Enteroscope quality determination method and device, equipment, storage medium and program product

Abstract

The invention relates to an enteroscope quality determination method and device, equipment, a storage medium and a program product, and the method comprises the steps: recognizing a plurality of enteroscope video clips of a detection object through a mirror retreating recognition model, determining a mirror retreating starting video clip, determining a plurality of target video clips in a mirror retreating stage according to the mirror retreating starting video clip, and determining the enteroscope quality of the detection object according to the target video clips. And performing scoring processing on each target video clip through the intestinal tract cleanliness scoring model, and determining a cleanliness quantification result of each target video clip, thereby determining an enteroscope quality result of the detection object according to the cleanliness quantification result of each target video clip. According to the embodiment, the enteroscope video clip is recognized through the retreating recognition model, the retreating recognition model can effectively combine the feature information on each enteroscope video frame and learn the time sequence information on the video frame, the continuity of the video clip is ensured, the determined retreating starting video clip is more accurate, and the user experience is improved. Therefore, the accuracy of determining the enteroscope quality result of the detected object is improved.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a method, device, equipment, storage medium and program product for determining the quality of a colonoscope. Background Art

[0002] Colonoscopy is an examination method in which a colonoscope is inserted into the ileocecal cavity through the anus to observe colon lesions from the mucosal side. For example, colorectal cancer is one of the most common cancers in the world. Colonoscopy can detect adenomas early, thereby reducing the incidence of colorectal cancer. It is the gold standard for diagnosing colorectal cancer. When the blinding rate reaches 95% and the intestinal cleanliness score is >6 points, the adenoma detection rate can be effectively improved. It can be seen that the blinding rate and intestinal cleanliness score are key indicators of colonoscopy quality control. The accuracy of the intestinal cleanliness score is of great significance to improving the accuracy of colonoscopy. Therefore, how to improve the accuracy of the intestinal cleanliness score has become an urgent problem to be solved in this field. Summary of the invention

[0003] Based on this, it is necessary to provide a colonoscopy quality determination method, device, equipment, storage medium and program product that can improve the accuracy of intestinal cleanliness scoring in response to the above-mentioned technical problems.

[0004] In a first aspect, the present application provides a method for determining the quality of a colonoscope, the method comprising:

[0005] Identify multiple colonoscopy video clips of the detection object through the withdrawal recognition model to determine the withdrawal start video clip;

[0006] Determining a plurality of target video segments in a phase of rewinding according to the rewinding start video segment;

[0007] Scoring each of the target video clips using an intestinal cleanliness scoring model to determine a quantified result of the cleanliness of each of the target video clips;

[0008] The colonoscopy quality result of the detection object is determined according to the cleanliness quantification result of each target video segment.

[0009] In one embodiment, the scope withdrawal recognition model includes an ileocecal recognition model; the number of frames of the colonoscopy video segment is a first number of frames, and the multiple colonoscopy video segments of the detection object are recognized by the scope withdrawal recognition model to determine the scope withdrawal start video segment, including:

[0010] The multiple colonoscopy video segments are respectively identified by the ileocecal recognition model, and the first video segment including the ileocecal segment is determined in the colonoscopy video segments, and the first video segment is used as the start video segment for withdrawing the image.

[0011] In one embodiment, the camera-retracting recognition model includes a screen motion recognition model; the number of frames of the colonoscopy video segment is a second number of frames; the first number of frames is less than the second number of frames; and the recognizing multiple colonoscopy video segments of the detection object by the camera-retracting recognition model to determine the camera-retracting start video segment includes:

[0012] The multiple colonoscopy video segments are respectively identified by the image motion recognition model to determine the second video segment in which the colonoscopy state is the withdrawn state for the first time, and the second video segment is used as the withdrawn start video segment.

[0013] In one of the embodiments, determining a plurality of target video segments in the fade-out stage according to the fade-out starting video segment includes:

[0014] Determining a candidate video segment in a fade-out phase according to the fade-out start video segment;

[0015] The candidate video segments are evenly divided according to the number of frames of the candidate video segments to obtain a plurality of the target video segments in the de-focusing stage.

[0016] In one of the embodiments, determining a plurality of target video segments in the fade-out stage according to the fade-out starting video segment includes:

[0017] Determining a candidate video segment in a fade-out phase according to the fade-out start video segment;

[0018] Extracting features from the candidate video clips to obtain intestinal cleaning features of the candidate video clips;

[0019] A video segment having the same intestinal cleaning feature among the candidate video segments is used as a target video segment to obtain a plurality of target video segments in the de-staging stage.

[0020] In one embodiment, before identifying the multiple colonoscopy video clips of the detection object by using the recognizing model, the method further includes:

[0021] Determining an information video segment in the colonoscopy video according to the colonoscopy video of the test object and the screening model;

[0022] The multiple colonoscopy video segments are determined according to the information video segments.

[0023] In one embodiment, determining the colonoscopy quality result of the detected object according to the cleanliness quantification result of each target video segment includes:

[0024] According to the clean quantization results of each of the target video segments, determining a third video segment whose clean quantization results meet a preset condition from each of the target video segments;

[0025] Determining a ratio of the number of the third video segments to the number of the target video segments;

[0026] The colonoscopy quality result of the detection object is determined according to the ratio and a preset threshold.

[0027] In one embodiment, determining the colonoscopy quality result of the detected object according to the cleanliness quantification result of each target video segment includes:

[0028] According to the clean quantization results of each of the target video segments, determining a fourth video segment whose clean quantization results meet a preset condition from each of the target video segments;

[0029] Determining a weight of each of the fourth video segments according to the number of frames of each of the fourth video segments;

[0030] Determining a ratio of a weight of the fourth video segment to a weight of the candidate video segment;

[0031] The colonoscopy quality result of the detection object is determined according to the ratio and a preset threshold.

[0032] In one embodiment, determining the colonoscopy quality result of the detected object according to the ratio and a preset threshold value includes:

[0033] If the ratio is greater than the preset threshold, determining that the colonoscope quality result of the detected object is a quality result of the first cleaning level;

[0034] If the ratio is not greater than the preset threshold, it is determined that the colonoscopy quality result of the detected object is a quality result of the second cleanliness level; and the first cleanliness level is lower than the second cleanliness level.

[0035] In one embodiment, a plurality of colonoscopy video samples are obtained;

[0036] Determine each first video segment sample in the multiple colonoscopy video samples according to the start time point and the end time point of the ileocecal region in the multiple colonoscopy video samples;

[0037] An initial ileocecal recognition model is trained according to each of the first video segment samples and the labels of each of the first video segment samples to obtain the ileocecal recognition model; the labels of the first video segment samples include ileocecal labels and non-ileocecal labels.

[0038] In one embodiment, the method further comprises:

[0039] Obtain multiple colonoscopy video samples;

[0040] Determine each second video segment sample in the plurality of colonoscopy video samples according to the start time point and the end time point corresponding to the colonoscopy state in the plurality of colonoscopy video samples; the colonoscopy state includes an enter-end state, a withdraw-end state, and a stop state;

[0041] The initial picture motion recognition model is trained according to each of the second video clip samples and the labels of each of the second video clip samples to obtain the picture motion recognition model; the labels of the second video clip samples include a camera-in phase label, a camera-out phase label and a camera-stop phase label.

[0042] In one embodiment, the method further comprises:

[0043] Obtain multiple colonoscopy video samples;

[0044] Determine each third video segment sample from the plurality of colonoscopy video samples;

[0045] According to each of the third video clip samples and the label of each of the third video clip samples, the initial intestinal cleanliness scoring model is trained to determine the intestinal cleanliness scoring model; the label of the third video clip sample includes the cleanliness quantification result of the third video clip sample.

