WO2023221513A1

WO2023221513A1 - Image partitioning method, apparatus and device, and readable storage medium

Info

Publication number: WO2023221513A1
Application number: PCT/CN2022/142223
Authority: WO
Inventors: 罗茜; 郭昂; 陈志宇; 李芳�; 李文波
Original assignee: 中国科学院深圳先进技术研究院
Priority date: 2022-05-18
Filing date: 2022-12-27
Publication date: 2023-11-23
Also published as: CN114841978A

Abstract

Disclosed in the present application are an image partitioning method, apparatus and device, and a readable storage medium. The method comprises: acquiring a biological tissue slice dyed image to be partitioned; determining image blocks to be processed in the biological tissue slice dyed image; performing feature extraction on each said image block, so as to obtain a tissue morphological feature vector of each said image block, wherein each element in the tissue morphological feature vector corresponds to the response intensity of a tissue morphological feature in said image block; on the basis of the tissue morphological feature vectors of said image blocks, clustering said image blocks, so as to obtain a clustering result of said image blocks; and on the basis of the clustering result of said image blocks, partitioning the biological tissue slice dyed image. By using the scheme, a biological tissue slice dyed image can be partitioned on the basis of clustering, and there is no need to perform supervised training to obtain a biological tissue slice dyed image partition model.

Description

Image partitioning method, device, equipment and readable storage medium

This application claims the priority of the Chinese patent application submitted to the China Patent Office on May 18, 2022, with the application number CN202210539255. incorporated herein by reference.

Technical field

The present application relates to the technical field of computer-assisted biological tissue section staining image analysis, and more specifically, to an image partitioning method, device, equipment and readable storage medium.

Background technique

With the digitization of biological tissue section staining images and the rapid growth of computer computing power, computer-assisted biological tissue section staining image analysis technology has gradually attracted the attention of researchers and clinical pathologists. Among them, the partitioning of biological tissue section staining images is an important branch of computer-aided biological tissue section staining image analysis. The partitioning of biological tissue section staining images can divide the biological tissue section staining images into regions corresponding to different tissue types, which is helpful to assist The pathologist reads the film and assists in computer diagnosis.

The current partitioning scheme for biological tissue section stained images generally obtains a biological tissue section stained image partitioning model through supervised training, and uses the biological tissue section stained image partitioning model to partition the biological tissue section stained image. However, the biological tissue section stained image partitioning model obtained through supervised training has poor versatility. It can only partition the biological tissue section stained images covered by the training data set. It can only partition the biological tissue section stained images not covered by the training data set. , cannot be processed, and the supervised training of the biological tissue section stained image partitioning model needs to rely on a large number of accurately annotated training data sets. However, since the annotation of biological tissue section stained images requires a deep histopathological background, there is no accurate annotation The data set is very small.

Therefore, how to provide a partitioning scheme for biological tissue section stained images that does not require supervised training of a biological tissue section stained image partitioning model has become an urgent technical problem to be solved by those skilled in the art.

technical problem

In view of the above problems, this application proposes an image partitioning method, device, equipment and readable storage medium to provide a partitioning scheme for biological tissue section staining images that does not require supervised training of a biological tissue section staining image partitioning model.

Technical solutions

An image partitioning method, the method includes:

Obtain stained images of biological tissue sections to be partitioned;

Determine the image block to be processed in the stained image of the biological tissue section;

Perform feature extraction on each image block to be processed to obtain a tissue morphology feature vector of each image block to be processed. Each element in the tissue morphology feature vector corresponds to a tissue morphology feature in the image block to be processed. response intensity;

Based on the histomorphological feature vector of each image block to be processed, cluster each image block to be processed to obtain a clustering result of each image block to be processed; and

Based on the clustering results of each image block to be processed, the biological tissue section stained image is partitioned.

