Method and system for interpreting er and PR expression in breast cancer by using artificial intelligence

Abstract

The present invention relates to the technical field of artificial intelligence. Provided are a method and system for interpreting ER and PR expression in breast cancer by using artificial intelligence. The method comprises: collecting annotated pathological section images of breast cancer tissue, and performing data enhancement to obtain a sample image set; performing color difference balancing on image data in the sample image set to obtain a preprocessed data set; using the preprocessed data set to train a preset initial segmentation network; inputting the pathological section images of breast cancer tissue into a trained cell nucleus segmentation model to obtain a cell nucleus region image; morphologically processing the cell nucleus region image to obtain a processed target region; and extracting color features and morphological features within the target region, and inputting same into a classification model to obtain an interpretation result. The present invention uses a deep learning model to perform cell nucleus segmentation and feature extraction, thereby reducing subjectivity and errors brought by manual interpretation and improving the accuracy of interpretation.

Description

A method and system for interpreting ER and PR expression in breast cancer using artificial intelligence. Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method and system for interpreting the expression of ER and PR in breast cancer using artificial intelligence. Background Technology

[0002] Breast cancer is one of the most common cancers that seriously threaten women's health and is the second leading cause of cancer death among women. According to the latest breast cancer data report for my country in 2019, approximately 304,000 new cases of breast cancer are diagnosed annually in my country, accounting for 11.2% of global cases and 9.2% of cancer deaths, ranking among the highest in the world. This number is increasing at a rate of about 2% annually, indicating that breast cancer has become the most serious malignant tumor threatening the health of women in my country. Over the past decade, exciting progress has been made in the detection and treatment of breast cancer. The average 5-year survival rate for breast cancer in the United States is 90%, while in China it is only 73.1%. Therefore, improving the survival rate of breast cancer patients in my country is a significant challenge. Providing high-quality early detection of breast cancer, precise individualized treatment services, and predictive assessment for patients with recurrence are key factors in improving the 5-10 year survival rate of breast cancer patients.

[0003] A method for interpreting breast cancer immunohistochemical staining using artificial intelligence is disclosed in patent application number 202110170428.0, comprising the following steps: First, a deep convolutional neural network is used to classify and identify breast cancer tissue from a digital panoramic image of HE microscopic histopathology. Second, based on a pyramid model, rapid and accurate registration is performed between the HE and IHC microscopic breast cancer histopathology digital panoramic images. Third, the nuclear positivity rate of immunohistochemistry is calculated. This patent can automatically and quantitatively calculate and analyze the nuclear positivity rate of breast cancer tissue regions using immunohistochemistry. While ensuring the accuracy of the nuclear positivity rate calculation results, it reduces the large amount of repetitive and tedious work of doctors manually counting cells, significantly improving the calculation efficiency of the nuclear positivity rate and efficiently assisting doctors in completing the immunohistochemical nuclear positivity analysis process. However, this patent does not consider that the statistical results may be inaccurate due to factors such as cell overlap and background noise when counting nuclear positivity cells and non-nucleus positivity cells.

[0004] In view of the shortcomings of the aforementioned technologies, a new technical solution is needed to improve the above-mentioned technical problems. Summary of the Invention

[0005] To overcome the shortcomings of existing technologies, the purpose of this invention is to provide a method and system for interpreting the expression of ER and PR in breast cancer using artificial intelligence. This method not only improves the accuracy and efficiency of interpretation but also enhances consistency and reproducibility, possessing broad clinical application value and research potential. Through automated and standardized processes, it can provide strong support for the early diagnosis and personalized treatment of breast cancer.

[0006] To achieve the above objectives, the present invention provides the following solution:

[0007] A method for interpreting the expression of ER and PR in breast cancer using artificial intelligence includes:

[0008] Collect labeled images of breast cancer tissue pathology sections; the images of breast cancer tissue pathology sections include IHC images stained with ER and PR;

[0009] Data augmentation was performed on the breast cancer tissue pathology slide images to obtain a sample image set;

[0010] Color difference balancing is performed on the image data in the sample image set to obtain a preprocessed dataset;

[0011] The preprocessed dataset is used to train the preset initial segmentation network to obtain a trained cell nucleus segmentation model;

[0012] The breast cancer tissue pathology slide image is input into a trained cell nucleus segmentation model to obtain a cell nucleus region image;

[0013] Morphological processing is performed on the image of the cell nucleus region to obtain the processed target region;

[0014] Color and morphological features are extracted within the target area, and these features are input into a trained classification model to determine the positive or negative expression of ER and PR in each cell nucleus, thus obtaining the interpretation result.

[0015] Preferably, color and morphological features are extracted within the target region, and the color and morphological features are input into a trained classification model to determine the positive or negative expression of ER and PR in each cell nucleus. After obtaining the determination result, the method further includes:

[0016] The proportion of ER and PR positive cell nuclei was statistically analyzed based on the interpretation results to calculate the overall positive rate;

[0017] Heatmaps were generated based on the proportion of ER and PR positive cell nuclei and the overall positivity rate to show the spatial distribution of ER and PR expression.

