Medical image processing method and device

Abstract

The embodiment of the present application discloses a medical image processing method and device, which relates to the field of artificial intelligence technology, wherein the method comprises: obtaining a medical image set; performing difference recognition on the reference medical image and the target medical image to obtain a candidate non-lesion area in the target medical image; determining the area size information of the candidate non-lesion area as the candidate area size information; if the candidate area size information does not match the annotated area size information, adjusting the candidate non-lesion area according to the annotated area size information to obtain a target non-lesion area of ​​the target medical image. The present application can improve the accuracy of identifying non-lesion areas.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a medical image processing method and device. Background Art

[0002] Image segmentation is the process of dividing an image into several specific regions with unique properties and identifying the regions of interest. With the development of computer technology and medical analysis technology, medical image segmentation has become a top priority in medical analysis technology. Medical image segmentation is a key issue that determines whether medical images can provide reliable evidence in clinical diagnosis and treatment. For example, in the process of breast cancer pathological analysis, non-lesion areas are obtained from medical images including the breast. That is, non-lesion areas refer to areas in the medical image that are used to reflect that no lesions have occurred, that is, areas where immune cells are located; further, treatment plans for breast cancer can be formulated based on non-lesion areas. However, at present, medical images are mainly read manually, and regions of interest (such as non-lesion areas) are marked from medical images. Due to the subjectivity of manual reading, the accuracy of the identified areas is relatively low. Summary of the invention

[0003] The technical problem to be solved by the embodiments of the present application is to provide a medical image processing method and device that can improve the accuracy of identifying non-lesion areas.

[0004] An embodiment of the present application provides a medical image processing method, including:

[0005] Acquire a medical image set; the medical image set includes a reference medical image, a target medical image to be identified, and annotated area size information of a non-lesion area of ​​the target medical image, the target medical image includes a lesion area and a non-lesion area, and the reference medical image includes a lesion area;

[0006] Performing difference recognition on the reference medical image and the target medical image to obtain a candidate non-lesion area in the target medical image;

[0007] Determining region size information of the candidate non-lesion region as candidate region size information;

[0008] If the candidate region size information does not match the annotated region size information, the candidate non-lesion region is adjusted according to the annotated region size information to obtain a target non-lesion region of the target medical image.

[0009] An embodiment of the present application provides a medical image processing device, including:

[0010] an acquisition module, configured to acquire a medical image set; the medical image set includes a reference medical image, a target medical image to be identified, and annotated area size information of a non-lesion area of ​​the target medical image, the target medical image includes a lesion area and a non-lesion area, and the reference medical image includes a lesion area;

[0011] an identification module, configured to perform difference identification on the reference medical image and the target medical image to obtain a candidate non-lesion region in the target medical image;

[0012] A determination module, used to determine the region size information of the candidate non-lesion region as the candidate region size information;

[0013] The adjustment module is used to adjust the candidate non-lesion region according to the annotated region size information to obtain the target non-lesion region of the target medical image if the candidate region size information does not match the annotated region size information.

[0014] Optionally, the recognition module performs difference recognition on the reference medical image and the target medical image to obtain a candidate non-lesion area in the target medical image in a manner that includes:

[0015] Converting the reference medical image into an image in a target color space to obtain a converted reference medical image, and converting the target medical image into an image in the target color space to obtain a converted target medical image;

[0016] Acquiring brightness information of pixels in the converted reference medical image and brightness information of pixels in the converted target medical image;

[0017] Based on the brightness information of the pixels in the converted reference medical image and the brightness information of the pixels in the converted target medical image, the reference medical image and the target medical image are differentially identified to obtain a candidate non-lesion area in the target medical image.

[0018] Optionally, the recognition model performs difference recognition on the reference medical image and the target medical image according to the brightness information of the pixels in the converted reference medical image and the brightness information of the pixels in the converted target medical image, and a method for obtaining the candidate non-lesion area in the target medical image includes:

[0019] Acquire the difference between the brightness information of the pixel points in the converted reference medical image and the brightness information of the corresponding pixel points in the converted target medical image;

[0020] Identify, from the converted target medical image, a pixel whose brightness information corresponding difference is greater than a difference threshold as a first target pixel;

[0021] The region where the first target pixel point is located in the target medical image is determined as a candidate non-lesion region in the target medical image.

[0022] Optionally, the candidate region size information includes the candidate region ratio of the candidate non-lesion region in the target medical image; the annotated region size information includes the annotated region ratio of the non-lesion region in the target medical image in the target medical image;

[0023] If the candidate region size information does not match the annotated region size information, the adjustment module adjusts the candidate non-lesion region according to the annotated region size information to obtain the target non-lesion region of the target medical image, including:

[0024] Obtaining a difference in area ratio between the candidate area ratio and the marked area ratio;

[0025] If the area proportion difference is greater than the proportion threshold, it is determined that the candidate area size information does not match the marked area size information;

[0026] The candidate non-lesion area is adjusted according to the proportion of the marked area to obtain the target non-lesion area.

[0027] Optionally, the adjustment module adjusts the candidate non-lesion area according to the proportion of the marked area to obtain the target non-lesion area in a manner including:

[0028] If the proportion of the candidate area is less than the proportion of the annotated area, the candidate non-lesion area is expanded according to the proportion of the annotated area to obtain the target non-lesion area;

[0029] If the proportion of the candidate area is greater than the proportion of the annotated area, the candidate non-lesion area is reduced according to the proportion of the annotated area to obtain the target non-lesion area.

[0030] Optionally, if the proportion of the candidate area is less than the proportion of the annotated area, the adjustment module performs expansion processing on the candidate non-lesion area according to the proportion of the annotated area to obtain the target non-lesion area, including:

[0031] If the proportion of the candidate area is smaller than the proportion of the marked area, obtaining an expansion parameter, wherein the expansion parameter includes an expansion shape and an expansion size;

[0032] Iteratively expanding the candidate non-lesion region according to the expansion parameter to obtain an expanded candidate non-lesion region;

[0033] Obtaining the area ratio of the expanded candidate non-lesion area in the target medical image as the expanded area ratio;

[0034] The expanded candidate non-lesion area whose area ratio difference between the expanded area ratio and the marked area ratio is less than the ratio threshold is determined as the target non-lesion area.

[0035] Optionally, if the proportion of the candidate area is less than the proportion of the annotated area, the adjustment module performs expansion processing on the candidate non-lesion area according to the proportion of the annotated area to obtain the target non-lesion area, including:

[0036] If the proportion of the candidate area is less than the proportion of the annotated area, then obtaining pixel values ​​of pixels adjacent to the candidate non-lesion area in the target medical image and pixel values ​​of pixels in the candidate non-lesion area;

[0037] According to the pixel values ​​of the pixel points in the candidate non-lesion area and the pixel values ​​of the pixel points adjacent to the candidate non-lesion area, clustering is performed on the pixel points in the candidate non-lesion area and the pixel points adjacent to the candidate non-lesion area to obtain a clustered area;

[0038] Obtaining the area ratio of the cluster area in the target medical image as the cluster area ratio;

[0039] If the area ratio difference between the cluster area ratio and the marked area ratio is less than the ratio threshold, the cluster area is determined as the target non-lesion area.

[0040] Optionally, the adjustment module clusters the pixels in the candidate non-lesion area and the pixels adjacent to the candidate non-lesion area according to the pixel values ​​of the pixels in the candidate non-lesion area and the pixel values ​​of the pixels adjacent to the candidate non-lesion area, and the method of obtaining the clustered area includes:

[0041] Obtaining a pixel difference between a pixel value of a pixel point adjacent to the candidate non-lesion area and a pixel value of a pixel point in the candidate non-lesion area;

[0042] Determine, from the pixel points adjacent to the candidate non-lesion area, a pixel point whose corresponding pixel difference is less than a pixel difference threshold value as a second target pixel point;

[0043] The region where the second target pixel point in the target medical image is located is merged with the candidate non-lesion region to obtain the cluster region.

[0044] Optionally, if the proportion of the candidate area is greater than the proportion of the annotated area, the adjustment module reduces the candidate non-lesion area according to the proportion of the annotated area, and the method of obtaining the target non-lesion area includes:

[0045] If the proportion of the candidate area is greater than the proportion of the marked area, then obtaining reduction processing parameters, wherein the reduction processing parameters include a shape of the reduction processing and a size of the reduction processing;

[0046] Iteratively shrink the candidate non-lesion region according to the shrinking processing parameters to obtain a candidate non-lesion region after shrinking processing;

[0047] Obtaining the area ratio of the candidate non-lesion area after the reduction processing in the target medical image as the area ratio after the reduction processing;

[0048] The candidate non-lesion area after the reduction processing, in which the difference between the area ratio after the reduction processing and the area ratio of the marked area is less than the ratio threshold, is determined as the target non-lesion area.

[0049] Optionally, the candidate region size information includes the candidate region size of the candidate non-lesion region; the annotated region size information includes the annotated region size of the non-lesion region in the target medical image;

[0050] If the candidate region size information does not match the annotated region size information, the adjustment module adjusts the candidate non-lesion region according to the annotated region size information to obtain the target non-lesion region of the target medical image, including:

[0051] Obtaining a region size difference between the candidate region size and the marked region size;

[0052] If the region size difference is greater than the size threshold, it is determined that the candidate region size information does not match the marked region size information;

[0053] The candidate non-lesion area is adjusted according to the size of the marked area to obtain the target non-lesion area.

[0054] Optionally, the determination module determines the region size information of the candidate non-lesion region, and the manner in which the candidate region size information is determined includes:

[0055] Acquire the area size of the candidate non-lesion area and the image size of the target medical image, and determine the ratio between the area size of the candidate non-lesion area and the image size of the target medical image as the candidate area ratio; or,

[0056] Obtaining the number of pixels in the candidate non-lesion area and the number of pixels in the target medical image, and determining the ratio between the number of pixels in the candidate non-lesion area and the number of pixels in the target medical image as the candidate area proportion;

[0057] The candidate area proportion is determined as the candidate area size information.

[0058] Optionally, the determination module determines the region size information of the candidate non-lesion region, and the manner in which the candidate region size information is determined includes:

[0059] Acquire the region size of the candidate non-lesion region to obtain the candidate region size;

[0060] The candidate region size is determined as the candidate region size information.

[0061] Optionally, the device further comprises:

[0062] a labeling module, used for labeling the target non-lesion area in the target medical image to obtain a labeled target medical image;

[0063] A prediction module, configured to predict the target medical image using an image segmentation model to obtain a predicted non-lesion area in the target medical image, and to mark the predicted non-lesion area in the target medical image to obtain a predicted target medical image;

[0064] The above adjustment module is also used to adjust the image segmentation model according to the labeled target medical image and the predicted target medical image to obtain a target medical image segmentation model.

[0065] Optionally, the adjustment module adjusts the image segmentation model according to the labeled target medical image and the predicted target medical image to obtain the target medical image segmentation model in a manner including:

[0066] Determining a prediction loss value of the image segmentation model according to the labeled target medical image and the predicted target medical image;

[0067] If the predicted loss value does not meet the convergence condition, the image segmentation model is adjusted according to the predicted loss value to obtain a target medical image segmentation model.

[0068] Optionally, the labeling module labels the target non-lesion area in the target medical image, and a method for obtaining the labeled target medical image includes:

[0069] Binarizing the target medical image according to the pixel points in the target non-lesion area;

[0070] The target medical image after the binarization process is determined as the labeled target medical image.

