Training method of video identification network, and video identification method and system

Abstract

The invention relates to the technical field of image processing, in particular to a training method of a video recognition network and a video recognition method and system.The training method of the video recognition network comprises the steps that an enteroscope video sample is obtained; the enteroscope video sample comprises video clips shot for a plurality of ileocecal parts in an ileocecal part area; obtaining label data corresponding to the enteroscope video sample; the label data is used for representing the ileocecal part contained in each video frame in the video clip, or is used for representing whether each video frame in the video clip contains the ileocecal part area or not; based on the enteroscope video sample and the label data, training an initial recognition network to obtain a trained video recognition network; and the video identification network is used for identifying the ileocecum region.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a video recognition network training method, a video recognition method and a system. Background Art

[0002] Whether the colonoscopy has successfully reached the ileocecal region is one of the key evaluation indicators of the quality of colonoscopy. At present, the main option is to identify the images taken during the colonoscopy to determine whether the colonoscopy has successfully reached the ileocecal region.

[0003] However, in actual applications, it is found that during colonoscopy, there are many interference factors that affect image capture, which often results in poor effectiveness of the captured images. Therefore, the accuracy of image-based recognition of the ileocecal region is low. Summary of the invention

[0004] Based on this, it is necessary to provide a video recognition network training method, a colonoscopy video recognition method and system to address the above technical problems.

[0005] In a first aspect, the present application provides a method for training a video recognition network, the method comprising:

[0006] Obtaining a colonoscopy video sample; the colonoscopy video sample includes video clips shot at multiple ileocecal areas in the ileocecal region;

[0007] Obtaining label data corresponding to the colonoscopy video sample; the label data is used to characterize the ileocecal region contained in each video frame in the video segment, or to characterize whether each video frame in the video segment contains the ileocecal region;

[0008] Based on the colonoscopy video samples and the label data, an initial recognition network is trained to obtain a trained video recognition network; the video recognition network is used to identify the ileocecal region.

[0009] In one of the embodiments, the shooting order of the multiple ileocecal parts in the ileocecal region in the video clip is related to the inspection order of the ileocecal parts during the colonoscopy.

[0010] In one embodiment, the ileocecal area includes at least two areas of the ileocecal valve, the ileocecal orifice, the appendix orifice, and the cecal fundus.

[0011] In one of the embodiments, when the label data is used to characterize the ileo-blind region contained in each video frame in the video segment, the label data is also used to characterize the region where the ileo-blind region is located in the video frame.

[0012] In one embodiment, the shooting angles for shooting the multiple ileocecal parts in the ileocecal region in the video clips include multiple angles, and the shooting angles for shooting the various ileocecal parts in different video clips are not completely the same.

[0013] In one embodiment, the changing process of the shooting angle includes at least one of from near to far, from far to near, from left to right, from right to left, from outside to inside, and from inside to outside.

[0014] In one embodiment, the shooting distances for shooting each ileocecal region in the video clips include multiple distances, and the shooting distances for shooting each ileocecal region in different video clips are not completely the same.

[0015] In one of the embodiments, the movement patterns of the ileocecal regions contained in different video clips are not completely the same.

[0016] In one embodiment, the movement state includes a physiological activity state in which the ileocecal region is deformed due to intestinal peristalsis or respiratory movement.

[0017] In one embodiment, the colonoscopy video sample is an original colonoscopy video sample that has been preprocessed.

[0018] In one embodiment, the pre-processing includes at least one of removing invalid areas, de-redundancy processing, and removing non-information frames.

[0019] In a second aspect, the present application also provides a video recognition method, the method comprising:

[0020] Get the video clip to be processed;

[0021] The video segment to be processed is processed based on a video recognition network to determine whether the video segment to be processed contains an ileocecal region; the video recognition network is a video segment that trains an initial recognition network based on a colonoscopy video sample and label data corresponding to the colonoscopy video sample to obtain a trained video recognition network; the label data is used to characterize the ileocecal region contained in each video frame in the video segment, or to characterize whether each video frame in the video segment contains the ileocecal region; the colonoscopy video sample contains video segments shot for multiple ileocecal regions in the ileocecal region.

[0022] In one of the embodiments, the method further includes: playing the video clip to be processed in a visual interface; after processing the video clip to be processed based on the video recognition network to determine whether the video clip to be processed includes the ileocecal region, it also includes: based on the playback progress of the video clip to be processed, displaying prompt information corresponding to the target video frame in the video clip to be processed; the prompt information is used to characterize the ileocecal region contained in the target video frame.

[0023] In a third aspect, the present application also provides a video recognition network training system, the system comprising:

[0024] A colonoscopy video sample acquisition unit, used to acquire a colonoscopy video sample; the colonoscopy video sample includes video clips shot at multiple ileocecal areas in the ileocecal region;

[0025] A label data acquisition unit, used to acquire label data corresponding to the colonoscopy video sample; the label data is used to characterize the ileocecal region contained in each video frame in the video segment, or to characterize whether each video frame in the video segment contains the ileocecal region;

[0026] The video recognition network training unit is used to train the initial recognition network based on the colonoscopy video sample and the label data to obtain a trained video recognition network; the video recognition network is used to identify the ileocecal region.

[0027] In a fourth aspect, the present application further provides a video recognition system, the system comprising:

[0028] A to-be-processed video segment acquisition unit, used to acquire the to-be-processed video segment;

[0029] The identification unit of the video segment to be processed is used to process the video segment to be processed based on a video recognition network to determine whether the video segment to be processed contains an ileocecal region; the video recognition network is a video recognition network that is trained based on colonoscopy video samples and label data corresponding to the colonoscopy video samples to obtain a trained video recognition network; the label data is used to characterize the ileocecal region contained in each video frame in the video segment, or to characterize whether each video frame in the video segment contains the ileocecal region; the colonoscopy video samples contain video segments shot for multiple ileocecal regions in the ileocecal region.

[0030] In a fifth aspect, the present application further provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the training method of the video recognition network or the steps of the video recognition method when executing the computer program.

