Image screening method and device, computer equipment, storage medium and program product

The image screening method improves surgical video analysis by incorporating region of interest scoring to correct initial classification results, enhancing accuracy and efficiency in screening medical images.

CN119091177APending Publication Date: 2024-12-06WUHAN UNITED IMAGING HEALTHCARE SURGICAL TECH CO LTD
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202310661573.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-06-05
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

Existing video structured analysis methods for surgical videos inaccurately screen out video frames with valid information due to the lack of comprehensive consideration of the surgical status at the moment, leading to inefficient data utilization and increased workload for doctors.

Method used

An image screening method that introduces the concept of a region of interest and employs algorithms to score effective information, correcting initial image classification results based on the region of interest and target regions to improve accuracy in screening medical images.

Benefits of technology

Enhances the accuracy of image screening by retaining frames with valuable information, optimizing data storage, reducing doctor workload, and facilitating quick surgical clip location and advanced surgical analysis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119091177A_ABST
    Figure CN119091177A_ABST
Patent Text Reader

Abstract

The invention relates to an image screening method and device, computer equipment, a storage medium and a program product. The method comprises the following steps: acquiring an initial image classification result of at least one frame of medical image in a medical video to be processed; for each frame of medical image, correcting the initial image classification result of the medical image according to the region of interest of the medical image and a target region corresponding to the initial image classification result of the medical image, and generating a target image classification result of the medical image; and according to the target image classification result of each frame of medical image and a preset image screening condition, screening out a target medical image meeting the preset image screening condition from each frame of medical image. When video structured analysis is carried out, the region of interest of the medical image is introduced, and the initial image classification result of the medical image is updated according to the influence degree of the target region corresponding to the initial image classification result on the region of interest of the medical image, so that the image screening accuracy is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to an image screening method, device, computer equipment, storage medium and program product. Background Art

[0002] With the continuous development of image processing technology, for a video, a structured processing method can be used to process the video to obtain the structured information of the video. Among them, structured processing is a technology for extracting video content information. Based on the structured information of the video obtained from the video, image frames of interest can be screened out from the video, so that the screened image frames of interest can meet the different usage needs of users. For example: in the medical field, for the image video collected by the endoscope, the image video can be subjected to video structured analysis to screen out video frames or video clips that perform relevant operations on the tissue of interest from the image video, so as to facilitate the doctor to conduct targeted analysis on the screened video frames or video clips of interest when reviewing.

[0003] However, using traditional structured processing methods, there is a problem of inaccurate screening when screening out image frames of interest from videos. Summary of the invention

[0004] Based on this, it is necessary to provide an image screening method, device, computer equipment, computer-readable storage medium and computer program product that can improve the accuracy of image screening in response to the above technical problems.

[0005] In a first aspect, the present application provides an image screening method. The method comprises:

[0006] Obtaining an initial image classification result of at least one frame of medical image in the medical video to be processed;

[0007] For each frame of the medical image, according to the region of interest of the medical image and the target region corresponding to the initial image classification result of the medical image, the initial image classification result of the medical image is corrected to generate a target image classification result of the medical image;

[0008] According to the target image classification results of each frame of medical images and the preset image screening conditions, the target medical images that meet the preset image screening conditions are screened out from each frame of medical images.

[0009] In one embodiment, the initial image classification result includes at least one initial image classification label; and correcting the initial image classification result of the medical image according to the region of interest of the medical image and the target region corresponding to the initial image classification result of the medical image to generate a target image classification result of the medical image includes:

[0010] Obtaining target regions corresponding to each initial image classification label from the medical image;

[0011] Obtaining regions of interest in medical images;

[0012] According to the region of interest of the medical image and the target region corresponding to each initial image classification label, each initial image classification label is corrected to generate a target image classification label of the medical image, and the target image classification label is used as the target image classification result.

[0013] In one embodiment, according to the region of interest of the medical image and the target region corresponding to each initial image classification label, each initial image classification label is corrected to generate a target image classification label of the medical image, including:

[0014] For each initial image classification label, calculating the overlap rate between the region of interest of the medical image and the target region corresponding to the initial image classification label; the overlap rate is the ratio of a first area of ​​the overlap region between the region of interest and the target region to a second area of ​​the region of interest;

[0015] According to the size relationship between each overlap rate and a preset overlap rate threshold, each initial image classification label is corrected to generate a target image classification label for the medical image.

[0016] In one embodiment, each initial image classification label is corrected according to the size relationship between each overlap rate and a preset overlap rate threshold value to generate a target image classification label of the medical image, including:

[0017] For each overlap rate, if the overlap rate is greater than or equal to a preset overlap rate threshold, the initial image classification label corresponding to the overlap rate is used as the target image classification label of the medical image;

[0018] If the overlap rate is less than a preset overlap rate threshold, the initial image classification labels corresponding to the overlap rate are deleted from each initial image classification label to generate a target image classification label for the medical image.

[0019] In one embodiment, according to the region of interest of the medical image and the target region corresponding to each initial image classification label, each initial image classification label is corrected to generate a target image classification label of the medical image, including:

[0020] For each initial image classification label, calculating the overlap rate between the region of interest of the medical image and the target region corresponding to the initial image classification label;

[0021] Calculating an interference rate according to a first center point of the region of interest of the medical image, a second center point of the target region corresponding to the initial image classification label, and a preset edge point of the medical image; the interference rate is used to characterize the degree of interference of the target region on the region of interest;

[0022] According to the overlap rate corresponding to the initial image classification label and the interference rate corresponding to the initial image classification label, the influence degree of the target area corresponding to the initial image classification label on the area of ​​interest is calculated;

[0023] According to the magnitude relationship between the influence degree and a preset influence degree threshold, each initial image classification label is corrected to generate a target image classification label for the medical image.

[0024] In one embodiment, the interference rate is calculated according to a first center point of a region of interest of a medical image, a second center point of a target region corresponding to an initial image classification label, and a preset edge point of the medical image, including:

[0025] Determine a first distance between the first center point and the second center point according to first position information of a first center point of the region of interest of the medical image and second position information of a second center point of the target region corresponding to the initial image classification label;

[0026] Determining a second distance between the first center point and the preset edge point according to first position information of a first center point of the region of interest of the medical image and third position information of a preset edge point of the medical image;

[0027] An interference rate is calculated according to a ratio between the first distance and the second distance.