[0046] In a second aspect, the present application further provides a device for determining the quality of a colonoscope, the device comprising:

[0047] A first determination module is used to identify multiple colonoscopy video clips of the detection object through a withdrawal recognition model to determine a withdrawal start video clip;

[0048] A second determining module is used to determine a plurality of target video segments in a fade-out stage according to the fade-out starting video segment;

[0049] A third determination module is used to score each of the target video clips using an intestinal cleanliness scoring model to determine a cleanliness quantification result of each of the target video clips;

[0050] The fourth determination module is used to determine the colonoscopy quality result of the detection object according to the cleanliness quantification result of each target video segment.

[0051] In a third aspect, the present application further provides a computer device, the computer device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the following steps when executing the computer program:

[0052] Identify multiple colonoscopy video clips of the detection object through the withdrawal recognition model to determine the withdrawal start video clip;

[0053] Determining a plurality of target video segments in a phase of rewinding according to the rewinding start video segment;

[0054] Scoring each of the target video clips using an intestinal cleanliness scoring model to determine a quantified result of the cleanliness of each of the target video clips;

[0055] The colonoscopy quality result of the detection object is determined according to the cleanliness quantification result of each target video segment.

[0056] In a fourth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0057] Identify multiple colonoscopy video clips of the detection object through the withdrawal recognition model to determine the withdrawal start video clip;

[0058] Determining a plurality of target video segments in a phase of rewinding according to the rewinding start video segment;

[0059] Scoring each of the target video clips using an intestinal cleanliness scoring model to determine a quantified result of the cleanliness of each of the target video clips;

[0060] The colonoscopy quality result of the detection object is determined according to the cleanliness quantification result of each target video segment.

[0061] In a fifth aspect, the present application further provides a computer program product, the computer program product comprising a computer program, which implements the following steps when executed by a processor:

[0062] Identify multiple colonoscopy video clips of the detection object through the withdrawal recognition model to determine the withdrawal start video clip;

[0063] Determining a plurality of target video segments in a phase of rewinding according to the rewinding start video segment;

[0064] Scoring each of the target video clips using an intestinal cleanliness scoring model to determine a quantified result of the cleanliness of each of the target video clips;

[0065] The colonoscopy quality result of the detection object is determined according to the cleanliness quantification result of each target video segment.

[0066] The above-mentioned colonoscopy quality determination method, device, equipment, storage medium and program product use the scope withdrawal recognition model to identify multiple colonoscopy video clips of the detection object, determine the scope withdrawal start video clip, and determine multiple target video clips in the scope withdrawal stage based on the scope withdrawal start video clip. The target video clips are scored and processed by the intestinal cleanliness scoring model to determine the cleanliness quantification result of each target video clip, thereby determining the colonoscopy quality result of the detection object based on the cleanliness quantification result of each target video clip. In this embodiment, the colonoscopy video clips are identified by the scope withdrawal recognition model. The scope withdrawal recognition model can effectively combine the feature information on each colonoscopy video frame and the timing information learned on the video frame, thereby ensuring the continuity of the video clips, making the determined scope withdrawal start video clip more accurate, thereby improving the accuracy of the colonoscopy quality result of the detection object. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 is a schematic diagram of a colonoscope in an embodiment;

[0068] Figure 2 is a schematic diagram of a colonoscope in another embodiment;

[0069] Figure 3 A diagram showing an application environment of a method for determining the quality of a colonoscope in an embodiment;

[0070] Figure 4 is a schematic flow chart of a method for determining the quality of a colonoscope in one embodiment;

[0071] Figure 5 is a schematic diagram of a colonoscope in yet another embodiment;

[0072] Figure 6 A schematic diagram of a flow chart of a method for determining a target video segment in one embodiment;

[0073] Figure 7 is a schematic flow chart of a method for determining a target video segment in another embodiment;

[0074] Figure 8 is a flow chart of a method for determining multiple colonoscopy video clips in one embodiment;

[0075] Fig. 9 It is a flowchart of a method for determining colonoscopy quality results in one embodiment;

[0076] Fig.10 A schematic diagram of a flow chart of a method for determining colonoscopy quality results in another embodiment;

[0077] Fig.11 A schematic diagram of a flow chart of a method for determining an ileocecal recognition model in one embodiment;

[0078] Fig.12 A schematic diagram of a flow chart of a method for determining a picture motion recognition model in one embodiment;

[0079] Fig.13 A schematic diagram of a flow chart of a method for determining a picture motion recognition model in one embodiment;

[0080] Fig.14 is a structural block diagram of a device for determining the quality of a colonoscope in one embodiment;

[0081] Fig.15 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0082] 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.

[0083] In related technologies, the quality of colonoscopy is determined by an intestinal cleanliness assessment model. However, the intestinal cleanliness assessment model is a 2D neural network, which is trained based on report images. These intestinal cleanliness assessment models have achieved excellent performance in image classification, with an accuracy rate of over 90%. However, when the test set is a real colonoscopy video, the same cleanliness level may be divided into different levels of cleanliness. For example, there are partial / small amounts of stains, turbid liquid, and residual feces on the intestinal mucosa. However, based on the image, it is impossible to determine "partial / small amounts", such as Figure 1 Shown: The same piece of feces residue in Figure 1 (the lens is close) can be "partial", and Figure 2 (the camera is far away) can be "a small amount", Figure 1 and Figure 2 Different lenses show different levels of cleanliness, but Figure 1 and Figure 2 In essence, they are at the same level of cleanliness. Therefore, the intestinal cleanliness assessment model is not applicable to real colonoscopy videos.

[0084] The colonoscope quality determination method provided in the embodiment of the present application can be applied to Figure 3 The application environment includes a computer device, which may be a server, and its internal structure diagram may be as shown in FIG. Figure 3As shown. The computer device includes a processor, a memory and a network interface 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, 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 related to the quality of the colonoscopy. The network 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 method for determining the quality of the colonoscopy is implemented. The server can be implemented as an independent server or a server cluster consisting of multiple servers.

[0085] In one embodiment, Figure 4 As shown, a method for determining the quality of a colonoscope is provided, and the method is applied to Figure 3 The computer device in the example is used to illustrate, including the following steps:

[0086] S401, identifying multiple colonoscopy video clips of the detection object through a withdrawal recognition model to determine a withdrawal start video clip.

[0087] In this embodiment, multiple colonoscopy video clips are input into the scope withdrawal recognition model to determine the starting video clip for scope withdrawal. In a real clinical scenario, a doctor operates an endoscope to obtain a colonoscopy video, divides the obtained colonoscopy video, obtains multiple colonoscopy video clips, and inputs multiple colonoscopy video clips into the scope withdrawal recognition model. The scope withdrawal recognition model can be used to identify the patient's ileocecal valve or appendix and other ileocecal parts to confirm whether to enter the scope withdrawal stage. For example, the ileocecal part can be identified in two consecutive colonoscopy video clips, and it is determined that the scope withdrawal stage has been entered, and any colonoscopy video clip in which the ileocecal part appears is used as the starting video clip for scope withdrawal.

[0088] In a possible implementation, the moving picture in the colonoscopy video clip sample can be identified to see whether the moving picture is in the advancing state, the stopping state or the retreating state, and the retreating start video clip is determined according to the moving picture. For example, if the state of multiple consecutive colonoscopy video clips is the retreating state, it is considered that the retreating stage has been entered, and the colonoscopy video clip corresponding to any retreating state is used as the retreating start video clip.

[0089] Among them, the mirror withdrawal recognition model recognizes the colonoscopy video clips, effectively combines the feature information on each video frame in the colonoscopy video clips, and can further learn the timing information, that is, the "continuity between the previous frame and the next frame". For example, the mirror withdrawal recognition model can capture the continuity feature of "the residual feces gradually becomes smaller" (the lens is farther away), or the shape change from "looking from a distance" to "close" and then to "next to the screen". It can also learn the features when the ileocecal valve is not clear or only partially photographed. Therefore, it can effectively suppress the recognition errors caused by this problem and is more suitable for application in real colonoscopy videos.

[0090] S402: Determine a plurality of target video segments in a fade-out phase according to the fade-out start video segment.