[0031] In a sixth aspect, the present application also provides an endoscope system, including an endoscope, a display, a light source device and the above-mentioned computer device.

[0032] In a seventh aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the training method of the video recognition network or the steps of the video recognition method.

[0033] In an eighth 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 computer program implements the steps of the training method of the video recognition network or the steps of the video recognition method.

[0034] The training method, video recognition method and system of the above-mentioned video recognition network first obtain a colonoscopy video sample containing a video clip shot for multiple ileocecal parts in the ileocecal region. Then, obtain the label data corresponding to the colonoscopy video sample, and the label data can be used to characterize the ileocecal parts contained in each video frame in the video clip, or whether each video frame in the video clip contains the ileocecal region. Finally, based on the aforementioned colonoscopy video sample and the aforementioned label data, the initial recognition network is trained to obtain a video recognition network for identifying the ileocecal region in each video frame. The present application effectively expands the number of ileocecal region features that can be recognized by the video recognition network by using a video clip containing a video clip shot for multiple ileocecal parts in the ileocecal region as a colonoscopy video sample for training the video recognition network, thereby effectively improving the video recognition network's ability to recognize the ileocecal region. Therefore, the accuracy of identifying the ileocecal region in each video frame of the input video clip based on the video recognition network can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 A schematic diagram of a flow chart of a method for training a video recognition network provided in an embodiment;

[0036] Figure 2 A schematic diagram of a specific composition structure of the ileocecal region provided in an embodiment;

[0037] Figure 3 A schematic diagram of a specific process of colonoscopy provided in an embodiment;

[0038] Figure 4 A schematic diagram of a flow chart of a video recognition method provided in an embodiment;

[0039] Figure 5A flowchart of a specific method for displaying prompt information corresponding to a target video frame in a video segment to be processed provided in an embodiment;

[0040] Figure 6 A structural block diagram of a video recognition network training system provided in one embodiment;

[0041] Figure 7 is a structural block diagram of a video recognition system provided in one embodiment;

[0042] Figure 8 An internal structure diagram of a computer device provided in one embodiment;

[0043] Fig. 9 An internal structure diagram of another computer device provided in an embodiment. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present application. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application.

[0045] When used herein, the singular forms "a", "an", and "said / the" may also include plural forms, unless the context clearly indicates otherwise. It should also be understood that the terms "include / comprise" or "have" and the like specify the presence of stated features, wholes, steps, operations, components, parts, or combinations thereof, but do not exclude the possibility of the presence or addition of one or more other features, wholes, steps, operations, components, parts, or combinations thereof. At the same time, the term "and / or" used in this specification includes any and all combinations of the relevant listed items.

[0046] At the same time, unless otherwise specified, the "multiple" in the embodiments of the present application refers to two or more.

[0047] The video recognition network training method and video recognition method provided in the embodiments of the present application can be applied to a terminal or a server for execution. The terminal can communicate with the server through a network; the data storage system can store data that the server needs to process; the data storage system can be integrated on the server, or placed on the cloud or other network servers; the terminal can be, but is not limited to, various personal computers, laptops, tablet computers, image processors, and AI boxes, etc.; the server can be implemented as an independent server or a server cluster consisting of multiple servers.

[0048] In one embodiment, Figure 1 As shown, a training method for a video recognition network is provided, comprising the following steps:

[0049] Step S110, obtaining a colonoscopy video sample; the aforementioned colonoscopy video sample includes video clips shot at multiple ileocecal areas in the ileocecal region.

[0050] In this step, the colonoscopy video sample refers to a video sample set containing video clips shot at multiple ileocecal areas in the ileocecal region. The ileocecal region refers to the area where the terminal ileum and the cecum meet, and the ileocecal region refers to the physiological area containing the ileocecal region. The ileocecal area refers to the characteristic area of ​​the ileocecal region. Generally speaking, the ileocecal region contains the ileocecal valve, ileocecal orifice, appendix orifice, cecal base and other anatomical characteristic areas (i.e., ileocecal area).

[0051] Specifically, the colonoscopy video sample may include one or more video clips, wherein each video clip captures multiple ileocecal regions of the ileocecal region. At the same time, the objects captured by each video clip may be the same or different. For example, the objects captured by multiple video clips may be different patients or the same patient.

[0052] As an embodiment of the present application, a specific form of the multiple video clips may be multiple video clips shot at multiple ileocecal areas within the ileocecal region of each patient during colonoscopy examinations of different patients.

[0053] In practical applications, the specific composition and structure of the ileocecal region can be as follows Figure 2 The form shown.

[0054] Step S120, obtaining label data corresponding to the colonoscopy video sample; the aforementioned label data is used to characterize the ileocecal region contained in each video frame in the video segment, or to characterize whether each video frame in the video segment contains the ileocecal region.

[0055] In this step, the label data corresponding to the colonoscopy video sample refers to the label data corresponding to the specific content represented by each video frame of the video clip. Specifically, the specific form of the label data corresponding to the colonoscopy video sample can be the first label data used to represent the ileocecal region contained in each video frame in the video clip. At this time, the first label data can include the specific name of the ileocecal region. The label data can also be the second label data used to represent whether each video frame in the video clip contains the ileocecal region.

[0056] In practical applications, the specific method of obtaining the label data corresponding to the colonoscopy video sample can be manual labeling (i.e., using manual labeling to label the ileocecal region contained in each video frame in the video clip, or whether each video frame in the video clip contains relevant information about the ileocecal region, respectively in the corresponding video frame), or labeling by a method other than manual labeling. The label data corresponding to each video frame in the video clip may include both the aforementioned first label data and the aforementioned second label data, or may include only any one of the aforementioned first label data and the aforementioned second label data.