[0028] In one embodiment, acquiring a region of interest of a medical image includes:

[0029] Determine whether the medical image contains the object of interest;

[0030] If the medical image includes an object of interest, determining a region of interest of the medical image based on the object of interest;

[0031] If the medical image does not include the object of interest, a preset area in the medical image is used as the region of interest of the medical image.

[0032] In one embodiment, determining a region of interest of a medical image based on an object of interest includes:

[0033] Determining an initial region of interest of the medical image based on a target position of the target of interest in the medical image;

[0034] The initial region of interest is enlarged according to a preset enlargement ratio to obtain the region of interest of the medical image.

[0035] In a second aspect, the present application also provides an image screening device. The device comprises:

[0036] An acquisition module, used to acquire an initial image classification result of at least one frame of medical image in the medical video to be processed;

[0037] A generating module, for correcting the initial image classification result of the medical image according to the region of interest of the medical image and the target region corresponding to the initial image classification result of the medical image for each frame of the medical image, and generating a target image classification result of the medical image;

[0038] The screening module is used to screen out target medical images that meet the preset image screening conditions from each frame of medical images according to the target image classification results of each frame of medical images and the preset image screening conditions.

[0039] In a third aspect, the present application further provides a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the image screening method in the first aspect when executing the computer program.

[0040] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the image screening method in the first aspect are implemented.

[0041] In a fifth aspect, the present application further provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the steps of the image screening method in the first aspect are implemented.

[0042] The above-mentioned image screening method, device, computer device, storage medium and computer program product, the computer device obtains the initial image classification result of at least one frame of medical image in the medical video to be processed; and for each frame of medical image, according to the region of interest of the medical image and the target region corresponding to the initial image classification result of the medical image, corrects the initial image classification result of the medical image to generate the target image classification result of the medical image; then, according to the target image classification result of each frame of medical image and the preset image screening condition, the target medical image that meets the preset image screening condition is screened out from each frame of medical image. That is to say, in the embodiment of the present application, when performing video structured analysis, by introducing the region of interest of the medical image, the initial image classification result of the medical image is corrected based on the region of interest of the medical image and the target region corresponding to the initial image classification result; that is, by the degree of influence of the target region corresponding to the initial image classification result on the region of interest of the medical image, the initial image classification result of the medical image is updated, so as to avoid the medical image that has little influence on the medical image by the initial image classification result, that is, the medical image that includes useful information is screened out, thereby improving the accuracy of image screening. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A diagram showing an application environment of an image screening method in one embodiment;

[0044] Figure 2 A schematic diagram of a process of an image screening method in one embodiment;

[0045] Figure 3 is a schematic flow chart of an image screening method in another embodiment;

[0046] Figure 4 is a schematic flow chart of an image screening method in another embodiment;

[0047] Figure 5 is a schematic flow chart of an image screening method in another embodiment;

[0048] Figure 6 is a schematic flow chart of an image screening method in another embodiment;

[0049] Figure 7 A schematic diagram of a complete process of an image screening method in one embodiment;

[0050] Figure 8 A schematic diagram of an algorithm for valid information probability scoring of a medical image in one embodiment;

[0051] Fig. 9 is a structural block diagram of an image screening device in one embodiment;

[0052] Fig.10FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0054] With the continuous development of image processing technology, for a video, a structured processing method can be used to process the video to obtain the structured information of the video. Among them, structured processing is a technology for extracting video content information. Based on the structured information of the video obtained from the video, image frames of interest can be screened out from the video, so that the screened image frames of interest can meet the different usage needs of users. For example: in the medical field, for the image video collected by the endoscope, the image video can be subjected to video structured analysis to screen out video frames or video clips that perform relevant operations on the tissue of interest from the image video, so as to facilitate the doctor to conduct targeted analysis on the screened video frames or video clips of interest when reviewing.

[0055] Among them, minimally invasive surgery is one of the mainstream surgical forms. Due to the limitation of operating space, the surgical field of view is often blocked, body fluids contaminate the lens, etc., resulting in blurred images, and the final video quality is uneven. Video structured analysis can remove video frames or video segments with invalid information by classifying video frames. It has the following values: (1) It greatly optimizes the utilization rate of hospital data storage space while retaining valuable surgical video information; (2) The structured video can be used as the result of rough editing. Doctors can perform personalized video precision editing based on the results of rough editing, reducing the workload of doctors; (3) Quickly locate surgical clips in surgical review, saving doctors the time spent browsing videos; (4) Provide data support for advanced applications such as subsequent surgical stage identification, surgical information statistics, and surgical safety event prediction.

[0056] However, existing video structured analysis methods all use classification algorithms to classify video frames without comprehensively considering the surgical status at the current moment. Therefore, when using existing video structured analysis methods to perform structured analysis on surgical videos, video frames with valid information are easily screened out, resulting in inaccurate screening.

[0057] Based on this, an embodiment of the present application proposes an image screening method, that is, a video structuring processing method. This method introduces the definition of a region of interest and innovatively proposes a method for scoring effective information of video frames, thereby realizing video structuring processing, which can improve the accuracy of video structuring and be closer to clinical application scenarios.

[0058] The image screening method provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown in the figure, the computer device 102 can be a terminal or a server; the terminal can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, and portable wearable devices, and the portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server can be implemented as an independent server or a server cluster composed of multiple servers.

[0059] In one embodiment, Figure 2 As shown, an image screening method is provided, which is applied to Figure 1 The computer device in the example is used to illustrate, including the following steps:

[0060] Step 201: Obtain an initial image classification result of at least one frame of medical image in a medical video to be processed.

[0061] Among them, the medical video to be processed may include multiple frames of medical images. When performing video structuring processing, all frames of medical images in the medical video to be processed can be processed, or part of the frames of medical images in the medical video to be processed can be processed. The part of the frames of medical images can be multiple frames of medical images corresponding to at least one video segment in the medical video to be processed.

