method comprises the steps of specific data set construction, image annotation and data division, feature learning and optimization by using an OCDet model, and model training and evaluation. The method comprises the following steps: firstly, establishing a Hamp containing PGCCs; and E, dyeing the image data set, then accurately marking the image and dividing the image into training, verification and test sets. The OCDet model takes CSPDarkNet as a core, combines an ECA mechanism, and focuses on deep learning and recoding of pathological semantic features. By training a data set, the model performs parameter updating and optimization through back propagation and gradient descent to identify PGCCs features. The verification set is used for model optimization, and the test set is used for evaluating model performance. According to the automatic detection technology, doctors can be helped to improve the diagnosis efficiency and reduce errors, important reference is provided for clinical treatment and prognosis evaluation, and remarkable clinical application value is shown.

Description

Technical Field

[0001] The present invention relates to tissue pathology image processing, and in particular to a method for detecting polyploid tumor giant cells of ovarian cancer in H&E images based on deep learning. Background Art

[0002] Ovarian cancer is the most dangerous gynecological malignancy that threatens women's lives. Pathological diagnosis is the gold standard for determining the occurrence and development of tumors, and hematoxylin and eosin (H&E) staining technology is a basic staining method widely used in histopathological diagnosis. This technology uses the alkaline properties of hematoxylin dye and the acidic properties of eosin dye to dye different parts of cells into blue-purple or red. Through this color contrast, pathologists can clearly observe subtle changes in cell structure, providing an important basis for disease diagnosis. Usually, pathologists need to use morphological hematoxylin and eosin (H&E) pathological images to identify various types of cells in ovarian tissue, including polyploid giant cancer cells (PGCCs). Polyploid giant cancer cells are a special type of cancer cell subpopulation characterized by single or multiple giant nuclei, and the nuclei are usually irregular, which is of great significance for the diagnosis and staging of tumors. However, the detection of polyploid giant cancer cells in H&E staining images is a complicated and time-consuming process.

[0003] As one of the commonly used staining techniques in pathological diagnosis, H&E staining images can provide rich information on cell morphology and structure. However, due to the significant morphological differences between PGCCs and traditional cells, their detection and analysis is a challenging task for pathologists.

[0004] Although the traditional diagnosis process of ovarian cancer has high accuracy, it has disadvantages such as strong subjectivity, complex operation, and long time consumption. With the rapid development of deep learning technology, computer-aided systems based on deep learning provide new possibilities for the automatic diagnosis of diseases. By training deep neural network models, it is possible to automatically detect and classify lesions in medical images, thereby improving the accuracy and efficiency of diagnosis. Deep learning models have been widely used in medical image detection tasks, which makes it possible to automatically and accurately detect target cells in various pathological images. Therefore, the application of deep learning technology to the detection of polyploid tumor giant cells in H&E stained images has important research value and development potential. Through deep learning algorithms, the characteristic information of polyploid tumor giant cells in H&E stained images can be automatically learned and identified, and these cells can be detected quickly and accurately, thereby assisting pathologists in more accurate tumor diagnosis and staging.

[0005] However, the current research on the detection of polyploid tumor giant cells in H&E stained images using deep learning technology is still in its infancy, and there are many challenges and problems to be solved. First, how to design a suitable deep learning model and optimize model parameters to improve the accuracy and stability of detecting polyploid tumor giant cells. Second, in H&E pathological images of ovarian cancer, the heterogeneity implied by morphology often leads to semantic ambiguity in the detection model. The morphological differences of polyploid tumor giant cells and the complexity of the staining images reduce the signal-to-noise ratio of the detection, requiring the detection model to construct a more complex feature space. Third, since the number and morphology of the nuclei of polyploid tumor giant cells vary greatly and the number is small, the model needs to learn pathological features with strong heterogeneity based on fewer samples.

[0006] It should be noted that the information disclosed in the above background technology section is only used for understanding the background of the present application, and therefore may include information that does not constitute prior art known to ordinary technicians in the field. Summary of the invention

[0007] The main purpose of the present invention is to solve the problems existing in the above-mentioned background technology and provide a method for detecting polyploid tumor giant cells in ovarian cancer in H&E images based on deep learning.