[0018] Preferably, the data enhancement includes: image rotation, image flipping, and image scaling.

[0019] Preferably, color difference balancing is performed on the image data in the sample image set to obtain a preprocessed dataset, including:

[0020] Calculate the pixel mean of the image data in the three RGB channels of the sample image set; wherein, the formula for calculating the pixel mean is:

[0021] Where D represents the pixel mean, This represents the pixel mean of the image data in the red channel within the sample image set. This represents the pixel mean of the image data in the green channel within the sample image set. This represents the pixel mean of the image data in the blue channel within the sample image set;

[0022] Calculate the balance coefficient for each color channel based on the pixel mean;

[0023] The color difference is balanced using the balance coefficient to obtain a preprocessed dataset.

[0024] Preferably, the image data in the sample image set is subjected to color difference balancing using the balancing coefficient to obtain a preprocessed dataset, including:

[0025] Formula used:

[0026] Color difference balancing is performed on the image data in the sample image set to obtain a color-balanced video surveillance image; where K R K represents the balance coefficient of the red channel. G K represents the balance coefficient of the green channel. B R' represents the balance coefficient of the blue channel, G' represents the pixel value of the red channel after balance, B' represents the pixel value of the green channel after balance, R represents the pixel value of the image data in the sample image set in the red channel, G represents the pixel value of the image data in the sample image set in the green channel, and B represents the pixel value of the image data in the sample image set in the blue channel.

[0027] Preferably, the initial segmentation network is a U-Net network.

[0028] Preferably, the image of the cell nucleus region is subjected to morphological processing to obtain the processed target region, including:

[0029] Connectivity analysis is performed on the cell nucleus region image to delete regions in the cell nucleus region image whose area is smaller than the noise area threshold, thus obtaining the first region;

[0030] The first region is expanded using a preset first structural element to connect the broken cell nuclear boundaries, thus obtaining the second region.

[0031] The second region is eroded using a pre-defined second structural element to restore the original shape of the cell nucleus, thus obtaining the third region.

[0032] An adaptive boundary smoothing algorithm is applied to the third region to smooth it, resulting in the processed target region.

[0033] Preferably, the expression for the processed target region is:

[0034] Where f(x,y) is the pixel value of the processed target region, indicating whether the pixel belongs to the cell nucleus region after the dilation or erosion operation; I(x+i,y+j) is the pixel value of the cell nucleus region image at position (x+i,y+j), indicating whether that position belongs to the cell nucleus; S is the neighborhood of the first and second structuring elements; T connect is the preset connection threshold, x is the horizontal coordinate of the image, y is the vertical coordinate of the image, and i and j both represent the relative offset in the structuring element.

[0035] A system for interpreting the expression of ER and PR in breast cancer using artificial intelligence includes:

[0036] The image collection module is used to collect labeled images of breast cancer tissue pathology sections; the images of breast cancer tissue pathology sections include ER and PR stained IHC images;

[0037] The data augmentation module is used to perform data augmentation on the breast cancer tissue pathology slide images to obtain a sample image set;

[0038] The color difference balance module is used to perform color difference balance on the image data in the sample image set to obtain a preprocessed dataset.

[0039] The model training module is used to train the preset initial segmentation network using the preprocessed dataset to obtain a trained cell nucleus segmentation model.

[0040] The detection module is used to input the breast cancer tissue pathology slide image into the trained cell nucleus segmentation model to obtain the cell nucleus region image;

[0041] The morphological processing module is used to perform morphological processing on the cell nucleus region image to obtain the processed target region.

[0042] The interpretation module is used to extract color and morphological features within the target area and input the color and morphological features into a trained classification model to interpret the positive or negative expression of ER and PR in each cell nucleus, thereby obtaining the interpretation result.

[0043] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0044] This invention provides a method and system for interpreting the expression of ER and PR in breast cancer using artificial intelligence. The method includes: collecting labeled breast cancer tissue pathological slide images; the breast cancer tissue pathological slide images include IHC images stained with ER and PR; performing data augmentation on the breast cancer tissue pathological slide images to obtain a sample image set; performing color difference balancing on the image data in the sample image set to obtain a preprocessed dataset; training a preset initial segmentation network using the preprocessed dataset to obtain a trained cell nucleus segmentation model; inputting the breast cancer tissue pathological slide images into the trained cell nucleus segmentation model to obtain cell nucleus region images; performing morphological processing on the cell nucleus region images to obtain processed target regions; extracting color features and morphological features within the target regions, and inputting the color features and morphological features into a trained classification model to interpret the positive or negative expression of ER and PR in each cell nucleus, obtaining the interpretation result. This invention uses a deep learning model for cell nucleus segmentation and feature extraction, which reduces the subjectivity and error of manual interpretation and improves the accuracy of interpretation. Furthermore, by using a trained cell nucleus segmentation model, cell nucleus regions can be accurately identified, ensuring more accurate feature extraction for each cell nucleus. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 is a flowchart of the method provided in an embodiment of the present invention;

[0047] Figure 2 is a schematic diagram of the system structure provided in an embodiment of the present invention. Detailed Implementation

[0048] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] The purpose of this invention is to provide a method and system for interpreting the expression of ER and PR in breast cancer using artificial intelligence. This method not only improves the accuracy and efficiency of interpretation but also enhances consistency and reproducibility, possessing broad clinical application value and research potential. Through automated and standardized processes, it can provide strong support for the early diagnosis and personalized treatment of breast cancer.

