CN115497092A

CN115497092A - Image processing method, device and equipment

Info

Publication number: CN115497092A
Application number: CN202210910919.9A
Authority: CN
Inventors: 叶虎; 韩骁; 周彦宁; 肖凯文; 蔡德; 马兆轩
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2022-07-29
Filing date: 2022-07-29
Publication date: 2022-12-20

Abstract

The application discloses an image processing method, device and equipment, which can be applied to various scenes such as artificial intelligence medical technology, machine learning and the like. The method comprises the following steps: acquiring a plurality of view images of a target digital slice; extracting cell image characteristics and positive probabilities corresponding to positive cells in each field image based on a cell detection model, and determining target positive cells in the target digital section based on the positive probabilities corresponding to the positive cells in the field images; when the positive probability of the target positive cell is matched with the suspicious probability interval, the slice type of the target digital slice is determined according to the cell image characteristics and the positive probability corresponding to the positive cell based on the cell detection model, so that a pathologist is assisted to carry out cytological diagnosis, the workload of the doctor is reduced, and the efficiency and the accuracy are improved.

Description

Image processing method, device and equipment

Technical Field

The present application relates to the field of computer technologies, and in particular, to an image processing method, apparatus, and device.

Background

In the medical health field, doctors can observe cell pictures through microscopic equipment to perform pathological diagnosis, however, because cytology and pathological doctors in primary hospitals are deficient, and because the cytology and pathological doctors are inconsistent and the cost of the doctors is high, extensive research is conducted to assist the pathologists in diagnosis through artificial intelligence technology.

However, in the related art, only most negative cells in the cell section can be excluded, the cell section still needs to be manually checked by a doctor for judgment, and the defect of low accuracy exists.

Disclosure of Invention

The embodiment of the application provides an image processing method, device and equipment, which improve the accuracy of pathological cell classification and reduce the workload of a pathologist.

In one aspect, an image processing method is provided, which is applied to a computer device, where the computer device is deployed with a cascaded cell detection model and a whole-slice classification model, and the method includes:

acquiring a plurality of view images of a target digital slice;

extracting cell image characteristics and positive probabilities corresponding to positive cells in each of the field images based on the cell detection model;

determining target positive cells in the target digital section based on the positive probability corresponding to the positive cells in the visual field image;

and when the positive probability corresponding to the target positive cell is matched with the suspicious probability interval, determining the slice type of the target digital slice according to the cell image characteristics and the positive probability corresponding to the positive cell based on the whole slice classification model.

In another aspect, an image processing apparatus applied to a computer device deployed with a cascaded cell detection model and a whole-slice classification model is provided, including:

a first acquisition module for acquiring a plurality of view images of a target digital slice;

the second acquisition module is used for extracting cell image characteristics and positive probability corresponding to positive cells in each field image based on the cell detection model;

a first determining module, configured to determine a target positive cell in the target digital slice based on a positive probability corresponding to a positive cell in the field image;

and the second determination module is used for determining the slice type of the target digital slice according to the cell image characteristics and the positive probability corresponding to the positive cells based on the whole slice classification model when the positive probability corresponding to the target positive cells is matched with the suspicious probability interval.

In another aspect, a computer-readable storage medium is provided, which stores a computer program, where the computer program is suitable for being loaded by a processor to execute the steps in the image processing method according to any one of the above embodiments.

In another aspect, a computer device is provided, the computer device includes a processor and a memory, the memory stores a computer program, and the processor is used for executing the steps in the image processing method according to any one of the above embodiments by calling the computer program stored in the memory.

In another aspect, a computer program product is provided, which comprises computer instructions for implementing the steps of the image processing method according to any one of the above embodiments when executed by a processor.

The embodiment of the application provides an image processing method, an image processing device and image processing equipment. According to the embodiment of the application, the cell detection model and the whole-slice classification model are cascaded, the positive cells in the target digital slice are detected based on the cell detection model, the characteristic map and the positive probability corresponding to the positive cells are obtained, the target positive cells are determined firstly, the slice type of the target digital slice is judged for the first time according to the positive probability corresponding to the target positive cells, if the positive probability corresponding to the target positive cells is matched with the probability interval, the slice type of the target digital slice is determined based on the whole-slice classification model, and therefore the pathological type of the target digital slice is judged, a pathologist is assisted to conduct cytological diagnosis, the workload of the doctor is reduced, and the efficiency and the accuracy are improved.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

Fig. 1 is a first flowchart of an image processing method according to an embodiment of the present disclosure.

Fig. 2 is a schematic view of a first application scenario of an image processing method according to an embodiment of the present application.

Fig. 3 is a schematic view of a second application scenario of the image processing method according to the embodiment of the present application.

Fig. 4 is a schematic view of a third application scenario of the image processing method according to the embodiment of the present application.

Fig. 5 is a second flowchart of the image processing method according to the embodiment of the present application.

Fig. 6 is a schematic structural diagram of an image processing apparatus according to an embodiment of the present application.

Fig. 7 is a schematic structural diagram of a computer device according to an embodiment of the present application.

Detailed Description

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making an invasive task, are within the scope of protection of the present application.

The embodiment of the application provides an image processing method, an image processing device, computer equipment and a storage medium. Specifically, the image processing method of the embodiment of the present application may be executed by a computer device, where the computer device may be a terminal or a server or the like. The terminal can be a smart phone, a tablet computer, a notebook computer, a smart television, a smart sound box, a wearable smart device, a smart vehicle-mounted terminal and the like, and can further comprise a client, wherein the client can be an application client, a browser client or an instant messaging client and the like. The server may be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, or a cloud server providing basic cloud computing services such as cloud service, a cloud database, cloud computing, a cloud function, cloud storage, network service, cloud communication, middleware service, domain name service, security service, content Delivery Network (CDN), big data and a man-made intelligent platform.

The embodiment of the application can be applied to various scenes, including but not limited to artificial intelligence medical technology, machine learning and the like. The artificial intelligence medical technology scene can include application scenes such as medical diagnosis and the like.

First, some terms or expressions appearing in the course of describing the embodiments of the present application are explained as follows:

artificial Intelligence (AI) is a theory, method, technique and application system that uses a digital computer or a machine controlled by a digital computer to simulate, extend and expand human Intelligence, perceive the environment, acquire knowledge and use the knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technique of computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a manner similar to human intelligence. Artificial intelligence is the research of the design principle and the realization method of various intelligent machines, so that the machines have the functions of perception, reasoning and decision making.

The artificial intelligence technology is a comprehensive subject and relates to the field of extensive technology, namely the technology of a hardware level and the technology of a software level. The artificial intelligence infrastructure generally includes technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and the like.