[0008] To achieve the above object, the present invention adopts the following technical solutions:

[0009] In a first aspect of the present invention, a method for detecting polyploid giant cells of ovarian cancer in H&E images based on deep learning comprises the following steps:

[0010] S1. Construct a H&E staining image dataset containing polyploid tumor giant cells PGCCs;

[0011] S2. Annotate the H&E stained images, mark the locations of polyploid tumor giant cells, and divide the annotated data set into training set, validation set, and test set;

[0012] S3. Construct an OCDet model based on deep feature-based channel attention recoding and fusion. The OCDet model uses CSPDarkNet as the feature extraction backbone and combines the attention mechanism ECA to achieve relationship modeling and recoding of pathological semantic features of different channels.

[0013] S4. The OCDet model was trained using the labeled training dataset. The model parameters were updated through the back-propagation algorithm and the gradient descent optimizer so that the model could learn the feature representation of PGCCs. The validation set was used to tune the model during the training process. After the training, the test set was used to evaluate the model. The trained model was used for the automatic detection of polyploid tumor giant cells in ovarian cancer in H&E images.

[0014] In a second aspect of the present invention, a computer-readable storage medium stores a computer program, characterized in that when the computer program is executed by a processor, the method for detecting polyploid tumor giant cells in ovarian cancer in H&E images based on deep learning is implemented.

[0015] In a third aspect of the present invention, a computer program product includes a computer program, characterized in that when the computer program is executed by a processor, it implements the method for detecting polyploid tumor giant cells in ovarian cancer in H&E images based on deep learning.

[0016] The present invention has the following beneficial effects:

[0017] The present invention proposes a new method for detecting polyploid giant cells PGCCs of ovarian cancer in hematoxylin-eosin H&E stained images using deep learning technology, achieving efficient and accurate automated detection, and having important clinical significance and application value. By constructing and training a target detection model, the present invention realizes automatic detection and recognition of PGCCs in H&E stained images. Specifically, the present invention uses the proposed detection model OCDet to learn the cell morphology and structural characteristics in H&E stained images, and especially models and optimizes the special morphology and structure of PGCCs. By training and optimizing the detection model, it can accurately distinguish PGCCs from other cells, and automatically mark the location and number of PGCCs in the image. Through automated detection, not only can the diagnostic efficiency be improved and human errors be reduced, but also more valuable reference information can be provided for subsequent clinical treatment and prognosis evaluation. In addition, the method of the present invention is also scalable and customizable, and can adapt to the detection needs of other target cells under different disease types and different staining conditions.

[0018] Compared with the existing polyploid tumor giant cell detection method, the main advantages of the present invention are:

[0019] 1. Without consuming a lot of manpower and material resources to construct a large data set, the special morphology and structure of polyploid tumor giant cells are precisely modeled and optimized, achieving accurate and efficient detection of polyploid tumor giant cells in H&E staining images.

[0020] 2. Once the OCDet model is trained and optimized, it can quickly process a large amount of H&E image data and realize the automatic detection of polyploid tumor giant cells. This greatly improves the efficiency of pathological diagnosis, reduces the workload of doctors, and enables more cases to be diagnosed in a shorter time.

[0021] 3. The designed deep learning model has good scalability. With the continuous accumulation of new H&E image data, the model can be continuously trained and optimized to further improve its ability to identify polyploid tumor giant cells. In addition, the model can also be applied to the detection of other types of cells or tissues, providing more possibilities and value for pathological diagnosis.

[0022] Other beneficial effects of the embodiments of the present invention will be further described below. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is an example diagram of polyploid tumor giant cells according to an embodiment of the present invention.

[0024] Figure 2 It is an overall network framework diagram of an embodiment of the present invention.

[0025] Figure 3 This is a flowchart of ECA-RER and ECA-MREF multi-scale feature extraction and fusion according to an embodiment of the present invention.

