A standard section recognition method, apparatus, ultrasonic equipment, and storage medium
By calculating the standard score of ultrasound images during obstetric examinations and accumulating the standard score using image continuity information, the problems of false alarms and missed detections in the identification of standard sections of ultrasound images are solved, and the identification accuracy is improved.
Patent Information
- Application Number
- CN202111679414.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-12-31
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2041-12-31
AI Technical Summary
In existing technologies, the identification of standard sections of ultrasound images during obstetric examinations suffers from false alarms and missed detections. In particular, when the fetal position varies, the classification model does not accurately predict blurred or uncovered images, resulting in a decrease in the accuracy of standard section image identification.
By obtaining the section category and prediction score of ultrasound images from the video stream, the standard score is calculated. The standard score is accumulated using the continuity information of the ultrasound images. The image is determined to be a standard section only when it reaches the standard score threshold, thus avoiding occasional false alarms and missed detections.
It improves the accuracy of standard section recognition, reduces the false alarm rate of non-standard sections and the false negative rate of standard sections, and ensures the reliability and accuracy of recognition.
Smart Images

Figure CN116433951B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of ultrasound technology, and more specifically, to a standard section recognition method, apparatus, ultrasound device, and computer-readable storage medium. Background Technology
[0002] During obstetric examinations, doctors need to operate the ultrasound probe while simultaneously pressing buttons on the machine to store ultrasound images from 35 sections. Related technologies utilize classification models to automatically identify the section categories of the ultrasound images. This involves inputting the acquired ultrasound images into a classification model to obtain a category prediction score and a standard section score. First, the section category corresponding to the maximum category prediction score is determined as the predicted section category of the ultrasound image. Then, whether the standard section score is greater than a threshold determines whether the ultrasound image corresponds to the standard section image of the predicted section category.
[0003] However, during the rapid movement of the ultrasound probe by the doctor, there will always be some blurry, special images not covered in the training set. The classification model's predictions for these images tend to be rather random, leading to some clearly non-standard images being detected as standard sections. In addition, because the fetal position varies when ultrasound images are acquired, even the most standard frame for some sections may not meet the requirements of the standard section images in the training set. As a result, the network's prediction output will not meet the threshold requirements, causing no standard section image to be detected when scanning that section category.
[0004] Therefore, how to avoid false alarms and missed detections of standard cross-section images and improve the accuracy of standard cross-section recognition is a technical problem that needs to be solved by those skilled in the art. Summary of the Invention
[0005] The purpose of this application is to provide a standard section recognition method, apparatus, ultrasonic device, and computer-readable storage medium, which avoids false alarms and missed detections of standard section images and improves the accuracy of standard section recognition.
[0006] To achieve the above objectives, this application provides a standard section recognition method, comprising:
[0007] Obtain a target ultrasound image from the video stream, and obtain the target section category and the target prediction score of the target section category from the target ultrasound image;
[0008] Calculate the standard score corresponding to the target section category based on the target prediction score of the target section category corresponding to the target ultrasound image;
[0009] If the standard score is greater than or equal to the standard score threshold, then the target ultrasound image is determined to be a standard section image of the target section category;
[0010] The calculation of the standard score corresponding to the target section category based on the target prediction score of the target section category corresponding to the target ultrasound image includes: if the section category of the previous frame ultrasound image of the target ultrasound image is not the target section category, then the standard score corresponding to the target section category is initialized; if the section category of the previous frame ultrasound image of the target ultrasound image is the target section category, then the standard score corresponding to the target section category calculated based on the target prediction score of the target section category corresponding to the previous frame ultrasound image of the target ultrasound image is obtained; and when the target prediction score of the target section category corresponding to the target ultrasound image is greater than or equal to a score threshold, the standard score is accumulated.
[0011] The target prediction score includes any one or a combination of several of the target standard section score, the target basic standard section score, and the target joint score. The joint score is used to comprehensively describe the standard degree and basic standard degree of the section.
[0012] Accordingly, when the target prediction score of the target section category corresponding to the target ultrasound image is greater than or equal to the score threshold, the standard score is accumulated, including:
[0013] Determine the score threshold corresponding to the predicted score of each target in the target section category corresponding to the target ultrasound image;
[0014] If the target prediction score of any target section category corresponding to the target ultrasound image is greater than or equal to the corresponding score threshold, then the standard score is accumulated by the corresponding score.
[0015] The step of acquiring the target section category of the target ultrasound image and the target prediction score of the target section category includes:
[0016] The target ultrasound image is input into a classification model, and the target section category of the target ultrasound image is obtained from the output of the classification model.
[0017] Obtain the target standard section score corresponding to the target section category from the output of the classification model, and / or obtain the target basic standard section score corresponding to the target section category from the output of the classification model, and / or calculate the target joint score based on the target standard section score, the target basic standard section score and the joint score parameter of the target section category corresponding to the target ultrasound image.
[0018] This also includes:
[0019] Determine the lower bound and upper bound of the joint fraction parameter corresponding to the target section category;
[0020] Based on the number of target frames in the ultrasound images preceding the target ultrasound image that have the target candidate section category, interpolation is performed between the lower bound of the joint score parameter and the upper bound of the joint score parameter to calculate the joint score parameter of the target section category corresponding to the target ultrasound image.
[0021] Wherein, determining the lower bound and upper bound of the joint score parameter corresponding to the target section category includes:
[0022] Obtain target slice samples whose sum of the standard slice score and the basic standard slice score corresponding to the candidate slice category is greater than or equal to a preset value;
[0023] In a coordinate system containing the coordinate points corresponding to the target section sample, a first target linear function satisfying a first condition and a second target linear function satisfying a second condition are determined; wherein, the coordinate points corresponding to the target section sample in the coordinate system are determined based on the standard section fraction and the basic standard section fraction of the target section sample, the first condition includes that the sum of the distances between all the coordinate points corresponding to the target section sample and the first target linear function is minimized and that the basic standard section fractions corresponding to all the non-standard section samples are located at the function value of the first target linear function, and the second condition includes that the sum of the distances between all the coordinate points corresponding to the target section sample and the second target linear function is minimized;
[0024] The parameters of the first objective linear function are determined as the upper bound of the joint fractional parameters, and the parameters of the second objective linear function are determined as the lower bound of the joint fractional parameters.
[0025] The determination of the score threshold corresponding to the predicted score of each target in the target section category corresponding to the target ultrasound image includes:
[0026] Determine the lower and upper threshold bounds of the target standard section score or the target basic standard section score corresponding to the target section category;
[0027] Based on the number of target frames in the ultrasound images preceding the target ultrasound image that have the same section category as the target section, interpolation is performed between the lower threshold and the upper threshold to calculate the score threshold corresponding to the target standard section score or the target basic standard section score; wherein, the score threshold is negatively correlated with the number of target frames.