[0062] In addition, the initial image classification result of the medical image may include at least one initial image classification label, including but not limited to normal, blurred, occluded, smoke, uneven lighting, in vitro, and without instrument, etc.; that is, the computer device can input each frame of medical image in the medical video to be processed into a preset image classification model for image classification processing, thereby obtaining the initial image classification result of each frame of medical image; wherein, the preset image classification model can be a multi-label classification model, and the preset image classification model can also include multiple image classification sub-models, and each image classification sub-model can be a single-label or multi-label classification model; the embodiment of the present application does not make specific limitations on this.

[0063] Furthermore, after performing image classification processing on each frame of medical images in the medical video to be processed through a preset image classification model to obtain the initial image classification label of each frame of medical images, the initial image classification label of each frame of medical images in the medical video to be processed can be stored in a local storage device of the computer device, or the initial image classification label of each frame of medical images can be stored in an external storage device such as a server or a database that is communicatively connected to the computer device; the embodiment of the present application does not make specific limitations on this.

[0064] That is to say, the computer device can obtain the initial image classification result of at least one frame of medical image in the medical video to be processed from the storage device, or input the at least one frame of medical image into a preset image classification model for image classification processing to obtain the initial image classification result of the at least one frame of medical image.

[0065] Step 202 , for each frame of medical image, correct the initial image classification result of the medical image according to the region of interest of the medical image and the target region corresponding to the initial image classification result of the medical image to generate a target image classification result of the medical image.

[0066] Among them, the target area corresponding to the initial image classification result of the medical image may include the local areas in the medical image corresponding to the various initial image classification labels of the medical image, and may also include the fusion area of ​​the local areas in the medical image corresponding to the various initial image classification labels of the medical image; for example: when the initial image classification labels of the medical image include smoke and occlusion, the target area corresponding to the initial image classification result of the medical image may include the local area in the medical image corresponding to the smoke and the local area corresponding to the occlusion; or, the target area may also include the fusion area in the medical image corresponding to the smoke and the occlusion.

[0067] In addition, it should be noted that when the initial image classification labels are normal, in vitro, without instruments, etc., since these labels have no corresponding target areas, the relevant operations corresponding to this step do not need to be performed for such labels; that is, for such labels, label correction processing does not need to be performed, and these labels of the medical image can be retained.

[0068] Exemplarily, for an initial image classification label with a corresponding target area, the initial image classification label in the initial image classification result of the medical image can be label-corrected according to the region of interest of the medical image and the target area corresponding to the initial image classification label, thereby generating a target image classification result of the medical image. It should be noted that the regions of interest of each frame of medical image can be the same or different; for example, the region of interest of the medical image can include the medical tissue region, lesion region, surgical instrument region, image center region, etc. in the medical image; when the medical image includes a no-instrument label, the region of interest of the medical image can be at least one of the medical tissue region, lesion region, image center region, etc.; when the medical image does not include a no-instrument label, the region of interest of the medical image can be at least one of the surgical instrument region, medical tissue region, lesion region, image center region, etc.; the embodiment of the present application does not specifically limit the region of interest of each medical image.

[0069] Exemplarily, when performing label correction, it can be determined whether to retain the initial image classification label based on the degree of influence of the target area corresponding to the initial image classification label on the area of ​​interest of the medical image; if the degree of influence of the target area corresponding to the initial image classification label on the area of ​​interest of the medical image is greater than or equal to a preset influence degree threshold, it means that the initial image classification label has a greater impact on the medical image, and the initial image classification label can be retained; for example: assuming that the target area corresponding to the smoke has a greater impact on the area of ​​interest of the medical image, it means that less useful information can be obtained from the area of ​​interest at this time; then, when performing a review analysis, the medical image can be screened out from the video to be analyzed based on the smoke label of the medical image, so that only the part of the medical image where the smoke has a smaller impact on the area of ​​interest of the medical image is retained, which is beneficial for doctors to obtain more useful information from the area of ​​interest of this part of the medical image.

[0070] On the contrary, if the impact of the target area corresponding to the initial image classification label on the area of ​​interest of the medical image is less than the preset impact threshold, it means that the initial image classification label has little impact on the medical image, and the initial image classification label can be deleted; based on the above example, that is, when the impact of the target area corresponding to the smoke on the area of ​​interest of the medical image is small, the smoke label of the medical image is deleted; thus, when the medical video to be processed is structured screened by the smoke label, the medical image can be avoided from being screened out. On the contrary, the medical image can be retained, so that the doctor can obtain more useful information from the area of ​​interest of the medical image.

[0071] Based on the same method as above, each initial image classification label in the initial image classification result corresponding to the medical image can be judged one by one, so as to realize the correction of the initial image classification result and generate the target image classification result of the medical image.

[0072] Step 203 , based on the target image classification results of each frame of medical image and the preset image screening condition, the target medical image that meets the preset image screening condition is screened out from each frame of medical image.

[0073] The target image classification result of the medical image may include at least one target image classification label.

[0074] Exemplarily, after correcting the initial image classification results of each frame of medical image in the medical video to be processed and obtaining the target image classification results of each frame of medical image, the target medical image that meets the preset image screening condition can be screened out from the medical video to be processed according to the preset image screening condition. The preset image screening condition may include at least one image classification label input by the user.

[0075] Exemplarily, when the target medical images satisfying the preset image screening conditions are screened out from each frame of medical images according to the preset image screening conditions, the medical images in each frame of medical images that include the image classification labels in the preset image screening conditions can be used as the screened target medical images, and other medical images in each frame of medical images except the medical images that include the image classification labels in the preset image screening conditions can also be used as the screened target medical images. For example: when the preset image screening conditions include no device labels, the medical images including no device labels can be screened out as target medical images, thereby realizing the detection and analysis of medical images without devices; the medical images including no device labels can also be screened out, and other medical images except the medical images including no device labels, that is, medical images with devices, can be screened out as target medical images, thereby realizing the detection and analysis of medical images with devices.

