Ultrasonic cardiology image section recognition method and ultrasonic cardiology video section quality control method

Abstract

The present invention provides an ultrasonic cardiogram section recognition method, an ultrasonic cardiogram video section quality control method, an electronic device and a readable storage medium, the method comprising: using a trained target detection model to detect the section inherent structure of the acquired ultrasonic cardiogram to be recognized, so as to obtain the corresponding first section inherent structure information; according to the first section inherent structure information corresponding to the ultrasonic cardiogram to be recognized and the multivariate Gaussian distribution model corresponding to each of the various preset section types acquired in advance, respectively calculating the final probability value of the ultrasonic cardiogram to be recognized belonging to each preset section type and other section types; taking the section type corresponding to the maximum final probability value as the recognition result of the section type of the ultrasonic cardiogram to be recognized. The present invention can effectively improve the accuracy of ultrasonic cardiogram section type recognition by making full use of the inherent structure information in the standard section of the ultrasonic cardiogram.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to an ultrasonic cardiology image section recognition method, an ultrasonic cardiology video section quality control method, an electronic device and a readable storage medium. Background Art

[0002] Medical imaging technology can realize a lot of cardiac examination methods, including SPECT, PET, CT, MRI, ultrasound and other modalities. Doctors can diagnose suspicious lesions found in the images. Although there are very mature motion artifact correction algorithms based on cardiac gating, which make the cardiac imaging of new SPECT, PET, CT and MRI clearer than traditional scanning; but because the heart is in motion, its nature cannot be fully represented by a single static image. If dynamic imaging is to be achieved based on this type of modality, the person being examined needs to bear more ionizing radiation dose or longer scanning time, which will burden him. Echocardiography is a commonly used cardiac ultrasound examination. The image represents the degree of reflection of ultrasound by the heart tissue. Through continuous ultrasound, the blood and myocardial tissue movement speed at any position are checked in real time. It has the characteristics of fast, real-time, no ionizing radiation, and low price. However, because the examination of echocardiography depends largely on the scanning technique of the operator, it requires considerable clinical experience to obtain an echocardiogram with a standard section.

[0003] In recent years, more and more deep learning applications have emerged in the field of medical imaging, which can realize applications including detection, classification, registration, segmentation, and noise reduction. Through convolutional neural networks (CNN), it is possible to adaptively learn how to extract high-order features from images that are suitable for the given task. Currently, more and more CNN-based image classification models are used for tasks such as tumor benign and malignant differentiation and pathological diagnosis, and their diagnostic efficiency is superior to traditional machine learning models.

[0004] The existing echocardiographic standard section recognition model is basically constructed using the following method: using the section category of the image as an annotation, a CNN model for image classification is selected for training.

[0005] However, due to the low signal-to-noise ratio and inter-operator differences of ultrasound images, the end-to-end classification model is affected by these factors in the data during the training process, and learns noise that is irrelevant to the annotation, thereby damaging the final recognition efficiency.

[0006] It should be noted that the information disclosed in the background technology section of the invention is only intended to deepen the understanding of the general background technology of the invention, and should not be regarded as an admission or suggestion in any form that the information constitutes prior art already known to those skilled in the art. Summary of the invention

[0007] The purpose of the present invention is to provide an ultrasonic cardiac image section recognition method, an ultrasonic cardiac video section quality control method, an electronic device and a readable storage medium, which can effectively improve the accuracy of ultrasonic cardiac image section type recognition by making full use of the inherent structural information in the standard section of the ultrasonic cardiac image.

[0008] To achieve the above object, the present invention provides an ultrasonic cardiogram section recognition method, comprising:

[0009] Using the trained target detection model to detect the inherent structure of the section of the acquired ultrasound cardiogram to be identified, so as to obtain the inherent structure information of the first section corresponding to the ultrasound cardiogram to be identified;

[0010] Calculating respectively the initial probability value of the to-be-identified ultrasound cardiac image belonging to each of the preset section types according to the inherent structural information of the first section and the pre-acquired multivariate Gaussian distribution models corresponding to the various preset section types;

[0011] Normalizing the initial probability values ​​of the to-be-identified ultrasound cardiac image belonging to each of the preset section types respectively, so as to obtain the normalized probability values ​​of the to-be-identified ultrasound cardiac image belonging to each of the preset section types;

[0012] According to the normalized probability values ​​of the ultrasound image to be identified belonging to each of the preset section types, respectively calculating the final probability value of the ultrasound image to be identified belonging to each of the preset section types and the final probability values ​​of other section types, wherein the other section types are section types other than all the preset section types;

[0013] The maximum final probability value is determined from all the final probability values, and the section type corresponding to the maximum final probability value is used as the recognition result of the section type of the ultrasonic cardiac image to be recognized.

[0014] Optionally, the target detection model is trained by the following steps:

[0015] Acquire a first training data set, wherein the first training data set includes a plurality of first training samples, wherein the first training samples include ultrasound cardiology training images of preset section types and corresponding section intrinsic structure labels, wherein the section intrinsic structure labels include intrinsic structure category labels and intrinsic structure position labels;

[0016] The first training data set is used to train a pre-built target detection model to obtain a trained target calibration model.