[0071] In one aspect, an embodiment of the present application provides a computer program product or a computer program, wherein the computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the following steps:

[0072] Acquire a medical image set; the medical image set includes a reference medical image, a target medical image to be identified, and annotated area size information of a non-lesion area of ​​the target medical image, the target medical image includes a lesion area and a non-lesion area, and the reference medical image includes a lesion area;

[0073] Performing difference recognition on the reference medical image and the target medical image to obtain a candidate non-lesion area in the target medical image;

[0074] Determining region size information of the candidate non-lesion region as candidate region size information;

[0075] If the candidate region size information does not match the annotated region size information, the candidate non-lesion region is adjusted according to the annotated region size information to obtain a target non-lesion region of the target medical image.

[0076] On one hand, the present application provides a computer device, including: a processor and a memory;

[0077] The processor is used to call the device control application stored in the memory to achieve:

[0078] Acquire a medical image set; the medical image set includes a reference medical image, a target medical image to be identified, and annotated area size information of a non-lesion area of ​​the target medical image, the target medical image includes a lesion area and a non-lesion area, and the reference medical image includes a lesion area;

[0079] Performing difference recognition on the reference medical image and the target medical image to obtain a candidate non-lesion area in the target medical image;

[0080] Determining region size information of the candidate non-lesion region as candidate region size information;

[0081] If the candidate region size information does not match the annotated region size information, the candidate non-lesion region is adjusted according to the annotated region size information to obtain a target non-lesion region of the target medical image.

[0082] In the present application, the computer device can perform difference identification on the target medical image and the reference medical image, obtain the candidate non-lesion area of ​​the target medical image, obtain the area size information of the candidate non-lesion area, and obtain the candidate area size information. If the candidate area size information does not match the annotated area size information of the target medical image, it indicates that there is a large difference between the identified candidate non-lesion area and the actual non-lesion area of ​​the target medical image, that is, the accuracy of the identified candidate non-lesion area is relatively low; therefore, the candidate non-lesion area can be adjusted according to the annotated area size information to obtain the target non-lesion area of ​​the target medical image. In other words, by automatically identifying the candidate non-lesion area in the target medical image based on the reference medical image and the target medical image, no manual intervention is required, and the accuracy and efficiency of identifying the non-lesion area can be improved; and the identified candidate non-lesion area can be adjusted according to the annotated area size information, which can further improve the accuracy of identifying the non-lesion area. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0084] Figure 1 It is a schematic diagram of the architecture of a medical image processing system provided by the present application;

[0085] Figure 2a This is a schematic diagram of a data interaction scenario provided by this application;

[0086] Figure 2b This is a schematic diagram of a data interaction scenario provided by this application;

[0087] Figure 3 It is a flowchart of a medical image processing method provided by the present application;

[0088] Figure 4 This is a schematic diagram of a scenario for obtaining a candidate non-lesion area provided by the present application;

[0089] Figure 5is a schematic diagram of a scenario for iteratively expanding a candidate non-lesion area provided by the present application;

[0090] Figure 6 It is a schematic diagram of a scenario for adjusting an image segmentation model provided by the present application;

[0091] Figure 7 is a schematic diagram of an image segmentation model provided by the present application;

[0092] Figure 8 is a schematic diagram of the structure of a medical image processing device provided in an embodiment of the present application;

[0093] Fig. 9 It is a structural diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0094] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0095] Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines so that machines have the functions of perception, reasoning and decision-making.

[0096] Artificial intelligence technology is a comprehensive discipline that covers a wide range of fields, including both hardware-level and software-level technologies. The basic technologies of artificial intelligence generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, large medical image processing technology, operation / interaction systems, mechatronics and other technologies. Artificial intelligence software technology mainly includes computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0097] Among them, Computer Vision (CV) is a science that studies how to make machines "see". To put it more specifically, it refers to machine vision that uses cameras and computers to replace human eyes to identify, track and measure targets, and further performs graphic processing to make computer processing into images that are more suitable for human eye observation or transmission to instruments for detection. As a scientific discipline, computer vision studies related theories and technologies, and attempts to establish an artificial intelligence system that can obtain information from images or multidimensional data. Computer vision technology usually includes image processing, image recognition, image semantic understanding, image retrieval, OCR, video processing, video semantic understanding, video content / behavior recognition, three-dimensional object reconstruction, 3D technology, virtual reality, augmented reality, simultaneous positioning and map construction, and other technologies, as well as common biometric recognition technologies such as face recognition and fingerprint recognition.

[0098] The medical image processing method provided in the embodiment of the present application mainly involves computer vision technology of artificial intelligence. Specifically, the non-lesion areas in the medical image are automatically identified by using computer vision technology, which can improve the accuracy of identifying the non-lesion areas.

[0099] First, a medical image processing system for implementing the medical image processing method of the present application is introduced. Figure 1 As shown, the medical image processing system includes a server and at least one terminal.

[0100] Among them, the server may refer to a device used to identify an area of ​​interest (such as a lesion area or a non-lesion area) in a medical image. This application takes the area of ​​interest as a non-lesion area as an example for explanation. The terminal may refer to a user-oriented front-end device. Specifically, the terminal may refer to a device used to obtain medical images. For example, the terminal may refer to a medical device that can scan a certain part of a human body or an animal to obtain a medical image; or, the terminal may refer to a non-medical device (such as a smart phone) that can obtain a medical image from a medical device.

[0101] Among them, the server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms. The terminal can be a medical device, a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, etc., but is not limited to this. Each terminal and server can be directly or indirectly connected via wired or wireless communication; at the same time, the number of terminals and the number of servers can be one or more, and this application does not limit this.

[0102] Among them, medical images can be obtained by scanning a certain part of the human body or animal, such as medical images including chest CT scan images, magnetic resonance imaging (MRI) images and other medical images. Medical images usually include lesion areas, non-lesion areas, lesion areas and non-lesion areas; lesion areas can refer to areas in medical images that reflect lesions, and non-lesion areas can refer to areas in medical images that reflect no lesions, such as the non-lesion area can refer to the area where immune cells are located in the medical image.

[0103] For further understanding, please refer to Figure 2a and Figure 2b , is a schematic diagram of a data interaction scenario provided in an embodiment of the present application, Figure 2a and Figure 2b In the example of the medical image of the breast, in the process of breast cancer treatment, it is necessary to obtain the non-lesion area in the medical image, and formulate a corresponding treatment plan for the breast cancer patient according to the non-lesion area. Therefore, the embodiment of the present application mainly relates to how to identify the non-lesion area from the medical image. Specifically, assuming that the terminal is a medical device, such as Figure 2aAs shown, the terminal can scan a user's breast to obtain a whole-view digital pathology slice 11 (Whole Slide Image, WSI) including the breast, and the whole-view digital pathology slice 11 (Whole Slide Image, WSI) includes a tumor area 12, and the tumor area 12 refers to an area used to reflect the occurrence of tumor lesions, and the tumor area also includes immune cells and lesion cells; the tumor area 12 can be divided to obtain multiple sub-areas, and each sub-area is used as a candidate medical image 13. Among them, each candidate medical image includes lesion cells, that is, each candidate medical image includes a lesion area; there are candidate medical images that also include immune cells, that is, there are candidate medical images that also include non-lesion areas. The size of the full-view digital pathology slice can be 50000mm*50000mm, 0.344μm / pixel (micrometer / pixel); the image size of each candidate medical image is the same, such as the size of the candidate medical image can be 2160mm*2160mm, 0.524μm / pixel. Of course, the size of the full-view digital pathology slice and the candidate medical image can also be other values, which are not limited in this application. Figure 2a In the present invention, the terminal can display a scoring interface about medical images, which includes candidate medical images and scoring options for candidate medical images. The user can score the candidate medical image through the scoring options to obtain a scoring result, which can include at least one of the area size of the non-lesion area in the candidate medical image (i.e., the size of the annotated area) and the area ratio of the non-lesion area in the candidate medical image in the candidate medical image (i.e., the annotated area ratio). Accordingly, the terminal can obtain the user's scoring result for the candidate medical image and use the scoring result as the annotated area size information of the candidate medical image. Further, the candidate medical image with an annotated area ratio of zero (or an annotated area size of zero) can be screened out from the candidate medical images as a reference medical image, that is, the reference medical image only includes the lesion area and does not include the non-lesion area; and the candidate medical image with an annotated area ratio of not zero (or annotated area size of not zero) can be screened out from the candidate medical images as a target medical image, that is, the target medical image includes the lesion area and the non-lesion area.

[0104] After the terminal obtains the target medical image, the reference medical image, and the annotated area size information of the target medical image, a medical image set can be generated based on the target medical image, the reference medical image, and the annotated area size information of the target medical image, that is, the target medical image, the reference medical image, and the annotated area size information of the target medical image are added to the medical image set, and the medical image set is sent to the server. Since the reference medical image only includes the lesion area, and the target medical image includes the lesion area and the non-lesion area, there is a similarity between the lesion area in the reference medical image and the lesion area in the target medical image, and there is a difference between the lesion area in the reference medical image and the non-lesion area in the target medical image. Therefore, if Figure 2b As shown, the server can perform difference recognition between the reference medical image and the target medical image to obtain the candidate non-lesion area of ​​the target medical image; that is, the reference medical image and the target medical image are compared to obtain the area in the target medical image that is different from the reference medical image, and the area with the difference is used as the candidate non-lesion area of ​​the target medical image. Figure 2b The target medical image includes multiple candidate non-lesion regions. Figure 2b The white area in the figure represents the candidate non-lesion area. Figure 2b Only three candidate non-lesion regions are illustrated. Usually, the target medical image includes three or more candidate regions, and the region sizes and shapes of the candidate non-lesion regions are different.

[0105] like Figure 2b As shown, after the server obtains the candidate non-lesion area, it can obtain the candidate area size information of the candidate non-lesion area; compare the candidate area size information with the annotated area size information. If the candidate area size information does not match the annotated area size information, it indicates that the difference between the candidate non-lesion area and the actual non-lesion area in the target medical image is relatively large. Therefore, the server can adjust the candidate non-lesion area according to the annotated area size information until the candidate non-lesion area is the same as the actual non-lesion area in the target medical image or the difference is not large, and obtain the adjusted candidate non-lesion area, and determine the adjusted candidate non-lesion area as the target non-lesion area. The target non-lesion area can be used to formulate a treatment plan for the patient, or it can be used to train the image segmentation model, that is, the target non-lesion area is used as the annotation data for training the image segmentation model. The image segmentation model can refer to a model for identifying non-lesion areas in medical images.

[0106] It should be noted that the above process of identifying the non-lesion area in the target medical image can be performed by the server or by the terminal. Of course, it can also be performed by the terminal and the server together, and this application does not limit this. Among them, the terminal performs the above process of identifying the non-lesion area in the target medical image, and the terminal and the server jointly perform the above process of identifying the non-lesion area in the target medical image. Both can refer to the server performing the above process of identifying the non-lesion area in the target medical image, and the repeated parts will not be repeated.

[0107] Based on the above description, please see Figure 3 , is a flowchart of a medical image processing method provided in an embodiment of the present application. The method can be executed by a computer device, which can refer to Figure 1 Alternatively, the computer device may refer to a terminal or server in Figure 1 The terminal and the server in the embodiment, that is, the method can be executed by the terminal and the server together. Figure 3 As shown, the medical image processing method may include the following steps S101 to S104.