[0050] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0051] Figure 1 is a flowchart of the method provided in an embodiment of the present invention. As shown in Figure 1, the present invention provides a method for interpreting the expression of ER and PR in breast cancer using artificial intelligence, including:

[0052] Step 100: Collect labeled images of breast cancer tissue pathology sections; the images of breast cancer tissue pathology sections include IHC images stained with ER and PR.

[0053] Step 200: Perform data augmentation on the breast cancer tissue pathology slide images to obtain a sample image set;

[0054] Step 300: Perform color difference balancing on the image data in the sample image set to obtain a preprocessed dataset;

[0055] Step 400: Train the pre-processed dataset on the preset initial segmentation network to obtain a trained cell nucleus segmentation model;

[0056] Step 500: Input the breast cancer tissue pathology slide image into the trained cell nucleus segmentation model to obtain the cell nucleus region image;

[0057] Step 600: Perform morphological processing on the cell nucleus region image to obtain the processed target region;

[0058] Step 700: Extract color and morphological features within the target area, and input the color and morphological features into the trained classification model to determine the positive or negative expression of ER and PR in each cell nucleus, and obtain the interpretation result.

[0059] Preferably, after step 700, the method further includes:

[0060] Step 800: Based on the interpretation results, the proportion of ER and PR positive cell nuclei is counted to calculate the overall positive rate;

[0061] Step 900: Generate a heatmap based on the proportion of ER and PR positive cell nuclei and the overall positivity rate to show the spatial distribution of ER and PR expression.

[0062] Specifically, step 100 in this embodiment includes:

[0063] 1. Determine data sources. First, collaborate with hospitals, pathology laboratories, or biobanks to obtain histopathological images of breast cancer patients. These images should be sections stained with immunohistochemistry (IHC), specifically for ER (estrogen receptor) and PR (progesterone receptor).

[0064] 2. Image Acquisition: The slides are scanned using a high-resolution digital pathology scanner to ensure image clarity and detail preservation. In this embodiment, an appropriate magnification (such as 20x or 40x) is selected during scanning to facilitate subsequent analysis.

[0065] 3. Annotation Process: Image annotation is performed using professional image annotation software (such as ImageJ, QuPath, or Labelbox). Annotation criteria for ER and PR positive cells are defined based on clinical standards and research needs. Positive cells typically exhibit specific staining intensities and distribution patterns. This embodiment requires image annotation by experienced pathologists to ensure accuracy and consistency. The positive or negative status of each cell nucleus should be clearly marked, and relevant metadata (such as patient information, sample origin, etc.) should be recorded.

[0066] 4. Data storage: The labeled images and their corresponding annotation information (such as cell nucleus location, positive / negative status) are stored in a database to ensure data security and accessibility. Images can be saved using standard image formats (such as TIFF or PNG), and annotation information can be stored using JSON or XML formats.

[0067] 5. Data quality check: After data collection is completed, a quality check is conducted to ensure that all images meet research standards and are accurately labeled. If necessary, a second review is performed to ensure the reliability of the data.

[0068] 6. Data preparation: Organize the collected images and annotation information into a standardized format to facilitate subsequent data augmentation and processing steps. Ensure the dataset has a clear structure to facilitate automated processing.

[0069] Through the above steps, this embodiment successfully collected labeled breast cancer tissue pathology slide images, laying the foundation for subsequent image processing and analysis.

[0070] Furthermore, both ER and PR proteins are expressed in the nuclei of tumor cells. Positive staining results in brown color. Tumor cell nuclei stained light brown are interpreted as positive 1+; stained brown are interpreted as positive 2+; stained dark brown are interpreted as positive 3+; and unstained nuclei are interpreted as negative.

[0071] Furthermore, this embodiment first determines the staining intensity (1–3+), then calculates the percentage of positive cells, not the sum of three positive cells. Positive internal controls can be selected from adjacent normal breast ductal epithelial cells, whose ER and PR staining intensities vary. Negative internal controls can be selected from myoepithelial cells and stromal cells in breast tissue, which do not express ER and PR. Exemplarily, this embodiment also calculates the percentage of positively expressed cells among tumor cells (e.g., ER (3+, 80%)).

[0072] Preferably, the data enhancement includes: image rotation, image flipping, and image scaling.