Computer Vision technology (CV for short) Computer Vision is a science for researching how to make a machine look, and in particular, it refers to that a camera and a Computer are used to replace human eyes to perform machine Vision such as identification and measurement on a target, and further perform graphic processing, so that the Computer processing becomes an image more suitable for human eyes to observe or to transmit to an instrument to detect. As a scientific discipline, computer vision research-related theories and techniques attempt to build artificial intelligence systems that can capture information from images or multidimensional data. Computer vision technologies generally include technologies such as image processing, image recognition, image semantic understanding, image retrieval, OCR, video processing, video semantic understanding, video content/behavior recognition, three-dimensional object reconstruction, 3D technologies, virtual reality, augmented reality, synchronous positioning, and map construction, and also include technologies of common biometric identification such as face recognition and fingerprint recognition.

Machine Learning (ML for short) is a multi-domain cross subject, and relates to multiple subjects such as probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and the like. The method specially researches how a computer simulates or realizes the learning behavior of human beings so as to acquire new knowledge or skills and reorganizes the existing knowledge structure to improve the performance of the computer. Machine learning is the core of artificial intelligence, is the fundamental approach for computers to have intelligence, and is applied throughout various fields of artificial intelligence. Machine learning and deep learning generally include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and formal education learning.

A Residual Network (ResNet) is a convolutional neural Network, and is characterized by easy optimization and can improve accuracy by adding considerable depth.

fast-RCNN (an Object Detection algorithm) was published in 2015 as fast R-CNN (forward read-Time Object Detection with Region proposed network for Real-Time Object Detection). The biggest innovation of the algorithm is to provide an RPN (Region probable Network) Network, and connect Region generation and a convolution Network together by using an Anchor mechanism.

The full-field digital slice Image (WSI) is a high-resolution digital pathological Image, and a WSI is formed after a pathological slice is digitalized, and the contained huge information provides a reliable basis for quantitative analysis of digital pathology.

Negative and positive: negative in a medical examination generally means normal, and positive means problematic. Yin and positive are used more medically and have become terms that generally refer to the presence or absence or to the results of a test. Generally, a positive majority indicates a disease or the occurrence of a certain pathology; negative, in most cases, negates or excludes the possibility of a certain lesion.

Cervical cancer, one of the most common malignancies threatening female health, is highly prevalent in developing countries with underdeveloped economic implications. The etiology of the cervical cancer is definite, and the morbidity and mortality can be effectively reduced through screening, early diagnosis and early treatment. Currently, effective screening protocols include modes such as high-risk HPV detection, cervical cytology screening and combined screening of the two. The cervical cytology screening is based on pathological diagnosis of cervical cell smear and belongs to the field of cell morphological diagnosis.

In the related art, a pathologist can be assisted by a neural network model to perform pathological diagnosis, however, at present, only most negative cells in a section can be eliminated, so that the workload of the pathologist is reduced, and the diagnosis is still required to be performed after the pathologist inspects each positive section. In the related full-scale analysis technology, the problem of low accuracy still exists.

Therefore, an embodiment of the present application provides an image processing method, in which a cell detection model and a whole slice classification model are cascaded, positive cells in a target digital slice are detected based on the cell detection model, a feature map of the positive cells and a positive probability of the positive cells are obtained, if the target digital slice is determined to be a suspicious positive slice for the first time according to the positive probability, the slice positive probability of the target digital slice is determined through the whole slice classification model, and a slice type of the target digital slice is determined for the second time according to the slice positive probability, so as to determine a pathology type of the target digital slice, thereby assisting a pathologist in cytology diagnosis, reducing workload of the physicians, and improving efficiency and accuracy.

The following are detailed below. It should be noted that the description sequence of the following embodiments is not intended to limit the priority sequence of the embodiments.

The embodiments of the present application provide an image processing method, which may be executed by a terminal or a server, or may be executed by both the terminal and the server; the embodiment of the present application is described by taking an example in which the image processing method is executed by a server.

Referring to fig. 1 to 5, fig. 1 and 5 are schematic flow diagrams of an image processing method according to an embodiment of the present disclosure, and fig. 2 to 4 are schematic application scenarios of the image processing method according to the embodiment of the present disclosure. The method can be applied to a computer device deployed with a cascaded cell detection model and a full sheet classification model, comprising:

a plurality of field-of-view images of a digital slice of an object is acquired 101.

The cell detection model is used for detecting cell images and positive probabilities corresponding to positive cells in the visual field images, and judging the section type of the target digital section for the first time according to the positive probabilities corresponding to the positive cells. The full-slice classification model is used for judging the slice type of the target digital slice for the second time if the target digital slice is judged to be a suspicious positive slice for the first time. The slice category may be considered to be the pathological category of the target digital slice. The pathological type determined for the first time may include one of positive, negative and suspected positive.

The target digital slice may be a full-field digital slice (WSI for short).

In this embodiment, step 101 may include: acquiring a target digital slice; performing foreground extraction based on the target digital slice to obtain a foreground area of the target digital slice; and segmenting the foreground region into a plurality of view images with preset sizes based on a grid division mode.

Specifically, one WSI can be divided into different view images by a grid segmentation method. Specifically, for a WSI, first, a foreground region (cell region) may be extracted by using a conventional image segmentation method, and then, the foreground region is segmented into a view image with a fixed size based on a mesh division manner, where the fixed size of the view image is not limited in the present application and may be determined according to a cell detection model, for example, the fixed size may be set to 1280x1280.

For example, referring to fig. 2, the WSI is input into a computer device, and the computer device extracts a foreground region except the WSI, i.e., a cell region, by using a conventional image segmentation method, and then segments the foreground region into a fixed-size visual field image based on a mesh segmentation method. Then, the segmented visual field image is input to a cell detection model.

It is easy to understand that, in general, the WSI includes cells with a very large magnitude, and in order to reduce the difficulty of performing the subsequent extraction of the cell features and improve the efficiency of performing the subsequent processing on the cells, the WSI may be firstly segmented into a plurality of field images, and then each field image is subjected to the subsequent processing.

And 102, extracting cell image characteristics and positive probability corresponding to the positive cells in each field image based on the cell detection model.

The cell detection model is obtained by training according to a positive visual field image marked with positive cells and a negative visual field image not marked with the positive cells, and can output cell image characteristics and positive probability corresponding to the positive cells in the visual field image by inputting the visual field image.

Specifically, the cell detection model may adopt a commonly used two-stage detection model, namely, a resenet 50 pre-trained by using fast RCNN and ImageNet data set, wherein the detection category of the model may be set to 1 category, that is, positive cells of the detection visual field image.