[0028] The determination of the lower and upper threshold bounds of the target standard section score corresponding to the target section category includes:
[0029] Obtain standard and non-standard section samples corresponding to the target section category, and determine the standard section score corresponding to the standard and non-standard section samples;
[0030] The standard segment score is used as the target standard segment score threshold. The weighted value of precision and recall is calculated according to the first weight ratio. The target standard segment score threshold corresponding to the maximum weighted value is determined as the upper bound of the target standard segment score threshold.
[0031] The standard segment score is used as the target standard segment score threshold. The weighted value of precision and recall is calculated according to the second weight ratio. The target standard segment score threshold corresponding to the maximum weighted value is determined as the lower bound of the target standard segment score threshold.
[0032] Wherein, the first weight ratio and the second weight ratio are the ratios of the weight of precision to the weight of recall, and the first weight ratio is greater than or equal to the second weight ratio.
[0033] The determination of the lower and upper threshold bounds of the basic target standard section score corresponding to the target section category includes:
[0034] Obtain basic standard section samples and non-standard section samples corresponding to the target section category, and determine the basic standard section scores corresponding to the basic standard section samples and non-standard section samples;
[0035] The basic standard segment score is used as the target basic standard segment score threshold. The weighted value of precision and recall is calculated according to the third weight ratio. The target basic standard segment score threshold corresponding to the maximum weighted value is determined as the upper bound of the threshold of the basic target standard segment score.
[0036] The basic standard segment score is used as the target basic standard segment score threshold. The weighted value of precision and recall is calculated according to the fourth weight ratio. The target basic standard segment score threshold corresponding to the maximum weighted value is determined as the lower bound of the threshold of the basic target standard segment score.
[0037] Wherein, the third weight ratio and the fourth weight ratio are the ratios of the weight of precision to the weight of recall, and the third weight ratio is greater than or equal to the fourth weight ratio.
[0038] To achieve the above objectives, this application provides a standard cross-section recognition device, comprising:
[0039] The acquisition module is used to acquire a target ultrasound image from a video stream, and to acquire the target section category and the target prediction score of the target section category of the target ultrasound image;
[0040] The calculation module is used to calculate the standard score corresponding to the target section category based on the target prediction score of the target section category corresponding to the target ultrasound image;
[0041] The determination module is used to determine that the target ultrasound image is a standard section image of the target section category when the standard score is greater than or equal to the standard score threshold.
[0042] Specifically, the calculation module is used to: if the section category of the previous frame ultrasound image of the target ultrasound image is not the target section category, initialize the standard score corresponding to the target section category; if the section category of the previous frame ultrasound image of the target ultrasound image is the target section category, obtain the standard score corresponding to the target section category calculated based on the target prediction score of the target section category corresponding to the previous frame ultrasound image of the target ultrasound image; and accumulate the standard score when the target prediction score of the target section category corresponding to the target ultrasound image is greater than or equal to the score threshold.
[0043] To achieve the above objectives, this application provides an ultrasonic device, comprising:
[0044] Memory, used to store computer programs;
[0045] A processor is used to implement the steps of the standard section recognition method described above when executing the computer program.
[0046] To achieve the above objectives, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the standard section recognition method described above.
[0047] As can be seen from the above scheme, the standard section recognition method provided in this application includes: acquiring a target ultrasound image from a video stream, acquiring the target section category of the target ultrasound image and the target prediction score of the target section category; calculating the standard score corresponding to the target section category based on the target prediction score of the target section category corresponding to the target ultrasound image; if the standard score is greater than or equal to the standard score threshold, then determining that the target ultrasound image is a standard section image of the target section category; wherein, calculating the standard score corresponding to the target section category based on the target prediction score of the target section category corresponding to the target ultrasound image includes: if the section category of the previous frame ultrasound image of the target ultrasound image is not the target section category, then initializing the standard score corresponding to the target section category; if the section category of the previous frame ultrasound image of the target ultrasound image is the target section category, then acquiring the standard score corresponding to the target section category calculated based on the target prediction score of the target section category corresponding to the previous frame ultrasound image of the target ultrasound image, then accumulating the standard score when the target prediction score of the target section category corresponding to the target ultrasound image is greater than or equal to the score threshold.
[0048] The standard section recognition method provided in this application maintains a corresponding standard score for each section category. For each frame of ultrasound image acquired by the ultrasound equipment, if the target prediction score of its corresponding section category is greater than or equal to a score threshold, the standard score corresponding to that section category is accumulated. For the currently acquired target ultrasound image, its corresponding section category is the target section category. The more consecutive ultrasound images with the target section category in the previous ultrasound images, the larger the currently accumulated standard score. When the standard score is greater than or equal to the standard score threshold, the target ultrasound image is determined to be a standard section image of the target section category. Therefore, the standard section recognition method provided in this application utilizes the continuity information in the ultrasound video to calculate the accumulated standard score of the current target ultrasound image, avoiding the detection of sporadic standard section images. Using the standard score to measure the relative standardization also avoids the inability to detect absolutely standard sections (e.g., using a predetermined threshold to measure whether it is standard), reducing the false alarm rate of non-standard sections and the false negative rate of standard sections, thus improving the accuracy of standard section recognition. This application also discloses a standard section recognition device, an ultrasonic device, and a computer-readable storage medium, which can achieve the same technical effects as described above.
[0049] It should be understood that the above general description and the following detailed description are merely exemplary and do not limit this application. Attached Figure Description
[0050] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. The drawings are used to provide a further understanding of this disclosure and constitute a part of the specification. They are used together with the following detailed description to explain this disclosure, but do not constitute a limitation of this disclosure. In the drawings:
[0051] Figure 1 This is a flowchart illustrating a standard section recognition method according to an exemplary embodiment;
[0052] Figure 2 A flowchart illustrating another standard section recognition method according to an exemplary embodiment;
[0053] Figure 3 This is a schematic diagram illustrating a method for determining the upper and lower bounds of a threshold corresponding to a long axis section of the femur according to an exemplary embodiment;
[0054] Figure 4 This is a structural diagram illustrating a standard section recognition device according to an exemplary embodiment;
[0055] Figure 5 This is a structural diagram of an ultrasonic device according to an exemplary embodiment. Detailed Implementation
[0056] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application. Furthermore, in the embodiments of this application, "first," "second," etc., are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.
[0057] This application discloses a standard section recognition method, which avoids false alarms and missed detections of standard section images and improves the accuracy of standard section recognition.