[0076] In the above-mentioned image screening method, the computer device obtains the initial image classification result of at least one frame of medical image in the medical video to be processed; and for each frame of medical image, according to the region of interest of the medical image and the target region corresponding to the initial image classification result of the medical image, corrects the initial image classification result of the medical image to generate the target image classification result of the medical image; and then, according to the target image classification result of each frame of medical image and the preset image screening condition, the target medical image that meets the preset image screening condition is screened out from each frame of medical image. That is to say, in the embodiment of the present application, when performing video structured analysis, by introducing the region of interest of the medical image, the initial image classification result of the medical image is corrected based on the region of interest of the medical image and the target region corresponding to the initial image classification result; that is, by the degree of influence of the target region corresponding to the initial image classification result on the region of interest of the medical image, the initial image classification result of the medical image is updated, so as to avoid screening out the medical image that has little influence on the medical image, that is, the medical image that includes useful information, thereby improving the accuracy of image screening.

[0077] In one embodiment, when the initial image classification result includes at least one initial image classification label, Figure 3 As shown, the above step 202 may also include:

[0078] Step 301: Acquire target regions corresponding to each initial image classification label from a medical image.

[0079] Exemplarily, for each initial image classification label corresponding to the medical image, the computer device can obtain the target region corresponding to each initial image classification label from the medical image. Optionally, for different initial image classification labels, the same or different region extraction algorithms can be used to obtain the target region corresponding to the initial image classification label from the medical image.

[0080] Step 302: Acquire a region of interest of a medical image.

[0081] Exemplarily, when the medical image does not include an instrument label, that is, there is a surgical instrument in the medical image, the area where the surgical instrument is located can be used as the area of ​​interest of the medical image, or the area where the tip of the surgical instrument is located can be used as the area of ​​interest of the medical image, etc.

[0082] Exemplarily, when a medical image includes no instrument labels, that is, there are no surgical instruments in the medical image, then the image center area of ​​the medical image can be used as the region of interest of the medical image, or the medical tissue area, lesion area, etc. in the medical image can be used as the region of interest of the medical image, etc.

[0083] That is, the computer device can determine the region of interest of the medical image by judging whether the medical image includes no device label, and then extract the region of interest from the medical image.

[0084] Step 303, correcting each initial image classification label according to the region of interest of the medical image and the target region corresponding to each initial image classification label, generating a target image classification label of the medical image, and using the target image classification label as the target image classification result.

[0085] Exemplarily, for each initial image classification label of a medical image, the initial image classification label is corrected according to the region of interest of the medical image and the target region corresponding to the initial image classification label; optionally, the region of interest of the medical image and the target region corresponding to the initial image classification label can be input into a preset influence degree quantification algorithm to obtain a quantified value of the influence degree of the target region on the region of interest; and then, based on the relationship between the influence degree quantification value and a preset influence degree threshold, it is determined whether to retain or delete the initial image classification label.

[0086] Two different influence degree quantification algorithms will be introduced below to achieve correction processing of the initial image classification labels.

[0087] The first type: Figure 4 As shown, the above step 303 may include:

[0088] Step 401, for each initial image classification label, calculate the overlap rate between the region of interest of the medical image and the target region corresponding to the initial image classification label; wherein the overlap rate is the ratio of the first area of ​​the overlap area between the region of interest and the target region to the second area of ​​the region of interest.

[0089] Step 402: Correct each initial image classification label according to the size relationship between each overlap rate and a preset overlap rate threshold value to generate a target image classification label for the medical image.

[0090] That is, in this embodiment, the degree of influence of the target area on the area of ​​interest is quantified by the overlap rate or coverage rate between the area of ​​interest and the target area; if the overlap rate between the area of ​​interest and the target area is less than the preset overlap rate threshold, it means that the influence of the target area on the area of ​​interest is small; if the overlap rate between the area of ​​interest and the target area is greater than or equal to the preset overlap rate threshold, it means that the influence of the target area on the area of ​​interest is large.

[0091] Accordingly, for the overlap rate corresponding to each initial image classification label, if the overlap rate is greater than or equal to the preset overlap rate threshold, that is, the target region has a greater impact on the region of interest, the initial image classification label corresponding to the overlap rate is used as the target image classification label of the medical image; that is, the initial image classification label of the medical image is retained. If the overlap rate is less than the preset overlap rate threshold, that is, the target region has a less impact on the region of interest, the initial image classification label corresponding to the overlap rate is deleted from each initial image classification label, thereby generating a target image classification label for the medical image, that is, deleting the initial image classification label of the medical image.

[0092] In one embodiment, based on the first quantification method described above, the medical image can be further scored for effective information according to the overlap rate between the region of interest of the medical image and the target region corresponding to the initial image classification label, to obtain the effective information score of the medical image; for the overlap rate between the region of interest of the medical image and the target region corresponding to the initial image classification label, the overlap rate can be used as the invalid information score, then, based on this, the effective information score of the medical image can be equal to 1 minus the invalid information score. Accordingly, a preset effective information score threshold can be set for the effective information score of the medical image; when the effective information score of the medical image is greater than the preset effective information score threshold, it means that there is more effective information in the medical image, and the initial image classification label of the medical image can be deleted at this time; on the contrary, when the effective information score of the medical image is less than or equal to the preset effective information score threshold, it means that there is less effective information in the medical image, and the initial image classification label of the medical image can be retained at this time, so that when the image is screened based on the initial image classification label, the medical image with less effective information can be screened out, and the medical image with more effective information can be screened out for subsequent image detection and analysis.

[0093] Optionally, the effective information scoring result of the medical image may be obtained, in addition to the effective information score, as an effective information probability, which may be equal to 1 minus the invalid information score of the medical image and then multiplied by 100%, thereby obtaining the effective information probability of the medical image. It should be noted that the form of the effective information scoring result obtained after the effective information scoring of the medical image is performed in the embodiments of the present application is not specifically limited.

[0094] The second type: Figure 5 As shown, the above step 303 may include:

[0095] Step 501 : for each initial image classification label, calculate the overlap rate between the region of interest of the medical image and the target region corresponding to the initial image classification label.

[0096] The overlap ratio is a ratio of a first area of ​​an overlap region between the region of interest and the target region to a second area of ​​the region of interest.

[0097] Step 502, calculate the interference rate based on the first center point of the region of interest of the medical image, the second center point of the target region corresponding to the initial image classification label and the preset edge point of the medical image; the interference rate is used to characterize the degree of interference of the target region on the region of interest.