[0026] Figure 4 The test results of the present invention are compared with those of other models. DETAILED DESCRIPTION

[0027] The following is a detailed description of the embodiments of the present invention. It should be emphasized that the following description is only exemplary and is not intended to limit the scope and application of the present invention.

[0028] The embodiment of the present invention provides a method for detecting polyploid giant cells of ovarian cancer in H&E images based on deep learning, comprising the following steps:

[0029] S1. Construct a H&E staining image dataset containing polyploid tumor giant cells PGCCs;

[0030] S2. Annotate the H&E stained images, mark the locations of polyploid tumor giant cells, and divide the annotated data set into training set, validation set, and test set;

[0031] S3. Construct an OCDet model based on deep feature-based channel attention recoding and fusion. The OCDet model uses CSPDarkNet as the feature extraction backbone and combines the attention mechanism ECA to achieve relationship modeling and recoding of pathological semantic features of different channels.

[0032] S4. The OCDet model was trained using the labeled training dataset. The model parameters were updated through the back-propagation algorithm and the gradient descent optimizer so that the model could learn the feature representation of PGCCs. The validation set was used to tune the model during the training process. After the training, the test set was used to evaluate the model. The trained model was used for the automatic detection of polyploid tumor giant cells in ovarian cancer in H&E images.

[0033] The inventive method proposes an innovative deep learning model, OCDet, for detecting polyploid tumor giant cells in ovarian cancer in hematoxylin-eosin (H&E) pathological images.

[0034] In a preferred embodiment, the OCDet model includes the following parts connected in sequence: a backbone network Backbone, including a feature extraction model CSPDarkNet, which extracts deep features from the input H&E stained image and constructs a mapping relationship from the input image to a multi-resolution feature map; a connection part Neck, which further processes and optimizes the features extracted by the backbone network Backbone, including enhancing significant features and removing redundant features through the ECA-RER module, and realizing efficient aggregation and fusion of multi-scale features through the ECA-MREF module; a detection head Head, which uses feature maps of different resolutions to regress the detection box, obtains the final detection result of polyploid tumor giant cells, and compares it with the Ground Truth through a loss function to optimize the model performance.

[0035] In a more preferred embodiment, Figure 2 and Figure 3 As shown in the figure, the processing of the backbone network Backbone of the OCDet model includes: performing data augmentation operations to improve the quality of pathological images and enhance the feature representation of images; adjusting the pathological images to a uniform size, such as 640 pixels × 640 pixels, to meet the input requirements of the feature extraction model; inputting the resized images into the CSPDarkNet model for feature extraction, which converts the images into a set of multi-resolution feature maps to capture pathological information at different levels.

[0036] like Figure 2 and Figure 3As shown in the figure, the processing of the connection part Neck of the OCDet model includes: strengthening the significant features and removing or suppressing the redundant features of the feature map extracted by Backbone through the ECA-RER module to optimize the feature representation and realize the refined processing process of feature embedding (Embedding Refinement); aggregating the refined features output by the ECA-RER module layer by layer through the ECA-MREF module to realize the effective fusion of multi-scale features (Feature Fusion), and obtaining a set of feature maps corresponding to different resolutions, thereby enhancing the model's recognition ability for polyploid tumor giant cells; using the two modules of the connection part Neck, the OCDet model can further improve the expression ability of features and provide richer and more accurate feature information for the final detection task.

[0037] like Figure 3 As shown, in a preferred embodiment, the specific designs of the ECA-RER module and the ECA-MREF module are as follows:

[0038] The feature maps from the backbone network (CSPDarkNet) are three feature maps with different resolutions, and our model needs to enhance the feature processing. The ECA-RER module includes multiple ECA mechanisms, which perform global average pooling operations on the three feature maps from the backbone. Each feature map is converted into a one-dimensional vector, which contains the global information of the channel. Then, a one-dimensional convolution is performed to capture the correlation between channels and generate weight coefficients for each channel. In particular, the kernel size of the one-dimensional convolution is adaptively determined by the nonlinear mapping of the channel dimension C, which ensures that feature maps with different numbers of channels can obtain the most suitable cross-channel interaction. Subsequently, these weight coefficients are applied to the original three feature maps respectively, and the weighted operation is realized by element-by-element multiplication. The weighted feature map retains the original information while enhancing the feature representation of important channels, suppressing unimportant channels, and realizing embedding refinement at a specific resolution. The ECA-MREF module includes multiple ECA mechanisms and bidirectional cross-resolution fusion operations. The feature maps of lower levels have higher resolutions and contain more detailed information, while the feature maps of higher levels have lower resolutions but have stronger semantic expression capabilities. ECA-MREF fuses feature maps of different resolutions in a bidirectional manner. This means that it can not only fuse high-level semantic information into low-level feature maps to enhance their semantic expression capabilities, but also pass low-level detail information to high-level feature maps to supplement their detail information. Multi-resolution embedding fusion is mainly achieved through two approaches - upsampling and downsampling. The low-resolution feature map is interpolated through the upsampling operation to match its size with the upper-level feature map, and then connected and fused. The input feature map is divided into two parts, and then cross-connected (multiple convolution residual blocks) are performed between the two parts, and finally merged (concatenation operation). This cross-stage connection method helps the model learn richer feature combinations and improve the generalization ability of the model.

[0039] The processing of the detection head Head of the OCDet model includes: using multiple detection heads to process feature maps of different resolutions obtained from the connection part Neck to achieve detection frame regression of polyploid tumor giant cells and obtain predicted detection results. Further, the obtained detection results are compared with the true annotation Ground Truth, and the error is calculated using a loss function, which is composed of a combination of detection loss and classification loss to optimize the accuracy of the detection frame and the accuracy of classification; wherein the detection loss quantifies the similarity between the predicted frame and the true frame by calculating the intersection of the predicted frame and the true frame and dividing it by the union, and at the same time introduces a difference loss to optimize the prediction probability in the form of cross entropy, so that the network can more accurately identify the target position; wherein the classification loss adopts different calculation methods according to the presence or absence of the label. For the case where the label exists, the weighted cross entropy loss is used, and for the case where the label does not exist, the logarithmic loss function adjusted by the weight coefficient is used. Through this processing method of the detection head Head, the OCDet model can effectively and accurately locate and classify polyploid tumor giant cells, and at the same time, the detection performance of the model is continuously optimized through the guidance of the loss function.

[0040] The important innovative design of the embodiment of the present invention includes the OCDet model and its components, which improve the detection performance of the model by optimizing feature representation and multi-scale feature fusion. The OCDet model achieves efficient and accurate detection of polyploid tumor giant cells by combining the three parts of Backbone, Neck and Head. In particular, in the Neck part, the present invention designs two modules, ECA-RER and ECA-MREF. The ECA-RER module optimizes feature representation by enhancing significant features and removing redundant features; the ECA-MREF module enhances the model's detection ability for polyploid tumor giant cells through multi-scale feature fusion.

[0041] The above embodiment proposes an innovative method for detecting polyploid giant cells of ovarian cancer in H&E images based on deep learning. The method significantly improves the efficiency and accuracy of pathological image diagnosis by constructing an automated OCDet model. The OCDet model cleverly integrates CSPDarkNet as the Backbone for deep feature extraction. Further, the ECA-RER module enhances feature significance and removes redundant features, and the ECA-MREF module realizes effective aggregation and fusion of multi-scale features. This structural design enables the model to not only accurately capture the special morphological and structural characteristics of polyploid giant cells, but also learn pathological features with strong heterogeneity based on a small number of samples, solving the problems of strong subjectivity, complex operation and long time consumption in the traditional detection process. The detection method of the present invention can effectively reduce the workload of doctors as an auxiliary diagnosis method, so that more cases can be diagnosed quickly, and has good scalability and customizability, and is adapted to the cell detection requirements under different disease types and staining conditions, and provides an efficient and accurate solution for the detection and auxiliary diagnosis of polyploid giant cells of ovarian cancer in H&E images.

[0042] Specific embodiments of the present invention are further described below.