[0058] See Figure 1 A flowchart illustrating a standard section recognition method according to an exemplary embodiment is shown below. Figure 1 As shown, it includes:
[0059] S101: Obtain a target ultrasound image from the video stream, and obtain the target section category of the target ultrasound image and the target prediction score of the target section category;
[0060] In this embodiment, the execution entity is an ultrasound device, and the purpose is to automatically identify the section category of the acquired target ultrasound image and determine whether it is a standard section image of that section category. In this embodiment, the section category of the target ultrasound image can be identified based on a classification model. The classification model can employ a convolutional neural network or other artificial intelligence algorithm models. Multiple section categories are predefined, and rules are established for standard sections, basic standard sections, and non-standard sections corresponding to each section category. Based on this, multiple standard sections, multiple basic standard sections, and multiple non-standard sections corresponding to each section category are determined as training samples to train the classification model. The trained classification model is used to identify section categories. In specific implementation, the ultrasound device acquires the target ultrasound image and inputs it into the trained classification model. The output of the classification model is the predicted score corresponding to each section category. For example, 13 key section categories in obstetric screening are predefined, and the classification model outputs the predicted scores corresponding to each of the 13 section categories, i.e., 13 predicted scores. Preferably, the predicted scores include standard section scores, basic standard section scores, and non-standard section scores; that is, the output of the classification model is the standard section score, basic standard section score, and non-standard section score corresponding to each section category. In the example above, the classification model outputs the standard section scores, basic standard section scores, and non-standard section scores corresponding to 13 section categories, i.e., 13 × 3 = 39 predicted scores.
[0061] Furthermore, the target section category of the target ultrasound image is obtained from the output of the classification model. As a feasible implementation, the section category corresponding to the maximum predicted score can be used as the target section category of the target ultrasound image predicted by the classification model. In a specific implementation, if the convolutional neural network outputs a single predicted score for each section category, the section category corresponding to the maximum predicted score is used as the target section category of the target ultrasound image predicted by the convolutional neural network. If the convolutional neural network outputs multiple predicted scores for each section category, the section categories corresponding to the maximum predicted scores of all values are used as the target section categories of the target ultrasound image predicted by the convolutional neural network.
[0062] As another feasible implementation, the section category corresponding to the maximum predicted score is determined as the target candidate category of the target ultrasound image; the state category of ultrasound images in a preset number of frames preceding the target ultrasound image is determined in the video stream; if the state category is consistent with the target candidate category, then the state category is determined as the target section category corresponding to the target ultrasound image. In a specific implementation, a target queue is maintained to store the predicted scores corresponding to multiple consecutive frames of ultrasound images predicted by the classification model. Each element in the target queue records the predicted score corresponding to one frame of ultrasound image, and the section category corresponding to the maximum predicted score in the target queue is determined as the pending state category. Further, it is determined that the number of frames in the preset number of ultrasound images whose candidate category is consistent with the pending state category meets a preset condition. If so, the pending state category is determined as the state category of the ultrasound images in the preset number of frames preceding the target ultrasound image. The preset condition here may include the number of frames in the preset number of ultrasound images whose candidate category is consistent with the pending state category being greater than or equal to a preset value, or the ratio of the number of frames in the preset number of ultrasound images whose candidate category is consistent with the pending state category to the preset number of frames being greater than or equal to a preset ratio. For example, the preset number of frames is 5, and the preset condition is that the number of frames in which the candidate category of the cross-section class matches the pending state class is greater than or equal to 3. If the target candidate category of the target ultrasound image matches the state category of the ultrasound image in the preset number of frames preceding the target ultrasound image, it indicates that the classification model has high consistency in predicting the cross-section class of multiple consecutive frames, and the target candidate category is determined as the target cross-section class corresponding to the target ultrasound image; otherwise, the output cannot determine the cross-section class of the target ultrasound image.
[0063] In this embodiment, the target ultrasound image is determined to be a standard section image of the target section category based on the target prediction score corresponding to the target section category. The target prediction score here may include the target standard section score and the target basic standard section score corresponding to the target section category, which can be obtained from the output of the classification model. The target prediction score may also include a joint target score calculated based on the target standard section score and the target basic standard section score; the joint target score is used to comprehensively describe the standardization and basic standardization of the section.
[0064] S102: Calculate the standard score corresponding to the target section category based on the target prediction score of the target section category corresponding to the target ultrasound image;
[0065] The calculation of the standard score corresponding to the target section category based on the target prediction score of the target section category corresponding to the target ultrasound image includes: if the section category of the previous frame ultrasound image of the target ultrasound image is not the target section category, then the standard score corresponding to the target section category is initialized; if the section category of the previous frame ultrasound image of the target ultrasound image is the target section category, then the standard score corresponding to the target section category calculated based on the target prediction score of the target section category corresponding to the previous frame ultrasound image of the target ultrasound image is obtained; and when the target prediction score of the target section category corresponding to the target ultrasound image is greater than or equal to a score threshold, the standard score is accumulated.
[0066] In this embodiment, a standard score is maintained for each section category. The initial value can be zero. For each frame of ultrasound image acquired by the ultrasound device, if the target prediction score of its corresponding section category is greater than or equal to the score threshold, the standard score is accumulated.
[0067] In specific implementation, if the target ultrasound image is the first frame of ultrasound image where the target section category is detected, the current value of the standard score corresponding to the target section category is an initial value, such as zero. If the target prediction score of the target section category corresponding to the target ultrasound image is greater than or equal to a score threshold, the standard score is accumulated; otherwise, it is initialized. If the section category of the previous frame of ultrasound image is also the target section category, the standard score of the target section category has already been calculated when recognizing that previous frame (if the target prediction score of the target section category corresponding to the previous frame is greater than or equal to a score threshold, the standard score of the target section category is accumulated; otherwise, it is initialized). Therefore, the standard score calculated based on the previous frame of ultrasound image is obtained. If the target prediction score of the target section category corresponding to the target ultrasound image is greater than or equal to a score threshold, the standard score is accumulated; otherwise, it is initialized. This recognition process is repeated when recognizing the next frame of ultrasound image.
[0068] For example, the first frame of the ultrasound image corresponds to section category A, and the initial standard score for section category A is 0. Since the target prediction score for section category A in the first frame is greater than the score threshold, the standard score is incremented by 20, so the standard score for section category A is now 20. The second frame of the ultrasound image corresponds to section category A, and its target prediction score is greater than the score threshold, so the standard score for section category A is now 40. The third frame of the ultrasound image corresponds to section category A, and its target prediction score is less than the score threshold, so the standard score for section category A is initialized to 0. The fourth frame of the ultrasound image corresponds to section category A, and its target prediction score is greater than the score threshold, so the standard score for section category A is incremented again from 0 to 20. The fifth frame of the ultrasound image corresponds to section category B. Since this is the first ultrasound image detected in section category B, the initial standard score for section category B is 0. Because the target prediction score for section category B in the fifth frame is greater than the score threshold, the standard score is incremented by 20, resulting in a standard score of 20 for section category B. The sixth frame of the ultrasound image reverts to section category A. Since the section category of its preceding frame (the fifth frame) is section category B, which is different from section category A, the standard score for section category A is incremented again from 0 to 20.