[0057] Exemplarily, in the case where the label data includes the aforementioned first label data, the initial recognition network is trained based on the colonoscopy video sample and the label data to obtain the trained video recognition network. It is possible to determine whether the input video segment contains the ileocecal region by inputting a video segment to the video recognition network and determining whether each video frame of the input video segment contains any ileocecal region (i.e., as long as each video frame of the input video segment contains one of the ileocecal regions, it can be determined that the input video segment contains the ileocecal region). In the case where the label data includes the aforementioned second label data, the initial recognition network is trained based on the colonoscopy video sample and the label data to obtain the trained video recognition network. It is possible to determine whether the input video segment contains the ileocecal region by inputting a video segment to the video recognition network and determining whether each video frame of the input video segment contains a certain ileocecal region in the ileocecal region (i.e., when each video frame of the input video segment contains any ileocecal region in the ileocecal region, it can be determined that the input video segment contains the ileocecal region).

[0058] Step S130, based on the colonoscopy video samples and label data, the initial recognition network is trained to obtain a trained video recognition network; the aforementioned video recognition network is used to identify the ileocecal region.

[0059] In this step, the initial recognition network refers to an untrained video recognition network; the trained video recognition network refers to a trained video recognition network obtained after the initial recognition network is trained based on colonoscopy video samples and label data.

[0060] Specifically, the specific form of the video recognition network can be a three-dimensional convolutional neural network model 3D-CNN (Three-Dimensional Convolutional Neural Networks, 3D-CNN) that can process three-dimensional input data, or a deep learning network model (CNN-LSTM) that combines convolutional neural networks (CNN) and long short-term memory networks (LSTM), or other forms of network models other than the aforementioned two network models.

[0061] In practical applications, when the existing technology is used to identify a single image taken during a colonoscopy to determine whether the colonoscopy has successfully reached the ileocecal region, since the effectiveness of a single image will be affected by interference factors such as the endoscope field of view, shooting angle, intestinal peristalsis, etc., errors often occur in the results of identifying the ileocecal region based on a single image (for example, misidentifying the diverticulum as the appendix orifice). When the technical solution of the present application is used, since the video recognition network is trained based on a colonoscopy video sample containing video clips taken for multiple ileocecal parts in the ileocecal region, the number of sample features that can be used to identify the ileocecal region is greater. Therefore, compared with the existing technology, the video recognition network obtained based on the technical solution of the present application has a stronger ability to recognize the ileocecal region. Therefore, the result of identifying the ileocecal region is more accurate.

[0062] The training method of the above-mentioned video recognition network, first, obtains a colonoscopy video sample containing video clips shot at multiple ileocecal parts in the ileocecal region. Then, obtain the label data corresponding to the colonoscopy video sample, and the label data can be used to characterize the ileocecal parts contained in each video frame in the video clip, or whether each video frame in the video clip contains the ileocecal region. Finally, based on the aforementioned colonoscopy video sample and the aforementioned label data, the initial recognition network is trained to obtain a video recognition network for identifying the ileocecal region in each video frame. The present application effectively expands the number of ileocecal region features that can be recognized by the video recognition network by using video clips containing multiple ileocecal parts shot at the ileocecal region as the colonoscopy video samples for training the video recognition network, thereby effectively improving the recognition ability of the obtained video recognition network for the ileocecal region. Therefore, the accuracy of identifying the ileocecal region in each video frame of the input video clip based on the video recognition network can be improved.

[0063] Regarding the shooting order of the video clips, in one embodiment, the shooting order of the multiple ileocecal parts in the ileocecal region in the above video clips is related to the inspection order of each ileocecal part during the colonoscopy.

[0064] Among them, the shooting order, that is, the shooting order when shooting multiple ileocecal parts in the ileocecal region, can be determined based on the inspection order corresponding to each ileocecal part during the colonoscopy. For example, the shooting order can be the same as or opposite to the inspection order of each ileocecal part during the colonoscopy. Each ileocecal part refers to each ileocecal part that needs to be inspected in the ileocecal region during the colonoscopy.

[0065] Specifically, during each colonoscopy, the order of examining the various ileocecal parts in the ileocecal region of the patient may be different (for example, after the endoscope reaches the ileocecal opening of the patient, it can choose to continue to move in the direction of the ileum, or choose to continue to move in the direction of the appendix opening. Choosing different directions of advancement will result in a certain difference in the order of examining the various ileocecal parts). Therefore, there may be multiple shooting orders for the ileocecal parts in each video clip. That is, in each video clip, the shooting order for shooting multiple ileocecal parts in the ileocecal region may also be different.

[0066] As a specific embodiment of the present application, the colonoscopy sequence can be specifically expressed as follows: Figure 3 At this time, the shooting order of the ileocecal part in the video clip can be the same as the inspection order, that is, the ileocecal valve, ileocecal opening, ileum, ileocecal opening, appendix opening and cecal base, or it can be the opposite of the inspection order, that is, the cecal base, appendix opening, ileocecal opening, ileum, ileocecal opening and ileocecal valve.

[0067] In the embodiment of the present application, a number of video clips shot when examining each ileocecal region during a colonoscopy are used as colonoscopy video samples for training a video recognition network, so that the shooting order of each ileocecal region in the video clips in the colonoscopy video samples is consistent with the actual endoscopic examination process, so that when the colonoscopy video samples are used to train the video recognition network, the video recognition network can learn more motion features that are consistent with the motion characteristics of the endoscope lens of the shooting sample, thereby effectively enhancing the accuracy of identifying the ileocecal region in each video frame of the input video clip based on the video recognition network in actual applications.

[0068] Regarding the specific parts included in the ileocecal area, in one embodiment, the ileocecal area includes at least two parts of the ileocecal valve, the ileocecal orifice, the appendix orifice, and the cecal fundus.

[0069] Among them, the ileocecal area refers to the various ileocecal areas that need to be examined in the ileocecal region during the colonoscopy. The ileocecal valve, ileocecal orifice, appendix orifice, and cecal base are all parts that can better reflect the anatomical structural characteristics of the ileocecal region.