[0098] For example, the first distance can be determined according to the first center point of the region of interest of the medical image and the second center point of the target region corresponding to the initial image classification label, and the second distance can be determined according to the first center point of the region of interest of the medical image and the preset edge point of the medical image; then, the ratio of the first distance to the second distance is calculated, and the ratio is used as the interference rate of the medical image. The preset edge point of the medical image can be an edge point corresponding to the foreground region excluding the background in the medical image, and the edge point corresponding to the foreground region can be the edge point in the foreground region farthest from the center of the field of view, or the edge point in the foreground region farthest from the center point of the region of interest, or any edge point in the foreground region, etc.

[0099] Exemplarily, the first position information of the first center point of the region of interest, the second position information of the second center point of the target region corresponding to the initial image classification label, and the third position information of the preset edge point of the medical image can be first determined from the medical image; then, based on the first position information of the first center point of the region of interest of the medical image and the second position information of the second center point of the target region corresponding to the initial image classification label, the first distance between the first center point and the second center point is determined; and based on the first position information of the first center point of the region of interest of the medical image and the third position information of the preset edge point of the medical image, the second distance between the first center point and the preset edge point is determined; and then based on the ratio between the first distance and the second distance, the interference rate is calculated.

[0100] Step 503: Calculate the influence of the target area corresponding to the initial image classification label on the region of interest according to the overlap rate corresponding to the initial image classification label and the interference rate corresponding to the initial image classification label.

[0101] Exemplarily, the effective information probability of the medical image can be determined based on the overlap rate corresponding to the initial image classification label, and then the degree of influence of the target area corresponding to the initial image classification label on the area of ​​interest can be calculated based on the effective information probability of the medical image and the interference rate corresponding to the initial image classification label.

[0102] Exemplarily, a first preset coefficient corresponding to the effective information probability and a second preset coefficient corresponding to the interference rate may be obtained; then, based on the effective information probability of the medical image, the first preset coefficient corresponding to the effective information probability, the interference rate corresponding to the initial image classification label, and the second preset coefficient corresponding to the interference rate, a weighted summation method is used to calculate the degree of influence of the target area corresponding to the initial image classification label on the region of interest. Optionally, after performing the weighted summation, the result of the weighted summation may be multiplied by 100% to obtain the degree of influence of the target area corresponding to the initial image classification label on the region of interest.

[0103] The numerical range of the first preset coefficient and the second preset coefficient is between 0 and 1, and the sum of the first preset coefficient and the second preset coefficient is equal to 1; for example, the first preset coefficient can be 0.5, and the second preset coefficient can be 0.5; or, the first preset coefficient can be 0.7, and the second preset coefficient can be 0.3, etc. The embodiment of the present application does not limit the specific numerical values ​​of the first preset coefficient and the second preset coefficient, and the magnitude relationship between the first preset coefficient and the second preset coefficient.

[0104] Step 504: Correct each initial image classification label according to the magnitude relationship between the influence degree and a preset influence degree threshold, and generate a target image classification label for the medical image.

[0105] Exemplarily, when the image degree is greater than the preset impact degree threshold, it means that there is more valid information in the medical image, and the initial image classification label of the medical image can be deleted at this time; on the contrary, when the image degree is less than or equal to the preset impact degree threshold, it means that there is less valid information in the medical image, and the initial image classification label of the medical image can be retained at this time. In the same way, each initial image classification label of the medical image can be corrected to obtain the target image classification label of the medical image.

[0106] Figure 6 FIG. 1 is a flow chart of an image screening method in another embodiment. This embodiment relates to an optional implementation process of a computer device acquiring a region of interest of a medical image. Based on the above embodiment, Figure 6 As shown, the above step 302 includes:

[0107] Step 601, determining whether the medical image includes an object of interest.

[0108] The target of interest may include at least one of a surgical instrument, a tissue, and a lesion.

[0109] Exemplarily, the computer device can identify the target of interest in the medical image and determine whether the medical image includes the target of interest based on the recognition result. For example, the computer device can input the medical image into a preset target detection model to detect and identify the target of interest in the medical image, thereby obtaining a recognition result; wherein the target detection model may include but is not limited to a target detection model based on deep learning, such as: a target detection model based on yolo v5, a target detection model based on SSD, a target detection model based on Faster R-CNN, etc.; the embodiment of the present application does not specifically limit this.

[0110] Step 602: If the medical image includes an object of interest, a region of interest of the medical image is determined based on the object of interest.

[0111] Exemplarily, when determining that a medical image includes an object of interest, the computer device may segment the object of interest in the medical image, thereby determining a region of interest corresponding to the object of interest; the region of interest may be a regular region determined based on the object of interest, such as an anchor box, or an irregular region determined based on the object of interest, such as a connected domain. When segmenting the object of interest in the medical image, a preset target segmentation model may be used to segment the object of interest in the medical image, and the preset target segmentation model may include a semantic segmentation model based on deep learning, a target segmentation model based on an FCN neural network, and the like.

[0112] In addition, it should be noted that when the medical image includes multiple targets of interest, the region of interest corresponding to each target of interest can be determined based on each target of interest, or the regions of interest corresponding to each target of interest can be fused to obtain the region of interest corresponding to the medical image, or one of the regions of interest can be selected from the regions of interest corresponding to each target of interest as the region of interest corresponding to the medical image.

[0113] Exemplarily, the computer device may also determine the target position of the target of interest in the medical image based on a preset target segmentation model, and then determine the region of interest of the medical image based on the target position of the target of interest in the medical image; or, the computer device may first determine the initial region of interest of the medical image based on the target position of the target of interest in the medical image, and then magnify the initial region of interest according to a preset magnification ratio, thereby obtaining the region of interest of the medical image.

[0114] Step 603: If the medical image does not include the object of interest, a preset area in the medical image is used as the area of ​​interest of the medical image.

[0115] Exemplarily, the preset area can be the central area of ​​the medical image, the central area of ​​the foreground area of ​​the medical image, or a user-defined image area, etc.; the embodiment of the present application does not specifically limit the position and size of the preset area.