[0069] Furthermore, if the target predicted score includes multiple predicted scores, different score thresholds can be set for different predicted scores, and different cumulative score values can be set for different predicted scores. If any target predicted score of the target section category corresponding to the target ultrasound image is greater than or equal to the corresponding score threshold, the standard score is accumulated by the corresponding score. For example, if the target predicted score includes a standard section score and a basic standard section score, if the standard section score is greater than or equal to the standard section score threshold, the standard score is increased by a first preset value; if the basic standard section score is greater than or equal to the basic standard section score threshold, the standard score is increased by a second preset value; if both the standard section score and the basic standard section score are greater than or equal to the basic standard section score threshold, the standard score is increased by both the first and second preset value scores simultaneously. Additionally, it should be noted that the standard score is initialized only when each target predicted score is less than the corresponding score threshold. For example, if the standard section score is less than the standard section score threshold and the basic standard section score is less than the basic standard section score threshold, the standard section score is set to zero. The specific values of the first and second preset values are not limited here and can be flexibly set according to the actual situation. Preferably, the score of the first preset value is greater than the score of the second preset value. For example, if the standard section score is greater than or equal to the standard section score threshold, the standard score is increased by 50; if the basic standard section score is greater than or equal to the basic standard section score threshold, the standard score is increased by 25.
[0070] In the following situations: if the state category of the ultrasound images in the preceding preset number of frames matches the target candidate category predicted by the classification model, then the state category is determined as the target section category corresponding to the target ultrasound image; otherwise, the output indicates that the section category of the target ultrasound image cannot be determined. Furthermore, if section categories of multiple consecutive ultrasound images are detected, the standard score calculation step is triggered. For example, if section categories of three consecutive ultrasound images are detected, the standard score calculation step is triggered; otherwise, the target ultrasound image is directly determined to be a non-standard image. This avoids detecting occasional standard section images while improving detection efficiency.
[0071] S103: If the standard score is greater than or equal to the standard threshold, then the target ultrasound image is determined to be a standard section image of the target section category.
[0072] In this step, it is determined whether the standard score is greater than or equal to the standard score threshold. If so, the target ultrasound image is determined to be a standard section image of the target section category; otherwise, the target ultrasound image is determined to be a non-standard section image.
[0073] The standard section recognition method provided in this application maintains a corresponding standard score for each section category. For each frame of ultrasound image acquired by the ultrasound device, if the target prediction score of its corresponding section category is greater than or equal to a score threshold, the standard score corresponding to that section category is accumulated. For the currently acquired target ultrasound image, its corresponding section category is the target section category. The more consecutive ultrasound images with the target section category in the previous ultrasound images, the larger the currently accumulated standard score. When the standard score is greater than or equal to the standard score threshold of that category, the target ultrasound image is determined to be a standard section image of the target section category. Therefore, the standard section recognition method provided in this application utilizes the continuity information in the ultrasound video to calculate the accumulated standard score of the current target ultrasound image, avoiding the detection of sporadic standard section images. Using the standard score to measure the relative standardization also avoids the inability to detect absolutely standard sections (e.g., using a predetermined threshold to measure whether it is standard), reducing the false alarm rate of non-standard sections and the false negative rate of standard sections, thus improving the accuracy of standard section recognition.
[0074] This application discloses a standard section recognition method. Compared to the first embodiment, this embodiment further explains and optimizes the technical solution. Specifically:
[0075] See Figure 2 A flowchart illustrating another standard section recognition method according to an exemplary embodiment, such as... Figure 2 As shown, it includes:
[0076] S201: Obtain a target ultrasound image from a video stream, input the target ultrasound image into a classification model, and obtain the target section category, the target standard section score, and the target basic standard section score of the target ultrasound image from the output of the classification model;
[0077] S202: Calculate the target joint score based on the target standard section score, the target basic standard section score, and the joint score parameter of the target section category corresponding to the target ultrasound image;
[0078] In practice, the joint objective score is used to comprehensively describe the standard and basic standard scores of the cross-section. Specifically, P = kPsp + Pbsp + b, where P is the joint objective score, Psp is the standard objective cross-section score, Pbsp is the basic objective cross-section score, and (k, b) are the joint score parameters.
[0079] In a preferred embodiment, this embodiment further includes: determining the lower bound and the upper bound of the joint fraction parameter corresponding to the target section category; and calculating the joint fraction parameter of the target section category corresponding to the target ultrasound image by interpolating between the lower bound and the upper bound of the joint fraction parameter based on the number of target frames of ultrasound images preceding the target ultrasound image whose section category is the target candidate section category.
[0080] In practice, the upper and lower bounds of the joint fraction parameter corresponding to each slice category are predetermined. After the target slice category of the target ultrasound image is determined, the ultrasound equipment determines the upper and lower bounds of the joint fraction parameter corresponding to the target slice category. Based on the number of target frames of ultrasound images preceding the target ultrasound image whose slice category is a candidate slice category, interpolation is performed between the upper and lower bounds of the joint fraction parameter to calculate the joint fraction parameter of the target slice category corresponding to the target ultrasound image.
[0081] It should be noted that the upper and lower bounds of the joint score parameters corresponding to each slice category can be manually set according to the actual situation, or they can be automatically searched using slice samples of each slice category. Using the aforementioned target section categories as an example, determining the lower bound and upper bound of the joint fraction parameter corresponding to the target section category includes: obtaining target section samples whose sum of standard section score and basic standard section score corresponding to the candidate section category is greater than or equal to a preset value; determining a first target linear function satisfying a first condition and a second target linear function satisfying a second condition in a coordinate system containing the coordinate points corresponding to the target section samples; wherein, based on the standard section score and basic standard section score of the target section samples, the coordinate points corresponding to the target section samples in the coordinate system are determined, the first condition includes that the sum of the distances between the coordinate points corresponding to all the target section samples and the first target linear function is minimized and that the basic standard section scores corresponding to all the non-standard section samples are located at the function value of the first target linear function, and the second condition includes that the sum of the distances between the coordinate points corresponding to all the target section samples and the second target linear function is minimized; the parameters of the first target linear function are determined as the upper bound of the joint fraction parameter, and the parameters of the second target linear function are determined as the lower bound of the joint fraction parameter.
[0082] In specific implementation, target slice samples are obtained where the sum of the standard slice score and the basic standard slice score corresponding to the candidate slice category is greater than or equal to a preset value. For example, target samples where the sum of the standard slice score and the basic standard slice score corresponding to the candidate slice category is greater than or equal to 0.4 are obtained. A coordinate system is constructed with the standard slice score on the horizontal axis and the basic standard slice score on the vertical axis. The coordinate points corresponding to all target slice samples and non-standard slice samples are plotted in the coordinate system. A first objective linear function is constructed such that the sum of the distances between the coordinate points corresponding to all target slice samples and the first objective linear function is minimized. Simultaneously, the constraint condition is that the coordinate points corresponding to all non-standard slice samples are located below the first objective linear function. The parameters of the first objective linear function are then solved as the upper bound of the joint score parameters. The objective function is:
[0083]
[0084] The constraint is: subject to [P′ sp 1][kb] T ≥P′ bsp .