[0043] The method of the present invention mainly includes the following processes: first, blocks are captured from whole slide images of ovarian cancer, and a data set is constructed through manual annotation; second, the OCDet model is trained by loading a training set to detect polyploid tumor giant cells in H&E stained images; finally, the model is evaluated through a test set, and optimization and fine-tuning are performed to achieve good generalization and stable practicality of the final model.

[0044] A method for detecting polyploid tumor giant cells of ovarian cancer in hematoxylin-eosin images using deep learning technology, comprising:

[0045] Step S1 constructs a rich H&E staining image dataset containing polyploid tumor giant cells, and randomly cuts out blocks from the whole slide images of the dataset. This process may require manual participation to screen and correct the data. Finally, data preprocessing is performed, including image enhancement, noise removal, color standardization, etc., to improve the quality and consistency of the image.

[0046] Step S2: Professional pathologists annotate the H&E images and clearly mark the location of polyploid tumor giant cells. This process uses the voc annotation format to meet the needs of the proposed model. Finally, the annotated data set is divided into a training set, a validation set, and a test set. The training set is used for model training, the validation set is used for model tuning, and the test set is used for the final evaluation of the model.

[0047] Step S3 designs a channel attention recoding and fusion model OCDet based on deep features according to the morphological and structural characteristics of polyploid tumor giant cells. The model uses CSPDarkNet as the feature extraction backbone and combines the attention mechanism ECA to achieve relationship modeling and recoding of pathological semantic features of different channels to improve the performance of the model.

[0048] Step S4 uses the labeled training dataset to train the OCDet model. The model parameters are updated through the back-propagation algorithm and the gradient descent optimizer so that the model can fully learn the feature representation of PGCCs. The trained model is evaluated using an independent test dataset to calculate the model's accuracy, recall, F1 score and other indicators. The model is performance tuned based on the evaluation results.

[0049] The specific implementation process is as follows:

[0050] Step S1 - 1 collects H&E pathological image data of polyploid tumor giant cells of ovarian cancer from Peking University Third Hospital. This batch of data contains 13 whole slide images (WSIs).

[0051] Step S1-2: WSIs were divided into 512*512 pixel tiles, and 197, 66, and 66 tiles were randomly selected for the training, validation, and test sets in S1-1. The tiles were manually screened and corrected to ensure that each tile contained target cells and had little background interference. These tiles covered the morphological diversity of polyploid tumor giant cells.

[0052] Step S2-1 Experienced pathologists annotate each H&E image block in turn, and try to annotate without missing any information. The annotation information includes the location, size and other characteristics of the polyploid tumor giant cells. Finally, XML files are used as the storage format of the annotation information to facilitate the subsequent reading and processing of the model.

[0053] Step S2-2 divides the labeled data set into training set, validation set and test set with a distribution ratio of 6:2:2. The training set is used for model training and optimization, the validation set is used to monitor the model training process and adjust the hyperparameters, and the test set is used to evaluate the performance of the model.

[0054] Step S3-1 proposes OCDet for the detection of polyploid tumor giant cells in morphological H&E images. OCDet consists of three parts: Backbone, Neck, and Head. In the Backbone processing flow, the pathological image is first data enhanced and then resized to 640*640 pixels, and finally input into the feature extraction model CSPDarkNet. This process constructs the following mapping:

[0055]

[0056] Where x represents a tile, represents the data set, f n Represents feature maps at three resolutions.

[0057] In step S3-2, in the Neck part, the three feature maps of different resolutions obtained from Backbone are processed by the ECA-RER and ECA-MREF modules to achieve efficient Embedding Refinement and Feature Fusion. The ECA-RER module performs the emphasis of the significant features extracted by Backbone and the removal of redundant features, thereby optimizing the feature representation. The refinement process is defined as the mapping from input features to weighted features:

[0058]

[0059] The ECA-MREF module performs efficient layer-by-layer aggregation of the updated features and obtains three feature maps corresponding to different resolutions.

