Intelligent identification and grading method for stomach intestinal metaplasia

Abstract

The embodiment of the invention discloses an intelligent recognition and grading method for gastric intestinal metaplasia, and the method comprises the steps: collecting a gastric mucosa HE staining digital pathological image, and building a universal part recognition model through a convolutional neural network according to the pathological classification standards of different parts of a stomach body, gastric antrum and gastric horn. In order to reduce judgment errors caused by other glands, a semantic segmentation model is trained, and the gastric mucosa is accurately found out. On the basis, a developed high-precision detection and recognition algorithm is used for recognizing the enteric glands, and quantitative determination is carried out on the enteric glands in the gastric mucosa of the stomach body, the antrum and the corner of the stomach. And the identified target glands are automatically counted, and intelligent OLGIM grading is carried out.

Description

Technical Field

[0001] This invention relates to the field of public health technology, specifically to an intelligent identification and classification method for gastric intestinal metaplasia. Background Technology

[0002] Gastric intestinal metaplasia (GIM) is a pathological change in which normal cells of the gastric mucosa are replaced by cells similar to those of the intestinal mucosa (such as goblet cells, Paneth cells, and absorptive cells).

[0003] Currently, the OLGIM (Operative Link on Gastritis Assessment Based on Intestinal Metaplasia) system is the mainstream method for grading gastritis based on histological morphology.

[0004] OLGIM grading is based on the extent and density of intestinal metaplastic glands in hematoxylin-eosin (HE) stained tissue sections. It relies on pathologists' visual identification and manual counting, which is time-consuming and labor-intensive. Furthermore, it primarily focuses on morphological changes, making it overly dependent on the reader's experience and prone to observational bias, empirical bias, and bias due to reader fatigue. Therefore, the current manual OLGIM grading system has limitations in terms of diagnostic efficiency and accurate quantitative assessment. Summary of the Invention

[0005] To address this, embodiments of the present invention provide an intelligent identification and grading method for gastric intestinal metaplasia, thereby solving the technical problem that existing technologies rely too heavily on the experience of the image reader, are prone to observational and empirical biases, and have limitations in terms of accurate quantitative evaluation.

[0006] To achieve the above objectives, the embodiments of the present invention provide the following technical solutions:

[0007] According to a first aspect of the present invention, a method for intelligent identification and grading of gastric intestinal metaplasia is provided, the method comprising:

[0008] Obtain HE-stained digital pathological images of gastric mucosa and preprocess the HE-stained digital pathological images of gastric mucosa;

[0009] The gastric mucosa HE-stained digital pathological images are classified into target regions such as the gastric body, gastric angle, or gastric antrum using a site recognition model; the site recognition model is obtained by constructing and training a Transformer.

[0010] The target region image is segmented using a segmentation model to obtain a gastric mucosa image; the segmentation model is constructed and trained using a variant of the convolutional neural network UNet, UTTransNet.

[0011] The intestinal metaplastic glands in the gastric mucosa image are identified using an intestinal metaplastic gland recognition model, and the proportion of intestinal metaplastic glands to the total glands is calculated. The intestinal metaplastic gland recognition model is obtained by constructing and training a convolutional neural network YOLO.

[0012] The current pathological image is graded based on the proportion of intestinal metaplasia glands and the definition of intestinal metaplasia, generating grading results.

[0013] Further, the gastric mucosa HE-stained digital pathological images are preprocessed, including:

[0014] The input gastric mucosa HE-stained digital pathological images were subjected to uniform size adjustment, normalization, and data enhancement processing.

[0015] Furthermore, the part recognition model is obtained by constructing and training using Transformer, and includes:

[0016] The training set is divided into three categories according to the stomach body, stomach antrum and stomach angle to train the site recognition model. The learning rate, optimizer and regularization are adjusted to optimize the performance of the site recognition model.

[0017] The hyperparameters are validated using a validation set and adjusted based on the validation results. This process is repeated multiple times until the optimal combination of hyperparameters is obtained.

[0018] The optimal hyperparameter combination is evaluated using an independent test set. The iteratively completed part recognition model is then fine-tuned and optimized using model pruning to generate a trained part recognition model.