[0108] S101. Acquire a medical image set; the medical image set includes a reference medical image, a target medical image to be identified, and annotated area size information of a non-lesion area of ​​the target medical image, the target medical image includes a lesion area and a non-lesion area, and the reference medical image includes a lesion area.

[0109] The computer device can obtain a medical image set; the medical image set includes a reference medical image, a target medical image to be identified, and annotated region size information, the target medical image includes a lesion region and a non-lesion region, and the reference medical image includes a lesion region; the annotated region size information includes at least one of the region size of the non-lesion region in the target medical image (i.e., the annotated region size) and the region ratio of the non-lesion region in the target medical image in the target medical image (i.e., the annotated region ratio). The annotated region size information may refer to the information obtained by annotating the non-lesion region in the target medical image (e.g., manually annotating), that is, the annotated region size information may be used as the standard region size information of the target medical image.

[0110] Optionally, the computer device may generate a task for obtaining the size information of the annotated region of the target medical image, publish the task to multiple users (such as multiple doctors), obtain the scores of the target medical image by multiple users, and obtain multiple scores corresponding to the target medical image. The scores corresponding to the target medical image are used to reflect the area proportion of the non-lesion area in the target medical image, or reflect the area size of the non-lesion area. The computer device may obtain the number of occurrences of each score corresponding to the target medical image, and determine the score with the largest number of occurrences as the annotated region size information of the target medical image; by determining the annotated region size information of the target medical image according to the number of occurrences of the scores corresponding to the target medical image, the accuracy of obtaining the annotated region size information may be improved. Optionally, the computer device may average the scores corresponding to the target medical image to obtain the averaged scores, and determine the averaged scores as the annotated region size information of the target medical image; by averaging the scores corresponding to the target medical image to determine the annotated region size information of the target medical image, the accuracy of obtaining the annotated region size information may be improved.

[0111] S102: Perform difference recognition on the reference medical image and the target medical image to obtain a candidate non-lesion region in the target medical image.

[0112] Since the target medical image generally includes a large number of non-lesion areas, it is time-consuming to manually annotate the non-lesion areas of the target medical image, and the manual annotation method is relatively subjective, resulting in a relatively low accuracy in obtaining the non-lesion areas. Therefore, the computer device can perform differential recognition on the reference medical image and the target medical image to obtain the candidate non-lesion area in the target medical image, that is, automatically recognize the non-lesion area in the target medical image based on the reference medical image and the target medical image, which can improve the accuracy and efficiency of identifying the non-lesion area.

[0113] S103: Determine the region size information of the candidate non-lesion region as the candidate region size information.

[0114] S104: If the size information of the candidate region does not match the size information of the annotated region, the candidate non-lesion region is adjusted according to the size information of the annotated region to obtain a target non-lesion region of the target medical image.

[0115] In steps S103 and S104, after the candidate non-lesion area of ​​the target medical image is obtained, it is usually identified that there is a difference between the candidate non-lesion area and the actual non-lesion area of ​​the target medical image, that is, there is a situation where the accuracy of the identified candidate non-lesion area is relatively low; therefore, the computer device needs to adjust the candidate non-lesion area. Specifically, the computer device can determine the area size information of the candidate non-lesion area to obtain the candidate area size information; the candidate area size information includes at least one of the area size of the candidate non-lesion area in the target medical image (i.e., the candidate area size) and the area proportion of the candidate non-lesion area in the target medical image (i.e., the candidate area proportion). For example, the area proportion of the candidate non-lesion area in the target medical image can be determined based on the number of pixels in the candidate non-lesion area and the number of pixels in the target medical image to obtain the candidate area proportion; or, the area proportion of the candidate non-lesion area in the target medical image can be determined based on the size of the candidate non-lesion area and the size of the target medical image to obtain the candidate area proportion; further, the candidate area size information can be determined based on the candidate area proportion.

[0116] Alternatively, the candidate region size information can be determined based on the region size of the candidate non-lesion region. If the candidate region size information does not match the annotated region size information, it indicates that there is a large difference between the identified candidate non-lesion region and the actual non-lesion region of the target medical image, that is, the accuracy of the identified candidate non-lesion region is relatively low; therefore, the candidate non-lesion region can be adjusted according to the annotated region size information to obtain the target non-lesion region of the target medical image. If the candidate region size information does not match the annotated region size information, it indicates that there is no difference between the identified candidate non-lesion region and the actual non-lesion region of the target medical image, or it indicates that there is a small difference between the identified candidate non-lesion region and the actual non-lesion region of the target medical image, that is, the accuracy of the identified candidate non-lesion region is relatively high; therefore, the candidate non-lesion region can be determined as the target non-lesion region of the target medical image.

[0117] In the present application, the computer device can perform difference identification on the target medical image and the reference medical image, obtain the candidate non-lesion area of ​​the target medical image, obtain the area size information of the candidate non-lesion area, and obtain the candidate area size information. If the candidate area size information does not match the annotated area size information of the target medical image, it indicates that there is a large difference between the identified candidate non-lesion area and the actual non-lesion area of ​​the target medical image, that is, the accuracy of the identified candidate non-lesion area is relatively low; therefore, the candidate non-lesion area can be adjusted according to the annotated area size information to obtain the target non-lesion area of ​​the target medical image. In other words, by automatically identifying the candidate non-lesion area in the target medical image based on the reference medical image and the target medical image, no manual intervention is required, and the accuracy and efficiency of identifying the non-lesion area can be improved; and the identified candidate non-lesion area can be adjusted according to the annotated area size information, which can further improve the accuracy of identifying the non-lesion area.

[0118] Optionally, the above step S102 may include the following steps s11 to s13.

[0119] s11. Convert the reference medical image into an image in a target color space to obtain a converted reference medical image, and convert the target medical image into an image in the target color space to obtain a converted target medical image.

[0120] s12. Obtaining brightness information of pixel points in the converted reference medical image and brightness information of pixel points in the converted target medical image.

[0121] s13. Based on the brightness information of the pixels in the converted reference medical image and the brightness information of the pixels in the converted target medical image, difference recognition is performed on the reference medical image and the target medical image to obtain a candidate non-lesion area in the target medical image.

[0122] In steps s11 to s13, since the corresponding brightness information of the same pixel in the medical image is inconsistent in different color spaces, the accuracy of identifying the non-lesion area is avoided due to the different color spaces of the medical image. The computer device can convert the reference medical image into an image in the target color space to obtain a converted reference medical image, and convert the target medical image into an image in the target color space to obtain a converted target medical image. The target color space may refer to a HED color space, an RGB color space, or other color space, which is not limited in this application. Further, the brightness information of the pixel points in the converted reference medical image and the brightness information of the pixel points in the converted target medical image can be obtained, and the brightness information may refer to a brightness value or a brightness level; the brightness information of the pixel points in the converted reference medical image is compared with the brightness information of the pixel points in the converted target medical image to perform difference identification on the reference medical image and the target medical image, and obtain a candidate non-lesion area in the target medical image. By converting the target medical image and the reference medical image, that is, converting the target medical image and the reference medical image to a unified color space, the accuracy of identifying non-lesion areas due to the difference in color space of medical images can be avoided, and the accuracy of identifying candidate non-lesion areas can be improved.

[0123] For example, Figure 4 As shown, the computer device can convert the target medical image into an image in the HED color space to obtain a converted target medical image, decompose the dab channel image from the converted target medical image, and obtain a target dab channel image. Similarly, the computer device can convert the reference medical image into an image in the HED color space to obtain a converted reference medical image, decompose the dab channel image from the converted reference medical image, and obtain a reference dab channel image ( Figure 4 Then a histogram is used to describe the brightness levels of pixels in the target dab channel image and the reference dab channel image. Figure 4 The horizontal axis represents the pixel value corresponding to the pixel point of the medical image, and the vertical axis represents the brightness level of the pixel point in the medical image. Figure 4 The dashed line in the middle represents the brightness level of the pixel points of the target dab channel image (i.e., the converted target medical image), and the solid line represents the brightness level of the pixel points of the reference dab channel image (i.e., the converted reference medical image). Figure 4It can be seen that the brightness level corresponding to the pixel point with a pixel value in [165,195] in the target dab channel image is significantly different from the brightness level corresponding to the pixel point with a pixel value in [165,195] in the reference dab channel image, and the brightness level difference between the pixels at other pixel values ​​is relatively small. Therefore, the pixel point with a pixel value in [165,195] in the converted target medical image can be used as the first target pixel point, and the area where the first target pixel point is located in the target medical image can be used as the candidate non-lesion area of ​​the target medical image.

[0124] Optionally, the above step s13 may include the following steps s21 to s23.

[0125] s21. Obtain the difference between the brightness information of the pixel points in the converted reference medical image and the brightness information of the corresponding pixel points in the converted target medical image.

[0126] s22. Identify, from the converted target medical image, pixels whose brightness information corresponding to a difference greater than a difference threshold as first target pixels.

[0127] s23. Determine the region where the first target pixel point is located in the target medical image as a candidate non-lesion region in the target medical image.

[0128] In steps s21 to s23, the computer device may obtain the difference between the brightness information of the pixel in the converted reference medical image and the brightness information of the corresponding pixel in the converted target medical image, that is, the difference may refer to the difference between the brightness value of the pixel in the converted reference medical image and the brightness value of the corresponding pixel in the converted target medical image, that is, the difference may refer to the difference between the brightness values ​​of two pixels located at the same position in the converted reference medical image and the converted target medical image. If both pixels are pixels in the lesion area, the difference between the brightness information of the two pixels is relatively small or the same; if one pixel is a pixel in the lesion area and the other pixel is a pixel in the non-lesion area, the difference between the brightness information of the two pixels is relatively large. Therefore, the computer device may identify the pixel whose brightness information corresponding difference is greater than the difference threshold from the converted target medical image as the first target pixel; and determine the area where the first target pixel is located in the target medical image as the candidate non-lesion area in the target medical image. By determining the candidate non-lesion area based on the difference between the brightness information of the pixel points in the converted reference medical image and the brightness information of the corresponding pixel points in the converted target medical image, the accuracy of identifying the non-lesion area can be improved.

[0129] Optionally, step S103 may include: obtaining the region size of the candidate non-lesion region and the image size of the target medical image, and determining the ratio between the region size of the candidate non-lesion region and the image size of the target medical image as the candidate region proportion. Alternatively, obtaining the number of pixels in the candidate non-lesion region and the number of pixels in the target medical image, and determining the ratio between the number of pixels in the candidate non-lesion region and the number of pixels in the target medical image as the candidate region proportion.

[0130] The computer device can obtain the region size of the candidate non-lesion region and the image size of the target medical image. The region size can refer to the area of ​​the candidate non-lesion region, that is, the region size is the cumulative sum of the areas of all candidate non-lesion regions in the target medical image, and the image size can refer to the area of ​​the target medical image. The ratio between the region size of the candidate non-lesion region and the image size of the target medical image is determined as the candidate region ratio. For example, the target medical image is a breast medical image of a breast cancer patient, and the candidate region ratio can be expressed by the following formula (1), where the IC value in formula (1) represents the candidate region ratio, the IC (immune cell) region area represents the area of ​​the candidate non-lesion region, and the tumor region area refers to the area of ​​the target medical image. Among them, since there is a certain error in the identified candidate non-lesion region, that is, the candidate non-lesion region includes stained immune cells (i.e., non-lesion region) and non-immune cell regions (lesion regions) surrounding the immune cells; therefore, the IC region area can be expressed as: IC region area = stained immune cell area + area of ​​the region surrounding the cells.