[0085] Where k and b are the slope and intercept of the first objective linear function, i.e., the joint fractional parameters, P sp P is the predicted score of the standard section or the basic standard section. bsp For the standard section or the basic standard section, the basic standard section prediction score, P' sp For non-standard sections or other standard sections, predict the score, P' bsp The base standard section prediction score for non-standard sections or other section models.
[0086] Furthermore, an unconstrained second objective linear function is constructed such that the sum of distances between the coordinate points corresponding to all target section samples and the second objective linear function is minimized. The parameters of the second objective linear function are then solved as the lower bound of the joint fraction parameters. The objective function is:
[0087]
[0088] Where k and b are the slope and intercept of the second objective linear function, i.e., the joint fractional parameters.
[0089] Taking the long axis section of the femur as an example, such as Figure 3As shown, the horizontal axis represents the standard section score (SP Prob), and the vertical axis represents the basic standard section score (BSP Prob). `sp` represents the standard section, `bsp` represents the basic standard section, `nsp` represents the non-standard section, and `other` represents other sections. The two lines perpendicular to the horizontal axis represent the upper and lower bounds of the standard section score threshold, with the left line representing the lower bound and the right line representing the upper bound. The two lines parallel to the horizontal axis represent the upper and lower bounds of the basic standard section score threshold, with the lower line representing the lower bound and the upper line representing the upper bound. The two diagonal lines represent the upper and lower bounds of the joint score parameter, with the upper right diagonal line representing the upper bound and the lower left diagonal line representing the lower bound.
[0090] S203: Calculate the standard score corresponding to the target section category based on the target standard section score, the target basic standard section score, and the target joint score of the target section category corresponding to the target ultrasound image;
[0091] The calculation of the standard score corresponding to the target section category based on the target standard section score, target basic standard section score, and target joint score of the target section category corresponding to the target ultrasound image includes: if the section category of the previous frame ultrasound image of the target ultrasound image is the target section category, then the standard score corresponding to the target section category calculated based on the target standard section score, target basic standard section score, and target joint score of the target section category corresponding to the previous frame ultrasound image of the target ultrasound image is obtained; if the section category of the previous frame ultrasound image of the target ultrasound image is not the target section category, then the standard score corresponding to the target section category is initialized; a score threshold corresponding to the predicted score of each target of the target section category corresponding to the target ultrasound image is determined; if any target predicted score of the target section category corresponding to the target ultrasound image is greater than or equal to the corresponding score threshold, then the standard score is accumulated by the corresponding score.
[0092] In a preferred embodiment, determining the score threshold corresponding to the predicted score of each target in the target section category corresponding to the target ultrasound image includes: determining a lower threshold and an upper threshold for the target standard section score or the target basic standard section score corresponding to the target section category; calculating the score threshold corresponding to the target standard section score or the target basic standard section score by interpolating between the lower threshold and the upper threshold based on the number of target frames of ultrasound images preceding the target ultrasound image with the section category of the target section; wherein the score threshold is negatively correlated with the number of target frames.
[0093] In practice, the upper and lower bounds of the prediction score threshold are pre-determined for each slice category. After determining the target slice category of the target ultrasound image, the ultrasound equipment determines the upper and lower bounds of the prediction score threshold corresponding to the target slice category. Further, based on the number of target frames in the ultrasound images preceding the target ultrasound image that have a candidate slice category, the score threshold is calculated by interpolation between the upper and lower bounds of the prediction score threshold. Linear interpolation can be used here. For example, if the interpolation frame range is 10 frames, the upper bound of the prediction score threshold for slice category A is A0, and the lower bound is A1; the upper bound of the prediction score threshold for slice category B is B0, and the lower bound is B1. If the predicted section category of the first frame of ultrasound image is section category A, then its corresponding score threshold is A0. If the predicted section category of the second frame of ultrasound image is also section category A, then its corresponding score threshold is A0+(A1-A0) / 10. If the predicted section category of the third frame of ultrasound image is also section category A, then its corresponding score threshold is A0+2(A1-A0) / 10, and so on. If the predicted section category of the tenth frame of ultrasound image is also section category A, then its corresponding score threshold is A1. If the predicted section category of the first frame of ultrasound image is section category A, then its corresponding score threshold is A0. If the predicted section category of the second frame of ultrasound image is also section category A, then its corresponding score threshold is A0+(A1-A0) / 10. However, if the predicted section category of the third frame of ultrasound image is section category B, then its corresponding score threshold is B0. If the predicted section category of the fourth frame of ultrasound image is still section category B, then its corresponding score threshold is B0+(B1-B0) / 10, and so on.
[0094] It is evident that the score threshold is negatively correlated with the number of target frames in ultrasound images preceding the target ultrasound image that belong to the target section category. That is, when the target section category is first predicted, the score threshold uses the upper bound of the predicted score threshold. As the number of accumulated ultrasound images with the target section category increases, the score threshold gradually decreases until it reaches the lower bound of the predicted score threshold. It is understandable that for section categories with similar structures, the predicted scores output by the classification model are similar and not very high. Therefore, in this embodiment, a score threshold for the target section category corresponding to the target ultrasound image is calculated. Subsequently, the predictive score of the target section category corresponding to the target ultrasound image is compared with this score threshold to determine whether the target section category predicted by the classification model can be trusted. Furthermore, the larger the number of target frames in ultrasound images preceding the target ultrasound image that belong to the target section category, the higher the trust level of the predicted target section category, thus reducing the score threshold.
[0095] It should be noted that the upper and lower bounds of the standard section score threshold and the basic standard section score threshold for each section category can be manually set according to the actual situation, or they can be automatically searched using section samples of each section category. Using the aforementioned target segment categories as an example, determining the lower and upper threshold bounds of the target standard segment score corresponding to the target segment category includes: obtaining standard segment samples and non-standard segment samples corresponding to the target segment category, and determining the standard segment scores corresponding to the standard segment samples and non-standard segment samples; using the standard segment score as the target standard segment score threshold, calculating the weighted value of precision and recall according to a first weight ratio, and determining the target standard segment score threshold corresponding to the maximum weighted value as the upper threshold bound of the target standard segment score; using the standard segment score as the target standard segment score threshold, calculating the weighted value of precision and recall according to a second weight ratio, and determining the target standard segment score threshold corresponding to the maximum weighted value as the lower threshold bound of the target standard segment score; wherein, the first weight ratio and the second weight ratio are the ratio of the weight of precision to the weight of recall, and the first weight ratio is greater than or equal to the second weight ratio.
[0096] In practice, for the segment samples corresponding to the target segment category, basic standard segment samples are removed. That is, the upper and lower bounds of the standard segment score threshold are determined using standard and non-standard segment samples. The standard segment score is used as the target standard segment score threshold. The weighted values of precision and recall are calculated according to a first weight ratio. The target standard segment score threshold corresponding to the maximum weighted value is determined as the upper bound of the standard segment score threshold. For example, the ratio of the precision weight to the recall weight (i.e., the first weight ratio) is 5:1. Similarly, the standard segment score is used as the target standard segment score threshold. The weighted values of precision and recall are calculated according to a second weight ratio. The target standard segment score threshold corresponding to the maximum weighted value is determined as the lower bound of the standard segment score threshold. For example, the ratio of the precision weight to the recall weight (i.e., the second weight ratio) is 1:1.