[0019] Furthermore, labeled gland images are acquired before training the segmentation model and the intestinal gland recognition model;

[0020] The labeled gland images are downsampled, and the downsampled gland images are used for training.

[0021] Furthermore, early stopping points are set during the training of the segmentation model and the intestinal gland recognition model, and the learning rate, batch size, and number of iterations are dynamically adjusted to optimize the training process.

[0022] Furthermore, the current pathological image is graded based on the proportion of intestinal metaplasia glands and the definition of intestinal metaplasia, generating grading results, including:

[0023] The expression for the proportion of intestinal glands is:

[0024]

[0025] Based on the definition and standards of gastric intestinal metaplasia, the proportion of the metaplastic glands is used to determine the mild, moderate, or severe level of intestinal metaplasia.

[0026] Furthermore, the method also includes setting up log recording to track metrics such as loss, accuracy, and mean precision during the training process.

[0027] According to a second aspect of the present invention, an intelligent identification and grading system for gastric intestinal metaplasia is provided, the system comprising:

[0028] The pathological image preprocessing module is used to acquire and preprocess the gastric mucosa HE-stained digital pathological images.

[0029] The site recognition module is used to build and train a site recognition model using Transformer, and to classify the HE-stained digital pathological images of the gastric mucosa into target area images of the gastric body, gastric angle, or gastric antrum using the trained site recognition model.

[0030] The segmentation module is used to establish a segmentation model and an intestinal gland recognition model using UTTransNet, a variant of the convolutional neural network UNet, and train them using a training set. The trained segmentation model is used to segment the target region image to obtain a gastric mucosa image.

[0031] The intestinal metaplastic gland recognition module is used to identify intestinal metaplastic glands in the gastric mucosa image using a trained intestinal metaplastic gland recognition model, and to calculate the proportion of intestinal metaplastic glands to the total glands.

[0032] The grading module is used to grade the current pathological image based on the proportion of intestinal metaplasia glands and the definition of intestinal metaplasia, and generate grading results.

[0033] According to a third aspect of the present invention, an intelligent identification and grading device for gastric intestinal metaplasia is provided, the device comprising: a processor and a memory;

[0034] The memory is used to store one or more program instructions;

[0035] The processor is configured to run one or more program instructions to perform the steps of a method for intelligent identification and grading of gastric intestinal metaplasia as described in any of the preceding claims.

[0036] According to a fourth aspect of the present invention, a computer program product is provided, wherein a computer program is stored on the computer program, and when the computer program is executed by a processor, the steps of an intelligent identification and grading method for gastric intestinal metaplasia as described in any of the preceding claims are implemented.

[0037] The embodiments of the present invention have the following advantages:

[0038] This invention acquires HE-stained digital pathological images of gastric mucosa and uses a convolutional neural network to build a general site recognition model based on pathological classification standards for different parts of the stomach, including the gastric body, antrum, and angular region. To reduce judgment errors caused by other glands, a semantic segmentation model is trained to accurately locate the gastric mucosa. Based on this, a developed high-precision detection and recognition algorithm is used to identify intestinal metaplastic glands, and quantitative measurements are performed on the intestinal metaplastic glands in the gastric mucosa of the gastric body, antrum, and angular region. The identified target glands are counted and then subjected to automated intelligent OLGIM grading. Attached Figure Description

[0039] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0040] The structures, proportions, sizes, etc. illustrated in this specification are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed herein, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0041] Figure 1 A schematic diagram of the logical structure of an intelligent identification and grading system for gastric intestinal metaplasia provided in an embodiment of the present invention;

[0042] Figure 2 A flowchart illustrating an intelligent identification and grading method for gastric intestinal metaplasia provided in an embodiment of the present invention;

[0043] Figure 3 A schematic diagram of gastric slice location identification for an intelligent identification and grading method for gastric intestinal metaplasia provided in an embodiment of the present invention;

[0044] Figure 4 This is a schematic diagram illustrating the detection and identification of target glands in an intelligent identification and grading method for gastric intestinal metaplasia provided in an embodiment of the present invention.

[0045] Figure 5 This is a schematic diagram of loss training in an intelligent identification and grading method for gastric intestinal metaplasia provided in an embodiment of the present invention.