[0116] In this embodiment, when obtaining the region of interest of a medical image, the computer device can first determine whether the medical image includes a target of interest, and when the medical image includes the target of interest, determine the region of interest of the medical image based on the target of interest, and when the medical image does not include the target of interest, use a preset area in the medical image as the region of interest of the medical image; in this way, the region of interest can be determined for each frame of the medical image in the medical video to be processed, so that when the position of the target of interest in the medical video to be processed changes, the region of interest of each frame of the medical image can be accurately obtained, thereby improving the accuracy of video structured analysis, and further improving the accuracy of image screening.

[0117] In one embodiment, Figure 7 As shown, a complete embodiment of an image screening method is provided, comprising the following steps:

[0118] Step 701, obtaining an initial image classification result of at least one frame of medical image in a medical video to be processed; the initial image classification result includes at least one initial image classification label.

[0119] Optionally, each frame of medical images in the medical video to be processed can be input into a preset deep learning network model for multi-target detection, so as to obtain an initial image classification label for each frame of medical images; the initial image classification label includes but is not limited to normal, blurred, occluded, smoke, uneven lighting, in vitro, no instrument, etc.

[0120] Step 702 : for each frame of medical image, obtaining from the medical image a target region corresponding to each initial image classification label corresponding to the medical image, and obtaining a region of interest of the medical image.

[0121] Optionally, when the computer device performs multi-target detection through the above-mentioned deep learning network model to obtain the initial image classification label of the medical image, it can further detect the invalid information area corresponding to each initial image classification label through the deep learning network model; Figure 8 As shown, the invalid information area corresponding to the initial image classification label can be Figure 8 The a1 area shown in is the target area corresponding to the initial image classification label.

[0122] Optionally, refer to Figure 8As shown, when acquiring the region of interest of a medical image, the target of interest in the medical image, such as the tip of a surgical instrument, can be identified first, thereby obtaining an initial region of interest a2; then, the initial region of interest a2 can be enlarged in equal proportion according to a preset magnification ratio to obtain a region of interest a3 of the medical image; the preset magnification ratio can include 1 times, 2 times, 3 times, etc. In the case that the target of interest is not included in the medical image, the central area of ​​the medical image can be used as the region of interest of the medical image.

[0123] Step 703 : For each initial image classification label of the medical image, calculate the overlap rate between the region of interest of the medical image and the target region corresponding to the initial image classification label.

[0124] The overlap rate is a ratio of a first area of ​​an overlap region between the region of interest and the target region to a second area of ​​the region of interest.

[0125] Step 704 , calculating the interference rate according to the first center point of the region of interest of the medical image, the second center point of the target region corresponding to the initial image classification label, and the preset edge points of the medical image.

[0126] The interference rate is used to characterize the degree of interference of the target area on the area of ​​interest.

[0127] Optionally, the first distance between the first center point and the second center point can be determined based on the first position information of the first center point of the region of interest of the medical image and the second position information of the second center point of the target region corresponding to the initial image classification label; and the second distance between the first center point and the preset edge point can be determined based on the first position information of the first center point of the region of interest of the medical image and the third position information of the preset edge point of the medical image; then, the interference rate is calculated based on the ratio between the first distance and the second distance.

[0128] Step 705 , calculating the influence of the target area corresponding to the initial image classification label on the region of interest according to the overlap rate corresponding to the initial image classification label and the interference rate corresponding to the initial image classification label.

[0129] Optionally, refer to Figure 8 As shown, formula (1) can be used to calculate the influence of the target area corresponding to the initial image classification label on the region of interest.

[0130]

[0131] Among them, score is used to characterize the degree of influence, and can also be used to characterize the probability score of valid information contained in the medical image; a4 is used to characterize the first area of ​​the overlapping area a4 between the region of interest a3 and the target area a1; a3 is used to characterize the second area of ​​the region of interest a3; a4 / a3 is used to characterize the degree to which the invalid information area occludes the region of interest, and k1 is used to characterize the occlusion coefficient; d1 is used to characterize the first distance between the first center point p2 of the region of interest a3 and the second center point p1 of the target area a1; d2 is used to characterize the second distance between the first center point p2 of the region of interest a3 and the preset edge point p3 of the medical image; d1 / d2 is used to characterize the degree to which the invalid information area interferes with the region of interest, and k2 is used to characterize the anti-interference coefficient.

[0132] It should be noted that the value range of the shielding coefficient k1 and the anti-interference coefficient k2 is between 0 and 1, and the sum of the shielding coefficient k1 and the anti-interference coefficient k2 is 1. In addition, Figure 8 The white area of ​​the traditional Chinese medical image is the foreground area, that is, the field of view area of ​​the captured image, and the black area is the background area. The preset edge point p3 of the medical image can be the edge point in the field of view area that is farthest from the first center point p2 of the area of ​​interest a3, and of course, it can also be other edge points in the field of view area.

[0133] Step 706, determining whether the impact degree is greater than a preset impact degree threshold.

[0134] Step 707: if the influence degree is greater than a preset influence degree threshold, the initial image classification labels corresponding to the overlap rate are deleted from each initial image classification label to generate a target image classification label for the medical image.

[0135] Optionally, for medical images with blur, occlusion, smoke, uneven illumination and other labels, the initial image classification labels of the medical images can be corrected in combination with the image levels (or effective information probability scores) corresponding to each label; for example, when a medical image includes labels of smoke and uneven illumination, the image level of the smoke can be determined. If the image level of the smoke exceeds the preset impact level threshold, it means that the smoke area does not block the surgical operation area, and the medical image is still a valid video frame and is not classified as a smoke frame; similarly, the image level of uneven illumination can be determined. If the image level of uneven illumination does not exceed the preset impact level threshold, it means that the uneven illumination area blocks the surgical operation area, and the medical image is an invalid video frame and should be classified as an uneven illumination frame. This allows the medical image frame to be removed when performing uneven illumination removal in the subsequent structured analysis.

[0136] Step 708: If the influence degree is less than or equal to the preset influence degree threshold, the initial image classification label corresponding to the overlap rate is used as the target image classification label of the medical image to generate the target image classification label of the medical image.