[0073] Specifically, in step 200, the data augmentation process for the breast cancer tissue pathology slide images is as follows: First, a subset of images is randomly selected from the labeled breast cancer tissue pathology slide images collected in step 100 as the basis for data augmentation. Next, image processing libraries (such as OpenCV or PIL) are used to apply various data augmentation techniques to each image to increase the diversity and quantity of samples. Specific augmentation methods include: image rotation (e.g., random rotation from 0 to 30 degrees), horizontal and vertical flipping, random cropping (preserving the integrity of cell nuclei), adjusting brightness and contrast, adding Gaussian noise, and color jittering. These augmentation operations can simulate different shooting conditions and sample variability, thereby improving the model's generalization ability. After each augmentation, the generated new image is saved along with the original image, along with corresponding annotation information, ensuring that the augmented sample retains the correct label. Finally, through these augmentation operations, an image set containing diverse samples is generated, providing a rich data foundation for subsequent model training.

[0074] Furthermore, in this embodiment, a series of data augmentation operations are defined to improve the robustness and generalization ability of the model, taking into account the characteristics of breast cancer tissue pathology slide images. Commonly used augmentation methods include: rotation, which simulates different slice orientations by randomly rotating the image (e.g., from 0 to 30 degrees) to ensure the model can adapt to different viewpoints; flipping, including horizontal and vertical flipping, to increase sample diversity, especially when the cell nucleus has strong symmetry; cropping, which randomly crops a portion of the image while maintaining the integrity of the cell nucleus to enhance the model's ability to learn local features; brightness and contrast adjustment, which simulates different staining and imaging conditions by randomly changing the brightness and contrast of the image; adding noise, such as Gaussian noise, to improve the model's robustness to image noise; color jittering, which randomly adjusts the hue, saturation, and brightness of the image to enhance the model's adaptability to different staining intensities; and affine transformation, which further increases sample diversity through operations such as translation, scaling, and tilting. These augmentation strategies aim to simulate various variations that may be encountered in the real world, ensuring that the trained model can maintain good performance under different conditions.

[0075] Preferably, color difference balancing is performed on the image data in the sample image set to obtain a preprocessed dataset, including:

[0076] Calculate the pixel mean of the image data in the three RGB channels of the sample image set; wherein, the formula for calculating the pixel mean is:

[0077] Where D represents the pixel mean, This represents the pixel mean of the image data in the red channel within the sample image set. This represents the pixel mean of the image data in the green channel within the sample image set. This represents the pixel mean of the image data in the blue channel within the sample image set;

[0078] Calculate the balance coefficient for each color channel based on the pixel mean;

[0079] The color difference is balanced using the balance coefficient to obtain a preprocessed dataset.

[0080] Preferably, the image data in the sample image set is subjected to color difference balancing using the balancing coefficient to obtain a preprocessed dataset, including:

[0081] Formula used:

[0082] Color difference balancing is performed on the image data in the sample image set to obtain a color-balanced video surveillance image; where K R K represents the balance coefficient of the red channel.G K represents the balance coefficient of the green channel. B R' represents the balance coefficient of the blue channel, G' represents the pixel value of the red channel after balance, B' represents the pixel value of the green channel after balance, R represents the pixel value of the image data in the sample image set in the red channel, G represents the pixel value of the image data in the sample image set in the green channel, and B represents the pixel value of the image data in the sample image set in the blue channel.

[0083] Specifically, this embodiment considers that when external lighting conditions change, the surface color of cells remains unchanged, but the color of objects captured by imaging devices will deviate with changes in lighting, gradually deviating from the true color of the object. This invention utilizes a color difference balance algorithm to compensate for this deficiency in imaging devices and correct color differences in the image.

[0084] Preferably, the initial segmentation network is a U-Net network.

[0085] Optionally, this embodiment trains the pre-processed initial segmentation network U-Net using a pre-processed dataset as follows: First, a pre-processed dataset, including the input image and corresponding labeled mask, is prepared after data augmentation and color difference balancing. Next, a U-Net model is built using a deep learning framework (such as TensorFlow or PyTorch). This model consists of an encoder and a decoder, capable of effectively extracting image features and performing pixel-level segmentation. During training, the pre-processed dataset is divided into a training set and a validation set to monitor model performance. Then, a loss function (such as cross-entropy loss or Dice loss) and an optimizer (such as Adam or SGD) are defined to optimize the model parameters. Next, U-Net is iteratively trained using the training set. The output is calculated through forward propagation and compared with the ground truth labels, while the model weights are updated through backpropagation. During training, model performance is periodically evaluated on the validation set, and hyperparameters (such as learning rate, batch size, etc.) are adjusted to improve segmentation accuracy. After training, the trained U-Net model is saved for use in subsequent steps for segmenting the cell nucleus region. Through this series of steps, the U-Net model can effectively learn the nuclear features in breast cancer tissue pathology slide images and achieve high-quality nuclear segmentation.