[0046] Figure 6 This is a first schematic diagram illustrating the identification of digital pathological images in an intelligent identification and grading method for gastric intestinal metaplasia according to an embodiment of the present invention.

[0047] Figure 7 This is a second schematic diagram illustrating the identification of digital pathological images in an intelligent identification and grading method for gastric intestinal metaplasia according to an embodiment of the present invention.

[0048] Figure 8 This is a third schematic diagram illustrating the identification of digital pathological images in an intelligent identification and grading method for gastric intestinal metaplasia according to an embodiment of the present invention.

[0049] Figure 9 This is a fourth schematic diagram illustrating the identification of digital pathological images in an intelligent identification and grading method for gastric intestinal metaplasia provided in an embodiment of the present invention. Detailed Implementation

[0050] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. 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.

[0051] To address the aforementioned technical issues that rely too heavily on the viewer's experience, which can easily lead to observational and empirical biases and limit the accuracy of quantitative evaluation.

[0052] refer to Figure 1 This invention discloses an intelligent identification and grading system for gastric intestinal metaplasia, which includes: a pathological image preprocessing module 1; a site identification module 2; a segmentation module 3; an intestinal metaplastic gland identification module 4; and a grading module 5.

[0053] Corresponding to the aforementioned intelligent identification and grading system for gastric intestinal metaplasia, this invention also discloses an intelligent identification and grading method for gastric intestinal metaplasia. The following details the intelligent identification and grading method for gastric intestinal metaplasia disclosed in this invention, in conjunction with the aforementioned intelligent identification and grading system for gastric intestinal metaplasia.

[0054] refer to Figure 2This invention discloses an intelligent identification and grading method for gastric intestinal metaplasia. The method includes: acquiring a hematoxylin and eosin (HE) stained digital pathological image of the gastric mucosa and preprocessing the image; classifying the image into target regions such as the gastric body, gastric angle, or gastric antrum using a site recognition model; the site recognition model is constructed and trained using Transformer; segmenting the target region image using a segmentation model to obtain a gastric mucosa image; the segmentation model is constructed and trained using a convolutional neural network; identifying intestinal metaplasia glands in the gastric mucosa image using an intestinal metaplasia gland recognition model and calculating the proportion of intestinal metaplasia glands to the total glands; the intestinal metaplasia gland recognition model is constructed and trained using convolutional neural networks UNet and YOLO; and grading the current pathological image according to the proportion of intestinal metaplasia glands and the definition of intestinal metaplasia to generate a grading result.

[0055] Furthermore, the gastric mucosa HE-stained digital pathological image is preprocessed, including: uniform size adjustment, normalization, and data enhancement processing of the input gastric mucosa HE-stained digital pathological image.

[0056] Furthermore, the site recognition model is constructed and trained using Transformer, including: training the site recognition model by dividing the training set into three categories according to the stomach body, antrum, and angle; adjusting the learning rate, optimizer, and regularization to optimize the performance of the site recognition model; validating the model using a validation set and adjusting the hyperparameters based on the validation results, iterating multiple times until the optimal hyperparameter combination is obtained; evaluating the optimal hyperparameter combination using an independent test set; fine-tuning the iterated site recognition model and optimizing it using model pruning to generate the trained site recognition model.

[0057] 1) Construction of a deep learning-based model for identifying the gastric body / angle / antrum:

[0058] a. Image feature processing: Adjust the size of the image after horizontal and vertical segmentation, normalize it, and enhance it.

[0059] b. Deep Learning Model Construction: The Transformer was used as the primary deep learning model to adapt to input pathological images of different sizes from various medical institutions. The training set was divided into three categories based on different locations of the stomach: body, antrum, and angle. The model was then trained to efficiently process large-scale, high-resolution images.

[0060] c. Model performance evaluation and parameter optimization:

[0061] During training, model performance is optimized by adjusting hyperparameters. A validation set is used to validate and evaluate model performance. Based on the validation results, hyperparameters such as learning rate, optimizer, and regularization are adjusted iteratively until the optimal hyperparameter combination is obtained. An independent test set is used to evaluate the optimal hyperparameter combination, allowing for further fine-tuning of the model. Simultaneously, training techniques such as model pruning are employed to further improve model performance.