[0060] Step S3-3 In the Head part, OCDet uses multiple detection heads to perform detection box regression on feature maps of different resolutions and obtains the final detection result of polyploid tumor giant cells. The result is used to calculate the loss with Ground Truth. The loss function used is composed of detection loss and classification loss. The detection loss is defined as:

[0061]

[0062]

[0063]

[0064] The classification loss is defined as:

[0065]

[0066] Where D box is the prediction box, G box is the Ground Truth, q is the label, α and γ are weight coefficients, and y is the general distribution value. dfl The significance of is to optimize the probability of the two positions (one left and one right) closest to the label y in the form of cross entropy, so that the network can focus on the target position faster.

[0067] Step S4-1 loads the labeled training dataset and the OCDet model with initialized weights. The training is divided into two stages, namely the freezing stage and the thawing stage, and the batch size of these two stages is set to 16. In the freezing stage, only some parameters of the network are fine-tuned, and the set epoch is 50. In the thawing stage, all parameters of OCDet will participate in the training to better fit the data, and the set epoch is 150. At the same time, the training of OCDet uses the SGD optimizer, and the initial learning rate is set to 0.01.

[0068] Step S4-2 inputs the training data into the model, performs forward propagation through each layer of the model, and obtains the prediction result. The loss is calculated based on the prediction result of the model and the true label, and the gradient of the loss is calculated, and the parameters of the model are updated using the back propagation algorithm and the optimizer strategy. The process of forward propagation, loss calculation, back propagation, and parameter update is repeated until the preset number of training rounds is reached or other stopping conditions are met (such as the performance on the validation set no longer improves).

[0069] Step S4-3 loads an independent test dataset and uses the trained model to predict the test dataset. Calculate the model's evaluation indicators, such as precision, recall, F1-Score, and mAP@0.5, based on the prediction results and true labels. These indicators are used to comprehensively evaluate the performance of the model in the target detection task of polyploid tumor giant cells. Optimize the performance of the model based on the evaluation results. First, analyze the types and causes of errors in the model on the test set; then, take targeted optimization measures for these problems, such as adjusting the model structure, modifying the loss function, adding data enhancement, etc.; finally, retrain the model and evaluate its performance until satisfactory results are achieved.

[0070] experiment

[0071] The superiority of the method is discussed from the perspective of quantitative and qualitative indicators. The quantitative indicators are precision, recall, F1-Score and mAP@0.5, which can be used to evaluate the performance of the model. To avoid data leakage, the test set tiles are used for evaluation. The results are shown in Table 1 (bold represents the best results).

[0072] Table 1

[0073]

[0074] As can be seen from the table, the model OCDet has shown significant advantages in multiple key indicators. For recall rate, OCDet scored 77.03%, ranking first among all models. This performance shows that the model of the present invention can not only accurately identify real targets, but also cover all real targets as comprehensively as possible to reduce missed detections. For F1-Score and mAP@0.5, the model of the present invention obtained a high score of 0.72 and reached 75.91% respectively, ranking first among all models. This indicator fully demonstrates the superiority of OCDet in detecting polyploid cancer giant cells.

[0075] Qualitative indicators can intuitively show the results, as shown in the figure. For the detection of polyploid tumor giant cells, although Faster RCNN did detect all polyploid cells, it mistakenly identified three targets as one and detected it again, which obviously did not meet the detection requirements. The number of missed targets for the other three models (YOLO V7, SSD, and RetinaNet) was inconsistent. Only OCDet can accurately identify all target cells. By comparison, it was found that OCDet detected target cells in morphological H&E pathology images of ovarian cancer more comprehensively and accurately than other methods.

[0076] In summary, the present invention proposes an innovative deep learning model OCDet for detecting polyploid tumor giant cells in ovarian cancer in hematoxylin-eosin (H&E) pathological images. Compared with existing methods, the present invention has the following advantages:

[0077] 1. Without consuming a lot of manpower and material resources to construct a large data set, the special morphology and structure of polyploid tumor giant cells are precisely modeled and optimized, achieving accurate and efficient detection of polyploid tumor giant cells in H&E staining images.