[0137] Step 709 , based on the target image classification results of each frame of medical image and the preset image screening condition, the target medical image that meets the preset image screening condition is screened out from each frame of medical image.

[0138] Among them, the target image classification result of the medical image may include at least one target image classification label. The user can arbitrarily select the label of each frame of the medical image in the medical video to be processed, so as to realize the screening and selection of specific segments in the medical video to be processed, which can be used for automatic rough editing of the video during surgery review or teaching communication. For example: the user can select a label, and the computer device can remove the medical image frames including the label in the medical video to be processed according to the label selected by the user, and only include the medical image frames that do not include the label; the user can also select multiple labels at the same time, and the computer device can remove the medical image frames including each label in the medical video to be processed according to the multiple labels selected by the user, and retain the medical image frames that do not include the multiple labels.

[0139] In the embodiment of the present application, the method based on deep learning target detection detects the invalid information area in each frame of the medical image in the medical video to be processed, and introduces the definition of the region of interest. The region of interest of each frame of the medical image is detected by the target detection method; then, the impact of the invalid information area detection result on the region of interest is evaluated by the effective information scoring method to achieve the purpose of video structured analysis. Compared with the traditional method, the method proposed in the embodiment of the present application analyzes the characteristics of the surgical process and innovates for the medical surgical scene. It is more in line with the needs of clinicians, the frame classification is more accurate, and the obtained video structured information is more comprehensive, which can improve the accuracy of medical image screening.

[0140] In addition, multi-label classification, invalid information area selection, and area of ​​interest selection of video frame images are realized based on deep learning target detection methods. Deep learning target detection methods have stronger feature extraction capabilities than machine learning (ML) technology, can discover deep features of images, and have higher accuracy in frame classification and label selection; at the same time, target detection methods can locate label positions compared to target recognition methods, and obtain richer video structured information. Compared with traditional single-label classification, multi-label classification allows a frame to have multiple labels, and there is no mutual exclusion of labels between frames, which can better handle the simultaneous occurrence of multiple events in complex surgical scenarios.

[0141] Furthermore, the definition of the region of interest is introduced. In medical images, the region of interest to doctors usually only occupies a small part of the medical image, which contains important surgical information. Due to the complexity of the surgical scene, it is inevitable that medical image frames with poor quality such as smoke, blur, and uneven lighting will appear, but these medical image frames may still retain important surgical information. For example, the appearance of smoke is often accompanied by important surgical actions such as hemostasis that the doctor is performing. If the invalid information area does not affect the region of interest, it cannot be used as a smoke frame and is filtered out together with the smoke frame with no valid information through the smoke label. When classifying frames, the influence of the region of interest is referred to, and the frame classification is more accurate and more in line with the needs of doctors.

[0142] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0143] Based on the same inventive concept, the embodiment of the present application also provides an image screening device for implementing the above-mentioned image screening method. The implementation solution provided by the device to solve the problem is similar to the implementation solution recorded in the above-mentioned method, so the specific limitations in one or more image screening device embodiments provided below can refer to the limitations on the image screening method above, and will not be repeated here.

[0144] In one embodiment, Fig. 9 As shown, an image screening device is provided, including: an acquisition module 901, a generation module 902 and a screening module 903, wherein:

[0145] The acquisition module 901 is used to acquire an initial image classification result of at least one frame of medical image in the medical video to be processed.

[0146] The generation module 902 is used to correct the initial image classification result of the medical image according to the region of interest of the medical image and the target region corresponding to the initial image classification result of the medical image for each frame of the medical image, and generate a target image classification result of the medical image.

[0147] The screening module 903 is used to screen out target medical images that meet the preset image screening conditions from each frame of medical images according to the target image classification results of each frame of medical images and the preset image screening conditions.

[0148] In one embodiment, the initial image classification result includes at least one initial image classification label; the above-mentioned generation module 902 includes a first acquisition submodule, a second acquisition submodule and a generation submodule; wherein the first acquisition submodule is used to acquire a target area corresponding to each initial image classification label from the medical image; the second acquisition submodule is used to acquire the region of interest of the medical image; the generation submodule is used to correct each initial image classification label according to the region of interest of the medical image and the target area corresponding to each initial image classification label, generate a target image classification label for the medical image, and use the target image classification label as the target image classification result.

[0149] In one embodiment, the above-mentioned generation submodule includes a calculation unit and a generation unit; wherein the calculation unit is used to calculate the overlap rate between the region of interest of the medical image and the target region corresponding to the initial image classification label for each initial image classification label; the overlap rate is the ratio of the first area of ​​the overlap area between the region of interest and the target area to the second area of ​​the region of interest; the generation unit is used to correct each initial image classification label according to the size relationship between each overlap rate and a preset overlap rate threshold, and generate a target image classification label for the medical image.

[0150] In one of the embodiments, a generation unit is used to, for each overlap rate, if the overlap rate is less than or equal to a preset overlap rate threshold, use the initial image classification label corresponding to the overlap rate as the target image classification label of the medical image; if the overlap rate is greater than the preset overlap rate threshold, delete the initial image classification label corresponding to the overlap rate from each initial image classification label to generate a target image classification label for the medical image.

[0151] In one of the embodiments, the above-mentioned calculation unit is also used to calculate the overlap rate between the region of interest of the medical image and the target region corresponding to the initial image classification label for each initial image classification label; calculate the interference rate based on the first center point of the region of interest of the medical image, the second center point of the target region corresponding to the initial image classification label and the preset edge point of the medical image; the interference rate is used to characterize the degree of interference of the target region on the region of interest; calculate the degree of influence of the target region corresponding to the initial image classification label on the region of interest based on the overlap rate corresponding to the initial image classification label and the interference rate corresponding to the initial image classification label; the generation unit is also used to correct each initial image classification label based on the size relationship between the influence degree and the preset influence degree threshold, and generate a target image classification label for the medical image.

[0152] In one embodiment, the above-mentioned calculation unit is used to determine the first distance between the first center point and the second center point based on the first position information of the first center point of the region of interest of the medical image and the second position information of the second center point of the target region corresponding to the initial image classification label; determine the second distance between the first center point and the preset edge point based on the first position information of the first center point of the region of interest of the medical image and the third position information of the preset edge point of the medical image; and calculate the interference rate based on the ratio between the first distance and the second distance.