[0062] Furthermore, labeled gland images are acquired before training the segmentation model and the intestinal gland recognition model; the labeled gland images are downsampled, and the downsampled gland images are used for training.

[0063] Furthermore, early stopping points are set during the training of the segmentation model and the intestinal gland recognition model to optimize the training process.

[0064] 2) Construction of deep learning-based segmentation and intestinal gland recognition models:

[0065] a. Image feature processing: The labeled glands are used for model training. Downsampling is employed to significantly compress the labeled images and then normalize them.

[0066] b. Deep learning model construction: using convolutional neural network models:

[0067] Prepare the training script: This includes preliminary data preparation, data loading, model definition, loss function, optimizer, etc. Create or modify the configuration file, set parameters such as input size, and ensure that the script can correctly read the configuration file and hyperparameters.

[0068] Prepare logs and visualizations: Set up logs to track metrics such as loss and accuracy during training. Use visualization tools to monitor the training process in real time.

[0069] Model training: The UNet convolutional neural network algorithm was used to segment the gastric mucosa layer; the YOLO convolutional neural network algorithm was used to identify and analyze the glands in the gastric mucosa.

[0070] c. Model performance evaluation and parameter optimization:

[0071] Setting early stopping points and dynamically adjusting hyperparameters such as learning rate, batch size, and number of iterations can optimize the training process.

[0072] Based on the model's gland identification results, the total number of glands and intestinal metaplasia glands in the gastric mucosa were counted comprehensively. The effectiveness was evaluated and corrected using the OLGIM rating of pathologists as the gold standard to improve the accuracy of the model.

[0073] Combining optimal gastric body / angle / antrum region recognition models, gastric segmentation models, and intestinal metaplasia gland recognition models, the pathological images to be tested are detected to accurately locate the gastric mucosa. The next step involves automated extraction of intestinal metaplasia glands from the gastric mucosa and gland counting. Furthermore, the current pathological image is graded based on the proportion of intestinal metaplasia glands and the definition of intestinal metaplasia, generating grading results, including:

[0074] The expression for the proportion of intestinal glands is:

[0075]

[0076] Based on the Sydney standard for gastric intestinal metaplasia, the proportion of the metaplastic glands is used to determine the mild, moderate, or severe level of intestinal metaplasia.

[0077] Furthermore, the method also includes setting up log recording to track metrics such as loss, accuracy, and mean precision during the training process.

[0078] In addition, embodiments of the present invention also provide an intelligent identification and grading device for gastric intestinal metaplasia, the device comprising: a processor and a memory; the memory for storing one or more program instructions; the processor for running one or more program instructions to perform the steps of an intelligent identification and grading method for gastric intestinal metaplasia as described in any of the preceding embodiments.

[0079] In addition, embodiments of the present invention also provide a computer program product, wherein the program product stores a computer program, and when the computer program is executed by a processor, it implements the steps of the intelligent identification and grading method for gastric intestinal metaplasia as described in any of the preceding claims.

[0080] In this embodiment of the invention, the processor can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0081] The various methods, steps, and logic diagrams disclosed in the embodiments of this invention can be implemented or executed. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The processor reads information from the storage medium and, in conjunction with its hardware, completes the steps of the above methods.

[0082] The storage medium can be memory, such as volatile memory or non-volatile memory, or may include both volatile and non-volatile memory.

[0083] Among them, non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory.

[0084] Volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (Synchlink DRAM, SLDRAM), and direct memory bus RAM (DRRAM).

[0085] The storage media described in the embodiments of the present invention are intended to include, but are not limited to, these and any other suitable types of memory.