Optionally, the determining the image block to be processed in the stained image of the biological tissue section includes:

Using a tissue detection algorithm to determine the tissue area and background area of the stained image of the biological tissue section;

Divide the stained image of the biological tissue section into image blocks of equal size;

For each image block, calculate the proportion of the tissue area in the image block based on the tissue area and background area of the image block; and

The image block whose tissue area proportion meets the preset conditions is determined as the image block to be processed.

Optionally, the feature extraction is performed on each image block to be processed to obtain the tissue morphology feature vector of each image block to be processed, including:

Use a feature extraction network to extract features from each image block to be processed to obtain the histomorphological feature vector of each image block to be processed. The feature extraction network is obtained by pre-training the convolutional neural network using natural image data sets. .

Optionally, clustering each image block to be processed based on the histomorphological feature vector of each image block to be processed, and obtaining a clustering result of each image block to be processed, including:

Obtain the position information of each image block to be processed in the stained image of the biological tissue section;

Generate a morphological feature data set based on the position information of each image block to be processed in the biological tissue section stained image and the tissue morphology feature vector of each image block to be processed;

Based on the morphological feature data set, generate a tissue morphology feature spatial distribution map, where each spatial location point in the tissue morphology feature spatial distribution map corresponds to a tissue morphology feature spectrum;

Perform dimensionality reduction on each tissue morphology feature spectrum to obtain the dimensionally reduced tissue morphology feature spectrum; and

A clustering algorithm is used to cluster each dimensionally reduced tissue morphology feature spectrum, and multiple clusters are obtained. Each cluster corresponds to image blocks to be processed with similar tissue morphology feature spectrum, and different clusters correspond to image blocks with different tissue morphology feature spectrum. Process image blocks.

Optionally, partitioning the biological tissue section stained image based on the clustering results of each image block to be processed includes:

The area corresponding to each image block to be processed in the stained image of the biological tissue section is colored according to the color information corresponding to the cluster where the image block to be processed is located, so as to achieve staining of the biological tissue section. Partitions of images.

Optionally, the dimensionality reduction of each tissue morphology characteristic spectrum is performed to obtain the dimensionally reduced tissue morphology characteristic spectrum, including:

Use a nonlinear dimensionality reduction algorithm or a linear dimensionality reduction algorithm to reduce the dimensionality of each tissue morphology feature spectrum to obtain the dimensionally reduced tissue morphology feature spectrum.

Optionally, the clustering algorithm includes:

Any of Kmeans, hierarchical clustering, density-based clustering method, maximum expectation clustering using Gaussian mixture model, graph group detection, mean shift clustering.

An image partitioning device, the device includes:

An acquisition unit is used to acquire stained images of biological tissue sections to be partitioned;

A determination unit configured to determine the image block to be processed in the stained image of the biological tissue section;

A feature extraction unit is used to extract features of each image block to be processed and obtain a tissue morphology feature vector of each image block to be processed. Each element in the tissue morphology feature vector corresponds to a tissue morphology feature at that location. Describe the response intensity of the image block to be processed;

A clustering unit is used to cluster each image block to be processed based on the histomorphological feature vector of each image block to be processed, and obtain a clustering result of each image block to be processed; and

A partitioning unit is used to partition the biological tissue section stained image based on the clustering results of each image block to be processed.

An image partitioning device including a memory and a processor;

The memory is used to store programs;

The processor is used to execute the program and implement the following steps:

Obtain stained images of biological tissue sections to be partitioned;

Perform feature extraction on each image block to be processed to obtain a tissue morphology feature vector of each image block to be processed. Each element in the tissue morphology feature vector corresponds to a tissue morphology feature in the image block to be processed. response strength;

A non-volatile computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

Obtain stained images of biological tissue sections to be partitioned;

beneficial effects

Through the above technical solutions, this application discloses an image partitioning method, device, equipment and non-volatile computer-readable storage medium. Obtain the stained image of the biological tissue section to be partitioned; determine the image blocks to be processed in the stained image of the biological tissue section; perform feature extraction on each image block to be processed, and obtain the tissue morphology feature vector of each image block to be processed, Each element in the tissue morphology feature vector corresponds to the response intensity of a tissue morphology feature in the image block to be processed; based on the tissue morphology feature vector of each image block to be processed, cluster each image block to be processed, and obtain Clustering results of each image block to be processed; based on the clustering results of each image block to be processed, the biological tissue section stained image is partitioned. Using the above scheme, biological tissue section stained images can be partitioned based on clustering, without the need for supervised training of the biological tissue section stained image partitioning model.