[0153] In one embodiment, the second acquisition submodule includes a judgment unit and a determination unit; wherein the judgment unit is used to judge whether the medical image includes a target of interest; the determination unit is used to determine the region of interest of the medical image based on the target of interest when the medical image includes the target of interest; and when the medical image does not include the target of interest, use a preset region in the medical image as the region of interest of the medical image.

[0154] In one of the embodiments, the determination unit is used to determine an initial region of interest of the medical image based on a target position of the target of interest in the medical image; and to magnify the initial region of interest according to a preset magnification ratio to obtain the region of interest of the medical image.

[0155] Each module in the above-mentioned image screening device can be implemented in whole or in part by software, hardware or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each module.

[0156] In one embodiment, a computer device is provided. The computer device may be a terminal or a server. Taking a terminal as an example, its internal structure diagram may be as follows: Fig.10As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, an image screening method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0157] Those skilled in the art will understand that Fig.10 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0158] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the steps of the image screening method in any of the above embodiments are implemented.

[0159] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the image screening method in any of the above embodiments are implemented.

[0160] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the steps of the image screening method in any of the above embodiments.

[0161] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0162] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0163] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0164] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. An image screening method, characterized in that: The method comprises: Obtaining an initial image classification result of at least one frame of medical image in the medical video to be processed; For each frame of the medical image, correcting the initial image classification result of the medical image according to the region of interest of the medical image and the target region corresponding to the initial image classification result of the medical image, and generating a target image classification result of the medical image; According to the target image classification results of the medical images of each frame and the preset image screening conditions, the target medical images that meet the preset image screening conditions are screened out from the medical images of each frame.

2. The method according to claim 1, characterized in that The initial image classification result includes at least one initial image classification label; and the correcting the initial image classification result of the medical image according to the region of interest of the medical image and the target region corresponding to the initial image classification result of the medical image to generate the target image classification result of the medical image includes: Acquire a target region corresponding to each of the initial image classification labels from the medical image; Acquiring a region of interest of the medical image; According to the region of interest of the medical image and the target region corresponding to each of the initial image classification labels, each of the initial image classification labels is corrected to generate a target image classification label of the medical image, and the target image classification label is used as the target image classification result.

3. The method according to claim 2, characterized in that The step of correcting each of the initial image classification labels according to the region of interest of the medical image and the target region corresponding to each of the initial image classification labels to generate a target image classification label of the medical image includes: For each of the initial image classification labels, calculating the overlap rate between the region of interest of the medical image and the target region corresponding to the initial image classification label; the overlap rate is a ratio of a first area of ​​an overlap region between the region of interest and the target region to a second area of ​​the region of interest; According to the size relationship between each of the overlap rates and the preset overlap rate threshold, each of the initial image classification labels is corrected to generate a target image classification label for the medical image.

4. The method according to claim 3, characterized in that The method of correcting each of the initial image classification labels according to the size relationship between each of the overlap rates and the preset overlap rate threshold to generate a target image classification label for the medical image includes: For each of the overlap rates, if the overlap rate is greater than or equal to the preset overlap rate threshold, taking the initial image classification label corresponding to the overlap rate as the target image classification label of the medical image; If the overlap rate is less than the preset overlap rate threshold, the initial image classification label corresponding to the overlap rate is deleted from each of the initial image classification labels to generate a target image classification label for the medical image.

5. The method according to claim 2, characterized in that: The step of correcting each of the initial image classification labels according to the region of interest of the medical image and the target region corresponding to each of the initial image classification labels to generate a target image classification label of the medical image includes: For each of the initial image classification labels, calculating the overlap rate between the region of interest of the medical image and the target region corresponding to the initial image classification label; Calculating an interference rate according to a first center point of the region of interest of the medical image, a second center point of the target region corresponding to the initial image classification label, and a preset edge point of the medical image; the interference rate is used to characterize the degree of interference of the target region on the region of interest; Calculating the influence of the target area corresponding to the initial image classification label on the region of interest according to the overlap rate corresponding to the initial image classification label and the interference rate corresponding to the initial image classification label; According to the magnitude relationship between the influence degree and a preset influence degree threshold, each of the initial image classification labels is corrected to generate a target image classification label for the medical image.

6. The method according to claim 5, characterized in that The calculating the interference rate according to the first center point of the region of interest of the medical image, the second center point of the target region corresponding to the initial image classification label, and the preset edge point of the medical image includes: Determining a first distance between the first center point and the second center point according to first position information of a first center point of the region of interest of the medical image and second position information of a second center point of the target region corresponding to the initial image classification label; Determining a second distance between the first center point and the preset edge point according to first position information of a first center point of the region of interest of the medical image and third position information of a preset edge point of the medical image; The interference rate is calculated according to a ratio between the first distance and the second distance.

7. The method according to claim 2, characterized in that The step of acquiring the region of interest of the medical image comprises: Determining whether the medical image includes an object of interest; If the medical image includes the object of interest, determining a region of interest of the medical image based on the object of interest; If the medical image does not include the object of interest, a preset area in the medical image is used as the area of ​​interest of the medical image.

8. The method according to claim 7, characterized in that The determining the region of interest of the medical image based on the object of interest comprises: Determining an initial region of interest of the medical image based on a target position of the target of interest in the medical image; The initial region of interest is enlarged according to a preset enlargement ratio to obtain the region of interest of the medical image.

9. An image screening device, characterized in that: The device comprises: An acquisition module, used to acquire an initial image classification result of at least one frame of medical image in the medical video to be processed; A generating module, configured to correct the initial image classification result of the medical image according to the region of interest of the medical image and the target region corresponding to the initial image classification result of the medical image for each frame of the medical image, and generate a target image classification result of the medical image; The screening module is used to screen out target medical images that meet the preset image screening conditions from the medical images of each frame according to the target image classification results of the medical images of each frame and the preset image screening conditions.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

12. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Citation Information

Patent Citations

  • Depth map processing method and device and electronic equipment

    CN110378945A

  • Image diagnosis support system and image diagnosis support method

    US20200342598A1

  • Medical image processing apparatus and method

    US20210350534A1