Wherein the positive probability is used to characterize the probability that the cell is positive. Specifically, if the probability of positive is greater than a preset positive threshold, the pathological category of the positive cells can be determined to be positive. If the positive probability is less than a preset negative threshold, the pathological type of the positive cells can be determined to be negative. If the positive probability is not less than the preset negative threshold and not less than the preset negative threshold, a whole-slice classification model cascaded with the cell detection model is required to determine the pathological type of the visual field image. Therefore, the accuracy of model case analysis can be greatly improved based on the cascade system of the two models.

And 103, determining target positive cells in the target digital section based on the positive probability corresponding to the positive cells in the visual field image.

The positive probability corresponding to the target positive cell can be used for determining the slice type of the target digital slice for the first time, and the slice type determined for the first time can include one of a positive slice, a negative slice and a suspicious positive slice.

In some embodiments, step 103 may generally comprise: and determining the positive cell with the highest positive probability in all the positive cells of the target digital section as the target positive cell.

Specifically, the maximum positive probability in the positive probabilities is determined as the positive probability corresponding to the target positive cell, and the positive probability corresponding to the target positive cell is matched with each probability interval, that is, the maximum positive probability in the positive probabilities is matched with each probability interval, if the maximum positive probability is matched with the positive probability interval, it can be determined that the pathological feature of the visual field image is positive, and the slice category of the target digital slice can be determined as a positive slice. If the maximum positive probability is matched with the negative probability interval, the positive probabilities of all the cells can be determined to be matched with the negative probability interval, and the slice type of the target digital slice can be determined to be a negative slice.

In some embodiments, step 103 may generally comprise: sequencing the positive cells in a descending order according to the positive probability; determining positive cells in the first preset number in the sequencing sequence as target positive cells in the target digital section; and averaging the positive probabilities corresponding to the positive cells of the first preset number in the sequencing sequence, and determining the average as the positive probability corresponding to the target positive cell.

The first preset number is not limited in the present application. Usually, the number of cells is orders of magnitude larger, and therefore, the first preset number of positive cells in the sorted order can be taken as the target positive cells. Specifically, the target positive cell can be considered as a cell with a more significant positive characteristic among all positive cells. Therefore, the positive probabilities corresponding to the first preset number of positive cells before, that is, the positive probabilities with more significant positive features, may be averaged, and the average value may be determined as the positive probability corresponding to the target positive cell, and then the slice type of the target digital slice may be determined according to the positive probability corresponding to the target positive cell. For example, the first three positive cells in the sorted order may be determined as the target positive cells, the positive probabilities corresponding to the first three positive cells in the sorted order may be averaged, and the average may be determined as the positive probability corresponding to the target positive cells, and if the positive probability corresponding to the target positive cells matches the positive probability interval, it may be determined that the pathological feature of the visual field image is positive, and the slice type of the target digital slice may be determined as positive. If the positive probability corresponding to the target positive cell is matched with the negative probability interval, the positive probabilities of all the positive cells can be determined to be matched with the negative probability interval, and the slice type of the target digital slice can be determined to be negative.

In this embodiment, after the cell detection model obtains the detection results of all the visual field images, the detection results may include cell image features and positive probabilities corresponding to positive cells in all the visual field images, the detection results of all the visual field images may be combined, all the positive cells are sorted according to the positive probabilities based on the combined detection results, a certain number of positive cells in the sorted order are taken, and the positive cell with the highest probability in the certain number of positive cells is determined as the target positive cell.

Similarly, in this embodiment, after the cell detection model obtains the detection results of all the visual field images, the detection results of all the visual field images may be merged, the positive cells with positive probability are sorted according to the merged detection results, a certain number of positive cells in the sorting order are taken, then, a first preset number of positive cells in the first preset number of positive cells are determined as the target positive cells, the average values of the positive probabilities corresponding to the first preset number of positive cells are obtained, and the average values are determined as the positive probabilities corresponding to the target positive cells.

Wherein the first predetermined number of positive cells may be the second predetermined number of positive cells. That is, the second preset number of positive cells may be taken from the merged detection results as suspicious positive cells, then the first preset number of suspicious positive cells in the second preset number of suspicious positive cells is determined as target positive cells, and the probability of the target positive cells is determined according to the positive probability corresponding to the first preset number of suspicious positive cells in the second preset number of suspicious positive cells.

In this embodiment, taking a diagnosis scenario of cervical liquid-based cells as an example for illustration, since the cervical liquid-based cell diagnosis is mainly for screening intraepithelial lesions, the positive cells may be divided into at least 6 types of positive cells, specifically as follows: atypical squamous cells (ASC-US), low-grade squamous intraepithelial lesions (LSIL), atypical Squamous Cells (ASCH), high-grade squamous intraepithelial lesions (HSIL), squamous carcinomas (SCC) and Atypical Glandular Cells (AGC). The above positive cells may be distinguished, but not limited, by being in different pathological cycles, such as atypical squamous cells prone to high grade lesions, being in high grade lesion cycles, etc.

In some embodiments, the method may further comprise: if the positive probability corresponding to the target positive cell is matched with the positive probability interval, judging the slice type of the target digital slice as a positive slice, wherein the minimum probability in the positive probability interval is greater than the maximum probability in the suspicious probability interval; and if the positive probability corresponding to the target positive cells is matched with the negative probability interval, determining that the slice type of the target digital slice is a negative slice, wherein the maximum probability in the negative probability interval is smaller than the minimum probability in the suspicious probability interval.

The suspicious probability interval is a probability interval between the positive probability interval and the negative probability interval, for example, the positive probability interval is greater than t2, the negative probability interval is less than t1, wherein t2 is greater than t1, and then t1 to t2 can be considered as the suspicious probability interval. If the positive probability corresponding to the target positive cell is greater than t2, namely the target positive cell is matched with the positive probability interval, the pathological type of the visual field image can be determined to be positive. If the positive probability corresponding to the target positive cell is smaller than t1, namely the target positive cell is matched with the negative probability interval, the pathological type of the visual field image can be determined to be negative.

Specifically, if the positive probability corresponding to the target positive cell matches the probability interval, the target digital slice may be considered as a suspicious positive slice, that is, it cannot be accurately determined whether the target digital slice is a positive slice or a negative slice at present.

And 104, when the positive probability corresponding to the target positive cell is matched with the suspicious probability interval, determining the slice type of the target digital slice according to the cell image characteristics and the positive probability corresponding to the positive cell based on the whole-slice classification model.