Description of the drawings

Figure 1 is a schematic flow chart of an image partitioning method disclosed in an embodiment of the present application;

Figure 2 is a schematic flow chart of a method disclosed in an embodiment of the present application for clustering each image block to be processed based on the histomorphological feature vector of each image block to be processed, and obtaining a clustering result of each image block to be processed;

Figure 3 is an example diagram of an image partitioning method disclosed in the embodiment of the present application;

Figure 4 is a schematic structural diagram of an image partitioning device disclosed in an embodiment of the present application;

Figure 5 is a hardware structure block diagram of an image partitioning device disclosed in the embodiment of the present application.

Detailed ways

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of this application.

Next, the image partitioning method provided by this application is introduced through the following embodiments.

Referring to Figure 1, Figure 1 is a schematic flowchart of an image partitioning method disclosed in an embodiment of the present application. The method may include:

Step S101: Obtain stained images of biological tissue sections to be partitioned.

The biological tissue section staining image to be partitioned can be a biological tissue section staining image with high spatial resolution (pixel resolution less than 1 micron), for example, it can be a H&E stained tissue microscopic image.

Step S102: Determine the image block to be processed in the stained image of the biological tissue section.

In this application, the stained image of the biological tissue section can be cut into grid-shaped image blocks corresponding to a smaller tissue area as image blocks to be processed. The specific implementation will be described in detail through the following embodiments.

As an implementation manner, a tissue detection algorithm can be used to determine the tissue area and background area of the biological tissue section stained image; the biological tissue section stained image is divided into equal-sized image blocks; for each image block, according to The tissue area and background area of the image block are calculated, and the tissue area ratio in the image block is calculated; the image block whose tissue area ratio meets the preset conditions is determined as the image block to be processed.

The method of dividing the biological tissue section stained image into equal-sized image blocks can be determined according to scene requirements. For example, the biological tissue section stained image to be partitioned contains 10 ⁴ × 10 ⁴ pixels. Divide it into grid-like square image blocks of equal size. Each image block is 10 ² × 10 ² pixels in size and covers an area of approximately 10 ² × 10 ² square microns.

The preset conditions for the tissue area ratio to be satisfied can be determined based on the scene requirements. For example, if the tissue area ratio exceeds the preset percentage (such as 90%), it can be determined that the tissue area ratio meets the preset conditions.

Step S103: Perform feature extraction on each image block to be processed to obtain a histomorphological feature vector of each image block to be processed. Each element in the histomorphological feature vector corresponds to a histomorphological feature in the to-be-processed image block. The response strength of the image patch.

In this application, the histomorphological feature vector of each image block to be processed can be various computer vision features, such as gray level co-occurrence matrix, threshold adjacency statistics, intensity statistical features, features output by the middle layer of the neural network model, etc. , this application does not make any limitations.

As an implementation method, a feature extraction network can be used to extract features of each image block to be processed to obtain the histomorphological feature vector of each image block to be processed. The feature extraction network uses natural image data sets to extract features. Pre-trained by convolutional neural network. For example, use convolutional neural networks (such as DenseNet, ResNet, VGG, etc.) pre-trained on natural image data sets (such as ImageNet) as feature extraction networks. The image block to be processed is used as input after resizing and normalization preprocessing, and is forward propagated in the convolutional neural network. In the pre-selected intermediate layer (such as the conv5_block32 layer of DenseNet201), the 3-channel image block is converted into a three-dimensional array with a large number of channels and a small number of pixels. Each channel corresponds to a visual feature, that is, a tissue morphological feature. Through a pooling operation (such as average global pooling), the three-dimensional array is reshaped into a tissue morphology feature vector. Each element in the tissue morphology feature vector corresponds to the response intensity of a morphological feature in the image block to be processed.