[0086] Preferably, the image of the cell nucleus region is subjected to morphological processing to obtain the processed target region, including:

[0087] Connectivity analysis is performed on the cell nucleus region image to delete regions in the cell nucleus region image whose area is smaller than the noise area threshold, thus obtaining the first region;

[0088] The first region is expanded using a preset first structural element to connect the broken cell nuclear boundaries, thus obtaining the second region.

[0089] The second region is eroded using a pre-defined second structural element to restore the original shape of the cell nucleus, thus obtaining the third region.

[0090] An adaptive boundary smoothing algorithm is applied to the third region to smooth it, resulting in the processed target region.

[0091] Specifically, in step 400, the morphological processing of the cell nucleus region image to obtain the processed target region is implemented as follows: First, a connected component analysis algorithm (such as the connectedComponents function in OpenCV) is used to analyze the cell nucleus region image, identify all connected components, and calculate the area of ​​each region. Based on a preset noise area threshold, regions with areas smaller than the threshold are deleted to obtain the first region. Next, a preset first structuring element (such as a 3x3 circular or cross-shaped structuring element) is used to dilate the first region (using the cv2.dilate function) to connect broken cell nucleus boundaries, forming a more complete cell nucleus shape, resulting in the second region. Then, a preset second structuring element (usually a smaller shape than the first structuring element) is applied to erode the second region (using the cv2.erode function) to remove excess noise and restore the original shape of the cell nucleus, resulting in the third region. Finally, an adaptive boundary smoothing algorithm (such as a smoothing method based on Gaussian blur or median filtering) is applied to smooth the third region to reduce boundary irregularities, resulting in the final processed target region. This series of morphological operations can effectively extract and optimize the cell nucleus region, providing a clear foundation for subsequent feature extraction and classification.

[0092] Optionally, the first structural element in this embodiment is used for the expansion operation to connect the broken nuclear boundaries. This embodiment selects a small circular or elliptical structural element because these shapes can effectively fill the gaps between cell nuclei. The size and shape of the structural element are adjusted according to the average size and shape of the cell nucleus.

[0093] Furthermore, the second structural element in this embodiment is used for the erosion operation to restore the original shape of the cell nucleus. In this embodiment, the structural element is the same as or slightly smaller than that used in the expansion operation to ensure that the connection effect is not compromised while restoring the shape. Selecting an appropriate structural element size is crucial; this embodiment determined experimentally to achieve optimal results.

[0094] Furthermore, this embodiment selects a suitable noise area threshold by analyzing the distribution of cell nucleus size in the training dataset. In this embodiment, it is set to a percentage (e.g., 10%) smaller than the average cell nucleus area to remove noise without affecting the real cell nucleus.

[0095] Preferably, the expression for the processed target region is:

[0096] Where f(x,y) is the pixel value of the processed target region, indicating whether the pixel belongs to the cell nucleus region after the dilation or erosion operation; I(x+i,y+j) is the pixel value of the cell nucleus region image at position (x+i,y+j), indicating whether that position belongs to the cell nucleus; S is the neighborhood of the first and second structuring elements; T connect is the preset connection threshold, x is the horizontal coordinate of the image, y is the vertical coordinate of the image, and i and j both represent the relative offset in the structuring element.

[0097] Optionally, the connection threshold in this embodiment is dynamically adjusted and set according to the characteristics of the image and the density of cell nuclei. This embodiment selects a threshold that can effectively connect broken boundaries by analyzing the connection effects of different images.

[0098] Furthermore, the method for determining the preset connection threshold is as follows:

[0099] 1. Image Feature Analysis: First, feature analysis is performed on the image of the cell nucleus region to be processed, including the shape, size, distribution density, and boundary features of the cell nuclei. By calculating the image's histogram, edge intensity, and texture features, preliminary information about the distribution and connectivity of the cell nuclei can be obtained.

[0100] 2. Preliminary Threshold Setting: Based on image features, a preliminary connectivity threshold is set. In this embodiment, this preliminary threshold is based on empirical rules or derived through statistical analysis of existing labeled data. For example, the average or median of cell nucleus areas in the image is selected as a reference for the preliminary threshold.

[0101] 3. Evaluation of the connection effect: A dilation operation is performed on the cell nucleus region image using a pre-set connection threshold, and the connection effect is evaluated. This embodiment judges the quality of the connection effect by calculating indicators such as the number of cell nuclei, area change, and boundary integrity after connection.

[0102] 4. Dynamically adjust the threshold: Based on the evaluation results, dynamically adjust the connection threshold. If the connection effect is not ideal (e.g., connecting too many or too few cell nuclei), the threshold can be gradually increased or decreased until an optimal value that can effectively connect broken boundaries is found. This process can be performed using cross-validation, i.e., testing different thresholds on different images and selecting the threshold that performs best in most images.