When the positive probability corresponding to the target positive cell is matched with the suspicious probability interval, the slice type of the target digital slice can be judged to be a suspicious positive slice for the first time, and then, based on the whole-slice classification model, the second judgment is carried out according to the cell image characteristics and the positive probability corresponding to the positive cell. The section type of the second judgment comprises one of a positive section and a negative section, and the questionable probability interval is a probability interval between the positive probability interval and the negative probability interval.

In this embodiment, step 104 may mainly include: determining slice positive probability corresponding to the target digital slice according to the cell image characteristics and the positive probability corresponding to the positive cells; and determining the slice type of the target digital slice according to the slice positive probability.

The whole-slice classification model is obtained by training according to cell image characteristics corresponding to suspicious positive cells of WSI (Wireless sensor interface) which are output by the cell detection model and labeled with slice types (positive slices or negative slices), and the slice types of the target digital slices can be obtained by the whole-slice classification model by inputting the cell image characteristics corresponding to the suspicious positive cells.

Wherein, the slice positive probability is used for representing the probability that the slice category of the target digital slice is a positive slice. From the slice positive probability, the slice class of the target digital slice can be predicted. For example, if the slice positive probability of the target digital slice is greater than a preset full positive threshold, determining that the slice type of the target digital slice is a positive slice; and if the slice positive probability of the target digital slice is not greater than the preset full-slice positive threshold value, determining that the slice type of the target digital slice is a negative slice.

In this embodiment, the step "determining the slice positive probability corresponding to the target digital slice according to the cell image feature and the positive probability corresponding to the positive cell" may mainly include: sequencing the positive cells in a descending order according to the positive probability, and determining the positive cells in the second preset number in the sequencing order as suspicious positive cells; and determining the slice positive probability corresponding to the target digital slice according to the cell image characteristics corresponding to the suspicious positive cells and the cell image characteristics corresponding to the target positive cells.

When the target positive cells are positive cells which are in the first preset number after the positive cells are sorted according to the positive probability in a descending order, the characteristics corresponding to the target positive cells can be fusion characteristics of the characteristics corresponding to the target positive cells.

Specifically, referring to fig. 2, after the visual field image is input into the cell detection model, the cell detection model outputs a positive feature map and a positive probability corresponding to positive cells in the visual field image, the positive feature maps are sorted according to the positive probability, and a second preset number of positive cell feature maps in the sorting can be selected as suspicious positive cell specific maps. For example, in the present embodiment, the second predetermined number may be set to 32, that is, the first 32 positive cell specific graphs in the ranking are selected as suspicious positive cell specific graphs, and then the suspicious positive cell specific graphs are input into the whole model.

In this embodiment, the step of "determining the slice positive probability corresponding to the target digital slice according to the cell image feature corresponding to the suspected positive cell and the cell image feature corresponding to the target positive cell" may include: fusing cell image features corresponding to the suspicious positive cells to obtain first fusion features; fusing the first fusion characteristic with a cell image characteristic corresponding to the target positive cell to obtain a second fusion characteristic; and determining the slice positive probability corresponding to the target digital slice according to the second fusion characteristic.

Specifically, referring to fig. 3, cell image features corresponding to 32 suspicious positive cells of the WSI are input into a global classification model, a self-attention module (self-attention module) of the global classification model fuses the cell image features corresponding to the suspicious positive cells, and all the cell features are fused by an attention mechanism (attention) to obtain a first fusion feature. And then, the whole-piece classification model further fuses the first fusion characteristic and the cell image characteristic corresponding to the target positive cell for enhancement, wherein the specific fusion mode is that the first fusion characteristic and the cell image characteristic corresponding to the target positive cell are spliced (concat), and then a linear layer (linear) is used for further fusion to obtain a second fusion characteristic. Then. The positive probability of the slice can be obtained by a Classifier (Classifier) of a full-scale classification model.

Specifically, the integrity of the corresponding features of the WSI can be enhanced by fusing the features of the 32 suspicious positive cells, the slice positive probability of the WSI is predicted according to the fused features, and the prediction accuracy can be improved.

In some embodiments, before 101, may further include: acquiring a sample image set, wherein the sample image set comprises a plurality of labeled sample digital slice images and a truth label of each labeled sample digital slice image, and the truth label comprises a labeling position of a sample positive cell on each labeled sample digital slice image and a truth positive probability of the sample positive cell on each sample digital slice image; and inputting the sample image set into an initial cell detection model to train to obtain the cell detection model.

It is easy to understand that the greatest characteristic of artificial intelligence is that learning ability is strong, training is carried out by inputting labeled sample data into an initial model with random parameters, parameters of the initial model are adjusted when errors occur, and a required model can be formed after a large amount of training.

In some embodiments, the step of inputting the sample image set into the initial cell detection model to train the cell detection model may mainly include: inputting the sample image set into an initial cell detection model, and acquiring the predicted position of the sample positive cells determined by the initial cell detection model according to the labeled sample digital slice image and the predicted positive probability of the sample positive cells; determining a first loss function according to the predicted position, the predicted positive probability, the labeled position and the true positive probability; and training the initial cell detection model according to the first loss function to obtain the cell detection model.

In this embodiment, the true label of the labeled sample digital slice image includes the labeled position of the sample positive cell and the true positive probability of the sample positive cell. Specifically, the positive cells of the sample can be labeled through a labeling frame, which mainly includes the coordinate information of the center point of the labeling frame, and the length and width information of the labeling frame. Specifically, the sample image set may further include an unlabeled sample digital slice image, which is a negative sample. The marked sample digital slice image can be marked as a positive sample, the unmarked sample digital slice image can be marked as a negative sample, and the positive sample and the negative sample can be distributed in proportion. And then, inputting the sample image set into an initial cell detection model, determining a loss function according to a result output by the initial cell detection model and a truth label of the labeled sample digital slice image, and training the initial cell detection model.

In this embodiment, the step of "determining the first loss function according to the predicted position, the predicted positive probability, and the annotated position and the true positive probability" may include: determining a first verification sample according to the first preset proportion and the marked sample digital slice image; and determining a first loss function according to the initial cell detection model according to the predicted position of the sample positive cells in the first verification sample, the predicted positive probability of the sample positive cells in the first verification sample, the annotated position of the sample positive cells in the first verification sample and the true positive probability of the sample positive cells in the first verification sample, which are determined by the first verification sample.

It is easily understood that, in order to improve the training efficiency, a part of the sample digital slice images in the sample image set can be proportionally selected as verification samples, and the initial detection model is trained by determining a loss function according to the verification samples. It should be noted that the first preset ratio is not limited in this application, and can be customized.