[0078] 2. Once the OCDet model is trained and optimized, it can quickly process a large amount of H&E image data and realize the automatic detection of polyploid tumor giant cells. This greatly improves the efficiency of pathological diagnosis, reduces the workload of doctors, and enables more cases to be diagnosed in a shorter time.

[0079] 3. The proposed deep learning model has good scalability. With the continuous accumulation of new H&E image data, the model can be continuously trained and optimized to further improve its ability to identify polyploid tumor giant cells. In addition, the model can also be applied to the detection of other types of cells or tissues, providing more possibilities and value for pathological diagnosis.

[0080] An embodiment of the present invention further provides a storage medium for storing a computer program, which at least performs the above method when executed.

[0081] An embodiment of the present invention further provides a control device, comprising a processor and a storage medium for storing a computer program; wherein the processor is configured to execute at least the method described above when executing the computer program.

[0082] An embodiment of the present invention further provides a processor, wherein the processor executes a computer program and at least executes the method described above.

[0083] The storage medium can be implemented by any type of non-volatile storage device, or a combination thereof. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a magnetic random access memory (FRAM), a ferromagnetic random access memory (Flash Memory), a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM); the magnetic surface memory can be a disk memory or a tape memory. The storage medium described in the embodiments of the present invention is intended to include, but is not limited to, these and any other suitable types of memory.

[0084] In the several embodiments provided by the present invention, it should be understood that the disclosed systems and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0085] The units described above as separate components may or may not be physically separated, and the components displayed 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 scheme of this embodiment.

[0086] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0087] A person skilled in the art can understand that: all or part of the steps of implementing the above method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiment; and the aforementioned storage medium includes: mobile storage devices, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), disks or optical disks, etc. Various media that can store program codes.

[0088] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0089] The methods disclosed in the several method embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new method embodiments.

[0090] The features disclosed in several product embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new product embodiments.

[0091] The features disclosed in several method or device embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new method embodiments or device embodiments.

[0092] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art of the present invention, several equivalent substitutions or obvious variations can be made without departing from the concept of the present invention, and the performance or use is the same, which should be regarded as belonging to the protection scope of the present invention.

Claims

1. A method for detecting polyploid giant cells in ovarian cancer H&E images based on deep learning, characterized in that: The steps include: S1. Construct a H&E staining image dataset containing polyploid tumor giant cells PGCCs; S2. Annotate the H&E stained images, mark the locations of polyploid tumor giant cells, and divide the annotated data set into training set, validation set, and test set; S3. Construct an OCDet model based on deep feature-based channel attention recoding and fusion. The OCDet model uses CSPDarkNet as the feature extraction backbone and combines the attention mechanism ECA to achieve relationship modeling and recoding of pathological semantic features of different channels. S4. The OCDet model was trained using the labeled training dataset. The model parameters were updated through the back-propagation algorithm and the gradient descent optimizer so that the model could learn the feature representation of PGCCs. The validation set was used to tune the model during the training process. After the training, the test set was used to evaluate the model. The trained model was used for the automatic detection of polyploid tumor giant cells in ovarian cancer in H&E images.

2. The method for detecting polyploid tumor giant cells in ovarian cancer based on deep learning in H&E images according to claim 1, characterized in that: In step S1, blocks are randomly cut from the whole slide image, and are manually screened and corrected; data preprocessing is performed on the blocks, including image enhancement, noise removal, and color standardization.

3. The method for detecting polyploid tumor giant cells in ovarian cancer based on deep learning in H&E images according to claim 1 or 2, characterized in that: The OCDet model consists of the following parts connected in sequence: The backbone network Backbone, including the feature extraction model CSPDarkNet, extracts deep features from the input H&E staining images and constructs a mapping relationship from the input image to the multi-resolution feature map; The connecting part Neck further processes and optimizes the features extracted by the backbone network Backbone, including enhancing the significant features and removing the redundant features through the ECA-RER module, and aggregating and fusing the multi-scale features through the ECA-MREF module; The detection head uses feature maps of different resolutions to regress the detection box to obtain the final detection result of polyploid tumor giant cells, and compares it with the true annotation Ground Truth through the loss function to optimize the model performance.