[0086] Those skilled in the art will recognize that, in one or more of the examples above, the functions described in this invention can be implemented using a combination of hardware and software. When applied as software, the corresponding functions can be stored in a computer-readable medium or transmitted as one or more instructions or code on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any medium that facilitates the transmission of computer programs from one place to another. Storage media can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0087] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A method for intelligent identification and grading of gastric intestinal metaplasia, characterized in that, The method includes: Obtain digital pathological images of gastric mucosa stained with hematoxylin and eosin (HE) and preprocess the images; The gastric mucosa HE-stained digital pathological images are classified into target regions such as the gastric body, gastric angle, or gastric antrum using a site recognition model; the site recognition model is obtained by constructing and training a Transformer. The target region image is segmented using a segmentation model to obtain a gastric mucosa image; the segmentation model is constructed and trained using the mosaic Transformer model in UNet. Intestinal metaplastic glands in the gastric mucosa image were identified using an intestinal metaplastic gland recognition model, which was constructed and trained using a convolutional neural network, YOLO. The current pathological image is graded based on the proportion of intestinal metaplasia glands and the definition of intestinal metaplasia, generating grading results.

2. The intelligent identification and grading method for gastric intestinal metaplasia as described in claim 1, characterized in that, Preprocessing of the HE-stained digital pathological images of the gastric mucosa includes: The input gastric mucosa HE-stained digital pathological images were subjected to uniform size adjustment, normalization, and data enhancement processing.

3. The intelligent identification and grading method for gastric intestinal metaplasia as described in claim 2, characterized in that, The part recognition model is obtained by building and training using Transformer, and includes: The training set is divided into three categories according to the stomach body, stomach antrum and stomach angle to train the site recognition model. The learning rate, optimizer and regularization are adjusted to optimize the performance of the site recognition model. The hyperparameters are validated using a validation set and adjusted based on the validation results. This process is repeated multiple times until the optimal combination of hyperparameters is obtained. The optimal hyperparameter combination is evaluated using an independent test set. The iteratively completed part recognition model is then fine-tuned and optimized using model pruning to generate a trained part recognition model.

4. The intelligent identification and grading method for gastric intestinal metaplasia as described in claim 3, characterized in that, Before training the segmentation model and the intestinal gland recognition model, the labeled gland images are obtained; The labeled gland images are downsampled, and the downsampled gland images are used for training.

5. The intelligent identification and grading method for gastric intestinal metaplasia as described in claim 4, characterized in that, Early stopping points were set during the training of the segmentation model and the intestinal gland recognition model, and the learning rate, batch size and number of iterations were dynamically adjusted to optimize the training process.

6. The intelligent identification and grading method for gastric intestinal metaplasia as described in claim 5, characterized in that, The current pathological image is graded based on the proportion of intestinal metaplasia glands and the definition of intestinal metaplasia, generating grading results, including: The expression for the proportion of intestinal glands is: Based on the Sydney standard for gastric intestinal metaplasia, the proportion of the metaplastic glands is used to determine the mild, moderate, or severe level of intestinal metaplasia.

7. The intelligent identification and grading method for gastric intestinal metaplasia as described in claim 6, characterized in that, The method also includes setting up log records to track metrics such as loss, accuracy, and mean precision during the training process.

8. A smart identification and grading system for gastric intestinal metaplasia, characterized in that, The system includes: The pathological image preprocessing module is used to acquire and preprocess the gastric mucosa HE-stained digital pathological images. The site recognition module is used to build and train a site recognition model using Transformer, and to classify the HE-stained digital pathological images of the gastric mucosa into target area images of the gastric body, gastric angle, or gastric antrum using the trained site recognition model. The segmentation module is used to establish a segmentation model and an intestinal gland recognition model using a convolutional neural network and train them using a training set. The trained segmentation model is used to segment the target region image to obtain a gastric mucosa image. The intestinal metaplastic gland recognition module is used to identify intestinal metaplastic glands in the gastric mucosa image using a trained intestinal metaplastic gland recognition model, and to calculate the proportion of intestinal metaplastic glands to the total glands. The grading module is used to grade the current pathological image based on the proportion of intestinal metaplasia glands and the definition of intestinal metaplasia, and generate grading results.

9. A smart identification and grading device for gastric intestinal metaplasia, characterized in that, The device includes: a processor and a memory; The memory is used to store one or more program instructions; The processor is configured to run one or more program instructions to perform the steps of a method for intelligent identification and grading of gastric intestinal metaplasia as described in any one of claims 1 to 7.

10. A computer program product, characterized in that, The computer program product stores a computer program, which, when executed by a processor, implements the steps of the intelligent identification and grading method for gastric intestinal metaplasia as described in any one of claims 1 to 7.

Images