Step S104: Cluster each image block to be processed based on the histomorphological feature vector of each image block to be processed, and obtain a clustering result of each image block to be processed.

In this application, the partitioning problem of stained images of biological tissue sections is transformed into an unsupervised clustering problem. Since unsupervised clustering does not require the construction of annotation data sets for specific tissue sample types, it not only reduces the huge cost of annotating a large number of stained images of biological tissue sections, but also greatly expands the scope of application of the method, which can solve the problem of obvious morphology of all different tissue types. Partitioning task of differentially stained images of biological tissue sections.

Step S105: Based on the clustering results of each image block to be processed, partition the stained image of the biological tissue section.

This embodiment discloses an image partitioning method. Obtain the stained image of the biological tissue section to be partitioned; determine the image blocks to be processed in the stained image of the biological tissue section; perform feature extraction on each image block to be processed, and obtain the tissue morphology feature vector of each image block to be processed, Each element in the tissue morphology feature vector corresponds to the response intensity of a tissue morphology feature in the image block to be processed; based on the tissue morphology feature vector of each image block to be processed, cluster each image block to be processed, and obtain Clustering results of each image block to be processed; based on the clustering results of each image block to be processed, the biological tissue section stained image is partitioned. Using the above scheme, biological tissue section stained images can be partitioned based on clustering, without the need for supervised training of the biological tissue section stained image partitioning model.

In another embodiment of the present application, step S104 is implemented by clustering each image block to be processed based on the histomorphological feature vector of each image block to be processed, and obtaining the clustering results of each image block to be processed. Explained.

Referring to Figure 2, Figure 2 is a method disclosed in an embodiment of the present application for clustering each image block to be processed based on the histomorphological feature vector of each image block to be processed, and obtaining a clustering result of each image block to be processed. The flow diagram of this method may include:

Step S201: Obtain the position information of each image block to be processed in the stained image of the biological tissue section.

Step S202: Generate a morphological feature data set based on the position information of each image block to be processed in the biological tissue section stained image and the tissue morphology feature vector of each image block to be processed.

Step S203: Generate a tissue morphology feature spatial distribution map based on the morphological feature data set. Each spatial position point in the tissue morphology feature spatial distribution map corresponds to a tissue morphology feature spectrum.

Step S204: Dimensionally reduce each tissue morphology feature spectrum to obtain a dimensionally reduced tissue morphology feature spectrum.

In this application, the number of morphological features is large, so the sample space is relatively sparse, and clustering will affect the effect. Therefore, based on the similarity between tissue morphological feature spectra (for example, measured by trigonometric distance), the feature dimension can be greatly reduced to three dimensions while retaining the structural relationship of the original data to the greatest extent.

In this application, a nonlinear dimensionality reduction algorithm or a linear dimensionality reduction algorithm can be used to reduce the dimensionality of each tissue morphology feature spectrum to obtain a dimensionally reduced tissue morphology feature spectrum.

Nonlinear dimensionality reduction algorithms include t-SNE (t-distributed stochastic neighbor embedding) algorithm, UMAP (Uniform Manifold Approximation and Projection, unified manifold approximation and projection) algorithm, etc.;

Linear dimensionality reduction algorithms include PCA (Principal Component Analysis, principal component analysis) algorithm, NMF (Non-negative matrix factorization, non-negative matrix decomposition), ICA (independent component analysis, independent component analysis), etc.

Step S205: Use a clustering algorithm to cluster each dimensionally reduced tissue morphology feature spectrum to obtain multiple clusters. Each cluster corresponds to an image block to be processed with similar tissue morphology feature spectrum, and different clusters correspond to different tissue morphology feature spectra. Different image blocks to be processed.