[0131]

[0132] Alternatively, the computer device may obtain the cumulative sum of the number of pixels in each candidate non-lesion area in the target medical image to obtain the number of pixels in the candidate non-lesion area (i.e., the total number of pixels in the candidate non-lesion area); and obtain the number of pixels in the target medical image, and determine the ratio between the number of pixels in the candidate non-lesion area and the number of pixels in the target medical image as the candidate area ratio. For example, if the target medical image is a breast medical image of a breast cancer patient, the candidate area ratio may be represented by the following formula (2), where the IC value represents the candidate area ratio, the number of pixels in the IC area represents the number of pixels in the candidate non-lesion area, and the number of pixels in the tumor area refers to the number of pixels in the target medical image.

[0133]

[0134] Further, after the candidate area ratio of the candidate non-lesion area is obtained, the candidate area ratio may be determined as the candidate area size information of the candidate non-lesion area.

[0135] Optionally, the candidate region size information includes the candidate region ratio of the candidate non-lesion region in the target medical image; the annotated region size information includes the annotated region ratio of the non-lesion region in the target medical image; step S104 may include the following steps s31 to s33.

[0136] s31. Obtain the area ratio difference between the candidate area ratio and the annotated area ratio.

[0137] s32. If the area proportion difference is greater than the proportion threshold, it is determined that the size information of the candidate area does not match the size information of the marked area.

[0138] s33. Adjust the candidate non-lesion area according to the proportion of the marked area to obtain the target non-lesion area.

[0139] In steps s31 to s33, the computer device may obtain the area ratio difference between the candidate area ratio and the annotated area ratio. If the area ratio difference is less than or equal to the area ratio threshold, it indicates that the difference between the candidate non-lesion area and the actual area of ​​the target medical image is relatively small or the same, and then it is determined that the candidate area size information matches the annotated area size information. If the area ratio difference is greater than the ratio threshold, it indicates that the difference between the candidate non-lesion area and the actual area of ​​the target medical image is relatively large, and then it is determined that the candidate area size information does not match the annotated area size information. Therefore, when the candidate area size information does not match the annotated area size information, it is necessary to reduce the candidate non-lesion area according to the annotated area ratio to obtain the target non-lesion area, even if the target non-lesion area is the same as or closer to the actual non-lesion area in the target medical image. By adjusting the candidate non-lesion area according to the annotated area ratio when the difference between the candidate non-lesion area and the actual area of ​​the target medical image is relatively large, the accuracy of identifying the non-lesion area can be improved.

[0140] Optionally, step s33 may include the following steps s41 to s42.

[0141] s41. If the proportion of the candidate region is less than the proportion of the annotated region, the candidate non-lesion region is expanded according to the proportion of the annotated region to obtain the target non-lesion region.

[0142] s42. If the proportion of the candidate area is greater than the proportion of the annotated area, the candidate non-lesion area is reduced according to the proportion of the annotated area to obtain the target non-lesion area.

[0143] In steps s41 to s42, when the size information of the candidate region does not match the size information of the annotated region, the computer device may reduce the candidate non-lesion region according to the proportion of the annotated region to obtain the target non-lesion region. Specifically, if the proportion of the candidate region is less than the proportion of the annotated region, it indicates that the candidate non-lesion region is smaller than the actual non-lesion region in the target medical image, then the candidate non-lesion region is expanded according to the proportion of the annotated region to obtain the target non-lesion region, that is, the candidate non-lesion region is expanded according to the proportion of the annotated region to obtain the target non-lesion region. If the proportion of the candidate region is greater than the proportion of the annotated region, it indicates that the candidate non-lesion region is larger than the actual non-lesion region in the target medical image; then the candidate non-lesion region may be reduced according to the proportion of the annotated region to obtain the target non-lesion region. By expanding or reducing the candidate non-lesion region according to the proportion of the annotated region, the accuracy of identifying the non-lesion region can be improved.

[0144] Optionally, step s41 may include the following steps s51 to s54.

[0145] s51. If the proportion of the candidate area is smaller than that of the marked area, obtain an expansion parameter, which includes an expansion shape and an expansion size.

[0146] s52. Iteratively expand the candidate non-lesion region according to the expansion parameter to obtain an expanded candidate non-lesion region.

[0147] s53. Obtain the area ratio of the expanded candidate non-lesion area in the target medical image as the expanded area ratio.

[0148] s54. The candidate non-lesion region after expansion whose area ratio difference between the expanded area ratio and the marked area ratio is less than the ratio threshold is determined as the target non-lesion region.

[0149] In steps s51 to s54, if the proportion of the candidate area is less than the proportion of the annotated area, it indicates that the candidate non-lesion area is smaller than the actual non-lesion area in the target medical image. Therefore, the computer device can obtain an expansion parameter, which includes an expansion shape and an expansion size; the expansion shape includes at least one of a rectangle, a trapezoid, a rhombus, etc., and the expansion size refers to the size of each expansion. Further, the candidate non-lesion area can be iteratively expanded according to the expansion parameter to obtain an expanded candidate non-lesion area, that is, the candidate non-lesion area can be expanded multiple times according to the expansion parameter to obtain multiple expanded candidate non-lesion areas, such as firstly expanding the candidate non-lesion area according to the expansion parameter to obtain a first expanded candidate non-lesion area; then expanding the first expanded candidate non-lesion area according to the expansion parameter to obtain a second expanded candidate non-lesion area, and so on. After obtaining the expanded candidate non-lesion area, the computer device can obtain the area ratio of the expanded candidate non-lesion area in the target medical image as the expanded area ratio; that is, the ratio between the area size of the expanded candidate non-lesion area and the image size of the target medical image can be obtained to obtain the expanded area ratio; or, the ratio between the number of pixels in the expanded candidate non-lesion area and the number of pixels in the target medical image can be obtained to obtain the expanded area ratio. After obtaining the expanded area ratio, the area ratio difference between the expanded area ratio and the annotated area ratio can be obtained, and the expanded candidate non-lesion area whose area ratio difference between the expanded area ratio and the annotated area ratio is less than the ratio threshold is determined as the target non-lesion area. By iteratively expanding the candidate non-lesion area and adjusting the candidate non-lesion area, that is, by fine-tuning the candidate non-lesion area multiple times, the accuracy of obtaining the non-lesion area can be improved.

[0150] For example, if the candidate area accounts for 0.6% and the annotated area accounts for 2.0%, the candidate area accounts for less than the annotated area. Figure 5 As shown, the computer device can adjust the candidate non-lesion area according to the extended parameters. For the convenience of distinction, before adjusting the candidate non-lesion area, the target medical image can be binarized according to the candidate non-lesion area, such as using black to represent the lesion area in the target medical image and using white to represent the non-lesion area in the target medical image, to obtain the processed target medical image 21. Assume that the iteration parameters are as follows:

[0151] kernel=disk(1).astype(np.uint8)

[0152] mask_ic=cv2.dilate(mask_init,kernel,iterations=ITERATION)

[0153] Among them, ITERATION represents the number of iterations, that is, the number of iterations can be 5, of course, it can be other values, mask_init represents the result of the last iteration, and mask_ic represents the result of this iteration, that is, the result of this iteration is obtained by iterating based on the result of the last iteration according to the expansion parameters. The iteration parameters also include the expansion parameters, that is, the expansion shape is a trapezoid, and the expansion size is the size of the area corresponding to 8 pixels. Figure 5 As shown, the candidate non-lesion area in the processed target medical image 21 is expanded according to the expansion parameters to obtain the target medical image 22 after the first expansion; further, the candidate non-lesion area in the target medical image 22 after the first expansion is expanded according to the expansion parameters to obtain the target medical image 23 after the second expansion; and so on, the target medical image 24 after the third expansion, the target medical image 25 after the fourth expansion, and the target medical image 26 after the fifth expansion can be obtained. Figure 5 It can be seen that the shape of the candidate non-lesion area in the expanded target medical image is close to a trapezoid, and the area size of the expanded candidate non-lesion area continues to increase, that is, the proportion of the expanded area continues to increase. Figure 5 After the first iteration, the proportion of the expanded area (that is, the proportion of the expanded candidate non-lesion area in the target medical image) is 1.1%, and after the second iteration, the proportion of the expanded area is 1.8%. After the third iteration, the proportion of the expanded area is 2.5%, after the fourth iteration, the proportion of the expanded area is 3.4%, and after the fifth iteration, the proportion of the expanded area is 4.3%. Since the proportion of the expanded area of ​​1.8% is closer to the proportion of the annotated area of ​​2%, the expanded candidate non-lesion area obtained after the second iteration can be used as the target non-lesion area.

[0154] It should be noted that, in the process of adjusting the candidate non-lesion region, the image size of the target medical image remains unchanged, and only the region size and shape of the candidate non-lesion region change, or only the region size of the candidate non-lesion region changes.

[0155] Optionally, step s41 may include the following steps s61 to s64.

[0156] s61. If the proportion of the candidate area is smaller than that of the annotated area, the pixel values ​​of the pixels adjacent to the candidate non-lesion area in the target medical image and the pixel values ​​of the pixels in the candidate non-lesion area are obtained.

[0157] s62. Cluster the pixels in the candidate non-lesion area and the pixels adjacent to the candidate non-lesion area according to the pixel values ​​of the pixels in the candidate non-lesion area and the pixel values ​​of the pixels adjacent to the candidate non-lesion area to obtain a clustered area.

[0158] s63, obtaining the area ratio of the cluster area in the target medical image as the cluster area ratio;

[0159] s64. If the difference between the proportion of the clustered area and the proportion of the marked area is less than the proportion threshold, the clustered area is determined as the target non-lesion area.

[0160] In steps s61 to s64, if the proportion of the candidate area is less than the proportion of the annotated area, it indicates that the candidate non-lesion area is smaller than the actual non-lesion area in the target medical image. Therefore, the computer device can obtain the pixel values ​​of the pixels adjacent to the candidate non-lesion area in the target medical image, as well as the pixel values ​​of the pixels in the candidate non-lesion area. Further, according to the pixel values ​​of the pixels in the candidate non-lesion area and the pixel values ​​of the pixels adjacent to the candidate non-lesion area, the pixels in the candidate non-lesion area and the pixels adjacent to the candidate non-lesion area can be clustered to obtain a clustering area; that is, the pixel values ​​of the adjacent pixels and the pixel values ​​of the pixels in the candidate non-lesion area whose difference is less than the pixel difference threshold are clustered into one category with the pixels in the candidate non-lesion area, and the pixels of the same category as the pixels in the candidate non-lesion area and the area where the pixels in the candidate non-lesion area are located are determined as clustering areas. Then, the area proportion of the cluster area in the target medical image can be obtained as the cluster area proportion, that is, the proportion between the area size of the cluster area and the image size of the target medical image can be obtained to obtain the cluster area proportion; or, the proportion between the number of pixels in the cluster area and the number of pixels in the target medical image can be obtained to obtain the cluster area proportion. After obtaining the cluster area proportion, the area proportion difference between the cluster area proportion and the annotated area proportion can be obtained; if the area proportion difference between the cluster area proportion and the annotated area proportion is less than the proportion threshold, the cluster area is determined as the target non-lesion area. If the area proportion difference between the cluster area proportion and the annotated area proportion is greater than or equal to the proportion threshold, the pixels in the candidate non-lesion area and the pixels adjacent to the candidate non-lesion area are re-clustered to obtain the cluster area.