4. The method for detecting polyploid tumor giant cells in ovarian cancer based on deep learning in H&E images according to claim 3, characterized in that: The processing of the backbone network Backbone of the OCDet model includes: Perform data augmentation operations to improve the quality of pathological images and enhance the feature representation of images; Resize pathology images to a uniform size; The resized images are input into the CSPDarkNet model for feature extraction, which converts the images into a set of multi-resolution feature maps to capture different levels of pathological information.

5. The method for detecting polyploid tumor giant cells in ovarian cancer based on deep learning in H&E images according to claim 3 or 4, characterized in that: The processing of the connection part Neck of the OCDet model includes: The ECA-RER module is used to enhance the significant features of the feature graph extracted by Backbone and remove or suppress redundant features to optimize the feature representation and realize the refined processing of feature embedding. The refined features output by the ECA-RER module are aggregated layer by layer through the ECA-MREF module to achieve the fusion of multi-scale features and obtain a set of feature maps corresponding to different resolutions to enhance the model's ability to recognize polyploid tumor giant cells. Preferably, for the ECA-RER module, a global average pooling operation is performed on each feature map from Backbone, the multidimensional feature map is converted into a one-dimensional feature vector, and the correlation between channels is captured by a one-dimensional convolution to generate a weight coefficient for each channel, wherein the kernel size of the one-dimensional convolution is adaptively determined by nonlinear mapping based on the channel dimension C of the feature map to achieve optimal interaction across channels; the weight coefficient obtained by the ECA-RER module is applied to the original feature map, and weighted by element-by-element multiplication to enhance the features of important channels and suppress unimportant channels, thereby achieving refined processing of feature embedding; Preferably, for the ECA-MREF module, a bidirectional cross-resolution fusion operation is adopted to fuse the low-level high-resolution feature map with the high-level low-resolution feature map, achieve size matching through upsampling and downsampling, and enhance the semantic expression ability and detail information of the feature map through cross-connection and merging operations, wherein the cross-connection is composed of multiple convolutional residual blocks to promote the model to learn richer feature combinations and improve the generalization ability of the model.

6. The method for detecting polyploid tumor giant cells in ovarian cancer based on deep learning in H&E images according to claim 3 or 4, characterized in that: The processing of the detection head Head of the OCDet model includes: Using multiple detection heads to process feature maps of different resolutions obtained from the connection part Neck, so as to achieve detection frame regression of polyploid tumor giant cells and obtain predicted detection results; Compare the obtained detection results with the ground truth, and use the loss function to calculate the error. The loss function is composed of the detection loss and the classification loss. The detection loss quantifies the similarity between the predicted box and the true box by calculating the intersection of the predicted box and the true box and dividing by the union, and introduces the difference loss to optimize the prediction probability in the form of cross entropy so that the network can more accurately identify the target position; The classification loss is calculated differently depending on whether a label exists or not. When a label exists, a weighted cross entropy loss is used, while when a label does not exist, a logarithmic loss function adjusted by a weight coefficient is used.

7. The method for detecting polyploid tumor giant cells in ovarian cancer based on deep learning in H&E images according to any one of claims 1 to 6, characterized in that: In step S4, the training process of the OCDet model includes two stages: Freezing stage: In this stage, only some parameters of the model are fine-tuned to stabilize the basic performance of the model. The number of training iterations in this stage is the set number of cycles; Thawing stage: After the freezing stage, all parameters of the model are trained to further improve the model's fit to the data and detection performance. The number of training iterations in this stage is more cycles.

8. The method for detecting polyploid giant cells of ovarian cancer in H&E images based on deep learning according to claim 7, characterized in that: In the freezing phase and the thawing phase, the model is trained using a uniform batch size, and the model parameters are updated through a back-propagation algorithm and a gradient descent optimizer.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, it implements the method for detecting polyploid tumor giant cells of ovarian cancer in H&E images based on deep learning as described in any one of claims 1 to 8.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, it implements the method for detecting polyploid tumor giant cells of ovarian cancer in H&E images based on deep learning as described in any one of claims 1 to 8.

Images