[0070] Specifically, on the premise that the ileocecal region includes at least two of the ileocecal valve, ileocecal opening, appendix opening, and cecal base, the specific expression form of each video clip can be a video clip shot at at least two of the ileocecal valve, ileocecal opening, appendix opening, and cecal base in the ileocecal region of the patient.

[0071] The embodiment of the present application uses video clips shot at at least two ileocecal areas including the ileocecal valve, ileocecal orifice, appendix orifice, and cecal fundus as colonoscopy video samples for training a video recognition network, thereby effectively ensuring the data validity of the colonoscopy video samples used for training a video recognition network, thereby effectively improving the accuracy of identifying the ileocecal region in each video frame of the input video clip based on the video recognition network.

[0072] Regarding the specific content represented by the label data, in one embodiment, when the label data is used to represent the ileo-blind part contained in each video frame in the video clip, the label data is also used to represent the area where the ileo-blind part is located in the video frame.

[0073] Specifically, the specific manifestation of the ileocecal region contained in each video frame in the video clip may include the ileocecal valve, ileocecal orifice, appendix orifice, and cecal base in the ileocecal region, which can better reflect the anatomical structural characteristics of the ileocecal region. The specific manifestation of the region where the ileocecal region is located in the video frame is not limited here.

[0074] Exemplarily, the area where the ileocecal region is located in the video frame may be any one of the upper left area of ​​the video frame, the lower left area of ​​the video frame, the upper right area of ​​the video frame, the lower right area of ​​the video frame, and the middle area of ​​the video frame, or may be the area where the ileocecal region is located recorded in the form of coordinates or the like.

[0075] The embodiment of the present application effectively improves the accuracy of the trained video recognition network in identifying the ileocecal region in each video frame of the input video clip by adding identification content used to characterize the area where the ileocecal region is located in the video frame to the label data corresponding to the colonoscopy video samples used to train the video recognition network.

[0076] Regarding the shooting angles of the video clips, in one embodiment, the shooting angles for shooting the multiple ileocecal parts in the ileocecal region in the above video clips include multiple angles, and the shooting angles for shooting the various ileocecal parts in different video clips are not completely the same.

[0077] The shooting angle refers to the angle at which the ileocecal area within the ileocecal region is shot.

[0078] Specifically, the specific manifestations of the ileocecal region may include the ileocecal valve, ileocecal orifice, appendix orifice, and cecal base in the ileocecal region, which are parts that can better reflect the anatomical structural characteristics of the ileocecal region. The different shooting angles for shooting multiple ileocecal parts in the ileocecal region may be caused by the different specific operation methods during each colonoscopy and the different motion characteristics of the endoscope lens when collecting video clips.

[0079] In practical applications, under the premise that the shooting angles for shooting multiple ileocecal parts in the ileocecal region include multiple angles, and the shooting angles for shooting each ileocecal part in different video clips are not completely the same, each video clip included in the colonoscopy video sample can be shot at multiple shooting angles (that is, the shooting angle can change multiple times during the shooting of each video clip). Moreover, in each video clip included in the colonoscopy video sample, the shooting angles for shooting different ileocecal parts in the ileocecal region can also be different.

[0080] Exemplarily, for the ileocecal region A, a plurality of different shooting angles of the ileocecal region A may be included in the video segment A. At the same time, in different video segments, the shooting angles of the same ileocecal region may not be completely the same. For example, the shooting angle of the ileocecal region A in the video segment B may not be completely the same as the shooting angle of the ileocecal region A in the video segment A (for example, the shooting angle of the ileocecal region A in the video segment B may be more or less than the shooting angle of the ileocecal region A in the video segment A, etc.). The ileocecal region A may be any ileocecal region, and the video segment A and the video segment B may be any two different video segments in the colonoscopy video sample.

[0081] The embodiment of the present application uses multiple video clips shot at multiple ileocecal areas in the ileocecal region at different shooting angles as colonoscopy video samples for training a video recognition network. This allows the video recognition network to learn more spatial information that is consistent with the motion characteristics of the endoscope lens that shot the samples when the colonoscopy video samples are used to train the video recognition network, thereby effectively ensuring the robustness of the video recognition network trained based on the colonoscopy video samples, and further effectively improving the accuracy of the video recognition network in identifying the ileocecal region in each video frame of the input video clip.

[0082] Regarding the change process of the shooting angle, in one embodiment, for the same ileocecal part, the change process of the shooting angle includes at least one of from near to far, from far to near, from left to right, from right to left, from outside to inside, and from inside to outside. In this change process, the shooting angle of the same ileocecal part can be changed continuously or discontinuously, so as to obtain multiple shooting angles for a single ileocecal part.

[0083] Among them, the multiple ileocecal parts may include the ileocecal valve, ileocecal orifice, appendix orifice, cecal base and other parts within the ileocecal region that can better reflect the anatomical structural characteristics of the ileocecal region.

[0084] In practical applications, during the change of shooting angle, including at least one of from near to far, from far to near, from left to right, from right to left, from outside to inside, and from inside to outside:

[0085] In the process of shooting each video clip included in the colonoscopy video sample, the process of changing the shooting angle for shooting multiple ileocecal parts in the ileocecal region may include at least one of from near to far, from far to near, from left to right, from right to left, from outside to inside, and from inside to outside.

[0086] The embodiment of the present application adopts at least one of the above-mentioned shooting angle change processes to shoot multiple ileocecal areas in the ileocecal region to obtain multiple video clips as colonoscopy video samples. When training a video recognition network based on colonoscopy video samples, the video recognition network can learn more spatial information consistent with the motion characteristics of the endoscope lens of the shooting sample, thereby effectively improving the robustness of the video recognition network trained based on the colonoscopy video samples.

[0087] Regarding the shooting distance of the video clips, in one embodiment, the shooting distance for shooting each ileocecal region in the above video clips includes multiple distances, and the shooting distances for shooting each ileocecal region in different video clips are not completely the same.

[0088] The shooting distance refers to the distance between the shooting lens and the ileocecal part when shooting the ileocecal part.