[0103] 5. Final Confirmation: After determining the optimal connection threshold, a final confirmation is performed. By comparing the results with expert annotations, it is ensured that the selected threshold can effectively connect the nuclear boundaries and maintain the morphological characteristics of the cell nucleus without excessive alteration.

[0104] Through the above steps, the selection of the connection threshold in this embodiment not only considers image features and cell nucleus density but also incorporates actual connection effect evaluation, ensuring optimal cell nucleus connection results under different image conditions. The dynamic adjustment method in this embodiment improves the algorithm's adaptability and accuracy, providing a solid foundation for subsequent cell nucleus analysis.

[0105] Optionally, in step 700, the specific implementation steps for feature extraction and classification of the processed target region image are as follows: First, color and morphological features are extracted from the target region image obtained after morphological processing. Color features are obtained by calculating the average color value, standard deviation, and color histogram of each cell nucleus region. Commonly used color spaces include RGB and HSV to better capture the staining intensity and distribution of the cell nucleus. Morphological features include the area, perimeter, shape factors (such as roundness and aspect ratio), and boundary complexity of the cell nucleus. These features are calculated using the contour analysis function in an image processing library (such as OpenCV). Next, the extracted color and morphological features are combined into a feature vector and input into a pre-trained classification model (such as a support vector machine, random forest, or deep learning model) to determine the positive or negative expression of ER and PR in each cell nucleus. The classification model makes predictions based on the feature vector and outputs the classification result for each cell nucleus. Finally, the judgment results are associated with the original image to generate annotation information containing the ER and PR expression status of each cell nucleus, providing a basis for subsequent analysis and clinical decision-making. This series of steps enables the effective classification and interpretation of cell nuclei, supporting in-depth research into breast cancer histopathology.

[0106] Furthermore, taking a Convolutional Neural Network (CNN) as an example, this embodiment constructs a classification model for cell nucleus classification. Its structure typically includes multiple convolutional layers, pooling layers, and fully connected layers. Specifically, the model's input layer receives the extracted cell nucleus feature vectors, then extracts local features through a series of convolutional layers (e.g., 3x3 convolutional kernels), with each convolutional layer followed by an activation function (e.g., ReLU) to introduce non-linearity. Subsequently, pooling layers (e.g., max pooling) are used for downsampling, reducing feature dimensionality and enhancing the model's translation invariance. After several convolutional and pooling layers, the feature map is flattened and input into the fully connected layer. Finally, an output layer (typically using a softmax activation function) performs classification, outputting the ER and PR expression status (positive or negative) of each cell nucleus. In implementing classification, this embodiment first inputs the processed cell nucleus image into the model, calculates the feature representation of each cell nucleus through forward propagation, and then uses the trained model for prediction, outputting the classification result. The model is trained using a labeled dataset, and the network parameters are optimized through backpropagation to minimize the loss between the predicted results and the true labels (such as cross-entropy loss), thereby improving classification accuracy and robustness. With this structure, CNNs can effectively capture the complex features of the cell nucleus, achieving efficient classification tasks.

[0107] Furthermore, the interpretation result of this embodiment may include the following information:

[0108] 1. Cell nucleus classification results: Each cell nucleus is classified as either positive or negative for ER (estrogen receptor) and PR (progesterone receptor) expression. This classification result in this embodiment is based on extracted feature vectors and predicted using a trained classification model, reflecting the cell nucleus's response to the corresponding hormones.

[0109] 2. Confidence Score: For each classification result, the model outputs a confidence score, representing the model's confidence level in that classification result. Typically, this score is calculated using a softmax layer, with values ​​ranging from 0 to 1. A score closer to 1 indicates that the model is more confident that the cell nucleus belongs to a particular category.

[0110] 3. The morphological characteristics of the cell nucleus: The interpretation results may also include the morphological characteristics of each cell nucleus, such as area, perimeter, shape factors (such as roundness, aspect ratio), etc. These characteristics in this embodiment help to further analyze the morphological changes of the cell nucleus and their biological significance.

[0111] 4. Image annotation information: The classification result and confidence score of each cell nucleus are annotated on the original image. Positive and negative cell nuclei are usually represented by different colors or symbols, which facilitates intuitive observation and analysis.

[0112] 5. Statistical analysis results: Statistical analysis is performed on the classification results of cell nuclei in the entire image, including the number and proportion of positive cell nuclei, and the distribution of different types of cell nuclei. These statistical data in this embodiment can provide important references for clinical decision-making.

[0113] These interpretations allow researchers and clinicians to better understand the biological characteristics of the cell nucleus and assess the hormone receptor status of breast cancer tissue, thus providing a basis for subsequent treatment planning.