[0097] Accordingly, determining the lower and upper threshold bounds of the basic target standard segment score corresponding to the target segment category includes: obtaining basic standard segment samples and non-standard segment samples corresponding to the target segment category, and determining the basic standard segment scores corresponding to the basic standard segment samples and non-standard segment samples; using the basic standard segment score as the target basic standard segment score threshold, calculating the weighted value of precision and recall according to the third weight ratio, and determining the target basic standard segment score threshold corresponding to the maximum weighted value as the upper threshold bound of the basic target standard segment score; using the basic standard segment score as the target basic standard segment score threshold, calculating the weighted value of precision and recall according to the fourth weight ratio, and determining the target basic standard segment score threshold corresponding to the maximum weighted value as the lower threshold bound of the basic target standard segment score; wherein, the third weight ratio and the fourth weight ratio are the ratio of the weight of precision to the weight of recall, and the third weight ratio is greater than or equal to the fourth weight ratio.
[0098] In practice, for the segment samples corresponding to the target segment category, standard segment samples are removed. That is, the upper and lower bounds of the basic standard segment score threshold are determined using basic standard segment samples and non-standard segment samples. The basic standard segment score is used as the target basic standard segment score threshold. The weighted values of precision and recall are calculated according to the third weight ratio. The target basic standard segment score threshold corresponding to the maximum weighted value is determined as the upper bound of the basic standard segment score threshold. For example, the ratio of the precision weight to the recall weight, i.e., the third weight ratio, is 10:1. Similarly, the basic standard segment score is used as the target basic standard segment score threshold. The weighted values of precision and recall are calculated according to the fourth weight ratio. The target basic standard segment score threshold corresponding to the maximum weighted value is determined as the lower bound of the basic standard segment score threshold. For example, the ratio of the precision weight to the recall weight, i.e., the fourth weight ratio, is 2:1.
[0099] Understandably, requirements will change constantly during the research and development process, and manually adjusting the upper and lower bounds of each score threshold would be a huge workload. Therefore, the above-mentioned method of automatically determining the upper and lower bounds of each score threshold can improve the algorithm iteration speed.
[0100] In this embodiment, if the standard section score of the target section category corresponding to the target ultrasound image is greater than or equal to the standard section score threshold, the standard score is accumulated by a first preset value. If the basic standard section score of the target section category corresponding to the target ultrasound image is greater than or equal to the basic standard section score threshold, the standard score is accumulated by a second preset value. If the joint score of the target section category corresponding to the target ultrasound image is greater than or equal to the joint score threshold, the standard score is accumulated by a third preset value. Here, the joint score threshold can be 0. When none of the above conditions are met, that is, when the standard section score is less than the standard section score threshold, the joint score is less than the joint score threshold, and the basic standard section score is less than the basic standard section score threshold, the standard score is initialized.
[0101] S204: If the standard score is greater than or equal to the standard threshold, then the target ultrasound image is determined to be a standard section image of the target section category.
[0102] Therefore, in this embodiment, the standard score of the section category simultaneously considers the standard section score, the basic standard section score, and the joint score. It uses the continuity information in the ultrasound video to calculate the cumulative standard score of the current target ultrasound image, which reduces the false alarm rate of non-standard sections and the false negative rate of standard sections, and further improves the accuracy of standard section recognition.
[0103] The following describes a standard section recognition device provided in an embodiment of this application. The standard section recognition device described below and the standard section recognition method described above can be referred to each other.
[0104] See Figure 4 A structural diagram of a standard section recognition device is shown according to an exemplary embodiment, as follows: Figure 4 As shown, it includes:
[0105] The acquisition module 401 is used to acquire a target ultrasound image from a video stream, and to acquire the target section category and the target prediction score of the target section category of the target ultrasound image.
[0106] Calculation module 402 is used to calculate the standard score corresponding to the target section category based on the target prediction score of the target section category corresponding to the target ultrasound image;
[0107] The determination module 403 is used to determine that the target ultrasound image is a standard section image of the target section category when the standard score is greater than or equal to the standard score threshold.
[0108] Specifically, the calculation module 402 is used to: if the section category of the preceding frame ultrasound image of the target ultrasound image is not the target section category, initialize the standard score corresponding to the target section category; if the section category of the preceding frame ultrasound image of the target ultrasound image is the target section category, obtain the standard score corresponding to the target section category calculated based on the target prediction score of the target section category corresponding to the preceding frame ultrasound image of the target ultrasound image. Then, when the target prediction score of the target section category corresponding to the target ultrasound image is greater than or equal to a score threshold, accumulate the standard score.
[0109] The standard section recognition device provided in this application maintains a corresponding standard score for each section category. For each frame of ultrasound image acquired by the ultrasound device, if the target prediction score of its corresponding section category is greater than or equal to a score threshold, the standard score corresponding to that section category is accumulated. For the currently acquired target ultrasound image, its corresponding section category is the target section category. The more consecutive ultrasound images with the target section category in the previous ultrasound images, the larger the currently accumulated standard score. When the standard score is greater than or equal to the standard score threshold, the target ultrasound image is determined to be a standard section image of the target section category. Therefore, the standard section recognition device provided in this application utilizes the continuity information in the ultrasound video to calculate the accumulated standard score of the current target ultrasound image, avoiding the detection of sporadic standard section images, reducing the false alarm rate of non-standard sections and the false negative rate of standard sections, and improving the accuracy of standard section recognition.
[0110] Based on the above embodiments, as a preferred implementation, the target prediction score includes any one or a combination of any of the target standard section score, the target basic standard section score, and the target joint score, wherein the joint score is used to comprehensively describe the standard degree and basic standard degree of the section;
[0111] Accordingly, the device also includes:
[0112] The first determining module is used to determine the score threshold corresponding to the predicted score of each target in the target section category corresponding to the target ultrasound image;
[0113] The calculation module 402 is specifically used to: when the target prediction score of any target section category corresponding to the target ultrasound image is greater than or equal to the corresponding score threshold, accumulate the corresponding score of the standard score.
[0114] Based on the above embodiments, as a preferred implementation, the acquisition module 401 is specifically used to acquire a target ultrasound image from a video stream, input the target ultrasound image into a classification model, acquire the target section category of the target ultrasound image from the output of the classification model, acquire the target standard section score corresponding to the target section category from the output of the classification model, and / or acquire the target basic standard section score corresponding to the target section category from the output of the classification model, and / or calculate the target joint score based on the joint score parameter of the target standard section score, the target basic standard section score, and the target section category corresponding to the target ultrasound image.