[0161] Optionally, step s62 may include the following steps s71 to s73.

[0162] s71. Obtain the pixel difference between the pixel value of the pixel point adjacent to the candidate non-lesion area and the pixel value of the pixel point in the candidate non-lesion area.

[0163] s72. Determine, from the pixel points adjacent to the candidate non-lesion area, a pixel point whose corresponding pixel difference is less than a pixel difference threshold value as a second target pixel point.

[0164] s73. Merge the region where the second target pixel point in the target medical image is located with the candidate non-lesion region to obtain the cluster region.

[0165] In steps s71 to s73, the computer device can obtain the pixel difference between the pixel value of the pixel point adjacent to the candidate non-lesion area and the pixel value of the pixel point in the candidate non-lesion area. If the pixel value between two pixel points is less than the pixel difference threshold, it indicates that the two pixel points have similarity; therefore, the pixel point whose corresponding pixel difference is less than the pixel difference threshold can be determined from the pixel points adjacent to the candidate non-lesion area as the second target pixel point, and the area where the second target pixel point is located in the target medical image is merged with the candidate non-lesion area to obtain the cluster area.

[0166] Optionally, the above step s42 may include the following steps s81 to s84.

[0167] s81. If the proportion of the candidate area is greater than that of the marked area, obtain the reduction processing parameters, which include the shape of the reduction processing and the size of the reduction processing.

[0168] s83. Perform iterative shrinking processing on the candidate non-lesion region according to the shrinking processing parameters to obtain the candidate non-lesion region after shrinking processing.

[0169] s84. Obtain the area ratio of the candidate non-lesion area after the reduction processing in the target medical image as the area ratio after the reduction processing.

[0170] s85. The candidate non-lesion region after the reduction processing whose area ratio difference between the area ratio after the reduction processing and the area ratio of the marked region is less than the ratio threshold is determined as the target non-lesion region.

[0171] In steps s81 to s84, if the proportion of the candidate area is greater than the proportion of the annotated area, it indicates that the candidate non-lesion area is larger than the actual non-lesion area in the target medical image. Therefore, the computer device can obtain the reduction processing parameters, which include the shape of the reduction processing and the size of the reduction processing. The size of the reduction processing can refer to the size of the reduction processing. Further, the candidate non-lesion area can be iteratively reduced according to the reduction processing parameters to obtain the candidate non-lesion area after the reduction processing; that is, the candidate non-lesion area can be reduced multiple times according to the reduction processing parameters to obtain multiple candidate non-lesion areas after the reduction processing, such as firstly, the candidate non-lesion area can be reduced according to the reduction processing parameters to obtain the first candidate non-lesion area after the reduction processing; then, the candidate non-lesion area after the first reduction processing is reduced according to the reduction processing parameters to obtain the second candidate non-lesion area after the reduction processing, and so on. The area proportion of the candidate non-lesion area after the reduction processing in the target medical image is obtained as the area proportion after the reduction processing. Then, the area ratio of the candidate non-lesion area after the reduction processing in the target medical image can be obtained as the area ratio after the reduction processing; that is, the ratio between the area size of the candidate non-lesion area after the reduction processing and the image size of the target medical image can be obtained to obtain the area ratio after the reduction processing; or, the ratio between the number of pixels in the candidate non-lesion area after the reduction processing and the number of pixels in the target medical image can be obtained to obtain the area ratio after the reduction processing. After obtaining the area ratio after the reduction processing, the area ratio difference between the area ratio after the reduction processing and the annotated area ratio can be obtained; the candidate non-lesion area after the reduction processing whose area ratio difference between the area ratio after the reduction processing and the annotated area ratio is less than the ratio threshold is determined as the target non-lesion area. By iteratively reducing the candidate non-lesion area according to the reduction processing parameters, the candidate non-lesion area is adjusted, that is, by fine-tuning the candidate non-lesion area multiple times, the accuracy of obtaining the non-lesion area can be improved.

[0172] Optionally, the above step s42 may include: if the proportion of the candidate area is greater than the proportion of the annotated area, the computer device may cluster the pixels of the candidate non-lesion area according to the pixel values ​​of the pixels in the candidate non-lesion area, and use the area where the pixels of the same class are located as the candidate non-lesion area after the reduction process. That is, the pixel difference between the pixel values ​​of each pixel in the candidate non-lesion area is obtained, and the area where the pixels whose pixel difference is greater than the pixel difference threshold are removed from the candidate non-lesion area to obtain the candidate non-lesion area after the reduction process. Further, the area proportion of the candidate non-lesion area after the reduction process in the target medical image can be obtained as the area proportion after the reduction process. The candidate non-lesion area after the reduction process whose area proportion difference between the area proportion after the reduction process and the annotated area proportion is less than the proportion threshold is determined as the target non-lesion area.

[0173] Optionally, step S103 may include: acquiring a region size of the candidate non-lesion region to obtain a candidate region size, and determining the candidate region size as the candidate region size information.

[0174] The computer device may acquire the region size of the candidate non-lesion region to obtain the candidate region size, that is, the candidate region size may refer to the area of ​​the candidate non-lesion region; the candidate region size may be determined as candidate region size information.

[0175] Optionally, the candidate region size information includes the candidate region size of the candidate non-lesion region; the annotated region size information includes the annotated region size of the non-lesion region in the target medical image; the above-mentioned step S104 may include the following steps s91 to s93.

[0176] s91. Obtain a region size difference between the candidate region size and the marked region size.

[0177] s92. If the region size difference is greater than the size threshold, it is determined that the candidate region size information does not match the marked region size information.

[0178] s93. Adjust the candidate non-lesion area according to the size of the marked area to obtain the target non-lesion area.

[0179] In steps s91 to s93, the computer device may obtain the region size difference between the candidate region size and the annotated region size. If the region size difference is less than or equal to the size threshold, indicating that the difference between the candidate non-lesion region and the actual non-lesion region of the target medical image is relatively small, then the candidate region size information is determined to match the annotated region size information, and the candidate non-lesion region may be determined as the target non-lesion region of the target medical image. If the region size difference is greater than the size threshold, indicating that the difference between the candidate non-lesion region and the actual non-lesion region of the target medical image is relatively large, then the candidate region size information is determined to not match the annotated region size information, and the candidate non-lesion region is adjusted according to the annotated region size to obtain the target non-lesion region.

[0180] Optionally, the above-mentioned adjusting the candidate non-lesion area according to the size of the annotated area to obtain the target non-lesion area includes: if the size of the candidate area is smaller than the size of the annotated area, expanding the candidate non-lesion area according to the size of the annotated area to obtain the target non-lesion area. If the size of the candidate area is larger than the size of the annotated area, shrinking the candidate non-lesion area according to the size of the annotated area to obtain the target non-lesion area.

[0181] Among them, the computer device expands the candidate non-lesion area according to the size of the marked area to obtain the implementation method of the target non-lesion area, which can refer to the above-mentioned expansion method according to the proportion of the marked area to obtain the target non-lesion area; the computer device shrinks the candidate non-lesion area according to the size of the marked area to obtain the implementation method of the target non-lesion area, which can refer to the above-mentioned shrinking method according to the proportion of the marked area to obtain the target non-lesion area, and the repeated parts will be omitted.

[0182] Optionally, the method may further include the following steps s111 to s113.

[0183] s111. Annotate the target non-lesion area in the target medical image to obtain an annotated target medical image.

[0184] s112. Use an image segmentation model to predict the target medical image to obtain a predicted non-lesion area in the target medical image, and mark the predicted non-lesion area in the target medical image to obtain a predicted target medical image.

[0185] s113. Adjust the image segmentation model according to the labeled target medical image and the predicted target medical image to obtain a target medical image segmentation model.

[0186] In steps s111 to s113, if Figure 6 As shown, the target non-lesion area of ​​the target medical image can be used to train the image segmentation model; specifically, the computer device can mark the target non-lesion area in the target medical image to obtain a marked target medical image (i.e., a binarized target medical image); that is, mark the target non-lesion area in the target medical image to distinguish the target non-lesion area from the lesion area in the target medical image to obtain a marked target medical image. Further, an image segmentation model can be used to predict the target medical image to obtain a predicted non-lesion area in the target medical image, and the predicted non-lesion area is marked in the target medical image to obtain a predicted target medical image; the image segmentation model may refer to a model used to identify non-lesion areas in medical images. From Figure 6 It can be seen that after training the image segmentation model with the labeled target medical image, the edges of the non-lesion areas in the predicted target medical image output by the image segmentation model are smoother and more natural, which can improve the accuracy of identifying the non-lesion areas.

[0187] Among them, Figure 7 As shown. The image segmentation model may refer to a FC-DenseNet model (image semantic segmentation model), which may include a downsampling path, an upsampling path, and residual connections skip connections; skip connections help the upsampling path restore spatial detail information by reusing feature maps. Among them, the downsampling path includes an input layer, a convolutional layer, a fully connected layer, an association layer, and a downward transition layer; the upsampling path includes an output layer, a convolutional layer, a fully connected layer, an association layer, and an upward transition layer. Optionally, the image segmentation model may also be other models, such as including but not limited to: a neural network (Fully Convolutional Network, FCN), a semantic segmentation network (SegNet), PSPNet, etc.

[0188] Optionally, the above step s113 may include the following steps s121 to s122.

[0189] s121. Determine a prediction loss value of the image segmentation model according to the labeled target medical image and the predicted target medical image.

[0190] s123. If the predicted loss value does not meet the convergence condition, the image segmentation model is adjusted according to the predicted loss value to obtain the target medical image segmentation model.

[0191] In steps s121 to s122, the computer device may use the annotated target medical image as the annotated data for training the image segmentation model, and determine the prediction loss value of the image segmentation model according to the annotated target medical image and the predicted target medical image. The prediction loss value is used to reflect the accuracy of the image segmentation model in predicting the non-lesion area of ​​the medical image, that is, the larger the prediction loss value, the lower the accuracy corresponding to the image segmentation model; the smaller the prediction loss value, the higher the accuracy corresponding to the image segmentation model. If the prediction loss value meets the convergence condition, it indicates that the image segmentation model has a relatively high accuracy in predicting the non-lesion area of ​​the medical image, and therefore, the image segmentation model can be used as the target medical image segmentation model. If the prediction loss value does not meet the convergence condition, it indicates that the image segmentation model has a relatively low accuracy in predicting the non-lesion area of ​​the medical image, and therefore, the image segmentation model can be adjusted according to the prediction loss value to obtain the target medical image segmentation model.

[0192] Optionally, the above step s111 may include: performing binarization processing on the target medical image according to the pixel points in the target non-lesion area; and determining the target medical image after the binarization processing as the labeled target medical image.

[0193] The computer device can perform binarization processing on the target medical image according to the pixel points in the target non-lesion area, that is, use a first color to mark the pixel points in the target non-lesion area of ​​the target medical image, and use a second color to mark the pixel points in the lesion area of ​​the target medical image; the first color is different from the second color, such as the first color can be white and the second color is black. Further, the target medical image after binarization processing can be determined as the labeled target medical image; by binarizing the target non-lesion area and the lesion area in the target medical image, it is helpful to distinguish the target non-lesion area and the lesion area in the target medical image.