[0091] In this embodiment, candidate video segments in the withdrawing stage are determined based on the withdrawing start video segment, and the candidate video segments are divided to obtain multiple target video segments. For example, the candidate video segments can be evenly divided according to the number of frames of the candidate video segments; the candidate video segments can also be arbitrarily divided to obtain target video segments, or they can be divided according to the intestinal cleaning characteristics of the candidate video segments to obtain multiple target video segments.

[0092] S403, scoring each target video segment using an intestinal cleanliness scoring model to determine a cleanliness quantification result of each target video segment.

[0093] In this embodiment, the target video clips are input into the intestinal cleanliness scoring model, and each target video clip is scored at different cleanliness levels to obtain the cleanliness quantification results of each target video clip. Figure 5 As shown, the intestinal cleanliness of the target video clip can be scored from 0 to 3 points to obtain the quantitative result of cleanliness. Among them, 0 points means that the entire intestinal mucosa cannot be observed due to unclear solid or liquid feces; 1 point means that part of the intestinal mucosa cannot be observed due to stains, turbid liquid, and residual feces; 2 points means that the intestinal mucosa is well observed, but a small amount of stains, turbid liquid, and feces remain; 3 points means that the intestinal mucosa is well observed, and there is basically no residual stains, turbid liquid, or feces. It can also be divided into other score segments for scoring, for example, 0-5 points, 1-10 points, etc.

[0094] S404, determining the colonoscopy quality result of the inspection object according to the cleanliness quantification result of each target video segment.

[0095] In this embodiment, the colonoscopy quality result of the detected object can be determined according to the number of target video segments of different cleanliness quantification results to the total number of target video segments. For example, the cleanliness quantification result is divided into 0 to 3 points. If the ratio of the target video segments with 0 points and 1 point to the total number of target video segments is greater than a preset threshold, it is considered that the colonoscopy quality result of the detected object is poor.

[0096] In a possible implementation, the number of target video clips with a score of 0 or 1 may be directly compared with a preset number. For example, if the number of target video clips with a score of 0 or 1 is greater than 10, it is considered that the colonoscopy quality result of the detected object is poor.

[0097] In the above-mentioned colonoscopy quality determination method, multiple colonoscopy video clips of the detection object are identified through the scope withdrawal recognition model to determine the withdrawal start video clip, and multiple target video clips in the withdrawal stage are determined based on the withdrawal start video clip, and each target video clip is scored and processed by the intestinal cleanliness scoring model to determine the cleanliness quantification result of each target video clip, thereby determining the colonoscopy quality result of the detection object based on the cleanliness quantification result of each target video clip. In this embodiment, the colonoscopy video clips are identified through the scope withdrawal recognition model, and the scope withdrawal recognition model can effectively combine the feature information on each colonoscopy video frame and the timing information learned on the video frame, thereby ensuring the continuity of the video clips, making the determined withdrawal start video clip more accurate, thereby improving the accuracy of the colonoscopy quality result of the detection object.

[0098] In one embodiment, the scope withdrawal recognition model includes an ileocecal recognition model; the number of frames of the colonoscopy video segment is a first number of frames, and multiple colonoscopy video segments of the detection object are recognized by the scope withdrawal recognition model to determine the starting video segment for scope withdrawal, including: respectively recognizing multiple colonoscopy video segments by the ileocecal recognition model, determining the first video segment including the ileocecal segment in the colonoscopy video segments, and using the first video segment as the starting video segment for scope withdrawal.

[0099] Among them, the ileocecal part includes the ileocecal valve, appendix orifice, terminal ileum, etc.

[0100] Optionally, the first frame number may be 10 frames, 12 frames, 15 frames, etc.

[0101] In this embodiment, it is assumed that the first number of frames is 12 frames, and the ileocecal recognition model recognizes 12 frames of colonoscopy video clips each time, and determines whether the "ileocecal valve", "appendix orifice", and "terminal ileum" parts appear. When any one of the three parts is recognized twice in a row, it is determined that the ileocecal part has been reached, and it is determined that the scope withdrawal stage has been entered, and the first video clip including the ileocecal part is used as the scope withdrawal starting video clip. For example, the ileocecal valve appears in the 10th and 11th colonoscopy video clips, and the 10th and 11th colonoscopy video clips are used as the scope withdrawal starting video clips, or any one of the 10th or 11th colonoscopy video clips is used as the scope withdrawal starting video clip.

[0102] In one embodiment, the withdrawal recognition model includes a screen motion recognition model; the number of frames of the colonoscopy video segment is a second number of frames; the first number of frames is less than the second number of frames; multiple colonoscopy video segments of the detection object are recognized by the withdrawal recognition model to determine the withdrawal starting video segment, including: respectively recognizing multiple colonoscopy video segments by the screen motion recognition model, determining the second video segment in which the colonoscopy state is the withdrawn state for the first time, and using the second video segment as the withdrawal starting video segment.

[0103] Among them, since the ileocecal part can be quickly identified, and it takes a certain amount of time for the movement of the picture to be observed, the first frame number is smaller than the second frame number. Optionally, the second frame number can be 50 frames, 60 frames, or 70 frames.

[0104] In this embodiment, the colonoscopy video clips of the second number of frames are input into the image motion recognition model, and the continuity of the colonoscopy video clips can be recognized. If it is a process of retracting the mirror, the picture is "gradually getting farther away", and for the process of entering the mirror, the picture is "gradually getting closer". The image motion recognition model recognizes 72 frames of colonoscopy video clips each time, and determines whether each colonoscopy video clip is in a retracting state. If 720 consecutive frames are in a retracting state, it is determined that the retracting stage has been entered, and the second video clip that enters the retracting state for the first time is used as the retracting start video clip.

[0105] In an embodiment of the present application, the scope withdrawal recognition model includes an ileocecal recognition model and a screen movement recognition model. The ileocecal recognition model is used to recognize the ileocecal part in multiple colonoscopy video clips of the first frame number. This method can quickly determine the starting video clip for scope withdrawal, and even when the ileocecal part is not obvious or not clear, the ileocecal part can be accurately identified. The screen movement recognition model is used to recognize the colonoscopy video clips of the second frame number to obtain the starting video clip for scope withdrawal, which can directly identify the scope withdrawal process without relying on the recognition of the ileocecal part, which is more reasonable. The above two scope withdrawal recognition models are suitable for different application scenarios, providing users with a variety of selection methods.

[0106] Figure 6 FIG. 1 is a flow chart of a method for determining a target video segment in an embodiment. Figure 6 As shown, the embodiment of the present application relates to a possible implementation method of how to determine multiple target video segments in the fade-out stage according to the fade-out starting video segment, including the following steps:

[0107] S601, determining a candidate video segment in a fade-out phase according to a fade-out start video segment.

[0108] In this embodiment, the retraction start video segment includes the ileocecal region, or the retraction start video segment is in a retraction state, and a video of the process of the endoscope being retracted from the ileocecal region to outside the body is used as a candidate video segment, or a video after the retraction start video segment in a retraction state is used as a candidate video segment.

[0109] S602: Evenly divide the candidate video segments according to the number of frames of the candidate video segments to obtain a plurality of target video segments in a de-focusing phase.

[0110] In this embodiment, the candidate video segments are evenly divided according to the total number of frames to obtain multiple target video segments in the de-focusing stage. Assuming that the candidate video segments are 2400 frames, the candidate video segments are evenly divided into 1000, and the number of frames of each target video segment is 24 frames.

[0111] In the embodiment of the present application, according to the video segment at the start of the withdrawal, the candidate video segment in the withdrawal stage is determined, and according to the number of frames of the candidate video segment, the candidate video segment is evenly divided to obtain multiple target video segments in the withdrawal stage. The embodiment of the present application can improve the efficiency of determining the target video segment by evenly dividing, thereby improving the efficiency of determining the test result of the colonoscopy quality of the test object.