[0115] Based on the above embodiments, as a preferred embodiment, it further includes:
[0116] The second determining module is used to determine the lower bound and the upper bound of the joint fraction parameter corresponding to the target section category;
[0117] The first calculation module is used to calculate the joint score parameter of the target section category corresponding to the target ultrasound image by interpolating between the lower bound of the joint score parameter and the upper bound of the joint score parameter based on the number of target frames of ultrasound images preceding the target ultrasound image whose section category is the target candidate section category.
[0118] Based on the above embodiments, as a preferred implementation, the second determining module is specifically used for: obtaining target section samples whose sum of standard section score and basic standard section score corresponding to the candidate section category is greater than or equal to a preset value; determining a first target linear function satisfying a first condition and a second target linear function satisfying a second condition in a coordinate system containing the coordinate points corresponding to the target section samples; wherein, the coordinate points corresponding to the target section samples in the coordinate system are determined based on the standard section score and basic standard section score of the target section samples, the first condition includes that the sum of the distances between the coordinate points corresponding to all the target section samples and the first target linear function is the smallest and that the basic standard section scores corresponding to all the non-standard section samples are located at the function value of the first target linear function, the second condition includes that the sum of the distances between the coordinate points corresponding to all the target section samples and the second target linear function is the smallest; the parameters of the first target linear function are determined as the upper bound of the joint fraction parameter, and the parameters of the second target linear function are determined as the lower bound of the joint fraction parameter.
[0119] Based on the above embodiments, as a preferred implementation, the first determining module is specifically used to: determine the lower threshold and upper threshold of the target standard section score or the target basic standard section score corresponding to the target section category; calculate the score threshold corresponding to the target standard section score or the target basic standard section score by interpolating between the lower threshold and the upper threshold based on the number of target frames of ultrasound images with the section category of the target section image preceding the target ultrasound image; wherein, the score threshold is negatively correlated with the number of target frames.
[0120] Based on the above embodiments, as a preferred implementation, the first determining module is specifically used for: obtaining standard and non-standard segment samples corresponding to the target segment category, and determining the standard segment scores corresponding to the standard and non-standard segment samples; using the standard segment score as a target standard segment score threshold, calculating the weighted value of precision and recall according to a first weight ratio, and determining the target standard segment score threshold corresponding to the maximum weighted value as the upper threshold of the target standard segment score; using the standard segment score as a target standard segment score threshold, calculating the weighted value of precision and recall according to a second weight ratio, and determining the target standard segment score threshold corresponding to the maximum weighted value as the lower threshold of the target standard segment score; wherein, the first weight ratio and the second weight ratio are the ratio of the weight of precision to the weight of recall, and the first weight ratio is greater than or equal to the second weight ratio.
[0121] Based on the above embodiments, as a preferred implementation, the first determining module is specifically used for: obtaining basic standard section samples and non-standard section samples corresponding to the target section category, and determining the basic standard section scores corresponding to the basic standard section samples and non-standard section samples; using the basic standard section scores as the target basic standard section score threshold, calculating the weighted value of precision and recall according to the third weight ratio, and determining the target basic standard section score threshold corresponding to the maximum weighted value as the upper threshold of the basic target standard section score; using the basic standard section scores as the target basic standard section score threshold, calculating the weighted value of precision and recall according to the fourth weight ratio, and determining the target basic standard section score threshold corresponding to the maximum weighted value as the lower threshold of the basic target standard section score; wherein, the third weight ratio and the fourth weight ratio are the ratio of the weight of precision to the weight of recall, and the third weight ratio is greater than or equal to the fourth weight ratio.
[0122] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.
[0123] Based on the hardware implementation of the above program modules, and in order to implement the method of the embodiments of this application, the embodiments of this application also provide an ultrasonic device. Figure 5 This is a structural diagram illustrating an ultrasonic device according to an exemplary embodiment, such as... Figure 5 As shown, the ultrasound equipment includes:
[0124] Communication interface 1 enables information exchange with other devices, such as network devices;
[0125] Processor 2 is connected to communication interface 1 to enable information interaction with other devices and, when running a computer program, executes the standard section recognition method provided by one or more of the above-mentioned technical solutions. The computer program is stored in memory 3.
[0126] Of course, in practical applications, the various components of the ultrasonic equipment are coupled together through bus system 4. It can be understood that bus system 4 is used to achieve communication and connection between these components. In addition to the data bus, bus system 4 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in... Figure 5 The general will label all buses as Bus System 4.
[0127] The memory 3 in this embodiment is used to store various types of data to support the operation of the ultrasound device. Examples of such data include any computer program used to operate the ultrasound device.
[0128] It is understood that memory 3 can be volatile memory or non-volatile memory, or both. Non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), ferromagnetic random access memory (FRAM), flash memory, magnetic surface memory, optical disc, or compact disc read-only memory (CD-ROM); magnetic surface memory can be disk storage or magnetic tape storage. Volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Synchronous Static Random Access Memory (SSRAM), Dynamic Random Access Memory (DRAM), Synchronous Dynamic Random Access Memory (SDRAM), Double Data Rate Synchronous Dynamic Random Access Memory (DDRSDRAM), Enhanced Synchronous Dynamic Random Access Memory (ESDRAM), SyncLink Dynamic Random Access Memory (SLDRAM), and Direct Rambus Random Access Memory (DRRAM).The memory 3 described in the embodiments of this application is intended to include, but is not limited to, these and any other suitable types of memory.
[0129] The methods disclosed in the embodiments of this application can be applied to processor 2, or implemented by processor 2. Processor 2 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in processor 2 or by instructions in the form of software. The processor 2 may be a general-purpose processor, DSP, or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 2 can implement or execute the methods, steps and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the methods disclosed in the embodiments of this application can be directly manifested as being executed by a hardware decoding processor, or being executed by a combination of hardware and software modules in the decoding processor. The software modules may be located in a storage medium, which is located in memory 3. Processor 2 reads the program in memory 3 and completes the steps of the aforementioned method in combination with its hardware.
[0130] When processor 2 executes the program, it implements the corresponding processes in the various methods of the embodiments of this application. For the sake of brevity, these will not be described in detail here.
[0131] In an exemplary embodiment, this application also provides a storage medium, namely a computer storage medium, specifically a computer-readable storage medium, such as a memory 3 that stores a computer program, which can be executed by a processor 2 to complete the steps described in the aforementioned method. The computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, Flash Memory, magnetic surface memory, optical disc, or CD-ROM.
[0132] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media that can store program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.
[0133] Alternatively, if the integrated units described above are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an ultrasound device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.