[0089] Specifically, the specific manifestations of the ileocecal area may include the ileocecal valve, ileocecal orifice, appendix orifice, and cecal base within the ileocecal region, which can better reflect the anatomical structural characteristics of the ileocecal region; the different shooting distances for multiple ileocecal parts within the ileocecal region may be due to the different specific operating methods during each colonoscopy and the different motion characteristics of the endoscope lens when collecting video clips.

[0090] In practical applications, under the premise that the shooting distance for shooting each ileocecal part includes multiple distances, and the shooting distances for shooting each ileocecal part in different video clips are not completely the same, each video clip included in the colonoscopy video sample can be shot using multiple shooting distances (that is, during the shooting process of each video clip, the shooting distance for the same ileocecal part can change multiple times). Moreover, in each video clip included in the colonoscopy video sample, the shooting distances for shooting multiple ileocecal parts in the ileocecal region can also be different.

[0091] Exemplarily, for the ileocecal region A, in the video segment A, a plurality of different shooting distances for the ileocecal region A may be included. At the same time, in different video segments, the shooting distances for the same ileocecal region may not be exactly the same. For example, the shooting distance for the ileocecal region A in the video segment B may not be exactly the same as the shooting distance for the ileocecal region A in the video segment A (for example, the shooting distance for the ileocecal region A in the video segment B may be greater than or less than the shooting distance for the ileocecal region A in the video segment A. Moreover, the number of changes in the shooting distance for the ileocecal region A in the video segment B may also be greater than or less than the number of changes in the shooting distance for the ileocecal region A in the video segment A). Among them, the ileocecal region A may be any ileocecal region, and the video segment A and the video segment B may be any two different video segments in the colonoscopy video sample.

[0092] The embodiment of the present application uses different shooting distances to shoot multiple ileocecal areas in the ileocecal region, and obtains multiple video clips of colonoscopy video samples for training the video recognition network. When training the video recognition network based on the colonoscopy video samples, the video recognition network can learn more spatial information consistent with the motion characteristics of the endoscope lens of the shooting sample, effectively ensuring the robustness of the video recognition network trained based on the colonoscopy video samples, thereby effectively improving the accuracy of the video recognition network in identifying the ileocecal region in each video frame of the input video clip.

[0093] Regarding the specific manifestations of the ileocecal regions contained in the video clips, in one embodiment, the movement forms of the ileocecal regions contained in different video clips are not completely the same.

[0094] The motion form refers to the motion form of the ileocecal area contained in the video clip.

[0095] Specifically, the specific manifestations of the ileocecal region may include the ileocecal valve, ileocecal orifice, appendix orifice, and cecal base in the ileocecal region, which are parts that can better reflect the anatomical structural characteristics of the ileocecal region. The differences in the motion forms of the various ileocecal parts contained in different video clips may be due to the different shooting objects of each video clip contained in the colonoscopy video sample, the different physiological activity states of the ileocecal region during each colonoscopy, and other reasons.

[0096] In practical applications, under the premise that the movement patterns of various ileocecal parts contained in different video clips are not exactly the same, each video clip contained in the colonoscopy video sample can include video images shot for multiple ileocecal parts with different movement patterns.

[0097] The embodiment of the present application uses video clips containing multiple ileocecal regions with different motion forms as colonoscopy video samples for training a video recognition network. This allows the video recognition network to learn more instantaneous time information consistent with the motion forms of the ileocecal region when training the video recognition network based on the colonoscopy video samples, thereby effectively ensuring the robustness of the video recognition network trained based on the colonoscopy video samples, and thereby effectively improving the accuracy of identifying the ileocecal region in each video frame of the input video clip based on the video recognition network.

[0098] Regarding the specific content represented by the movement morphology of each ileocecal region, in one embodiment, the above-mentioned movement state includes a physiological activity state in which the ileocecal region is deformed due to intestinal peristalsis or respiratory movement.

[0099] The motion form refers to the motion form of the ileocecal area contained in the video clip.

[0100] Specifically, under the premise that the movement forms of various ileocecal parts contained in different video clips are not exactly the same, each video clip contained in the colonoscopy video sample may include video images shot for the physiological activity states of multiple ileocecal parts that undergo deformation due to intestinal peristalsis or respiratory movement.

[0101] The embodiment of the present application uses several video clips of video images of physiological activity states in which the ileocecal region is deformed due to intestinal peristalsis or respiratory movement as colonoscopy video samples for training a video recognition network, thereby effectively ensuring the robustness of the video recognition network against deformation of the ileocecal region caused by intestinal peristalsis, respiratory movement, etc., and thereby effectively improving the accuracy of identifying the ileocecal region in each video frame of the input video clip based on the video recognition network.

[0102] In one embodiment, in the above video clips, the shooting angles for shooting the multiple ileocecal parts in the ileocecal region include multiple angles, and the shooting angles for shooting the various ileocecal parts in different video clips are not completely the same. At the same time, the shooting distances for shooting the various ileocecal parts in the above video clips include multiple distances, and the shooting distances for shooting the various ileocecal parts in different video clips are not completely the same. Moreover, the motion forms of the various ileocecal parts included in different video clips are not completely the same.

[0103] Among them, by adopting a variety of different shooting angles and a variety of different shooting distances, video shooting is performed on multiple ileocecal parts with different movement forms in the ileocecal region, so that the dynamic information and static information contained in the multiple ileocecal parts in the ileocecal region can be collected simultaneously in the time and space dimensions.

[0104] Specifically, static information can be specifically manifested as the ileocecal part contained in the video clip, the motion form of each ileocecal part (instantaneous time information), and the shooting angle and shooting distance (spatial information) when shooting each ileocecal part. Dynamic information can be specifically manifested as the time sequence when each video clip is actually shot. The static information contained in the video clip can effectively improve the tolerance of the video recognition network to the input data, so that in actual application, the video recognition network can be compatible with various possible image acquisition situations. The dynamic information contained in the video clip can enable the video recognition network to learn more motion features that are consistent with the motion characteristics of the endoscope lens during the training process, thereby effectively improving the recognition ability of the video recognition network for the ileocecal region in actual application.