[0112] Figure 7 FIG. 1 is a flow chart of a method for determining a target video segment in another embodiment. Figure 7 As shown, the embodiment of the present application relates to a possible implementation method of how to determine multiple target video segments in the fade-out stage according to the fade-out starting video segment, including the following steps:

[0113] S701, determining a candidate video segment in a fade-out phase according to a fade-out start video segment.

[0114] The specific implementation method can be found in the above S601, which will not be elaborated in this embodiment of the present application.

[0115] S702: extract features from the candidate video clips to obtain intestinal cleaning features of the candidate video clips.

[0116] In this embodiment, a network model may be used to extract features from the candidate video clips to obtain the intestinal cleansing features of the candidate video clips. Feature extraction algorithms such as principal component analysis may also be used to extract features from the candidate video clips to obtain the intestinal cleansing features of the candidate video clips.

[0117] S703: taking a video segment with the same intestinal cleaning feature in the candidate video segments as a target video segment, so as to obtain a plurality of target video segments in the de-staging stage.

[0118] In this embodiment, video clips with the same intestinal cleaning features are taken as a target video clip. For example, the candidate video clips include intestinal cleaning features A, intestinal cleaning features B, intestinal cleaning features C, and intestinal cleaning features D. There is no feces in the video clips of frames 1-100 in the candidate video clips, which belong to the intestinal cleaning feature A. There is feces a in the video clips of frames 101-350, which belong to the intestinal cleaning feature B. There is feces b in the video clips of frames 351-450, which belong to the intestinal cleaning feature D. There is feces a in the video clips of frames 451-500, which belong to the intestinal cleaning feature B. There is feces c in the video clips of frames 501-600, which belong to the intestinal cleaning feature C. The video clips with the same intestinal cleaning features are taken as a target video clip, and 5 target video clips can be obtained.

[0119] In an embodiment of the present application, candidate video segments in the withdrawing stage are determined based on the withdrawing starting video segment, and features are extracted from the candidate video segments to obtain intestinal cleaning features of the candidate video segments. Video segments with the same intestinal cleaning features in the candidate video segments are used as a target video segment to obtain multiple target video segments in the withdrawing stage, so that the determination of the target video segments is more targeted.

[0120] Figure 8 FIG. 1 is a flow chart of a method for determining multiple colonoscopy video clips in one embodiment. Figure 8 As shown, the following steps are included:

[0121] S801, determining an information video segment in the colonoscopy video according to the colonoscopy video of the detection object and the screening model.

[0122] In this embodiment, a pre-trained screening model is first used to screen the colonoscopy video of the test object frame by frame using the screening model to screen out non-information frames of the colonoscopy video, and the non-information frames are removed to obtain information video segments. Considering the fault tolerance of the screening model, when the number of consecutive non-information frames is greater than or equal to 5 frames, the non-information frames greater than or equal to 5 frames are deleted, leaving the information video segments. Optionally, the non-information frames include endoscopic examination, motion blur, out-of-focus blur, biopsy, surgery, flushing, etc.

[0123] Optionally, the screening model may be a convolutional neural network, for example, a two-dimensional convolutional neural network such as ResNet, VGG16, or AlexNet.

[0124] S802: Determine a plurality of colonoscopy video segments according to the information video segment.

[0125] In this embodiment, the information video segments are evenly divided according to the number of video frames to obtain a plurality of colonoscopy video segments. Alternatively, any division method may be used to obtain a plurality of colonoscopy video segments.

[0126] In the embodiment of the present application, the information video segments in the colonoscopy video are determined based on the colonoscopy video of the detected object and the screening model, and multiple colonoscopy video segments are determined based on the information video segments. In the present embodiment, non-information video frames are deleted to avoid information redundancy in the colonoscopy video segments, thereby improving the efficiency and accuracy of colonoscopy quality determination.

[0127] Fig. 9 FIG. 1 is a flow chart of a method for determining colonoscopy quality results in one embodiment. Fig. 9 As shown, the embodiment of the present application relates to a possible implementation method of how to determine the colonoscopy quality result of the detection object according to the cleanliness quantification result of each target video clip, including the following steps:

[0128] S901 , according to the clean quantization results of each target video segment, determining a third video segment whose clean quantization results meet a preset condition from each target video segment.

[0129] In this embodiment, the cleanliness quantification results of each target video clip are divided into 0 points, 1 point, 2 points and 3 points, where 0 points has the worst cleanliness and 3 points has the highest cleanliness. The target video clips corresponding to 0 points and 1 point are determined to be the third video clip, or the target video clip corresponding to 0 points may be the third video clip.

[0130] S902: Determine a ratio of the number of third video segments to the number of target video segments.

[0131] In this embodiment, each target video segment is evenly divided according to the number of frames of the candidate video segment, so the length of each target video segment is consistent, and the ratio of the number of third video segments to the number of target video segments can be directly determined. Assuming that the third video segment is a target video segment with a score of 0 and a score of 1, the number of third video segments is 3, and the number of target video segments is 10, then the ratio of the number of third video segments to the number of target video segments is 0.3. Alternatively, if the third video segment is a target video segment with a score of 0, the number of third video segments is 2, and the number of target video segments is 10, then the ratio of the number of third video segments to the number of target video segments is 0.2.

[0132] S903, determining the colonoscopy quality result of the detection object according to the ratio and the preset threshold.

[0133] The preset threshold value may be 0.2, 0.3, 0.4, etc., and may be determined according to the third video segment determination method and the cleanliness quantification result. For example, if the third video segment is a target video segment corresponding to 0 and 1 points, the preset threshold value may be selected as 0.3; if the third video segment is a target video segment corresponding to 0 points, the preset threshold value may be selected as 0.2.

[0134] In this embodiment, the ratio is compared with a preset threshold to determine the colonoscopy quality result of the detected object. When the ratio is greater than the preset threshold, the colonoscopy quality result is poor colonoscopy quality, and when the ratio is not greater than the preset threshold, the colonoscopy quality result is good colonoscopy quality. Alternatively, the difference between the ratio and the preset threshold is obtained. If the difference is greater than the threshold, the colonoscopy quality result is poor colonoscopy quality. If the difference is not greater than the threshold, the colonoscopy quality result is good colonoscopy quality.

[0135] Furthermore, the colonoscopy quality result of the detection object is determined according to the ratio and the preset threshold, including: if the ratio is greater than the preset threshold, the colonoscopy quality result of the detection object is determined to be a quality result of a first cleanliness level; if the ratio is not greater than the preset threshold, the colonoscopy quality result of the detection object is determined to be a quality result of a second cleanliness level; the first cleanliness level is less than the second cleanliness level.

[0136] In this embodiment, assuming that the preset threshold is 0.3, if the ratio is 0.2, it is determined that the colonoscopy quality result of the detection object is the quality result of the second cleanliness level, and if the ratio is 0.4, it is determined that the colonoscopy quality result of the detection object is the quality result of the first cleanliness level.

[0137] Fig.10 FIG. 1 is a flow chart of a method for determining colonoscopy quality results in another embodiment. Fig.10As shown, the embodiment of the present application relates to a possible implementation method of how to determine the colonoscopy quality result of the detection object according to the cleanliness quantification result of each target video clip, including the following steps:

[0138] S1001, according to the clean quantization results of each target video segment, determining a fourth video segment whose clean quantization results meet a preset condition from each target video segment.

[0139] In this embodiment, a fourth video segment whose cleanliness quantization result meets a preset condition is determined from the target video segments divided based on the cleanliness feature. For example, the cleanliness quantization result of the target video segment (1 frame-100 frames) is 0 points, the cleanliness quantization result of the target video segment (101 frames-350 frames) is 1 point, the cleanliness quantization result of the target video segment (351 frames-450 frames) is 2 points, the intestinal cleaning feature of the target video segment (451 frames-500 frames) is 3 points, and the cleanliness quantization result of the target video segment (501 frames-600 frames) is 2 points. The target video segments with cleanliness quantization results of 0 points and 1 point are used as the fourth video segments, and the fourth video segments are the target video segments (1 frame-100 frames) and the target video segments (101 frames-350 frames).