For example, please refer to fig. 4, wherein the digital slice image of the sample with the label frame on the left is a positive sample, and the digital slice image of the sample without the label on the right is a negative sample.

In this embodiment, the method may further include: extracting cell image characteristics corresponding to suspicious positive cells in the plurality of sample digital slice images based on the trained cell detection model; and inputting cell image characteristics corresponding to the suspicious positive cells into the initial full-sheet classification model so as to train and obtain the full-sheet classification model.

In particular, the truth label of the annotated sample digital slice image also includes the true slice positive probability of the sample digital slice image. Inputting the characteristics of suspicious positive cells into an initial full-wafer classification model, determining a loss function according to the result output by the initial full-wafer classification model and the true-value slice positive probability corresponding to the sample digital slice image, and training the initial full-wafer classification model to obtain the full-wafer classification model.

In some embodiments, the step of inputting cell image features corresponding to suspicious positive cells of the plurality of samples into the initial whole-slice classification model to train the whole-slice classification model may include: inputting cell image characteristics corresponding to a plurality of sample suspicious positive cells into an initial full-wafer classification model, and acquiring the predicted slice positive probability of the marked sample digital slice image determined by the initial full-wafer classification model according to the cell image characteristics corresponding to the plurality of sample suspicious positive cells; determining a second loss function according to the predicted slice positive probability and the true slice positive probability of the marked sample digital slice image; and training the initial full-scale classification model according to the second loss function to obtain the full-scale classification model.

Similarly, in order to improve the training efficiency, a part of the sample digital slice images may be selected from the sample image set in proportion as verification samples, a second loss function is determined according to the verification samples, and the initial full-slice classification model is trained. The step of determining a second loss function based on the predicted slice positive probability and the true slice positive probability of the labeled sample digital slice image may comprise: determining a second verification sample according to a second preset proportion and the marked sample digital slice image; and determining a second loss function according to the predicted slice positive probability of the second verification sample determined by the full-scale classification model according to the suspicious positive cells of the plurality of samples corresponding to the second verification sample and the true slice positive probability of the second verification sample.

Specifically, the whole slice classification model obtained by training can output slice positive probability of WSI according to the characteristics of a plurality of suspicious positive cells of the WSI. The slice positive probability is used to characterize the positive probability of the WSI slice.

For better explaining the image processing method provided by the embodiment of the present application, referring to fig. 5, the flow of the image processing method provided by the embodiment of the present application can be summarized and summarized as the following steps:

step 201, acquiring a plurality of visual field images of the target digital section through the cell detection model.

Wherein the target digital slice may be a full field of view digital slice (WSI). One WSI can be divided into different view images by a grid splitting method. Specifically, for a WSI, first, a foreground region (cell region) may be extracted by using a conventional image segmentation method, and then, the foreground region is segmented into a view image with a fixed size based on a grid division manner, where the fixed size of the view image is not limited in the present application and may be determined according to a cell detection model, for example, the fixed size may be set to 1280x1280.

Step 202, extracting cell image characteristics and positive probability corresponding to positive cells in each visual field image based on the cell detection model.

And step 203, determining target positive cells in the target digital section based on the positive probability corresponding to the positive cells in the visual field image, and determining the positive probability corresponding to the target positive cells.

In some embodiments, step 203 may generally comprise: and determining the positive cell with the highest positive probability in all the positive cells of the target digital section as the target positive cell.

The first preset number is not limited in the present application. Usually, the number of cells is orders of magnitude larger, and therefore, the first preset number of positive cells in the sorted order can be taken as the target positive cells. Specifically, the target positive cell can be considered as a cell with a more significant positive characteristic among all positive cells. Therefore, the positive probabilities corresponding to the first preset number of positive cells before, that is, the positive probabilities with more significant positive features, can be averaged, the average is determined as the positive probability corresponding to the target positive cell, and then the slice type of the target digital slice is determined according to the positive probability corresponding to the target positive cell. For example, the first three positive cells in the ranking order may be determined as the target positive cells, the average of the positive probabilities corresponding to the first three positive cells in the ranking order may be determined as the positive probability corresponding to the target positive cells, and if the positive probability corresponding to the target positive cells matches the positive probability interval, it may be determined that the pathological feature of the field image is positive, and the slice type of the target digital slice may be determined as positive. If the positive probability corresponding to the target positive cell is matched with the negative probability interval, the positive probabilities of all the positive cells can be determined to be matched with the negative probability interval, and the slice type of the target digital slice can be determined to be negative.

And step 204, judging the section type of the target digital section for the first time according to the positive probability corresponding to the target positive cell.

The slice category of the first determination may include one of a positive slice, a negative slice, and a suspicious positive slice.

Specifically, if the positive probability corresponding to the target positive cell matches with the positive probability interval, the target digital slice is determined to be a positive slice, wherein the minimum probability in the positive probability interval is greater than the maximum probability in the suspicious probability interval. And if the positive probability corresponding to the target positive cells is matched with the negative probability interval, determining the target digital slice as a negative slice, wherein the maximum probability in the negative probability interval is smaller than the minimum probability in the suspicious probability interval.

Step 205, if the positive probability corresponding to the target positive cell matches the suspicious probability interval, determining the slice type of the target digital slice as a suspicious positive slice, determining the slice positive probability of the target digital slice according to the cell image characteristics corresponding to the suspicious positive cell and the cell image characteristics corresponding to the target positive cell based on a full-slice classification model, and determining the slice type of the target digital slice for the second time according to the slice positive probability.

The suspicious positive cells are positive cells which are arranged in the positive cells according to a positive probability descending order and are positioned in the front second preset number in the arrangement order.

Wherein the slice type of the second determination includes one of a positive slice and a negative slice.

The whole-slice classification model is obtained by training according to cell image characteristics corresponding to suspicious positive cells of WSI (Wireless sensor interface) marked with pathological types (positive or negative) output by the cell detection model, and the slice positive probability of the target digital slice can be determined by the whole-slice classification model by inputting the cell image characteristics corresponding to the suspicious positive cells. Wherein the slice positive probability is used to characterize the probability that the target digital slice is a positive slice. According to the slice positive probability, the pathological category of the target digital slice can be predicted. For example, if the slice positive probability of the target digital slice is greater than a preset full positive threshold, determining that the target digital slice is a positive slice; and if the slice positive probability of the target digital slice is not greater than the preset full-slice positive threshold value, determining that the target digital slice is a negative slice.