[0105] In the embodiment of the present application, video clips containing both static information and dynamic information are used as colonoscopy video samples to train the video recognition network, which not only effectively improves the tolerance of the video recognition network to input data, but also effectively improves the recognition accuracy of the video recognition network for the ileocecal region. As for the specific form of the colonoscopy video sample, in one embodiment, the colonoscopy video sample is an original colonoscopy video sample that has been preprocessed.

[0106] The colonoscopy video sample refers to the original colonoscopy video sample that has been preprocessed and contains video clips shot at multiple ileocecal sites in the ileocecal region. The original colonoscopy video sample refers to the original colonoscopy video sample consisting of multiple video clips shot at multiple ileocecal sites in the ileocecal region of each patient during the colonoscopy examination of different patients.

[0107] The embodiment of the present application pre-processes the original colonoscopy video samples to obtain colonoscopy video samples, thereby effectively ensuring the recognition accuracy and operation speed of the video recognition network trained based on the colonoscopy video samples.

[0108] Regarding the specific method of preprocessing the original colonoscopy video samples, in one embodiment, the preprocessing includes at least one of removing invalid areas, removing redundancy, and removing non-information frames.

[0109] Among them, preprocessing refers to preprocessing of each video frame in the original colonoscopy video sample. By preprocessing the original colonoscopy video sample, the original colonoscopy video sample can be converted into a "colonoscopy video sample" that can be used to execute the above steps S110 to S130.

[0110] Specifically, a specific method for removing invalid areas may be to remove the black borders originally contained in each video frame by cropping each video frame (for example, the black borders that occupy a large proportion of the image in the endoscopic image and are used to display invalid information such as the shooting time, hospital, and patient age), so that the above-mentioned video recognition network can more accurately identify the features in the colonoscopy video samples. A specific method for de-redundancy processing may be to remove video frames containing repeated information in each video frame (for example, using interval sampling, similarity-based measurement, etc. to remove video frames containing repeated information in each video frame) to reduce the repeatability between adjacent frames, thereby improving the computing speed of the above-mentioned video recognition network when identifying colonoscopy video samples. A specific method for removing non-information frames may be to remove non-information frames (i.e., video frames that do not contain valid information) that contain blur, artifacts, and external content that obviously do not contain the ileocecal region in each video frame, so as to improve the recognition accuracy and computing speed of the above-mentioned video recognition network when identifying colonoscopy video samples.

[0111] In practical applications, in the process of preprocessing the original colonoscopy video samples, the three preprocessing methods of removing invalid areas, de-redundant processing, and removing non-information frames can be used in sequence, or only any one of the preprocessing methods of removing invalid areas, de-redundant processing, and removing non-information frames can be used, or any two of the preprocessing methods of removing invalid areas, de-redundant processing, and removing non-information frames can be used. Here, the order and combination of the three preprocessing methods of removing invalid areas, de-redundant processing, and removing non-information frames in the process of preprocessing the original colonoscopy video samples are not strictly limited.

[0112] The embodiment of the present application preprocesses the original colonoscopy video samples by adopting at least one preprocessing method including removing invalid areas, removing redundancy, and removing non-information frames to obtain colonoscopy video samples, thereby effectively ensuring the recognition accuracy and operation speed of the video recognition network trained based on the colonoscopy video samples.

[0113] In one embodiment, Figure 4 As shown, a video recognition method is provided, which uses a pre-trained video recognition network to synchronously recognize real-time video images during colonoscopy, so as to determine whether the colonoscopy has successfully reached the ileocecal region based on the synchronous recognition results of the video images. Among them, the video recognition network can be trained in the cloud or using offline devices. Offline devices can be computer terminal devices in the form of personal computers, laptops, tablets, image processors, AI boxes, etc.

[0114] The method comprises the following steps:

[0115] Step S410: obtaining a video segment to be processed.

[0116] In this step, the video segments to be processed refer to a number of video segments to be processed consisting of endoscopic video images acquired in real time during the colonoscopy process.

[0117] Step S420, processing the video clip to be processed based on the video recognition network to determine whether the video clip to be processed includes the ileocecal region; the aforementioned video recognition network is based on the colonoscopy video sample and the label data corresponding to the colonoscopy video sample, and the initial recognition network is trained to obtain the trained video recognition network; the aforementioned label data is used to characterize the ileocecal part contained in each video frame in the video clip, or to characterize whether each video frame in the video clip includes the ileocecal region; the aforementioned colonoscopy video sample includes video clips shot for multiple ileocecal parts in the ileocecal region.

[0118] In this step, the video recognition network refers to the Figures 1 to 3, and embodiments related to the training method of the above-mentioned video recognition network, and the network model obtained by training.

[0119] Specifically, the specific method of processing the video clip to be processed based on the video recognition network to determine whether the video clip to be processed contains the ileocecal region can be to use the video recognition network that has been trained above to identify the specific content represented by each video frame of the video clip to be processed, so that the video recognition network that has been trained above can output a recognition result that can be used to determine whether the video clip to be processed contains the ileocecal region.

[0120] In practical applications, the specific form of the video recognition network can be a three-dimensional convolutional neural network model 3D-CNN (Three-Dimensional Convolutional Neural Networks, 3D-CNN) that can process three-dimensional input data, or a deep learning network model (CNN-LSTM) that combines convolutional neural networks (CNN) and long short-term memory networks (LSTM), or other forms of network models other than the aforementioned two network models.

[0121] The embodiment of the present application processes the video segment to be processed based on a video recognition network trained with colonoscopy video samples and label data to determine whether the video segment to be processed contains the ileocecal region, thereby effectively ensuring the accuracy of identifying the ileocecal region in each video frame of the input video segment based on the video recognition network.