[0140] S1002: Determine a weight of each fourth video segment according to the number of frames of each fourth video segment.

[0141] In this embodiment, each target video segment is divided based on the cleanliness feature in the candidate video segment, so the number of frames of each target video segment is different. The weight of each fourth video segment can be determined according to the number of frames of each fourth video segment. For example, the fourth video segment includes a target video segment of 1-100 frames and a target video segment of 101-350 frames. Taking 50 frames as a reference, the weight of the fourth video segment of 1-100 frames is 2, and the weight of the video segment of 101-350 frames is 5.

[0142] S1003: Determine a ratio of a weight of the fourth video segment to a weight of the candidate video segments.

[0143] In this embodiment, 50 frames are also used as a benchmark to determine the weights of the candidate video segments. Assuming that the number of frames of the candidate video segment is 1000 frames, the weight of the candidate video segment is 20, and the weight of the fourth video segment is 7. The ratio of the weight of the fourth video segment to the weight of the candidate video segment is 0.35.

[0144] S1004, determining the colonoscopy quality result of the test object according to the ratio and a preset threshold.

[0145] For the specific implementation method, please refer to the above S903, which will not be elaborated in this embodiment of the present application.

[0146] In an embodiment of the present application, the ratio of the number of third video segments to the number of target video segments, and the ratio of the weight of the fourth video segment to the weight of the candidate video segments are determined respectively, so as to determine the colonoscopy quality result of the detection object according to the ratio and the preset threshold. The method is simple and efficient, and the colonoscopy quality result of the detection object can be quickly obtained.

[0147] Fig.11 FIG. 1 is a flow chart of a method for determining an ileocecal recognition model in an embodiment. Fig.11 As shown, the following steps are included:

[0148] S1101, obtaining multiple colonoscopy video samples.

[0149] S1102, determining each first video segment sample in the plurality of colonoscopy video samples according to the start time point and the end time point of the ileocecal region in the plurality of colonoscopy video samples.

[0150] S1103, training an initial ileocecal recognition model according to each first video segment sample and a label of each first video segment sample to obtain an ileocecal recognition model; the label of the first video segment sample includes an ileocecal label and a non-ileocecal label.

[0151] In this embodiment, 200 real colonoscopy videos are collected, and two junior doctors mark the start and end time points of the ileocecal valve, appendix orifice, and terminal ileum in the colonoscopy video samples. The first video clip sample is determined according to the start and end time points, and the first video clip sample is respectively labeled with ileocecal labels and non-ileocecal labels, and the ileocecal labels include "ileocecal valve", "appendix orifice", and "terminal ileum". The intersection of the start and end time points marked by the two junior doctors is taken, that is, "overlapping time", thereby improving the quality of the annotation, and then reviewed by a senior doctor. The first video clip sample is divided into a training set, a validation set, and a test set, which are input into the initial ileocecal recognition model, and the initial ileocecal recognition model is trained to obtain the ileocecal recognition model. The ileocecal recognition model can recognize the features when the ileocecal valve is not clearly photographed or only partially photographed, so the ileocecal recognition model can learn more features, thereby having better robustness. Optionally, the initial ileocecal recognition model includes a 3D convolutional neural network. For example, 3D convolutional neural networks such as 3DResNet.

[0152] Optionally, after obtaining the first video segment, the first video segment may be preprocessed, such as data desensitization, black frame cropping, video standardization, etc. The first video segment sample may be a video segment sample after removing the non-information frame, or may be a video segment sample without removing the non-information frame.

[0153] Fig.12 FIG. 1 is a flow chart of a method for determining a picture motion recognition model in an embodiment. Fig.12 As shown, the following steps are included:

[0154] S1201, obtaining multiple colonoscopy video samples.

[0155] S1202, determining each second video segment sample in the plurality of colonoscopy video samples according to the start time point and the end time point corresponding to the colonoscopy status in the plurality of colonoscopy video samples; the colonoscopy status includes the scope advancing state, the scope withdrawing state and the stop state.

[0156] S1203, training an initial picture motion recognition model according to each second video segment sample and a label of each second video segment sample to obtain a picture motion recognition model; the label of the second video segment sample includes a camera-in phase label, a camera-out phase label, and a camera-stop phase label.

[0157] In this embodiment, 200 real colonoscopy videos can also be collected, and two junior doctors will mark the starting time and ending time points of the "endoscope stage", "endoscope stage" and "stop stage" in the colonoscopy video samples. According to the starting time and ending time points, the second video clip samples are determined, and the second video clip samples are labeled "endoscope stage", "endoscope stage" and "stop stage" respectively. The intersection of the starting time points and the ending time points marked by the two junior doctors is taken, that is, the "overlapping time", so as to improve the quality of the marking, and then it is reviewed by a senior doctor. The second video clip samples and the labels of the second video clip samples are input into the initial picture motion recognition model, and the initial picture motion recognition model is trained to obtain the picture motion recognition model.

[0158] Similarly, after obtaining the second video clip, the first video clip can also be preprocessed, such as data desensitization, cropping black frames, video standardization, etc. The second video clip sample can be a video clip sample after removing non-information frames, or a video clip sample without removing non-information frames. The initial picture motion recognition model includes a 3D convolutional neural network, for example, a 3D convolutional neural network such as 3DResNet.

[0159] Fig.13 FIG. 1 is a flow chart of a method for determining a picture motion recognition model in an embodiment. Fig.13 As shown, the following steps are included:

[0160] S1301, obtaining multiple colonoscopy video samples.

[0161] S1302: Determine each third video segment sample from a plurality of colonoscopy video samples.

[0162] S1303, training an initial intestinal cleanliness scoring model according to each third video clip sample and the label of each third video clip sample to determine the intestinal cleanliness scoring model; the label of the third video clip sample includes a cleanliness quantification result of the third video clip sample.

[0163] In this embodiment, 200 real colonoscopy video samples are collected, and two junior doctors mark the starting time points and ending time points of different cleanliness quantification results in the colonoscopy video samples, and determine each third video segment sample based on the marked starting time points and ending time points. For example, the starting time points and ending time points of "0 points", "1 point", "2 points", and "3 points" appear in the colonoscopy video samples and are marked. When the colonoscopy video strictly cannot change from "very clean" to "a small amount of feces", the first frame where feces appears is used as the boundary, and "3 points" is marked before the boundary, and "2 points" is marked after the boundary. When "a small amount of feces" changes to "very clean" again, the first frame where a small amount of feces disappears from the screen is used as the boundary, and "2 points" is marked before the boundary, and "3 points" is marked after the boundary.

[0164] Among them, the standards of "0 points", "1 point", "2 points" and "3 points" are based on the Boston Bowel Preparation Scale (BBPS).

[0165] Input each third video clip sample and the label of each third video clip sample into the initial intestinal cleanliness scoring model, train the initial intestinal cleanliness scoring model, and obtain the intestinal cleanliness scoring model. Optionally, the initial intestinal cleanliness scoring model includes a 3D convolutional neural network, for example, 3DResNet and other 3D convolutional neural networks. Similar to the above embodiment, after obtaining the third video clip, the first video clip can also be preprocessed, such as data desensitization, cropping black frames, video standardization, etc. The first video clip sample can be a video clip sample after removing non-information frames, or a video clip sample without removing non-information frames.

[0166] It should be understood that, although the steps in the flowcharts involved in the above 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 is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may 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 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.

[0167] Based on the same inventive concept, the embodiment of the present application also provides a device for determining the quality of a colonoscope for implementing the above-mentioned method for determining the quality of a colonoscope. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in one or more embodiments of the device for determining the quality of a colonoscope provided below can refer to the limitations of the method for determining the quality of a colonoscope above, and will not be repeated here.