In this application, the clustering algorithm may include: any one of Kmeans, hierarchical clustering, density-based clustering method, maximum expectation clustering using Gaussian mixture model, graph group detection, and mean shift clustering. .

In this application, the tissue morphological characteristic spectra in each cluster are similar to each other, while the clusters are quite different from each other. Based on the assumption that image blocks with similar tissue morphological characteristics belong to the same tissue type and images with different tissue morphology characteristics belong to different tissue types, the tissue type classification at the image block level is realized.

It should be noted that, after using a clustering algorithm to cluster each dimensionally reduced tissue morphological feature spectrum to obtain multiple clusters, based on the clustering results of each image block to be processed, the stained image of the biological tissue section is Partitioning may include: coloring the area corresponding to each image block to be processed in the stained image of the biological tissue section, and coloring the image block to be processed according to the color information corresponding to the cluster where the image block to be processed is located, so as to achieve Partitions of stained images of biological tissue sections.

It should be understood that although the steps in the flowcharts of FIG. 1 and FIG. 2 are shown in sequence as indicated by arrows, these steps are not necessarily executed in the order indicated by arrows. Unless explicitly stated in this article, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least some of the steps in Figure 1 and Figure 2 may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but may be executed at different times. These sub-steps or The execution order of the stages is not necessarily sequential, but may be performed in turn or alternately with other steps or sub-steps of other steps or at least part of the stages.

For ease of understanding, please refer to Figure 3 , which is an example diagram of an image partitioning method disclosed in an embodiment of the present application. As shown in Figure 3, after the stained tissue microscopic image is divided into image blocks, a deep convolutional neural network is used to extract the tissue morphological features of the image blocks, and then a morphological feature data set is generated based on the position information of each image block. Based on the morphology The tissue morphology feature spatial distribution map is generated from the tissue morphology feature data set. Each spatial point in the tissue morphology feature spatial distribution map corresponds to a tissue morphology feature spectrum. Dimension reduction and clustering are then performed based on each tissue morphology feature spectrum, and finally the stained tissue microscopic image is obtained. partition.

The image partitioning device disclosed in the embodiment of the present application will be described below. The image partitioning device described below and the image partitioning method described above can be referred to each other correspondingly.

Referring to Figure 4, Figure 4 is a schematic structural diagram of an image partitioning device disclosed in an embodiment of the present application. As shown in Figure 4, the image partitioning device may include:

The acquisition unit 11 is used to acquire stained images of biological tissue sections to be partitioned;

Determining unit 12, used to determine the image block to be processed in the stained image of the biological tissue section;

The feature extraction unit 13 is used to extract features of each image block to be processed and obtain a tissue morphology feature vector of each image block to be processed. Each element in the tissue morphology feature vector corresponds to a tissue morphology feature. The response intensity of the image block to be processed;

The clustering unit 14 is used to cluster each image block to be processed based on the histomorphological feature vector of each image block to be processed, and obtain a clustering result of each image block to be processed;

The partitioning unit 15 is used to partition the biological tissue section stained image based on the clustering results of each image block to be processed.

Referring to Figure 5, Figure 5 is a hardware structure block diagram of an image partitioning device provided by an embodiment of the present application. Referring to Figure 5, the hardware structure of the image partitioning device may include: at least one processor 1, at least one communication interface 2, and at least one memory 3 and at least one communication bus 4;

In the embodiment of the present application, the number of processor 1, communication interface 2, memory 3, and communication bus 4 is at least one, and processor 1, communication interface 2, and memory 3 complete communication with each other through communication bus 4;

Processor 1 may be a central processing unit (CPU) or a specific integrated circuit (ASIC). Specific Integrated Circuit), or one or more integrated circuits, etc. configured to implement embodiments of the present invention;

Memory 3 may include high-speed RAM memory, and may also include non-volatile memory (non-volatile memory), etc., such as at least one disk memory;

The memory stores a program, and the processor can call the program stored in the memory. The program is used for:

Obtain stained images of biological tissue sections to be partitioned;

Based on the histomorphological feature vector of each image block to be processed, cluster each image block to be processed to obtain the clustering result of each image block to be processed;

Optionally, the detailed functions and extended functions of the program may refer to the above description.