[0194] See also Figure 8 , is a schematic diagram of the structure of a medical image processing device provided in an embodiment of the present application. The above-mentioned medical image processing device may be a computer program (including program code) running in a computer device, for example, the medical image processing device is an application software; the device may be used to execute the corresponding steps in the method provided in an embodiment of the present application. Figure 8 As shown, the medical image processing apparatus may include: an acquisition module 801 , a recognition module 802 , a determination module 803 , an adjustment module 804 , a labeling module 805 and a prediction module 806 .

[0195] An acquisition module 801 is used to acquire a medical image set; the medical image set includes a reference medical image, a target medical image to be identified, and annotated area size information of a non-lesion area of ​​the target medical image, the target medical image includes a lesion area and a non-lesion area, and the reference medical image includes a lesion area;

[0196] An identification module 802 is used to perform difference identification on the reference medical image and the target medical image to obtain a candidate non-lesion area in the target medical image;

[0197] A determination module 803 is used to determine the region size information of the candidate non-lesion region as the candidate region size information;

[0198] The adjustment module 804 is configured to adjust the candidate non-lesion region according to the annotated region size information to obtain a target non-lesion region of the target medical image if the candidate region size information does not match the annotated region size information.

[0199] Optionally, the recognition module 802 performs difference recognition on the reference medical image and the target medical image to obtain a candidate non-lesion area in the target medical image in a manner including:

[0200] Converting the reference medical image into an image in a target color space to obtain a converted reference medical image, and converting the target medical image into an image in the target color space to obtain a converted target medical image;

[0201] Acquiring brightness information of pixels in the converted reference medical image and brightness information of pixels in the converted target medical image;

[0202] Based on the brightness information of the pixels in the converted reference medical image and the brightness information of the pixels in the converted target medical image, the reference medical image and the target medical image are differentially identified to obtain a candidate non-lesion area in the target medical image.

[0203] Optionally, the recognition model performs difference recognition on the reference medical image and the target medical image according to the brightness information of the pixels in the converted reference medical image and the brightness information of the pixels in the converted target medical image, and a method for obtaining the candidate non-lesion area in the target medical image includes:

[0204] Acquire the difference between the brightness information of the pixel points in the converted reference medical image and the brightness information of the corresponding pixel points in the converted target medical image;

[0205] Identify, from the converted target medical image, a pixel whose brightness information corresponding difference is greater than a difference threshold as a first target pixel;

[0206] The region where the first target pixel point is located in the target medical image is determined as a candidate non-lesion region in the target medical image.

[0207] Optionally, the candidate region size information includes the candidate region ratio of the candidate non-lesion region in the target medical image; the annotated region size information includes the annotated region ratio of the non-lesion region in the target medical image in the target medical image;

[0208] If the candidate region size information does not match the annotated region size information, the adjustment module 804 adjusts the candidate non-lesion region according to the annotated region size information to obtain the target non-lesion region of the target medical image, including:

[0209] Obtaining a difference in area ratio between the candidate area ratio and the marked area ratio;

[0210] If the area proportion difference is greater than the proportion threshold, it is determined that the candidate area size information does not match the marked area size information;

[0211] The candidate non-lesion area is adjusted according to the proportion of the marked area to obtain the target non-lesion area.

[0212] Optionally, the adjusting module 804 adjusts the candidate non-lesion area according to the proportion of the annotated area to obtain the target non-lesion area in the following manner:

[0213] If the proportion of the candidate area is less than the proportion of the annotated area, the candidate non-lesion area is expanded according to the proportion of the annotated area to obtain the target non-lesion area;

[0214] If the proportion of the candidate area is greater than the proportion of the annotated area, the candidate non-lesion area is reduced according to the proportion of the annotated area to obtain the target non-lesion area.

[0215] Optionally, if the proportion of the candidate area is less than the proportion of the annotated area, the adjustment module 804 performs expansion processing on the candidate non-lesion area according to the proportion of the annotated area to obtain the target non-lesion area in the following manners:

[0216] If the proportion of the candidate area is smaller than the proportion of the marked area, obtaining an expansion parameter, wherein the expansion parameter includes an expansion shape and an expansion size;

[0217] Iteratively expanding the candidate non-lesion region according to the expansion parameter to obtain an expanded candidate non-lesion region;

[0218] Obtaining the area ratio of the expanded candidate non-lesion area in the target medical image as the expanded area ratio;

[0219] The expanded candidate non-lesion area whose area ratio difference between the expanded area ratio and the marked area ratio is less than the ratio threshold is determined as the target non-lesion area.

[0220] Optionally, if the proportion of the candidate area is less than the proportion of the annotated area, the adjustment module 804 performs expansion processing on the candidate non-lesion area according to the proportion of the annotated area to obtain the target non-lesion area in the following manners:

[0221] If the proportion of the candidate area is less than the proportion of the annotated area, then obtaining pixel values ​​of pixels adjacent to the candidate non-lesion area in the target medical image and pixel values ​​of pixels in the candidate non-lesion area;

[0222] According to the pixel values ​​of the pixel points in the candidate non-lesion area and the pixel values ​​of the pixel points adjacent to the candidate non-lesion area, clustering is performed on the pixel points in the candidate non-lesion area and the pixel points adjacent to the candidate non-lesion area to obtain a clustered area;

[0223] Obtaining the area ratio of the cluster area in the target medical image as the cluster area ratio;

[0224] If the area ratio difference between the cluster area ratio and the marked area ratio is less than the ratio threshold, the cluster area is determined as the target non-lesion area.

[0225] Optionally, the adjustment module 804 clusters the pixels in the candidate non-lesion area and the pixels adjacent to the candidate non-lesion area according to the pixel values ​​of the pixels in the candidate non-lesion area and the pixel values ​​of the pixels adjacent to the candidate non-lesion area, and the manner of obtaining the clustered area includes:

[0226] Obtaining a pixel difference between a pixel value of a pixel point adjacent to the candidate non-lesion area and a pixel value of a pixel point in the candidate non-lesion area;

[0227] Determine, from the pixel points adjacent to the candidate non-lesion area, a pixel point whose corresponding pixel difference is less than a pixel difference threshold value as a second target pixel point;

[0228] The region where the second target pixel point in the target medical image is located is merged with the candidate non-lesion region to obtain the cluster region.

[0229] Optionally, if the proportion of the candidate area is greater than the proportion of the annotated area, the adjustment module 804 reduces the candidate non-lesion area according to the proportion of the annotated area, and obtains the target non-lesion area in the following manners:

[0230] If the proportion of the candidate area is greater than the proportion of the marked area, then obtaining reduction processing parameters, wherein the reduction processing parameters include a shape of the reduction processing and a size of the reduction processing;

[0231] Iteratively shrink the candidate non-lesion region according to the shrinking processing parameters to obtain a candidate non-lesion region after shrinking processing;

[0232] Obtaining the area ratio of the candidate non-lesion area after the reduction processing in the target medical image as the area ratio after the reduction processing;

[0233] The candidate non-lesion area after the reduction processing, in which the difference between the area ratio after the reduction processing and the area ratio of the marked area is less than the ratio threshold, is determined as the target non-lesion area.

[0234] Optionally, the candidate region size information includes the candidate region size of the candidate non-lesion region; the annotated region size information includes the annotated region size of the non-lesion region in the target medical image;

[0235] If the candidate region size information does not match the annotated region size information, the adjustment module 804 adjusts the candidate non-lesion region according to the annotated region size information to obtain the target non-lesion region of the target medical image, including:

[0236] Obtaining a region size difference between the candidate region size and the marked region size;

[0237] If the region size difference is greater than the size threshold, it is determined that the candidate region size information does not match the marked region size information;

[0238] The candidate non-lesion area is adjusted according to the size of the marked area to obtain the target non-lesion area.

[0239] Optionally, the determination module 803 determines the region size information of the candidate non-lesion region, and the manner in which the candidate region size information is determined includes:

[0240] Acquire the area size of the candidate non-lesion area and the image size of the target medical image, and determine the ratio between the area size of the candidate non-lesion area and the image size of the target medical image as the candidate area ratio; or,

[0241] Obtaining the number of pixels in the candidate non-lesion area and the number of pixels in the target medical image, and determining the ratio between the number of pixels in the candidate non-lesion area and the number of pixels in the target medical image as the candidate area proportion;

[0242] The candidate area proportion is determined as the candidate area size information.

[0243] Optionally, the determination module 803 determines the region size information of the candidate non-lesion region, and the manner in which the candidate region size information is determined includes:

[0244] Acquire the region size of the candidate non-lesion region to obtain the candidate region size;

[0245] The candidate region size is determined as the candidate region size information.

[0246] Optionally, the device further comprises:

[0247] A labeling module 805 is used to label the target non-lesion area in the target medical image to obtain a labeled target medical image;

[0248] A prediction module 806 is used to predict the target medical image using an image segmentation model to obtain a predicted non-lesion area in the target medical image, and to mark the predicted non-lesion area in the target medical image to obtain a predicted target medical image;

[0249] The adjustment module 804 is further configured to adjust the image segmentation model according to the labeled target medical image and the predicted target medical image to obtain a target medical image segmentation model.

[0250] Optionally, the adjustment module 804 adjusts the image segmentation model according to the labeled target medical image and the predicted target medical image to obtain the target medical image segmentation model in a manner including:

[0251] Determining a prediction loss value of the image segmentation model according to the labeled target medical image and the predicted target medical image;

[0252] If the predicted loss value does not meet the convergence condition, the image segmentation model is adjusted according to the predicted loss value to obtain a target medical image segmentation model.

[0253] Optionally, the labeling module 805 labels the target non-lesion area in the target medical image, and a method of obtaining the labeled target medical image includes:

[0254] Binarizing the target medical image according to the pixel points in the target non-lesion area;

[0255] The target medical image after binarization processing is determined as the labeled target medical image. `

[0256] According to one embodiment of the present application, Figure 3 The steps involved in the medical image processing method shown can be represented by Figure 8 The various modules in the medical image processing device shown are executed. For example, Figure 3 The step S101 shown in FIG. 1 can be performed by Figure 8 The acquisition module 801 in is executed, Figure 3 The step S102 shown in FIG. 1 can be performed by Figure 8 The identification module 802 is executed; Figure 3 The step S103 shown in FIG. 1 can be performed by Figure 8 The determination module 803 is executed; Figure 3 The step S104 shown in FIG. 1 can be performed by Figure 8 The adjustment module 804 in is used to execute.

[0257] According to one embodiment of the present application, Figure 8 The various modules in the medical image processing device shown can be separately or completely combined into one or several units to constitute, or one (some) of the units can be further divided into multiple functionally smaller sub-units, which can achieve the same operation without affecting the realization of the technical effects of the embodiments of the present application. The above modules are divided based on logical functions. In practical applications, the functions of one module can also be implemented by multiple units, or the functions of multiple modules can be implemented by one unit. In other embodiments of the present application, the medical image processing device may also include other units. In practical applications, these functions can also be implemented with the assistance of other units, and can be implemented by the collaboration of multiple units.

[0258] According to one embodiment of the present application, the program can be executed by running on a general-purpose computer device such as a computer including a central processing unit (CPU), a random access memory medium (RAM), a read-only memory medium (ROM), and other processing elements and storage elements. Figure 3 A computer program (including program code) for each step involved in the corresponding method shown in Figure 8The medical image processing apparatus shown in and the medical image processing method of the embodiment of the present application are implemented. The above computer program can be recorded on, for example, a computer readable recording medium, and loaded into the above computing device through the computer readable recording medium and run therein.