[0122] Regarding the specific method of displaying the prompt information corresponding to the target video frame in the video segment to be processed, in one embodiment, Figure 5 As shown, the above video recognition method further includes:

[0123] Step S510: Play the video clip to be processed in the visual interface.

[0124] In this step, the visualization interface refers to a visualization interface with data, image, and video display functions; the video clips to be processed refer to a number of video clips to be processed consisting of endoscopic video images acquired in real time during the colonoscopy process.

[0125] In practical applications, the specific form of the visualization interface may be a video stream playback interface.

[0126] After the above step S420, the above video recognition method further includes:

[0127] Step S520: Based on the playback progress of the video segment to be processed, display the prompt information corresponding to the target video frame in the video segment to be processed; the above prompt information is used to represent the ileo-blind area contained in the target video frame.

[0128] In this step, the playback progress of the video segment to be processed refers to the playback progress of the video segment to be processed in the visualization interface; the target video frame refers to the video frame containing the ileocecal part among the video frames of the video segment to be processed; the ileocecal part refers to the ileocecal part that needs to be examined in the ileocecal area during the colonoscopy; the prompt information refers to the prompt information corresponding to the ileocecal part contained in the target video frame.

[0129] Specifically, the specific manifestation of the ileocecal region may include the ileocecal valve, ileocecal orifice, appendix orifice, and cecal base in the ileocecal region, which can better reflect the anatomical structural characteristics of the ileocecal region. The specific manifestation of the prompt information may be which ileocecal region is included in a certain target video frame, or the image content displayed in a certain target video frame is shot in the ileocecal region.

[0130] The embodiment of the present application realizes following the video playback progress and synchronously displaying the recognition results of each video frame by displaying the prompt information corresponding to the target video frame in the video segment to be processed based on the playback progress of the video segment to be processed in the visual interface, so that the operator can know the progress of the colonoscopy in real time.

[0131] 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.

[0132] Based on the same inventive concept, the embodiment of the present application also provides a video recognition network training system for implementing the video recognition network training method involved above. The implementation scheme for solving the problem provided by the system is similar to the implementation scheme recorded in the above method, so the specific limitations in the one or more video recognition network training system embodiments provided below can refer to the limitations of the video recognition network training method above, and will not be repeated here.

[0133] In one embodiment, Figure 6 As shown, a training system for a video recognition network is provided, the system comprising:

[0134] The colonoscopy video sample acquisition unit 610 is used to acquire a colonoscopy video sample; the colonoscopy video sample includes video clips shot at multiple ileocecal areas in the ileocecal region;

[0135] The label data acquisition unit 620 is used to acquire label data corresponding to the colonoscopy video sample; the label data is used to characterize the ileocecal region contained in each video frame in the video segment, or to characterize whether each video frame in the video segment contains the ileocecal region;

[0136] The video recognition network training unit 630 is used to train the initial recognition network based on the colonoscopy video sample and the label data to obtain a trained video recognition network; the video recognition network is used to identify the ileocecal region.

[0137] In one of the embodiments, in the training system of the video recognition network, the shooting order of the multiple ileocecal parts in the ileocecal region in the video clip is related to the inspection order of each ileocecal part during the colonoscopy.

[0138] In one of the embodiments, in the training system of the video recognition network, the ileocecal region includes at least two of the ileocecal valve, the ileocecal orifice, the appendix orifice, and the cecal base.

[0139] In one of the embodiments, in the training system of the video recognition network, when the label data is used to characterize the ileo-bacilli contained in each video frame in the video clip, the label data is also used to characterize the area where the ileo-bacilli are located in the video frame.

[0140] In one of the embodiments, in the training system of the video recognition network, the shooting angles for shooting the multiple ileocecal parts in the ileocecal region in the video clips include multiple angles, and the shooting angles for shooting the various ileocecal parts in different video clips are not exactly the same.

[0141] In one of the embodiments, in the training system of the above-mentioned video recognition network, the changing process of the shooting angle includes at least one of from near to far, from far to near, from left to right, from right to left, from outside to inside, and from inside to outside.

[0142] In one of the embodiments, in the training system of the video recognition network, the shooting distances for shooting each of the ileocecal regions in the video clips include multiple distances, and the shooting distances for shooting each of the ileocecal regions in different video clips are not exactly the same.

[0143] In one of the embodiments, in the training system of the video recognition network, the movement forms of the ileocecal areas contained in different video clips are not completely the same.

[0144] In one of the embodiments, in the training device of the video recognition network, the movement state includes a physiological activity state in which the ileocecal region is deformed due to intestinal peristalsis or respiratory movement.

[0145] In one of the embodiments, in the training system of the video recognition network, the colonoscopy video sample is an original colonoscopy video sample that has been preprocessed.

[0146] In one of the embodiments, in the training system of the video recognition network, the preprocessing includes at least one of removing invalid areas, de-redundancy processing, and removing non-information frames.

[0147] Based on the same inventive concept, the embodiment of the present application also provides a video recognition system for implementing the video recognition method involved above. The implementation scheme for solving the problem provided by the system is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more video recognition system embodiments provided below can refer to the limitations on the video recognition method above, and will not be repeated here.

[0148] In one embodiment, Figure 7 As shown, a video recognition system is provided, the system comprising:

[0149] The to-be-processed video segment acquisition unit 710 is used to acquire the to-be-processed video segment;

[0150] The video segment to be processed identification unit 720 is used to process the video segment to be processed based on a video recognition network to determine whether the video segment to be processed includes the ileocecal region; the video recognition network is based on the colonoscopy video sample and the label data corresponding to the colonoscopy video sample, and trains the initial recognition network to obtain the trained video recognition network; the label data is used to characterize the ileocecal region included in each video frame in the video segment, or to characterize whether each video frame in the video segment includes the ileocecal region; the colonoscopy video sample includes video segments shot for multiple ileocecal regions in the ileocecal region.