Embodiments of the present application also provide a non-volatile computer-readable storage medium. The non-volatile computer-readable storage medium can store a program suitable for execution by a processor. The program is used for:

Obtain stained images of biological tissue sections to be partitioned;

Finally, it should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that these entities or any such actual relationship or sequence between operations. Furthermore, the terms "comprises," "comprises," or any other variations thereof are intended to cover a non-exclusive inclusion such that a process, method, article, or apparatus that includes a list of elements includes not only those elements, but also those not expressly listed other elements, or elements inherent to the process, method, article or equipment. Without further limitation, an element defined by the statement "comprises a..." does not exclude the presence of additional identical elements in a process, method, article, or apparatus that includes the stated element.

Each embodiment in this specification is described in a progressive manner. Each embodiment focuses on its differences from other embodiments. The same and similar parts between the various embodiments can be referred to each other.

The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be practiced in other embodiments without departing from the spirit or scope of the application. Therefore, the present application is not to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

An image partitioning method, characterized in that the method includes:

Obtain stained images of biological tissue sections to be partitioned;

Determine the image block to be processed in the stained image of the biological tissue section;

Perform feature extraction on each image block to be processed to obtain a tissue morphology feature vector of each image block to be processed. Each element in the tissue morphology feature vector corresponds to a tissue morphology feature in the image block to be processed. response strength; and

Based on the histomorphological feature vector of each image block to be processed, cluster each image block to be processed to obtain the clustering result of each image block to be processed;

Based on the clustering results of each image block to be processed, the biological tissue section stained image is partitioned.
The method of claim 1, wherein determining the image block to be processed in the stained image of the biological tissue section includes:

Using a tissue detection algorithm to determine the tissue area and background area of the stained image of the biological tissue section;

Divide the stained image of the biological tissue section into image blocks of equal size;

For each image block, calculate the proportion of the tissue area in the image block based on the tissue area and background area of the image block; and

The image block whose tissue area proportion meets the preset conditions is determined as the image block to be processed.
The method according to claim 1, characterized in that, performing feature extraction on each image block to be processed to obtain the histomorphological feature vector of each image block to be processed includes:

Use a feature extraction network to extract features from each image block to be processed to obtain the histomorphological feature vector of each image block to be processed. The feature extraction network is obtained by pre-training the convolutional neural network using natural image data sets. .
The method according to claim 1, characterized in that, based on the histomorphological feature vector of each image block to be processed, each image block to be processed is clustered to obtain a clustering result of each image block to be processed, including :

Obtain the position information of each image block to be processed in the stained image of the biological tissue section;

Generate a morphological feature data set based on the position information of each image block to be processed in the biological tissue section stained image and the tissue morphology feature vector of each image block to be processed;

Based on the morphological feature data set, generate a tissue morphology feature spatial distribution map, where each spatial location point in the tissue morphology feature spatial distribution map corresponds to a tissue morphology feature spectrum;

Perform dimensionality reduction on each tissue morphology feature spectrum to obtain the dimensionally reduced tissue morphology feature spectrum; and

A clustering algorithm is used to cluster each dimensionally reduced tissue morphology feature spectrum, and multiple clusters are obtained. Each cluster corresponds to image blocks to be processed with similar tissue morphology feature spectrum, and different clusters correspond to image blocks with different tissue morphology feature spectrum. Process image blocks.
The method according to claim 4, characterized in that, based on the clustering results of each image block to be processed, partitioning the stained image of the biological tissue section includes:

The area corresponding to each image block to be processed in the stained image of the biological tissue section is colored according to the color information corresponding to the cluster where the image block to be processed is located, so as to achieve staining of the biological tissue section. Partitions of images.
The method according to claim 4, characterized in that: performing dimensionality reduction on each tissue morphology characteristic spectrum to obtain a dimensionally reduced tissue morphology characteristic spectrum, including:

Use a nonlinear dimensionality reduction algorithm or a linear dimensionality reduction algorithm to reduce the dimensionality of each tissue morphology feature spectrum to obtain the dimensionally reduced tissue morphology feature spectrum.
The method according to claim 4, characterized in that the clustering algorithm includes:

Any of Kmeans, hierarchical clustering, density-based clustering method, maximum expectation clustering using Gaussian mixture model, graph group detection, mean shift clustering.
An image partitioning device, characterized in that the device includes:

An acquisition unit is used to acquire stained images of biological tissue sections to be partitioned;

A determination unit configured to determine the image block to be processed in the stained image of the biological tissue section;

A feature extraction unit is used to extract features of each image block to be processed and obtain a tissue morphology feature vector of each image block to be processed. Each element in the tissue morphology feature vector corresponds to a tissue morphology feature at that location. Describe the response intensity of the image block to be processed;

A clustering unit is used to cluster each image block to be processed based on the histomorphological feature vector of each image block to be processed, and obtain a clustering result of each image block to be processed; and

A partitioning unit is used to partition the biological tissue section stained image based on the clustering results of each image block to be processed.
The device according to claim 8, characterized in that the determining unit is specifically used to:

Using a tissue detection algorithm to determine the tissue area and background area of the stained image of the biological tissue section;

Divide the stained image of the biological tissue section into image blocks of equal size;

For each image block, calculate the proportion of the tissue area in the image block based on the tissue area and background area of the image block; and

The image block whose tissue area proportion meets the preset conditions is determined as the image block to be processed.
The device according to claim 8, characterized in that the feature extraction unit is specifically used to:

Use a feature extraction network to extract features from each image block to be processed to obtain the histomorphological feature vector of each image block to be processed. The feature extraction network is obtained by pre-training the convolutional neural network using natural image data sets. .
The device according to claim 8, characterized in that the clustering unit is specifically used for:

Obtain the position information of each image block to be processed in the stained image of the biological tissue section;

Generate a morphological feature data set based on the position information of each image block to be processed in the biological tissue section stained image and the tissue morphology feature vector of each image block to be processed;

Based on the morphological feature data set, generate a tissue morphology feature spatial distribution map, where each spatial location point in the tissue morphology feature spatial distribution map corresponds to a tissue morphology feature spectrum;

Perform dimensionality reduction on each tissue morphology feature spectrum to obtain the dimensionally reduced tissue morphology feature spectrum; and

A clustering algorithm is used to cluster each dimensionally reduced tissue morphology feature spectrum, and multiple clusters are obtained. Each cluster corresponds to image blocks to be processed with similar tissue morphology feature spectrum, and different clusters correspond to image blocks with different tissue morphology feature spectrum. Process image blocks.
The device according to claim 11, characterized in that the partition unit is specifically used for:

The area corresponding to each image block to be processed in the stained image of the biological tissue section is colored according to the color information corresponding to the cluster where the image block to be processed is located, so as to achieve staining of the biological tissue section. Partitions of images.
An image partitioning device, characterized by including a memory and a processor;

The memory is used to store programs;

The processor is used to execute the program and implement the following steps:

Obtain stained images of biological tissue sections to be partitioned;

Determine the image block to be processed in the stained image of the biological tissue section;

Perform feature extraction on each image block to be processed to obtain a tissue morphology feature vector of each image block to be processed. Each element in the tissue morphology feature vector corresponds to a tissue morphology feature in the image block to be processed. response strength;

Based on the histomorphological feature vector of each image block to be processed, cluster each image block to be processed to obtain a clustering result of each image block to be processed; and

Based on the clustering results of each image block to be processed, the biological tissue section stained image is partitioned.
The device according to claim 13, characterized in that when the processor executes the program, it also performs the following steps:

Using a tissue detection algorithm to determine the tissue area and background area of the stained image of the biological tissue section;