[0168] In one embodiment, Fig.14 As shown, a device for determining the quality of a colonoscope is provided, comprising: a first determining module 11, a second determining module 12, a third determining module 13 and a fourth determining module 14, wherein:

[0169] The first determination module 11 is used to identify multiple colonoscopy video clips of the detection object through a withdrawal recognition model to determine the withdrawal start video clip;

[0170] A second determining module 12 is used to determine a plurality of target video segments in a fade-out phase according to a fade-out start video segment;

[0171] The third determination module 13 is used to score each target video segment using an intestinal cleanliness scoring model to determine a cleanliness quantification result of each target video segment;

[0172] The fourth determination module 14 is used to determine the colonoscopy quality result of the detection object according to the cleanliness quantification result of each target video segment.

[0173] In one embodiment, the first determining module includes:

[0174] The first determination unit is used to respectively identify the multiple colonoscopy video segments through the ileocecal recognition model, determine the first video segment including the ileocecal segment in the colonoscopy video segments, and use the first video segment as the start video segment for exiting the image.

[0175] In one embodiment, the first determining module further includes:

[0176] The second determination unit is used to respectively identify the multiple colonoscopy video segments through the picture movement recognition model, determine the second video segment in which the colonoscopy state is the withdrawn state for the first time, and use the second video segment as the withdrawn start video segment.

[0177] In one embodiment, the second determining module includes:

[0178] A third determining unit is used to determine a candidate video segment in a fade-out stage according to the fade-out starting video segment;

[0179] The fourth determining unit is used to evenly divide the candidate video segments according to the number of frames of the candidate video segments to obtain a plurality of target video segments in the de-focusing stage.

[0180] In one embodiment, the second determining module further includes:

[0181] A third determining unit is used to determine a candidate video segment in a fade-out stage according to the fade-out starting video segment;

[0182] An extraction unit, used for extracting features from the candidate video segments to obtain intestinal cleaning features of the candidate video segments;

[0183] The fifth determining unit is configured to use a video segment with the same intestinal cleaning feature in the candidate video segments as a target video segment to obtain a plurality of target video segments in a withdrawal stage.

[0184] In one embodiment, the colonoscope quality determination device further comprises:

[0185] A fifth determination module is used to determine the information video segment in the colonoscopy video according to the colonoscopy video of the detection object and the screening model;

[0186] The sixth determination module is used to determine multiple colonoscopy video segments based on the information video segments.

[0187] In one embodiment, the fourth determining module includes:

[0188] a sixth determining unit, configured to determine, from the target video segments, a third video segment whose cleanliness quantization result satisfies a preset condition according to the cleanliness quantization result of each target video segment;

[0189] a seventh determining unit, configured to determine a ratio of the number of the third video segments to the number of the target video segments;

[0190] The eighth determination unit is used to determine the colonoscopy quality result of the detection object according to the ratio and a preset threshold.

[0191] In one embodiment, the fourth determining module further includes:

[0192] The ninth determining unit is further configured to determine, from the target video segments, a fourth video segment whose cleanliness quantization result meets a preset condition according to the cleanliness quantization result of each target video segment;

[0193] a tenth determining unit, configured to determine a weight of each fourth video segment according to the number of frames of each fourth video segment;

[0194] an eleventh determining unit, configured to determine a ratio of a weight of the fourth video segment to a weight of the candidate video segment;

[0195] The eighth determination unit is used to determine the colonoscopy quality result of the detection object according to the ratio and a preset threshold.

[0196] In one embodiment, the eighth determination unit is further used to determine that the colonoscopy quality result of the detection object is a quality result of a first cleanliness level if the ratio is greater than a preset threshold; if the ratio is not greater than the preset threshold, then determine that the colonoscopy quality result of the detection object is a quality result of a second cleanliness level; the first cleanliness level is less than the second cleanliness level.

[0197] In one embodiment, the colonoscope quality determination device further comprises:

[0198] An acquisition module, used for acquiring multiple colonoscopy video samples;

[0199] A seventh determination module is used to determine each first video segment sample in the plurality of colonoscopy video samples according to the start time point and the end time point of the ileocecal part in the plurality of colonoscopy video samples;

[0200] The first training module is used to train the initial ileocecal recognition model according to each first video segment sample and the label of each first video segment sample to obtain the ileocecal recognition model; the label of the first video segment sample includes an ileocecal label and a non-ileocecal label.

[0201] In one embodiment, the colonoscope quality determination device further comprises:

[0202] An acquisition module, used for acquiring multiple colonoscopy video samples;

[0203] An eighth determination module is used to determine each second video segment sample in the plurality of colonoscopy video samples according to the start time point and the end time point corresponding to the colonoscopy state in the plurality of colonoscopy video samples; the colonoscopy state includes the enter state, the withdraw state and the stop state;

[0204] The second training module is used to train the initial picture motion recognition model according to each second video clip sample and the label of each second video clip sample to obtain a picture motion recognition model; the label of the second video clip sample includes a camera entry phase label, a camera exit phase label and a camera stop phase label.

[0205] In one embodiment, the colonoscope quality determination device further comprises:

[0206] An acquisition module, used for acquiring multiple colonoscopy video samples;

[0207] A ninth determination module, used to determine each third video segment sample from a plurality of colonoscopy video samples;

[0208] The third training module is used to train the initial intestinal cleanliness scoring model according to each third video clip sample and the label of each third video clip sample to determine the intestinal cleanliness scoring model; the label of the third video clip sample includes the cleanliness quantification result of the third video clip sample.

[0209] Each module in the above-mentioned colonoscope quality determination device can be implemented in whole or in part by software, hardware, or a combination thereof. Each of the above-mentioned modules 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 operations corresponding to each of the above modules.

[0210] Those skilled in the art will understand that Fig.15 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.

[0211] 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 following steps are implemented:

[0212] Identify multiple colonoscopy video clips of the detection object through the withdrawal recognition model to determine the withdrawal start video clip;

[0213] Determining a plurality of target video segments in a defocusing phase according to the defocusing starting video segment;

[0214] Scoring each target video clip using an intestinal cleanliness scoring model to determine a quantified result of the cleanliness of each target video clip;

[0215] According to the quantification results of the cleanliness of each target video segment, the colonoscopy quality results of the test object are determined.

[0216] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0217] The multiple colonoscopy video segments are respectively identified by the ileocecal recognition model, and the first video segment including the ileocecal segment in the colonoscopy video segments is determined, and the first video segment is used as the start video segment for withdrawing the image.

[0218] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0219] The plurality of colonoscopy video segments are respectively identified by the picture motion recognition model, the second video segment in which the colonoscopy state is the withdrawn state for the first time is determined, and the second video segment is used as the withdrawn start video segment.

[0220] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0221] Determine, according to the defocusing start video clip, a candidate video clip in the defocusing stage;

[0222] According to the number of frames of the candidate video segments, the candidate video segments are evenly divided to obtain a plurality of target video segments in the de-focusing stage.

[0223] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0224] Determine, according to the defocusing start video clip, a candidate video clip in the defocusing stage;

[0225] Extracting features from the candidate video clips to obtain intestinal cleaning features of the candidate video clips;

[0226] A video segment with the same intestinal cleaning feature in the candidate video segments is taken as a target video segment to obtain a plurality of target video segments in the de-staging stage.

[0227] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0228] Determine the information video clip in the colonoscopy video according to the colonoscopy video and the screening model of the test object;

[0229] Based on the information video clip, a plurality of colonoscopy video clips are determined.

[0230] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0231] According to the clean quantization results of each target video segment, determining a third video segment whose clean quantization results meet a preset condition from each target video segment;

[0232] determining a ratio of the number of third video segments to the number of target video segments;

[0233] The colonoscopy quality result of the test object is determined based on the ratio and the preset threshold.

[0234] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0235] According to the clean quantization results of each target video segment, determining a fourth video segment whose clean quantization results meet a preset condition from each target video segment;

[0236] Determining a weight of each fourth video segment according to the number of frames of each fourth video segment;

[0237] determining a ratio of a weight of the fourth video segment to a weight of the candidate video segments;

[0238] The colonoscopy quality result of the test object is determined based on the ratio and the preset threshold.