[0114] Optionally, in step 800, the specific steps for calculating the overall positivity rate by statistically analyzing the proportions of ER and PR positive cell nuclei based on the interpretation results are as follows: First, in this embodiment, the classification information of all cell nuclei, including the ER and PR expression status of each cell nucleus, is extracted from the cell nucleus classification results obtained in step 700. Next, the number of ER-positive cell nuclei and PR-positive cell nuclei are counted separately, and the total number of cell nuclei is calculated. The proportion of ER-positive cell nuclei is obtained by dividing the number of ER-positive cell nuclei by the total number of cell nuclei; similarly, the proportion of PR-positive cell nuclei is calculated. Finally, the overall positivity rate can be obtained by adding the number of ER-positive cell nuclei and PR-positive cell nuclei together and then dividing by the total number of cell nuclei. This embodiment uses the NumPy or Pandas library in Python to efficiently process data, ensuring the accuracy and reliability of the statistical results.

[0115] Further, in step 900, the specific steps for generating a heatmap based on the proportion of ER and PR positive cell nuclei and the overall positivity rate to illustrate the spatial distribution of ER and PR expression are as follows: First, using the spatial location information of cell nuclei (such as coordinates) and classification results, a two-dimensional grid is constructed, with each cell representing a region in the image. Next, based on the number of cell nuclei and their positivity status within each grid cell, the proportion of ER and PR positive cell nuclei in that region is calculated. Then, using a heatmap generation tool (such as the Matplotlib or Seaborn library), these proportions are mapped to color values ​​to generate a heatmap, where the color intensity represents the strength of ER and PR expression. Finally, the heatmap is overlaid on the original image to visually illustrate the spatial distribution of ER and PR, helping researchers and clinicians better understand the expression patterns of hormone receptors in cell nuclei and their distribution characteristics in tissues.

[0116] Corresponding to the above method, as shown in Figure 2, this embodiment also provides a system for interpreting the expression of ER and PR in breast cancer using artificial intelligence, including:

[0117] The image collection module is used to collect labeled images of breast cancer tissue pathology sections; the images of breast cancer tissue pathology sections include ER and PR stained IHC images;

[0118] The data augmentation module is used to perform data augmentation on the breast cancer tissue pathology slide images to obtain a sample image set;

[0119] The color difference balance module is used to perform color difference balance on the image data in the sample image set to obtain a preprocessed dataset.

[0120] The model training module is used to train the preset initial segmentation network using the preprocessed dataset to obtain a trained cell nucleus segmentation model.

[0121] The detection module is used to input the breast cancer tissue pathology slide image into the trained cell nucleus segmentation model to obtain the cell nucleus region image;

[0122] The morphological processing module is used to perform morphological processing on the cell nucleus region image to obtain the processed target region.

[0123] The interpretation module is used to extract color and morphological features within the target area and input the color and morphological features into a trained classification model to interpret the positive or negative expression of ER and PR in each cell nucleus, thereby obtaining the interpretation result.

[0124] The beneficial effects of this invention are as follows:

[0125] (1) This invention uses a deep learning model to segment cell nuclei and extract features, which reduces the subjectivity and error of manual interpretation and improves the accuracy of interpretation. With the trained cell nucleus segmentation model, cell nucleus regions can be accurately identified, ensuring more accurate feature extraction for each cell nucleus.

[0126] (2) Compared with traditional manual interpretation, the automated AI method of this invention can process a large number of images in a short time, significantly improving work efficiency; it can analyze multiple samples at the same time, adapt to clinical needs, and support large-scale screening and research.

[0127] (3) This invention ensures consistent interpretation results between different samples and different operators through a unified algorithm and model, reducing the influence of human factors; the training and application process of the AI ​​model can be repeated to ensure similar results under different experimental conditions.

[0128] (4) By extracting color and morphological features, this invention can comprehensively assess the state of the cell nucleus and provide richer information for interpretation; by utilizing the feature learning capabilities of deep learning models, it can automatically discover and utilize complex feature relationships, thereby improving the sensitivity and specificity of interpretation.

[0129] (5) Through data augmentation technology, the model can adapt to different sample variability and improve the generalization ability of the model; color difference balance processing ensures the consistency of image data under different staining conditions, so that the model can still maintain good performance under different experimental conditions.

[0130] (6) This invention provides pathologists with reliable interpretation results, assists in clinical decision-making, and helps doctors develop personalized treatment plans; through rapid and accurate interpretation, it can help achieve early diagnosis of breast cancer and improve the survival rate of patients.

[0131] (7) The application of this invention can promote the development of breast cancer-related research and drive the research and development of new drugs and treatment plans; with the accumulation of more data, the model can be continuously optimized to improve its applicability in different clinical scenarios.

[0132] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0133] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for interpreting the expression of ER and PR in breast cancer using artificial intelligence, characterized in that, include: Collect labeled images of breast cancer tissue pathology sections; the images of breast cancer tissue pathology sections include IHC images stained with ER and PR; Data augmentation was performed on the breast cancer tissue pathology slide images to obtain a sample image set; Color difference balancing is performed on the image data in the sample image set to obtain a preprocessed dataset; The preprocessed dataset is used to train the preset initial segmentation network to obtain a trained cell nucleus segmentation model; The breast cancer tissue pathology slide image is input into a trained cell nucleus segmentation model to obtain a cell nucleus region image; Morphological processing is performed on the image of the cell nucleus region to obtain the processed target region; Color and morphological features are extracted within the target area, and these features are input into a trained classification model to determine the positive or negative expression of ER and PR in each cell nucleus, thus obtaining the interpretation result.