[0259] In the present application, the computer device can perform difference identification on the target medical image and the reference medical image, obtain the candidate non-lesion area of ​​the target medical image, obtain the area size information of the candidate non-lesion area, and obtain the candidate area size information. If the candidate area size information does not match the annotated area size information of the target medical image, it indicates that there is a large difference between the identified candidate non-lesion area and the actual non-lesion area of ​​the target medical image, that is, the accuracy of the identified candidate non-lesion area is relatively low; therefore, the candidate non-lesion area can be adjusted according to the annotated area size information to obtain the target non-lesion area of ​​the target medical image. In other words, by automatically identifying the candidate non-lesion area in the target medical image based on the reference medical image and the target medical image, no manual intervention is required, and the accuracy and efficiency of identifying the non-lesion area can be improved; and the identified candidate non-lesion area can be adjusted according to the annotated area size information, which can further improve the accuracy of identifying the non-lesion area.

[0260] See also Fig. 9 , is a schematic diagram of the structure of a computer device provided in an embodiment of the present application. Fig. 9 As shown, the above-mentioned computer device 1000 may include: a processor 1001, a network interface 1004 and a memory 1005. In addition, the above-mentioned computer device 1000 may also include: a user interface 1003, and at least one communication bus 1002. Among them, the communication bus 1002 is used to realize the connection and communication between these components. Among them, the user interface 1003 may include a display screen (Display), a keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface). The memory 1005 may be a high-speed RAM memory, or it may be a non-volatile memory (non-volatile memory), such as at least one disk storage. The memory 1005 may optionally also be at least one storage device located away from the aforementioned processor 1001. As Fig. 9 As shown, the memory 1005 as a computer-readable storage medium may include an operating system, a network communication module, a user interface module, and a device control application program.

[0261] exist Fig. 9In the computer device 1000 shown, the network interface 1004 can provide a network communication function; the user interface 1003 is mainly used to provide an input interface for the user; and the processor 1001 can be used to call the device control application stored in the memory 1005 to achieve:

[0262] Acquire a medical image set; the medical image set includes a reference medical image, a target medical image to be identified, and annotated area size information of a non-lesion area of ​​the target medical image, the target medical image includes a lesion area and a non-lesion area, and the reference medical image includes a lesion area;

[0263] Performing difference recognition on the reference medical image and the target medical image to obtain a candidate non-lesion area in the target medical image;

[0264] Determining region size information of the candidate non-lesion region as candidate region size information;

[0265] If the candidate region size information does not match the annotated region size information, the candidate non-lesion region is adjusted according to the annotated region size information to obtain a target non-lesion region of the target medical image.

[0266] Optionally, the processor is used to call the computer program to perform difference recognition on the reference medical image and the target medical image to obtain a candidate non-lesion area in the target medical image, including:

[0267] Converting the reference medical image into an image in a target color space to obtain a converted reference medical image, and converting the target medical image into an image in the target color space to obtain a converted target medical image;

[0268] Acquiring brightness information of pixels in the converted reference medical image and brightness information of pixels in the converted target medical image;

[0269] Based on the brightness information of the pixels in the converted reference medical image and the brightness information of the pixels in the converted target medical image, the reference medical image and the target medical image are differentially identified to obtain a candidate non-lesion area in the target medical image.

[0270] Optionally, the processor is used to call the computer program to perform difference recognition on the reference medical image and the target medical image according to the brightness information of the pixels in the converted reference medical image and the brightness information of the pixels in the converted target medical image, and a method for obtaining the candidate non-lesion area in the target medical image includes:

[0271] Acquire the difference between the brightness information of the pixel points in the converted reference medical image and the brightness information of the corresponding pixel points in the converted target medical image;

[0272] Identify, from the converted target medical image, a pixel whose brightness information corresponding difference is greater than a difference threshold as a first target pixel;

[0273] The region where the first target pixel point is located in the target medical image is determined as a candidate non-lesion region in the target medical image.

[0274] Optionally, the candidate region size information includes the candidate region ratio of the candidate non-lesion region in the target medical image; the annotated region size information includes the annotated region ratio of the non-lesion region in the target medical image in the target medical image;

[0275] The processor is used to call the computer program to execute the method of adjusting the candidate non-lesion region according to the annotated region size information if the candidate region size information does not match the annotated region size information, and obtaining the target non-lesion region of the target medical image includes:

[0276] Obtaining a difference in area ratio between the candidate area ratio and the marked area ratio;

[0277] If the area proportion difference is greater than the proportion threshold, it is determined that the candidate area size information does not match the marked area size information;

[0278] The candidate non-lesion area is adjusted according to the proportion of the marked area to obtain the target non-lesion area.

[0279] Optionally, the processor is used to call the computer program to adjust the candidate non-lesion area according to the proportion of the marked area to obtain the target non-lesion area, including:

[0280] If the proportion of the candidate area is less than the proportion of the annotated area, the candidate non-lesion area is expanded according to the proportion of the annotated area to obtain the target non-lesion area;

[0281] If the proportion of the candidate area is greater than the proportion of the annotated area, the candidate non-lesion area is reduced according to the proportion of the annotated area to obtain the target non-lesion area.

[0282] Optionally, the processor is used to call the computer program to execute, if the proportion of the candidate area is less than the proportion of the annotated area, then expand the candidate non-lesion area according to the proportion of the annotated area, to obtain the target non-lesion area in a manner including:

[0283] If the proportion of the candidate area is smaller than the proportion of the marked area, obtaining an expansion parameter, wherein the expansion parameter includes an expansion shape and an expansion size;

[0284] Iteratively expanding the candidate non-lesion region according to the expansion parameter to obtain an expanded candidate non-lesion region;

[0285] Obtaining the area ratio of the expanded candidate non-lesion area in the target medical image as the expanded area ratio;

[0286] The expanded candidate non-lesion area whose area ratio difference between the expanded area ratio and the marked area ratio is less than the ratio threshold is determined as the target non-lesion area.

[0287] Optionally, if the proportion of the candidate area is less than the proportion of the annotated area, the processor is used to call the computer program to perform expansion processing on the candidate non-lesion area according to the proportion of the annotated area, and the method for obtaining the target non-lesion area includes:

[0288] If the proportion of the candidate area is less than the proportion of the annotated area, then obtaining pixel values ​​of pixels adjacent to the candidate non-lesion area in the target medical image and pixel values ​​of pixels in the candidate non-lesion area;

[0289] According to the pixel values ​​of the pixel points in the candidate non-lesion area and the pixel values ​​of the pixel points adjacent to the candidate non-lesion area, clustering is performed on the pixel points in the candidate non-lesion area and the pixel points adjacent to the candidate non-lesion area to obtain a clustered area;

[0290] Obtaining the area ratio of the cluster area in the target medical image as the cluster area ratio;

[0291] If the area ratio difference between the cluster area ratio and the marked area ratio is less than the ratio threshold, the cluster area is determined as the target non-lesion area.

[0292] Optionally, the processor is used to call the computer program to perform clustering processing on the pixels in the candidate non-lesion area and the pixels adjacent to the candidate non-lesion area according to the pixel values ​​of the pixels in the candidate non-lesion area and the pixel values ​​of the pixels adjacent to the candidate non-lesion area, and the method for obtaining the clustered area includes:

[0293] Obtaining a pixel difference between a pixel value of a pixel point adjacent to the candidate non-lesion area and a pixel value of a pixel point in the candidate non-lesion area;

[0294] Determine, from the pixel points adjacent to the candidate non-lesion area, a pixel point whose corresponding pixel difference is less than a pixel difference threshold value as a second target pixel point;

[0295] The region where the second target pixel point in the target medical image is located is merged with the candidate non-lesion region to obtain the cluster region.

[0296] Optionally, if the proportion of the candidate area is greater than the proportion of the annotated area, the processor is used to call the computer program to perform a reduction process on the candidate non-lesion area according to the proportion of the annotated area, and the method for obtaining the target non-lesion area includes:

[0297] If the proportion of the candidate area is greater than the proportion of the marked area, then obtaining reduction processing parameters, wherein the reduction processing parameters include a shape of the reduction processing and a size of the reduction processing;

[0298] Iteratively shrink the candidate non-lesion region according to the shrinking processing parameters to obtain a candidate non-lesion region after shrinking processing;

[0299] Obtaining the area ratio of the candidate non-lesion area after the reduction processing in the target medical image as the area ratio after the reduction processing;

[0300] The candidate non-lesion area after the reduction processing, in which the difference between the area ratio after the reduction processing and the area ratio of the marked area is less than the ratio threshold, is determined as the target non-lesion area.

[0301] Optionally, the candidate region size information includes the candidate region size of the candidate non-lesion region; the annotated region size information includes the annotated region size of the non-lesion region in the target medical image;

[0302] If the candidate region size information does not match the annotated region size information, the processor is used to call the computer program to adjust the candidate non-lesion region according to the annotated region size information to obtain the target non-lesion region of the target medical image, including:

[0303] Obtaining a region size difference between the candidate region size and the marked region size;

[0304] If the region size difference is greater than the size threshold, it is determined that the candidate region size information does not match the marked region size information;

[0305] The candidate non-lesion area is adjusted according to the size of the marked area to obtain the target non-lesion area.

[0306] Optionally, the processor is used to call the computer program to execute the determination of the region size information of the candidate non-lesion region, and the manner of determining the region size information as the candidate region size information includes:

[0307] Acquire the area size of the candidate non-lesion area and the image size of the target medical image, and determine the ratio between the area size of the candidate non-lesion area and the image size of the target medical image as the candidate area ratio; or,

[0308] Obtaining the number of pixels in the candidate non-lesion area and the number of pixels in the target medical image, and determining the ratio between the number of pixels in the candidate non-lesion area and the number of pixels in the target medical image as the candidate area proportion;

[0309] The candidate area proportion is determined as the candidate area size information.

[0310] Optionally, the processor is used to call the computer program to execute the determination of the region size information of the candidate non-lesion region, and the manner of determining the region size information as the candidate region size information includes:

[0311] Acquire the region size of the candidate non-lesion region to obtain the candidate region size;

[0312] The candidate region size is determined as the candidate region size information.

[0313] The processor is used to call the computer program to execute:

[0314] Annotating the target non-lesion area in the target medical image to obtain an annotated target medical image;

[0315] Using an image segmentation model to predict the target medical image to obtain a predicted non-lesion area in the target medical image, and marking the predicted non-lesion area in the target medical image to obtain a predicted target medical image;

[0316] The image segmentation model is adjusted according to the labeled target medical image and the predicted target medical image to obtain a target medical image segmentation model.

[0317] Optionally, the processor is used to call the computer program to adjust the image segmentation model according to the labeled target medical image and the predicted target medical image to obtain the target medical image segmentation model, including:

[0318] Determining a prediction loss value of the image segmentation model according to the labeled target medical image and the predicted target medical image;

[0319] If the predicted loss value does not meet the convergence condition, the image segmentation model is adjusted according to the predicted loss value to obtain a target medical image segmentation model.

[0320] Optionally, the processor is used to call the computer program to execute the process of labeling the target non-lesion area in the target medical image, and the method of obtaining the labeled target medical image includes:

[0321] Binarizing the target medical image according to the pixel points in the target non-lesion area;

[0322] The target medical image after the binarization process is determined as the labeled target medical image.