[0239] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0240] If the ratio is greater than a preset threshold, it is determined that the colonoscope quality result of the test object is a quality result of the first cleaning level;

[0241] If the ratio is not greater than the preset threshold, it is determined that the colonoscopy quality result of the detected object is a quality result of the second cleanliness level; and the first cleanliness level is less than the second cleanliness level.

[0242] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0243] Obtain multiple colonoscopy video samples;

[0244] Determine each first video segment sample in the multiple colonoscopy video samples according to the start time point and the end time point of the ileocecal region in the multiple colonoscopy video samples;

[0245] An initial ileocecal recognition model is trained according to each first video segment sample and a label of each first video segment sample to obtain an ileocecal recognition model; the label of the first video segment sample includes an ileocecal label and a non-ileocecal label.

[0246] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0247] Obtain multiple colonoscopy video samples;

[0248] Determine each second video segment sample in the multiple colonoscopy video samples according to the start time point and the end time point corresponding to the colonoscopy state in the multiple colonoscopy video samples; the colonoscopy state includes the enter state, the withdraw state and the stop state;

[0249] The initial picture motion recognition model is trained according to each second video clip sample and the label of each second video clip sample to obtain a picture motion recognition model; the label of the second video clip sample includes a camera-in phase label, a camera-out phase label and a camera-stop phase label.

[0250] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0251] Obtain multiple colonoscopy video samples;

[0252] Determine each third video segment sample from a plurality of colonoscopy video samples;

[0253] According to each third video clip sample and the label of each third video clip sample, the initial intestinal cleanliness scoring model is trained to determine the intestinal cleanliness scoring model; the label of the third video clip sample includes the cleanliness quantification result of the third video clip sample.

[0254] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0255] Identify multiple colonoscopy video clips of the detection object through the withdrawal recognition model to determine the withdrawal start video clip;

[0256] Determining a plurality of target video segments in a defocusing phase according to the defocusing starting video segment;

[0257] Scoring each target video clip using an intestinal cleanliness scoring model to determine a quantified result of the cleanliness of each target video clip;

[0258] According to the quantification results of the cleanliness of each target video segment, the colonoscopy quality results of the test object are determined.

[0259] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0260] The multiple colonoscopy video segments are respectively identified by the ileocecal recognition model, and the first video segment including the ileocecal segment in the colonoscopy video segments is determined, and the first video segment is used as the start video segment for withdrawing the image.

[0261] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0262] The plurality of colonoscopy video segments are respectively identified by the picture motion recognition model, the second video segment in which the colonoscopy state is the withdrawn state for the first time is determined, and the second video segment is used as the withdrawn start video segment.

[0263] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0264] Determine, according to the defocusing start video clip, a candidate video clip in the defocusing stage;

[0265] According to the number of frames of the candidate video segments, the candidate video segments are evenly divided to obtain a plurality of target video segments in the de-focusing stage.

[0266] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0267] Determine, according to the defocusing start video clip, a candidate video clip in the defocusing stage;

[0268] Extracting features from the candidate video clips to obtain intestinal cleaning features of the candidate video clips;

[0269] A video segment with the same intestinal cleaning feature in the candidate video segments is taken as a target video segment to obtain a plurality of target video segments in the de-staging stage.

[0270] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0271] Determine the information video clip in the colonoscopy video according to the colonoscopy video and the screening model of the test object;

[0272] Based on the information video clip, a plurality of colonoscopy video clips are determined.

[0273] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0274] According to the clean quantization results of each target video segment, determining a third video segment whose clean quantization results meet a preset condition from each target video segment;

[0275] determining a ratio of the number of third video segments to the number of target video segments;

[0276] The colonoscopy quality result of the test object is determined based on the ratio and the preset threshold.

[0277] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0278] According to the clean quantization results of each target video segment, determining a fourth video segment whose clean quantization results meet a preset condition from each target video segment;

[0279] Determining a weight of each fourth video segment according to the number of frames of each fourth video segment;

[0280] determining a ratio of a weight of the fourth video segment to a weight of the candidate video segments;

[0281] The colonoscopy quality result of the test object is determined based on the ratio and the preset threshold.

[0282] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0283] If the ratio is greater than a preset threshold, it is determined that the colonoscope quality result of the test object is a quality result of the first cleaning level;

[0284] If the ratio is not greater than the preset threshold, it is determined that the colonoscopy quality result of the detected object is a quality result of the second cleanliness level; and the first cleanliness level is less than the second cleanliness level.

[0285] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0286] Obtain multiple colonoscopy video samples;

[0287] Determine each first video segment sample in the multiple colonoscopy video samples according to the start time point and the end time point of the ileocecal region in the multiple colonoscopy video samples;

[0288] An initial ileocecal recognition model is trained according to each first video segment sample and a label of each first video segment sample to obtain an ileocecal recognition model; the label of the first video segment sample includes an ileocecal label and a non-ileocecal label.

[0289] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0290] Obtain multiple colonoscopy video samples;

[0291] Determine each second video segment sample in the multiple colonoscopy video samples according to the start time point and the end time point corresponding to the colonoscopy state in the multiple colonoscopy video samples; the colonoscopy state includes the enter state, the withdraw state and the stop state;

[0292] The initial picture motion recognition model is trained according to each second video clip sample and the label of each second video clip sample to obtain a picture motion recognition model; the label of the second video clip sample includes a camera-in phase label, a camera-out phase label and a camera-stop phase label.

[0293] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0294] Obtain multiple colonoscopy video samples;

[0295] Determine each third video segment sample from a plurality of colonoscopy video samples;

[0296] According to each third video clip sample and the label of each third video clip sample, the initial intestinal cleanliness scoring model is trained to determine the intestinal cleanliness scoring model; the label of the third video clip sample includes the cleanliness quantification result of the third video clip sample.

[0297] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0298] Identify multiple colonoscopy video clips of the detection object through the withdrawal recognition model to determine the withdrawal start video clip;

[0299] Determining a plurality of target video segments in a defocusing phase according to the defocusing starting video segment;

[0300] Scoring each target video clip using an intestinal cleanliness scoring model to determine a quantified result of the cleanliness of each target video clip;

[0301] According to the quantification results of the cleanliness of each target video segment, the colonoscopy quality results of the test object are determined.

[0302] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0303] The multiple colonoscopy video segments are respectively identified by the ileocecal recognition model, and the first video segment including the ileocecal segment in the colonoscopy video segments is determined, and the first video segment is used as the start video segment for withdrawing the image.

[0304] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0305] The plurality of colonoscopy video segments are respectively identified by the picture motion recognition model, the second video segment in which the colonoscopy state is the withdrawn state for the first time is determined, and the second video segment is used as the withdrawn start video segment.

[0306] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0307] Determine, according to the defocusing start video clip, a candidate video clip in the defocusing stage;

[0308] According to the number of frames of the candidate video segments, the candidate video segments are evenly divided to obtain a plurality of target video segments in the de-focusing stage.

[0309] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0310] Determine, according to the defocusing start video clip, a candidate video clip in the defocusing stage;

[0311] Extracting features from the candidate video clips to obtain intestinal cleaning features of the candidate video clips;

[0312] A video segment with the same intestinal cleaning feature in the candidate video segments is taken as a target video segment to obtain a plurality of target video segments in the de-staging stage.

[0313] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0314] Determine the information video clip in the colonoscopy video according to the colonoscopy video and the screening model of the test object;

[0315] Based on the information video clip, a plurality of colonoscopy video clips are determined.

[0316] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0317] According to the clean quantization results of each target video segment, determining a third video segment whose clean quantization results meet a preset condition from each target video segment;

[0318] determining a ratio of the number of third video segments to the number of target video segments;

[0319] The colonoscopy quality result of the test object is determined based on the ratio and the preset threshold.

[0320] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0321] According to the clean quantization results of each target video segment, determining a fourth video segment whose clean quantization results meet a preset condition from each target video segment;

[0322] Determining a weight of each fourth video segment according to the number of frames of each fourth video segment;

[0323] determining a ratio of a weight of the fourth video segment to a weight of the candidate video segments;

[0324] The colonoscopy quality result of the test object is determined based on the ratio and the preset threshold.