The suspicious probability interval is a probability interval between the positive probability interval and the negative probability interval, for example, the positive probability interval is greater than t2, the negative probability interval is less than t1, wherein t2 is greater than t1, and then t1 to t2 can be considered as the suspicious probability interval. If the positive probability corresponding to the target positive cell is greater than t2, namely the positive probability is matched with the positive probability interval, the pathological type of the visual field image can be determined to be positive. If the positive probability corresponding to the target positive cell is smaller than t1, namely the target positive cell is matched with the negative probability interval, the pathological type of the visual field image can be determined to be negative.

In this embodiment, the image processing method may be applied, but not limited to, in a cervical fluid-based cell diagnosis scenario, for example, the cervical fluid-based cell is scanned into a WSI, then the WSI is segmented into a fixed-size visual field image, and the visual field image is input into the cell detection model. The cell detection model extracts cell image features and positive probabilities corresponding to positive cells in a visual field image according to the visual field image, then inputs the cell image features corresponding to suspicious positive cells into a whole-slice classification model if the positive probability (the maximum positive probability in the positive probabilities) corresponding to target positive cells is matched with a suspicious probability interval, the whole-slice classification model determines the slice positive probability of the WSI according to the cell image features corresponding to the suspicious positive cells and the cell image features corresponding to the target positive cells, and the pathological category (positive slices or negative slices) of the WSI can be determined according to the slice positive probability. It is easy to understand that the image processing method can be but not limited to assist a pathologist in cervical fluid-based cell diagnosis, reduce the workload of the pathologist and improve the work efficiency of the pathologist, and can also be but not limited to replace the pathologist to perform cervical fluid-based cell diagnosis so as to determine the pathological type of pathological sections and record corresponding positive cell areas in the positive sections, thereby achieving the effect of improving the diagnosis efficiency of cervical fluid-based cells.

All the above technical solutions can be combined arbitrarily to form the optional embodiments of the present application, and are not described herein again.

According to the method and the device, the multiple visual field images of the target digital slice are obtained, the cell image characteristics and the positive probability corresponding to the positive cells in each visual field image are extracted based on the cell detection model, the target positive cells in the target digital slice are determined based on the positive probability corresponding to the positive cells in the visual field images, and when the positive probability corresponding to the target positive cells is matched with the suspicious probability interval, the slice type of the target digital slice is determined based on the whole classification model according to the cell image characteristics and the positive probability corresponding to the positive cells. According to the embodiment of the application, the cell detection model and the whole-chip classification model are cascaded, the positive cells in the target digital slice are detected based on the cell detection model, the characteristic map and the positive probability corresponding to the positive cells are obtained, the target positive cells are determined firstly, the slice type of the target digital slice is judged for the first time according to the positive probability corresponding to the target positive cells, if the positive probability corresponding to the target positive cells is matched with the probability interval, the slice type of the target digital slice is determined based on the whole-chip classification model, and therefore the pathological type of the target digital slice is judged, a pathologist is assisted to conduct cytological diagnosis, the workload of the doctor is reduced, and the efficiency and the accuracy are improved.

In order to better implement the image processing method according to the embodiment of the present application, an embodiment of the present application further provides an image processing apparatus. Referring to fig. 6, fig. 6 is a first structural schematic diagram of an image processing apparatus according to an embodiment of the present disclosure. The image processing apparatus 10 may be applied to a computer device, where the computer device is deployed with a cascaded cell detection model and a full-scale classification model, and includes:

a first acquisition module 11, configured to acquire a plurality of view images of a target digital slice;

the second obtaining module 12 is configured to extract, based on the cell detection model, cell image features and positive probabilities corresponding to positive cells in each field image;

a first determining module 13, configured to determine a target positive cell in the target digital slice according to a positive probability corresponding to the positive cell in the base view image;

and a second determining module 14, configured to determine, based on the full-scale classification model, a slice category of the target digital slice according to the feature and the positive probability corresponding to the positive cell when the positive probability corresponding to the target positive cell matches the suspicious probability region.

Optionally, the first determining module 13 may be configured to: sequencing the positive cells in a descending order according to the positive probability; determining positive cells in the first preset number in the sequencing sequence as target positive cells in the target digital section; and averaging the positive probabilities corresponding to the positive cells of the first preset number in the sequencing sequence, and determining the average as the positive probability corresponding to the target positive cell.

Optionally, the second determining module 14 may be configured to: determining slice positive probability corresponding to the target digital slice according to the cell image characteristics and the positive probability corresponding to the positive cells; and determining the slice type of the target digital slice according to the slice positive probability.

Optionally, the second determining module 14 may be configured to: sequencing the positive cells in a descending manner according to the positive probability, and determining the positive cells in the second preset number in the sequencing order as suspicious positive cells; and determining the slice positive probability of the target digital slice according to the cell image characteristics corresponding to the suspicious positive cells and the cell image characteristics corresponding to the target positive cells.

Optionally, the second determining module 14 may be configured to: fusing cell image characteristics corresponding to the suspicious positive cells to obtain first fusion characteristics; fusing the first fusion characteristic with a cell image characteristic corresponding to the target positive cell to obtain a second fusion characteristic; and determining the slice positive probability of the target digital slice according to the second fusion characteristic.

Optionally, the image processing apparatus 10 may further include a third determining module, which may be configured to: if the positive probability corresponding to the target positive cell is matched with the positive probability interval, judging the slice type of the target digital slice as a positive slice, wherein the minimum probability in the positive probability interval is greater than the maximum probability in the suspicious probability interval; and if the positive probability corresponding to the target positive cells is matched with the negative probability interval, judging that the slice type of the target digital slice is a negative slice, wherein the maximum probability in the negative probability interval is smaller than the minimum probability in the suspicious probability interval.

Optionally, the first obtaining module 11 may be configured to: acquiring a target digital slice; performing foreground extraction based on the target digital slice to obtain a foreground area of the target digital slice; and segmenting the foreground region into a plurality of view images with preset sizes based on a grid division mode.

Optionally, the image processing apparatus 10 may further include: a training module operable to: obtaining a sample image set, wherein the sample image set comprises a plurality of marked sample digital slice images and a truth label of each marked sample digital slice image, and the truth label comprises a marking position of a sample positive cell on each marked sample digital slice image and a truth positive probability of the sample positive cell on each sample digital slice image; and inputting the sample image set into an initial cell detection model to train to obtain the cell detection model.

Optionally, the training module may be specifically configured to: inputting the multi-sample image set into an initial cell detection model, and acquiring the predicted position of sample positive cells determined by the initial cell detection model according to the marked digital slice image and the predicted positive probability of the sample positive cells; determining a first loss function according to the predicted position, the predicted positive probability, the labeled position and the true positive probability; and training the initial cell detection model according to the first loss function to obtain the cell detection model.