Divide the stained image of the biological tissue section into image blocks of equal size;

For each image block, calculate the proportion of the tissue area in the image block based on the tissue area and background area of the image block; and

The image block whose tissue area proportion meets the preset conditions is determined as the image block to be processed.
The device according to claim 13, characterized in that when the processor executes the program, it also performs the following steps:

The feature extraction network is used to extract features of each image block to be processed to obtain the histomorphological feature vector of each image block to be processed. The feature extraction network is obtained by pre-training the convolutional neural network using natural image data sets. .
The device according to claim 13, characterized in that when the processor executes the program, it also performs the following steps:

Obtain the position information of each image block to be processed in the stained image of the biological tissue section;

Generate a morphological feature data set based on the position information of each image block to be processed in the biological tissue section stained image and the tissue morphology feature vector of each image block to be processed;

Based on the morphological feature data set, generate a tissue morphology feature spatial distribution map, where each spatial location point in the tissue morphology feature spatial distribution map corresponds to a tissue morphology feature spectrum;

Perform dimensionality reduction on each tissue morphology feature spectrum to obtain the dimensionally reduced tissue morphology feature spectrum; and

A clustering algorithm is used to cluster each dimensionally reduced tissue morphology feature spectrum, and multiple clusters are obtained. Each cluster corresponds to image blocks to be processed with similar tissue morphology feature spectrum, and different clusters correspond to image blocks with different tissue morphology feature spectrum. Process image blocks.
A non-volatile computer-readable storage medium on which a computer program is stored, characterized in that when the computer program is executed by a processor, the following steps are implemented:

Obtain stained images of biological tissue sections to be partitioned;

Determine the image block to be processed in the stained image of the biological tissue section;

Perform feature extraction on each image block to be processed to obtain a tissue morphology feature vector of each image block to be processed. Each element in the tissue morphology feature vector corresponds to a tissue morphology feature in the image block to be processed. response strength;

Based on the histomorphological feature vector of each image block to be processed, cluster each image block to be processed to obtain a clustering result of each image block to be processed; and

Based on the clustering results of each image block to be processed, the biological tissue section stained image is partitioned.
The non-volatile computer-readable storage medium according to claim 17, wherein when the computer-readable instructions are executed by the processor, the following steps are also performed:

Using a tissue detection algorithm to determine the tissue area and background area of the stained image of the biological tissue section;

Divide the stained image of the biological tissue section into image blocks of equal size;

For each image block, calculate the proportion of the tissue area in the image block based on the tissue area and background area of the image block; and

The image block whose tissue area proportion meets the preset conditions is determined as the image block to be processed.
The non-volatile computer-readable storage medium according to claim 17, wherein when the computer-readable instructions are executed by the processor, the following steps are also performed:

Use a feature extraction network to extract features from each image block to be processed to obtain the histomorphological feature vector of each image block to be processed. The feature extraction network is obtained by pre-training the convolutional neural network using natural image data sets. .
The non-volatile computer-readable storage medium according to claim 17, wherein when the computer-readable instructions are executed by the processor, the following steps are also performed:

Obtain the position information of each image block to be processed in the stained image of the biological tissue section;

Generate a morphological feature data set based on the position information of each image block to be processed in the biological tissue section stained image and the tissue morphology feature vector of each image block to be processed;

Based on the morphological feature data set, generate a tissue morphology feature spatial distribution map, where each spatial location point in the tissue morphology feature spatial distribution map corresponds to a tissue morphology feature spectrum;

Perform dimensionality reduction on each tissue morphology feature spectrum to obtain the dimensionally reduced tissue morphology feature spectrum; and

A clustering algorithm is used to cluster each dimensionally reduced tissue morphology feature spectrum, and multiple clusters are obtained. Each cluster corresponds to image blocks to be processed with similar tissue morphology feature spectrum, and different clusters correspond to image blocks with different tissue morphology feature spectrum. Process image blocks.