[0323] In the present application, the computer device can perform difference identification on the target medical image and the reference medical image, obtain the candidate non-lesion area of ​​the target medical image, obtain the area size information of the candidate non-lesion area, and obtain the candidate area size information. If the candidate area size information does not match the annotated area size information of the target medical image, it indicates that there is a large difference between the identified candidate non-lesion area and the actual non-lesion area of ​​the target medical image, that is, the accuracy of the identified candidate non-lesion area is relatively low; therefore, the candidate non-lesion area can be adjusted according to the annotated area size information to obtain the target non-lesion area of ​​the target medical image. In other words, by automatically identifying the candidate non-lesion area in the target medical image based on the reference medical image and the target medical image, no manual intervention is required, and the accuracy and efficiency of identifying the non-lesion area can be improved; and the identified candidate non-lesion area can be adjusted according to the annotated area size information, which can further improve the accuracy of identifying the non-lesion area.

[0324] It should be understood that the computer device 1000 described in the embodiment of the present application can execute the above Figure 3 The description of the medical image processing method in the corresponding embodiment can also be performed as described above. Figure 8 The description of the medical image processing device in the corresponding embodiment will not be repeated here. In addition, the description of the beneficial effects of the same method will not be repeated here either.

[0325] The present invention provides a computer program product or a computer program, the computer program product or the computer program comprising computer instructions, the computer instructions being stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the above-mentioned Figure 3 The description of the medical image processing method in the corresponding embodiment will not be repeated here. In addition, the description of the beneficial effects of the same method will not be repeated. For technical details not disclosed in the computer-readable storage medium embodiment involved in this application, please refer to the description of the method embodiment of this application.

[0326] As an example, the above program instructions may be deployed on a computer device for execution, or deployed on multiple computer devices located at one location for execution, or executed on multiple computer devices distributed at multiple locations and interconnected by a communication network. Multiple computer devices distributed at multiple locations and interconnected by a communication network may constitute a blockchain network.

[0327] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the above-mentioned program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The above-mentioned storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0328] The above disclosure is only the preferred embodiment of the present application, which certainly cannot be used to limit the scope of rights of the present application. Therefore, equivalent changes made according to the claims of the present application are still within the scope covered by the present application.

Claims

1. A medical image processing method, It is characterized in that include: Acquire a medical image set; the medical image set includes a reference medical image, a target medical image to be identified, and annotated area size information of a non-lesion area of ​​the target medical image, the target medical image includes a lesion area and a non-lesion area, and the reference medical image includes a lesion area; Performing difference recognition on the reference medical image and the target medical image to obtain a candidate non-lesion area in the target medical image; Determining region size information of the candidate non-lesion region as candidate region size information; If the candidate region size information does not match the annotated region size information, the candidate non-lesion region is adjusted according to the annotated region size information to obtain a target non-lesion region of the target medical image.

2. The method according to claim 1, It is characterized in that The step of performing difference recognition on the reference medical image and the target medical image to obtain a candidate non-lesion area in the target medical image includes: Converting the reference medical image into an image in a target color space to obtain a converted reference medical image, and converting the target medical image into an image in the target color space to obtain a converted target medical image; Acquiring brightness information of pixels in the converted reference medical image and brightness information of pixels in the converted target medical image; Based on the brightness information of the pixels in the converted reference medical image and the brightness information of the pixels in the converted target medical image, the reference medical image and the target medical image are differentially identified to obtain a candidate non-lesion area in the target medical image.

3. The method according to claim 2, It is characterized in that The step of performing difference recognition on the reference medical image and the target medical image according to the brightness information of the pixels in the converted reference medical image and the brightness information of the pixels in the converted target medical image to obtain a candidate non-lesion area in the target medical image includes: Acquire the difference between the brightness information of the pixel points in the converted reference medical image and the brightness information of the corresponding pixel points in the converted target medical image; Identify, from the converted target medical image, a pixel whose brightness information corresponding difference is greater than a difference threshold as a first target pixel; The region where the first target pixel point is located in the target medical image is determined as a candidate non-lesion region in the target medical image.

4. The method according to claim 1, It is characterized in that The candidate region size information includes the candidate region ratio of the candidate non-lesion region in the target medical image; the annotated region size information includes the annotated region ratio of the non-lesion region in the target medical image in the target medical image; If the size information of the candidate region does not match the size information of the annotated region, adjusting the candidate non-lesion region according to the size information of the annotated region to obtain the target non-lesion region of the target medical image includes: Obtaining a difference in area ratio between the candidate area ratio and the marked area ratio; If the area proportion difference is greater than the proportion threshold, it is determined that the candidate area size information does not match the marked area size information; The candidate non-lesion area is adjusted according to the proportion of the marked area to obtain the target non-lesion area.

5. The method according to claim 4, It is characterized in that The step of adjusting the candidate non-lesion area according to the proportion of the annotated area to obtain the target non-lesion area includes: If the proportion of the candidate area is less than the proportion of the annotated area, the candidate non-lesion area is expanded according to the proportion of the annotated area to obtain the target non-lesion area; If the proportion of the candidate area is greater than the proportion of the annotated area, the candidate non-lesion area is reduced according to the proportion of the annotated area to obtain the target non-lesion area.

6. The method according to claim 5, It is characterized in that If the proportion of the candidate area is less than the proportion of the annotated area, the candidate non-lesion area is expanded according to the proportion of the annotated area to obtain the target non-lesion area, including: If the proportion of the candidate area is smaller than the proportion of the marked area, obtaining an expansion parameter, wherein the expansion parameter includes an expansion shape and an expansion size; Iteratively expanding the candidate non-lesion region according to the expansion parameter to obtain an expanded candidate non-lesion region; Obtaining the area ratio of the expanded candidate non-lesion area in the target medical image as the expanded area ratio; The expanded candidate non-lesion area whose area ratio difference between the expanded area ratio and the marked area ratio is less than the ratio threshold is determined as the target non-lesion area.

7. The method according to claim 5, It is characterized in that If the proportion of the candidate area is less than the proportion of the annotated area, the candidate non-lesion area is expanded according to the proportion of the annotated area to obtain the target non-lesion area, including: If the proportion of the candidate area is less than the proportion of the annotated area, then obtaining pixel values ​​of pixels adjacent to the candidate non-lesion area in the target medical image and pixel values ​​of pixels in the candidate non-lesion area; According to the pixel values ​​of the pixel points in the candidate non-lesion area and the pixel values ​​of the pixel points adjacent to the candidate non-lesion area, clustering is performed on the pixel points in the candidate non-lesion area and the pixel points adjacent to the candidate non-lesion area to obtain a clustered area; Obtaining the area ratio of the cluster area in the target medical image as the cluster area ratio; If the area ratio difference between the cluster area ratio and the marked area ratio is less than the ratio threshold, the cluster area is determined as the target non-lesion area.

8. The method according to claim 7, It is characterized in that The clustering process is performed on the pixels in the candidate non-lesion area and the pixels adjacent to the candidate non-lesion area according to the pixel values ​​of the pixels in the candidate non-lesion area and the pixel values ​​of the pixels adjacent to the candidate non-lesion area to obtain the clustered area, including: Obtaining a pixel difference between a pixel value of a pixel point adjacent to the candidate non-lesion area and a pixel value of a pixel point in the candidate non-lesion area; Determine, from the pixel points adjacent to the candidate non-lesion area, a pixel point whose corresponding pixel difference is less than a pixel difference threshold value as a second target pixel point; The region where the second target pixel point in the target medical image is located is merged with the candidate non-lesion region to obtain the cluster region.

9. The method according to claim 5, It is characterized in that If the proportion of the candidate area is greater than the proportion of the annotated area, the candidate non-lesion area is reduced according to the proportion of the annotated area to obtain the target non-lesion area, including: If the proportion of the candidate area is greater than the proportion of the marked area, then obtaining reduction processing parameters, wherein the reduction processing parameters include a shape of the reduction processing and a size of the reduction processing; Iteratively shrink the candidate non-lesion region according to the shrinking processing parameters to obtain a candidate non-lesion region after shrinking processing; Obtaining the area ratio of the candidate non-lesion area after the reduction processing in the target medical image as the area ratio after the reduction processing; The candidate non-lesion area after the reduction processing, in which the difference between the area ratio after the reduction processing and the area ratio of the marked area is less than the ratio threshold, is determined as the target non-lesion area.

10. The method according to claim 1, It is characterized in that The candidate region size information includes the candidate region size of the candidate non-lesion region; the annotated region size information includes the annotated region size of the non-lesion region in the target medical image; If the size information of the candidate region does not match the size information of the annotated region, adjusting the candidate non-lesion region according to the size information of the annotated region to obtain the target non-lesion region of the target medical image includes: Obtaining a region size difference between the candidate region size and the marked region size; If the region size difference is greater than the size threshold, it is determined that the candidate region size information does not match the marked region size information; The candidate non-lesion area is adjusted according to the size of the marked area to obtain the target non-lesion area.

11. The method according to claim 4, It is characterized in that The determining the region size information of the candidate non-lesion region as the candidate region size information includes: Acquire the area size of the candidate non-lesion area and the image size of the target medical image, and determine the ratio between the area size of the candidate non-lesion area and the image size of the target medical image as the candidate area ratio; or, Obtaining the number of pixels in the candidate non-lesion area and the number of pixels in the target medical image, and determining the ratio between the number of pixels in the candidate non-lesion area and the number of pixels in the target medical image as the candidate area proportion; The candidate area proportion is determined as the candidate area size information.

12. The method of claim 1, It is characterized in that The method further comprises: Annotating the target non-lesion area in the target medical image to obtain an annotated target medical image; Using an image segmentation model to predict the target medical image to obtain a predicted non-lesion area in the target medical image, and marking the predicted non-lesion area in the target medical image to obtain a predicted target medical image; The image segmentation model is adjusted according to the labeled target medical image and the predicted target medical image to obtain a target medical image segmentation model.

13. The method according to claim 12, It is characterized in that The step of adjusting the image segmentation model according to the labeled target medical image and the predicted target medical image to obtain a target medical image segmentation model includes: Determining a prediction loss value of the image segmentation model according to the labeled target medical image and the predicted target medical image; If the predicted loss value does not meet the convergence condition, the image segmentation model is adjusted according to the predicted loss value to obtain a target medical image segmentation model.

14. The method according to claim 12, It is characterized in that The step of labeling the target non-lesion area in the target medical image to obtain the labeled target medical image includes: Binarizing the target medical image according to the pixel points in the target non-lesion area; The target medical image after the binarization process is determined as the labeled target medical image.

15. A medical image processing device, It is characterized in that include: An acquisition module, used for acquiring a medical image collection; The medical image set includes a reference medical image, a target medical image to be identified, and annotated area size information of a non-lesion area of ​​the target medical image, the target medical image includes a lesion area and a non-lesion area, and the reference medical image includes a lesion area; an identification module, configured to perform difference identification on the reference medical image and the target medical image to obtain a candidate non-lesion region in the target medical image; A determination module, used to determine the region size information of the candidate non-lesion region as the candidate region size information; The adjustment module is used to adjust the candidate non-lesion region according to the annotated region size information to obtain the target non-lesion region of the target medical image if the candidate region size information does not match the annotated region size information.

Images