Optionally, the training module may be further configured to: extracting cell image characteristics corresponding to sample suspicious positive cells in the plurality of sample digital slice images based on the trained cell detection model; and inputting cell image characteristics corresponding to the suspicious positive cells of the plurality of samples into the initial full-scale classification model so as to train and obtain the full-scale classification model.

Optionally, the annotated sample digital slice image is further annotated with a true-value slice positive probability of the annotated sample digital slice image, and the training module may be specifically configured to: inputting cell image characteristics corresponding to a plurality of sample suspicious positive cells into an initial full-wafer classification model, and acquiring the predicted slice positive probability of the marked sample digital slice image determined by the initial full-wafer classification model according to the cell image characteristics corresponding to the plurality of sample suspicious positive cells; determining a second loss function according to the positive probability of the predicted section and the true value section positive probability of the labeled sample digital section image; and training the initial full-scale classification model according to the second loss function to obtain the full-scale classification model.

It should be noted that, for the functions of each module in the image processing apparatus 10 in the embodiment of the present application, reference may be made to the specific implementation manner in each method embodiment described above, and details are not described here again.

The respective units in the image processing apparatus 10 described above may be wholly or partially implemented by software, hardware, and a combination thereof. The units may be embedded in a hardware form or independent from a processor in the computer device, or may be stored in a memory in the computer device in a software form, so that the processor calls to execute operations corresponding to the units.

The image processing apparatus 10 may be integrated into a terminal or a server having a memory and a processor and having computing capability, or the image processing apparatus 10 may be the terminal or the server.

The image processing apparatus 10 provided in the embodiment of the present application is applied to a computer device, the computer device is deployed with a cascaded cell detection model and a whole slice classification model, a plurality of visual field images of a target digital slice are acquired through a first acquisition module 11, then a second acquisition module 12 extracts cell image features and positive probabilities corresponding to positive cells in each visual field image based on the cell detection model, then a first determination module 13 determines, based on the cell image features and positive probabilities corresponding to the positive cells in the visual field images, slice categories of the target digital slice when the positive probabilities corresponding to the target positive cells in the target digital slice match with a suspicious probability interval, and a second determination module 14 determines, based on the whole slice classification model, slice categories of the target digital slice according to the cell image features and positive probabilities corresponding to the positive cells, thereby assisting a pathologist in performing cytological diagnosis, reducing workload of doctors, and improving efficiency and accuracy.

Optionally, the present application further provides a computer device, which includes a memory and a processor, where the memory stores a computer program, and the processor implements the steps in the foregoing method embodiments when executing the computer program.

Fig. 7 is a schematic structural diagram of a computer device according to an embodiment of the present application, where the computer device may be the terminal or the server shown in fig. 1. As shown in fig. 7, the computer device 20 may include: a communication interface 21, a memory 22, a processor 23 and a communication bus 24. The communication interface 21, the memory 22 and the processor 23 realize mutual communication through a communication bus 24. The communication interface 21 is used for the computer device 20 to perform data communication with an external device. The memory 22 may be used for storing software programs and modules, and the processor 23 may operate the software programs and modules stored in the memory 22, for example, the software programs of the corresponding operations in the foregoing method embodiments.

Alternatively, the processor 23 may call the software programs and modules stored in the memory 22 to perform the following operations: acquiring a plurality of view images of a target digital slice; extracting cell image characteristics and positive probability corresponding to positive cells in each field image based on a cell detection model; determining target positive cells in the target digital section based on the positive probability corresponding to the positive cells in the visual field image; and when the positive probability corresponding to the target positive cell is matched with the suspicious probability interval, determining the slice type of the target digital slice according to the cell image characteristics and the positive probability corresponding to the positive cell based on the whole-slice classification model.

The present application also provides a computer-readable storage medium for storing a computer program. The computer-readable storage medium is applicable to a computer apparatus, and the computer program causes the computer apparatus to execute the respective flows in the image processing method in the embodiment of the present application, which will not be redundantly described here for the sake of brevity.

The present application also provides a computer program product comprising computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instruction from the computer-readable storage medium, and executes the computer instruction, so that the computer device executes the corresponding process in the image processing method in the embodiment of the present application, which is not described herein again for brevity.

The present application also provides a computer program comprising computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instruction from the computer readable storage medium, and executes the computer instruction, so that the computer device executes the corresponding flow in the image processing method in the embodiment of the present application, which is not described herein again for brevity.

It should be understood that the processor of the embodiments of the present application may be an integrated circuit chip having signal processing capabilities. In implementation, the steps of the above method embodiments may be performed by integrated logic circuits of hardware in a processor or instructions in the form of software. The Processor may be a general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic device, or discrete hardware components. The various methods, steps, and logic blocks disclosed in the embodiments of the present application may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of the method disclosed in connection with the embodiments of the present application may be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software module may be located in ram, flash memory, rom, prom, or eprom, registers, etc. storage media as is well known in the art. The storage medium is located in a memory, and a processor reads information in the memory and completes the steps of the method in combination with hardware of the processor.

It will be appreciated that the memory in the embodiments of the subject application can be either volatile memory or nonvolatile memory, or can include both volatile and nonvolatile memory. The non-volatile Memory may be a Read-Only Memory (ROM), a Programmable ROM (PROM), an Erasable PROM (EPROM), an Electrically Erasable PROM (EEPROM), or a flash Memory. The volatile Memory may be a Random Access Memory (RAM) which serves as an external cache. By way of example, and not limitation, many forms of RAM are available, such as Static random access memory (Static RAM, SRAM), dynamic random access memory (Dynamic RAM, DRAM), synchronous Dynamic random access memory (Synchronous DRAM, SDRAM), double Data Rate Synchronous Dynamic random access memory (DDR SDRAM), enhanced Synchronous SDRAM (ESDRAM), synchronous Link DRAM (SLDRAM), and Direct bus RAM (DR RAM). It should be noted that the memory of the systems and methods described herein is intended to comprise, without being limited to, these and any other suitable types of memory.

It should be understood that the above memories are exemplary but not limiting illustrations, for example, the memories in the embodiments of the present application may also be Static Random Access Memory (SRAM), dynamic random access memory (dynamic RAM, DRAM), synchronous Dynamic Random Access Memory (SDRAM), double data rate SDRAM (double data rate SDRAM), enhanced SDRAM (enhanced SDRAM, ESDRAM), synchronous Link DRAM (SLDRAM), direct Rambus RAM (DR RAM), and so on. That is, the memory in the embodiments of the present application is intended to comprise, without being limited to, these and any other suitable types of memory.

Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.