[0151] In one of the embodiments, the above-mentioned video recognition system also includes: a prompt information display unit, which is used to play the video segment to be processed in a visual interface; after the video segment to be processed is processed based on the video recognition network to determine whether the video segment to be processed contains the ileocecal region, it also includes: based on the playback progress of the video segment to be processed, displaying prompt information corresponding to the target video frame in the video segment to be processed; the prompt information is used to characterize the ileocecal region contained in the target video frame.

[0152] The training system of the video recognition network and each part of the video recognition system can be implemented in whole or in part by software, hardware, or a combination thereof. The above parts can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above parts.

[0153] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 8As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. 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, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data such as training-related data of the video recognition network, or data such as video recognition-related data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a training method for a video recognition network or a video recognition method is implemented.

[0154] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Fig. 9 As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface, the display unit and the input device are connected to the system bus via the input / output interface. 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 input / output interface of the computer device is used to exchange information between the processor and an external device. 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 implemented through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a training method for a video recognition network or a video recognition method is implemented. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen 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 covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0155] Those skilled in the art will understand that Figure 8 and Fig. 9The 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.

[0156] 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 in the above-mentioned video recognition network training method embodiment are implemented, or the steps in the above-mentioned video recognition method embodiment are implemented.

[0157] In one embodiment, an endoscope system is provided, including an endoscope, a display, a light source device, and the above-mentioned computer device.

[0158] 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 in the above-mentioned video recognition network training method embodiment are implemented, or the steps in the above-mentioned video recognition method embodiment are implemented.

[0159] In one embodiment, a computer program product is provided, including a computer program, which, when executed by a processor, implements the steps in the above-mentioned video recognition network training method embodiment, or implements the steps in the above-mentioned video recognition method embodiment.

[0160] 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, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0161] 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.

[0162] 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.

[0163] 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. A video recognition network training method, characterized in that: The method comprises: Obtaining a colonoscopy video sample; the colonoscopy video sample includes video clips shot at multiple ileocecal areas in the ileocecal region; Obtaining label data corresponding to the colonoscopy video sample; the label data is used to characterize the ileocecal region contained in each video frame in the video segment, or to characterize whether each video frame in the video segment contains the ileocecal region; Based on the colonoscopy video samples and the label data, an initial recognition network is trained to obtain a trained video recognition network; the video recognition network is used to identify the ileocecal region.

2. The method according to claim 1, characterized in that The shooting order of the multiple ileocecal parts in the ileocecal region in the video clip is related to the inspection order of each ileocecal part during the colonoscopy.

3. The method according to claim 1, characterized in that The ileocecal area includes at least two areas of the ileocecal valve, the ileocecal orifice, the appendix orifice, and the cecal fundus.

4. The method according to any one of claims 1 to 3, characterized in that When the label data is used to characterize the ileo-blind part contained in each video frame in the video segment, the label data is also used to characterize the area where the ileo-blind part is located in the video frame.

5. The method according to any one of claims 1 to 3, characterized in that The shooting angles for shooting the multiple ileocecal parts in the ileocecal region in the video clips include multiple angles, and the shooting angles for shooting the various ileocecal parts in different video clips are not completely the same.

6. The method according to any one of claims 1 to 3, characterized in that The shooting distances for shooting each ileocecal region in the video clips include multiple distances, and the shooting distances for shooting each ileocecal region in different video clips are not completely the same.

7. The method according to any one of claims 1 to 3, characterized in that The movement patterns of the ileocecal regions contained in different video clips are not completely the same.

8. The method according to claim 7, characterized in that The movement state includes a physiological activity state in which the ileocecal region is deformed due to intestinal peristalsis or respiratory movement.

9. A video recognition method, characterized in that: The method comprises: Get the video clip to be processed; The video segment to be processed is processed based on a video recognition network to determine whether the video segment to be processed contains an ileocecal region; the video recognition network is a video segment that trains an initial recognition network based on a colonoscopy video sample and label data corresponding to the colonoscopy video sample to obtain a trained video recognition network; the label data is used to characterize the ileocecal region contained in each video frame in the video segment, or to characterize whether each video frame in the video segment contains the ileocecal region; the colonoscopy video sample contains video segments shot for multiple ileocecal regions in the ileocecal region.

10. The method according to claim 9, characterized in that The method further comprises: In the visual interface, playing the video clip to be processed; After the video recognition network is used to process the video segment to be processed and determines whether the video segment to be processed includes the ileocecal region, the method further includes: Based on the playback progress of the video segment to be processed, prompt information corresponding to the target video frame in the video segment to be processed is displayed; the prompt information is used to characterize the ileo-blind area contained in the target video frame.

11. A video recognition network training system, characterized in that: The system comprises: A colonoscopy video sample acquisition unit, used to acquire a colonoscopy video sample; the colonoscopy video sample includes video clips shot at multiple ileocecal areas in the ileocecal region; A label data acquisition unit, used to acquire label data corresponding to the colonoscopy video sample; the label data is used to characterize the ileocecal region contained in each video frame in the video segment, or to characterize whether each video frame in the video segment contains the ileocecal region; The video recognition network training unit is used to train the initial recognition network based on the colonoscopy video sample and the label data to obtain a trained video recognition network; the video recognition network is used to identify the ileocecal region.

12. A video recognition system, characterized in that: The system comprises: A to-be-processed video segment acquisition unit, used to acquire the to-be-processed video segment; The identification unit of the video segment to be processed is used to process the video segment to be processed based on a video recognition network to determine whether the video segment to be processed contains an ileocecal region; the video recognition network is a video recognition network that is trained based on colonoscopy video samples and label data corresponding to the colonoscopy video samples to obtain a trained video recognition network; the label data is used to characterize the ileocecal region contained in each video frame in the video segment, or to characterize whether each video frame in the video segment contains the ileocecal region; the colonoscopy video samples contain video segments shot for multiple ileocecal regions in the ileocecal region.

13. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the processor implements the steps of the method according to any one of claims 1 to 8, or implements the steps of the method according to claim 9 or 10.

14. An endoscope system comprising an endoscope, a display, a light source device and the computer device according to claim 13.

Images