[0325] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0326] If the ratio is greater than a preset threshold, it is determined that the colonoscope quality result of the test object is a quality result of the first cleaning level;

[0327] If the ratio is not greater than the preset threshold, it is determined that the colonoscopy quality result of the detected object is a quality result of the second cleanliness level; and the first cleanliness level is less than the second cleanliness level.

[0328] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0329] Obtain multiple colonoscopy video samples;

[0330] Determine each first video segment sample in the multiple colonoscopy video samples according to the start time point and the end time point of the ileocecal region in the multiple colonoscopy video samples;

[0331] An initial ileocecal recognition model is trained according to each first video segment sample and a label of each first video segment sample to obtain an ileocecal recognition model; the label of the first video segment sample includes an ileocecal label and a non-ileocecal label.

[0332] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0333] Obtain multiple colonoscopy video samples;

[0334] Determine each second video segment sample in the multiple colonoscopy video samples according to the start time point and the end time point corresponding to the colonoscopy state in the multiple colonoscopy video samples; the colonoscopy state includes the enter state, the withdraw state and the stop state;

[0335] The initial picture motion recognition model is trained according to each second video clip sample and the label of each second video clip sample to obtain a picture motion recognition model; the label of the second video clip sample includes a camera-in phase label, a camera-out phase label and a camera-stop phase label.

[0336] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0337] Obtain multiple colonoscopy video samples;

[0338] Determine each third video segment sample from a plurality of colonoscopy video samples;

[0339] According to each third video clip sample and the label of each third video clip sample, the initial intestinal cleanliness scoring model is trained to determine the intestinal cleanliness scoring model; the label of the third video clip sample includes the cleanliness quantification result of the third video clip sample.

[0340] 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.

[0341] 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.

[0342] 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 method for determining the quality of a colonoscope, characterized in that: The method comprises: Identify multiple colonoscopy video clips of the detection object through the withdrawal recognition model to determine the withdrawal start video clip; Determining a plurality of target video segments in a phase of rewinding according to the rewinding start video segment; Scoring each of the target video clips using an intestinal cleanliness scoring model to determine a quantified result of the cleanliness of each of the target video clips; The colonoscopy quality result of the detection object is determined according to the cleanliness quantification result of each target video segment.

2. The method according to claim 1, characterized in that The scope withdrawal recognition model includes an ileocecal recognition model; the number of frames of the colonoscopy video segment is a first number of frames, and the multiple colonoscopy video segments of the detection object are recognized by the scope withdrawal recognition model to determine the scope withdrawal start video segment, including: The multiple colonoscopy video segments are respectively identified by the ileocecal recognition model, and the first video segment including the ileocecal segment is determined in the colonoscopy video segments, and the first video segment is used as the start video segment for withdrawing the image.

3. The method according to claim 2, characterized in that The camera-retracting recognition model includes a screen movement recognition model; the number of frames of the colonoscopy video segment is a second number of frames; the first number of frames is less than the second number of frames; and the multiple colonoscopy video segments of the detection object are identified by the camera-retracting recognition model to determine the camera-retracting starting video segment, including: The multiple colonoscopy video segments are respectively identified by the image motion recognition model to determine the second video segment in which the colonoscopy state is the withdrawn state for the first time, and the second video segment is used as the withdrawn start video segment.

4. The method according to claim 1, characterized in that: The step of determining a plurality of target video segments in a fade-out phase according to the fade-out start video segment comprises: Determining a candidate video segment in a fade-out phase according to the fade-out start video segment; The candidate video segments are evenly divided according to the number of frames of the candidate video segments to obtain a plurality of the target video segments in the de-focusing stage.

5. The method according to claim 1, characterized in that The step of determining a plurality of target video segments in a fade-out phase according to the fade-out start video segment comprises: Determining a candidate video segment in a fade-out phase according to the fade-out start video segment; Extracting features from the candidate video clips to obtain intestinal cleaning features of the candidate video clips; A video segment having the same intestinal cleaning feature among the candidate video segments is used as a target video segment to obtain a plurality of target video segments in the de-staging stage.

6. The method according to any one of claims 1 to 5, characterized in that: Before identifying the multiple colonoscopy video clips of the detection object by using the recognizing model, the method further includes: Determining an information video segment in the colonoscopy video according to the colonoscopy video of the test object and the screening model; The multiple colonoscopy video segments are determined according to the information video segments.

7. The method according to claim 4, characterized in that Determining the colonoscopy quality result of the detected object according to the cleanliness quantification result of each target video segment includes: According to the clean quantization results of each of the target video segments, determining a third video segment whose clean quantization results meet a preset condition from each of the target video segments; Determining a ratio of the number of the third video segments to the number of the target video segments; The colonoscopy quality result of the detection object is determined according to the ratio and a preset threshold.

8. The method according to claim 5, characterized in that Determining the colonoscopy quality result of the detected object according to the cleanliness quantification result of each target video segment includes: According to the clean quantization results of each of the target video segments, determining a fourth video segment whose clean quantization results meet a preset condition from each of the target video segments; Determining a weight of each of the fourth video segments according to the number of frames of each of the fourth video segments; Determining a ratio of a weight of the fourth video segment to a weight of the candidate video segment; The colonoscopy quality result of the detection object is determined according to the ratio and a preset threshold.

9. The method according to claim 7 or 8, characterized in that: The method of determining the colonoscopy quality result of the detected object according to the ratio and a preset threshold value includes: If the ratio is greater than the preset threshold, determining that the colonoscope quality result of the detected object is a quality result of the first cleaning level; If the ratio is not greater than the preset threshold, it is determined that the colonoscopy quality result of the detected object is a quality result of the second cleanliness level; and the first cleanliness level is lower than the second cleanliness level.

10. The method according to claim 2, characterized in that The method further comprises: Obtain multiple colonoscopy video samples; Determine each first video segment sample in the multiple colonoscopy video samples according to the start time point and the end time point of the ileocecal region in the multiple colonoscopy video samples; An initial ileocecal recognition model is trained according to each of the first video segment samples and the labels of each of the first video segment samples to obtain the ileocecal recognition model; the labels of the first video segment samples include ileocecal labels and non-ileocecal labels.

11. The method according to claim 3, characterized in that The method further comprises: Obtain multiple colonoscopy video samples; Determine each second video segment sample in the plurality of colonoscopy video samples according to the start time point and the end time point corresponding to the colonoscopy state in the plurality of colonoscopy video samples; the colonoscopy state includes an enter-end state, a withdraw-end state, and a stop state; The initial picture motion recognition model is trained according to each of the second video clip samples and the labels of each of the second video clip samples to obtain the picture motion recognition model; the labels of the second video clip samples include a camera-in phase label, a camera-out phase label and a camera-stop phase label.

12. The method according to claim 1, characterized in that The method further comprises: Obtain multiple colonoscopy video samples; Determine each third video segment sample from the plurality of colonoscopy video samples; According to each of the third video clip samples and the label of each of the third video clip samples, the initial intestinal cleanliness scoring model is trained to determine the intestinal cleanliness scoring model; the label of the third video clip sample includes the cleanliness quantification result of the third video clip sample.

13. A device for determining the quality of a colonoscope, characterized in that: The device comprises: A first determination module is used to identify multiple colonoscopy video clips of the detection object through a withdrawal recognition model to determine a withdrawal start video clip; A second determining module is used to determine a plurality of target video segments in a fade-out stage according to the fade-out starting video segment; A third determination module is used to score each of the target video clips using an intestinal cleanliness scoring model to determine a cleanliness quantification result of each of the target video clips; The fourth determination module is used to determine the colonoscopy quality result of the detection object according to the cleanliness quantification result of each target video segment.

14. 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 steps of the method according to any one of claims 1 to 12 are implemented.

15. 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 12 are implemented.

16. 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 12 are implemented.

Images