It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.

In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one logical functional division, and in actual implementation, there may be other divisions, for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit.

The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application or portions thereof that substantially contribute to the prior art may be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer or a server) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: various media capable of storing program codes, such as a U disk, a removable hard disk, a ROM, a RAM, a magnetic disk, or an optical disk.

The above description is only for the specific embodiments of the present application, but the scope of the present application is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present application, and shall be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. An image processing method applied to a computer device deployed with a cascaded cell detection model and a whole slice classification model, the method comprising:

acquiring a plurality of view images of a target digital slice;

extracting cell image characteristics and positive probability corresponding to positive cells in each field image based on the cell detection model;

determining target positive cells in the target digital section based on the positive probability corresponding to the positive cells in the field image;

and when the positive probability corresponding to the target positive cell is matched with the suspicious probability interval, determining the slice type of the target digital slice according to the cell image characteristics and the positive probability corresponding to the positive cell based on the full-slice classification model.

2. The image processing method according to claim 1, wherein the determining the target positive cells in the target digital section based on the positive probability corresponding to the positive cells in the visual field image comprises:

sorting the positive cells in descending order according to the positive probability;

determining a first preset number of positive cells in the sequencing sequence as the target positive cells in the target digital section; and

and averaging the positive probabilities corresponding to the first preset number of positive cells in the sequencing sequence, and determining the average as the positive probability corresponding to the target positive cell.

3. The image processing method of claim 1, wherein the determining the slice class of the target digital slice according to the cell image feature and the positive probability corresponding to the positive cell based on the global classification model comprises:

determining slice positive probability corresponding to the target digital slice according to the cell image characteristics and the positive probability corresponding to the positive cells;

and determining the slice type of the target digital slice according to the slice positive probability.

4. The image processing method according to claim 3, wherein the determining the slice positive probability corresponding to the target digital slice according to the cell image feature and the positive probability corresponding to the positive cell comprises:

sequencing the positive cells in a descending order according to the positive probability, and determining the positive cells with the first second preset number in the sequencing order as suspicious positive cells;

and determining the slice positive probability corresponding to the target digital slice according to the cell image characteristics corresponding to the suspicious positive cells and the cell image characteristics corresponding to the target positive cells.

5. The image processing method according to claim 4, wherein the determining the slice positive probability corresponding to the target digital slice according to the cell image feature corresponding to the suspected positive cell and the cell image feature corresponding to the target positive cell comprises:

fusing cell image features corresponding to the suspicious positive cells to obtain first fusion features;

fusing the first fusion characteristic with a cell image characteristic corresponding to the target positive cell to obtain a second fusion characteristic;

and determining the slice positive probability corresponding to the target digital slice according to the second fusion characteristic.

6. The image processing method of claim 1, wherein the method further comprises:

if the positive probability corresponding to the target positive cell is matched with the positive probability interval, determining the slice type of the target digital slice as a positive slice, wherein the minimum probability in the positive probability interval is greater than the maximum probability in the suspicious probability interval;

and if the positive probability corresponding to the target positive cell is matched with the negative probability interval, determining the slice type of the target digital slice as a negative slice, wherein the maximum probability in the negative probability interval is smaller than the minimum probability in the suspicious probability interval.

7. The image processing method of claim 1, wherein said acquiring a plurality of field-of-view images of a digital slice of a target comprises:

acquiring a target digital slice;

performing foreground extraction based on the target digital slice to obtain a foreground area of the target digital slice;

and segmenting the foreground region into a plurality of view images with preset sizes based on a grid division mode.

8. The image processing method of any of claims 1 to 7, prior to said acquiring a plurality of field-of-view images of a digital slice of a target, comprising:

obtaining a sample image set, wherein the sample image set comprises a plurality of annotated sample digital slice images and a truth label of each annotated sample digital slice image, and the truth label comprises an annotation position of a sample positive cell on each annotated sample digital slice image and a truth positive probability of the sample positive cell on each sample digital slice image;

and inputting the sample image set into an initial cell detection model to train to obtain the cell detection model.

9. The image processing method of claim 8, wherein the inputting the sample image set into an initial cell detection model to train the cell detection model comprises:

inputting the sample image set into an initial cell detection model, and acquiring the predicted position of the sample positive cells determined by the initial cell detection model according to the labeled sample digital slice image and the predicted positive probability of the sample positive cells;

determining a first loss function according to the predicted position, the positive probability of prediction, the labeled position and the positive probability of truth;

and training the initial cell detection model according to the first loss function to obtain the cell detection model.

10. The image processing method of claim 8, wherein the method further comprises:

extracting cell image characteristics corresponding to sample suspicious positive cells in the plurality of sample digital slice images based on the cell detection model obtained by training;

inputting the cell image characteristics corresponding to the suspicious positive cells of the plurality of samples into an initial full-scale classification model so as to train and obtain the full-scale classification model.

11. The image processing method of claim 10, wherein the truth label of the labeled sample digital slice image further comprises a truth slice positive probability of the labeled sample digital slice image, and the inputting the cell image features corresponding to the suspicious positive cells of the plurality of samples into an initial full-scale classification model for training to obtain the full-scale classification model comprises:

inputting cell image characteristics corresponding to the plurality of sample suspicious positive cells into an initial full-wafer classification model, and acquiring the predicted slice positive probability of the labeled sample digital slice image, which is determined by the initial full-wafer classification model according to the cell image characteristics corresponding to the plurality of sample suspicious positive cells;

determining a second loss function according to the predicted slice positive probability and the true slice positive probability of the annotated sample digital slice image;

and training the initial full-scale classification model according to the second loss function to obtain the full-scale classification model.

12. An image processing apparatus applied to a computer device deployed with a cascaded cell detection model and a whole slice classification model, comprising:

a first determination module, configured to determine a target positive cell in the target digital slice based on a positive probability corresponding to a positive cell in the visual field image;

and the second determination module is used for determining the category of the target digital section according to the cell image characteristics and the positive probability corresponding to the positive cells based on the whole-sheet classification model when the positive probability corresponding to the target positive cells is matched with the suspicious probability interval.

13. A computer-readable storage medium, characterized in that it stores a computer program adapted to be loaded by a processor for performing the steps in the image processing method according to any one of claims 1 to 11.

14. A computer arrangement, characterized in that the computer arrangement comprises a processor and a memory, in which a computer program is stored, the processor being adapted to carry out the steps in the image processing method of any of claims 1-11 by invoking the computer program stored in the memory.

15. A computer program product comprising computer instructions, characterized in that said computer instructions, when executed by a processor, implement the steps in the image processing method of any of claims 1-11.