2. The method for interpreting the expression of ER and PR in breast cancer using artificial intelligence according to claim 1, characterized in that, Color and morphological features are extracted within the target region, and these features are input into a trained classification model to determine the positive or negative expression of ER and PR in each cell nucleus. After obtaining the determination result, the model further includes: The proportion of ER and PR positive cell nuclei was statistically analyzed based on the interpretation results to calculate the overall positive rate; Heatmaps were generated based on the proportion of ER and PR positive cell nuclei and the overall positivity rate to show the spatial distribution of ER and PR expression.

3. The method for interpreting the expression of ER and PR in breast cancer using artificial intelligence according to claim 1, characterized in that, The data augmentations include: image rotation, image flipping, and image scaling.

4. The method for interpreting the expression of ER and PR in breast cancer using artificial intelligence according to claim 1, characterized in that, Color difference balancing is performed on the image data in the sample image set to obtain a preprocessed dataset, including: Calculate the pixel mean of the image data in the three RGB channels of the sample image set; wherein, the formula for calculating the pixel mean is: Where D represents the pixel mean, This represents the pixel mean of the image data in the red channel within the sample image set. This represents the pixel mean of the image data in the green channel within the sample image set. This represents the pixel mean of the image data in the blue channel within the sample image set; Calculate the balance coefficient for each color channel based on the pixel mean; The color difference is balanced using the balance coefficient to obtain a preprocessed dataset.

5. The method for interpreting the expression of ER and PR in breast cancer using artificial intelligence according to claim 4, characterized in that, The color difference is balanced using the balance coefficient on the image data in the sample image set to obtain a preprocessed dataset, including: Formula used: Color difference balancing is performed on the image data in the sample image set to obtain a color-balanced video surveillance image; where K R K represents the balance coefficient of the red channel. G K represents the balance coefficient of the green channel. B R' represents the balance coefficient of the blue channel, G' represents the pixel value of the red channel after balance, B' represents the pixel value of the green channel after balance, R represents the pixel value of the image data in the sample image set in the red channel, G represents the pixel value of the image data in the sample image set in the green channel, and B represents the pixel value of the image data in the sample image set in the blue channel.

6. The method for interpreting the expression of ER and PR in breast cancer using artificial intelligence according to claim 1, characterized in that, The initial segmentation network is a U-Net network.

7. The method for interpreting the expression of ER and PR in breast cancer using artificial intelligence according to claim 1, characterized in that, Morphological processing is performed on the image of the cell nucleus region to obtain the processed target region, including: Connectivity analysis is performed on the cell nucleus region image to delete regions in the cell nucleus region image whose area is smaller than the noise area threshold, thus obtaining the first region; The first region is expanded using a preset first structural element to connect the broken cell nuclear boundaries, thus obtaining the second region. The second region is eroded using a pre-defined second structural element to restore the original shape of the cell nucleus, thus obtaining the third region. An adaptive boundary smoothing algorithm is applied to the third region to smooth it, resulting in the processed target region.

8. The method for interpreting the expression of ER and PR in breast cancer using artificial intelligence according to claim 1, characterized in that, The expression for the processed target region is: Where f(x,y) is the pixel value of the processed target region, indicating whether the pixel belongs to the cell nucleus region after the dilation or erosion operation; I(x+i,y+j) is the pixel value of the cell nucleus region image at position (x+i,y+j), indicating whether that position belongs to the cell nucleus; S is the neighborhood of the first and second structuring elements; T connect is the preset connection threshold, x is the horizontal coordinate of the image, y is the vertical coordinate of the image, and i and j both represent the relative offset in the structuring element.

9. A system for interpreting the expression of ER and PR in breast cancer using artificial intelligence, characterized in that, include: The image collection module is used to collect labeled images of breast cancer tissue pathology sections; the images of breast cancer tissue pathology sections include ER and PR stained IHC images; The data augmentation module is used to perform data augmentation on the breast cancer tissue pathology slide images to obtain a sample image set; The color difference balance module is used to perform color difference balance on the image data in the sample image set to obtain a preprocessed dataset. The model training module is used to train the preset initial segmentation network using the preprocessed dataset to obtain a trained cell nucleus segmentation model. The detection module is used to input the breast cancer tissue pathology slide image into the trained cell nucleus segmentation model to obtain the cell nucleus region image; The morphological processing module is used to perform morphological processing on the cell nucleus region image to obtain the processed target region. The interpretation module is used to extract color and morphological features within the target area and input the color and morphological features into a trained classification model to interpret the positive or negative expression of ER and PR in each cell nucleus, thereby obtaining the interpretation result.

Images