[0134] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A standard section recognition method, characterized in that, include: Obtain a target ultrasound image from the video stream, and obtain the target section category and the target prediction score of the target section category from the target ultrasound image; Calculate the standard score corresponding to the target section category based on the target prediction score of the target section category corresponding to the target ultrasound image; If the standard score is greater than or equal to the standard score threshold, then the target ultrasound image is determined to be a standard section image of the target section category; The calculation of the standard score corresponding to the target section category based on the target prediction score of the target section category corresponding to the target ultrasound image includes: if the section category of the previous frame ultrasound image of the target ultrasound image is not the target section category, then the standard score corresponding to the target section category is initialized; if the section category of the previous frame ultrasound image of the target ultrasound image is the target section category, then the standard score corresponding to the target section category calculated based on the target prediction score of the target section category corresponding to the previous frame ultrasound image of the target ultrasound image is obtained; when the target prediction score of the target section category corresponding to the target ultrasound image is greater than or equal to a score threshold, the standard score is accumulated; when the target prediction score of the target section category corresponding to the target ultrasound image is less than the score threshold, the standard score is initialized.
2. The standard section recognition method according to claim 1, characterized in that, The target prediction score includes any one or a combination of several of the target standard section score, the target basic standard section score, and the target joint score. The target joint score is used to comprehensively describe the standard degree and basic standard degree of the section. Accordingly, when the target prediction score of the target section category corresponding to the target ultrasound image is greater than or equal to the score threshold, the standard score is accumulated, including: Determine the score threshold corresponding to the predicted score of each target in the target section category corresponding to the target ultrasound image; If the target prediction score of any target section category corresponding to the target ultrasound image is greater than or equal to the corresponding score threshold, then the standard score is accumulated by the corresponding score.
3. The standard section recognition method according to claim 2, characterized in that, The acquisition of the target section category and the target prediction score of the target section category from the target ultrasound image includes: The target ultrasound image is input into a classification model, and the target section category of the target ultrasound image is obtained from the output of the classification model. Obtain the target standard section score corresponding to the target section category from the output of the classification model, and / or obtain the target basic standard section score corresponding to the target section category from the output of the classification model, and / or calculate the target joint score based on the target standard section score, the target basic standard section score and the joint score parameter of the target section category corresponding to the target ultrasound image.
4. The standard section recognition method according to claim 3, characterized in that, Also includes: Determine the lower bound and upper bound of the joint fraction parameter corresponding to the target section category; Based on the number of target frames in the ultrasound images preceding the target ultrasound image that have the target candidate section category, interpolation is performed between the lower bound of the joint score parameter and the upper bound of the joint score parameter to calculate the joint score parameter of the target section category corresponding to the target ultrasound image.
5. The standard section recognition method according to claim 4, characterized in that, Determining the lower bound and upper bound of the joint score parameter corresponding to the target section category includes: Obtain target slice samples whose sum of the standard slice score and the basic standard slice score corresponding to the candidate slice category is greater than or equal to a preset value; In a coordinate system containing the coordinate points corresponding to the target section sample, a first target linear function satisfying a first condition and a second target linear function satisfying a second condition are determined; wherein, the coordinate points corresponding to the target section sample in the coordinate system are determined based on the standard section fraction and the basic standard section fraction of the target section sample; the first condition includes that the sum of the distances between all the coordinate points corresponding to the target section sample and the first target linear function is minimized and that the basic standard section fractions corresponding to all non-standard section samples are located at the function value of the first target linear function; the second condition includes that the sum of the distances between all the coordinate points corresponding to the target section sample and the second target linear function is minimized. The parameters of the first objective linear function are determined as the upper bound of the joint fractional parameters, and the parameters of the second objective linear function are determined as the lower bound of the joint fractional parameters.
6. The standard section recognition method according to claim 2, characterized in that, Determining the score threshold corresponding to the predicted score of each object in the target section category corresponding to the target ultrasound image includes: Determine the lower and upper threshold bounds of the target standard section score or the target basic standard section score corresponding to the target section category; Based on the number of target frames in the ultrasound images preceding the target ultrasound image that have the same section category as the target section, interpolation is performed between the lower threshold and the upper threshold to calculate the score threshold corresponding to the target standard section score or the target basic standard section score; wherein, the score threshold is negatively correlated with the number of target frames.
7. The standard section recognition method according to claim 6, characterized in that, Determining the lower and upper threshold bounds of the target standard section score corresponding to the target section category includes: Obtain standard and non-standard section samples corresponding to the target section category, and determine the standard section score corresponding to the standard and non-standard section samples; The standard segment score is used as the target standard segment score threshold. The weighted value of precision and recall is calculated according to the first weight ratio. The target standard segment score threshold corresponding to the maximum weighted value is determined as the upper bound of the target standard segment score threshold. The standard segment score is used as the target standard segment score threshold. The weighted value of precision and recall is calculated according to the second weight ratio. The target standard segment score threshold corresponding to the maximum weighted value is determined as the lower bound of the target standard segment score threshold. Wherein, the first weight ratio and the second weight ratio are the ratios of the weight of precision to the weight of recall, and the first weight ratio is greater than or equal to the second weight ratio.
8. The standard section recognition method according to claim 6, characterized in that, Determining the lower and upper threshold bounds of the target basic standard section score corresponding to the target section category includes: Obtain basic standard section samples and non-standard section samples corresponding to the target section category, and determine the basic standard section scores corresponding to the basic standard section samples and non-standard section samples; The basic standard segment score is used as the target basic standard segment score threshold. The weighted value of precision and recall is calculated according to the third weight ratio. The target basic standard segment score threshold corresponding to the maximum weighted value is determined as the upper bound of the target basic standard segment score threshold. The basic standard segment score is used as the target basic standard segment score threshold. The weighted value of precision and recall is calculated according to the fourth weight ratio. The target basic standard segment score threshold corresponding to the maximum weighted value is determined as the lower bound of the target basic standard segment score threshold. Wherein, the third weight ratio and the fourth weight ratio are the ratios of the weight of precision to the weight of recall, and the third weight ratio is greater than or equal to the fourth weight ratio.
9. A standard cross-section recognition device, characterized in that, include: The acquisition module is used to acquire a target ultrasound image from a video stream, and to acquire the target section category and the target prediction score of the target section category of the target ultrasound image; The calculation module is used to calculate the standard score corresponding to the target section category based on the target prediction score of the target section category corresponding to the target ultrasound image; The determination module is used to determine that the target ultrasound image is a standard section image of the target section category when the standard score is greater than or equal to the standard score threshold. Specifically, the calculation module is used to: if the section category of the previous frame ultrasound image of the target ultrasound image is not the target section category, then initialize the standard score corresponding to the target section category; if the section category of the previous frame ultrasound image of the target ultrasound image is the target section category, then obtain the standard score corresponding to the target section category calculated based on the target prediction score of the target section category corresponding to the previous frame ultrasound image of the target ultrasound image; when the target prediction score of the target section category corresponding to the target ultrasound image is greater than or equal to a score threshold, accumulate the standard score; when the target prediction score of the target section category corresponding to the target ultrasound image is less than the score threshold, initialize the standard score.
10. An ultrasonic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the steps of the standard section recognition method as described in any one of claims 1 to 8.
11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the standard section recognition method as described in any one of claims 1 to 8.
Citation Information
Patent Citations
Method and device for determining optimal standard section of fetal
CN112155602A