[0017] Optionally, obtain the multivariate Gaussian distribution model corresponding to each preset slice type through the following steps:

[0018] For each of the preset slice types:

[0019] Selecting a first training sample belonging to the preset section type from the first training data set as a target sample;

[0020] For each of the target samples belonging to the preset section type, the trained target detection model is used to detect the section inherent structure of the ultrasound cardiology training image of the target sample to obtain a first detection result of the corresponding section inherent structure, and for each inherent structure corresponding to the preset section type, the first detection frame with the highest confidence among all the first detection frames corresponding to the inherent structure in the first detection result is used as the first target detection frame of the inherent structure, and the first section inherent structure information of the preset section type corresponding to the target sample is obtained according to the position information of each of the first target detection frames corresponding to the target sample;

[0021] The first section inherent structural information of the preset section type corresponding to each of the target samples belonging to the preset section type is combined to obtain a first section inherent structural information set of the preset section type, and a multivariate Gaussian distribution model corresponding to the preset section type is constructed based on the first section inherent structural information set of the preset section type.

[0022] Optionally, the position information of the first object detection frame includes a horizontal coordinate and a vertical coordinate of a center point of the first object detection frame, and a width and a length of the first object detection frame;

[0023] The acquiring, according to the position information of each of the first target detection frames corresponding to the target sample, the first section intrinsic structure information of the preset section type corresponding to the target sample comprises:

[0024] Acquire a first position information matrix according to the position information of each of the first target detection frames;

[0025] The first position information matrix is ​​reconstructed into a corresponding first position information vector to obtain the first section intrinsic structure information of the preset section type corresponding to the target sample.

[0026] Optionally, constructing a multivariate Gaussian distribution model corresponding to the preset section type according to the first section intrinsic structure information set of the preset section type includes:

[0027] Calculate the slice inherent structure mean and the slice inherent structure covariance matrix corresponding to the preset slice type according to the first slice inherent structure information set of the preset slice type;

[0028] According to the mean and covariance matrix corresponding to the preset section type, a multivariate Gaussian distribution model corresponding to the preset section type is constructed.

[0029] Optionally, constructing a multivariate Gaussian distribution model corresponding to the preset section type according to the mean and covariance matrix corresponding to the preset section type includes:

[0030] The multivariate Gaussian distribution model is constructed according to the following formula:

[0031]

[0032] Among them, f i (x) is the probability density function of the multivariate Gaussian distribution corresponding to the i-th preset section type, Σ i is the inherent structural covariance matrix of the slice corresponding to the i-th preset slice type, k i is the inherent structural covariance matrix Σ of the slice corresponding to the i-th preset slice type i rank, μ i is the mean of the inherent structure of the slice corresponding to the i-th preset slice type, n i is the number of inherent structures corresponding to the i-th preset section type, and x is an independent variable related to the first section inherent structure information of the i-th preset section type.

[0033] Optionally, the detecting of the inherent structure of the section of the acquired ultrasound image to be identified by using the trained target detection model to obtain the inherent structure information of the first section corresponding to the ultrasound image to be identified includes:

[0034] Using the trained target detection model to detect the inherent structure of the section of the acquired ultrasound cardiac image to be identified, so as to obtain a corresponding second detection result;

[0035] For each inherent structure of each preset section type, the second detection frame with the highest confidence among all the second detection frames corresponding to the inherent structure of the preset section type in the second detection result is used as the second target detection frame of the inherent structure of the preset section type, and the first section inherent structure information of the preset section type corresponding to the ultrasonic cardiac image to be identified is obtained according to the position information of each second target detection frame corresponding to the preset section type.

[0036] Optionally, normalizing the initial probability values ​​of the to-be-identified ultrasound image belonging to each preset section type respectively includes:

[0037] The following formula is used to normalize the initial probability value of the ultrasound image to be identified belonging to each preset section type:

[0038]

[0039] Among them, g i is the initial probability value of the ultrasound image to be identified belonging to the i-th preset section type, g i ' is g i The corresponding normalized probability value, i is a positive integer.

[0040] Optionally, the calculating, according to the normalized probability value of the to-be-identified ultrasound cardiac image belonging to each of the preset section types, respectively the final probability value of the to-be-identified ultrasound cardiac image belonging to each of the preset section types and the final probability values ​​of other section types, comprises:

[0041] The following formula is used to calculate the final probability value of the to-be-identified ultrasound image belonging to each of the preset section types:

[0042]

[0043] The final probability value of the ultrasound image to be identified belonging to other section types is calculated using the following formula:

[0044]

[0045] Among them, g″ i is the final probability value of the ultrasound image to be identified belonging to the i-th preset section type, N is the total number of preset section types, g′ i is the normalized probability value that the ultrasound image to be identified belongs to the i-th preset section type, g' h is the normalized probability value that the ultrasound image to be identified belongs to the hth preset section type, g' j is the normalized probability value that the ultrasound image to be identified belongs to the jth preset section type, g' m is the initial probability value of the ultrasound image to be identified belonging to the mth preset section type, g' k is the normalized probability value that the ultrasound image to be identified belongs to the kth preset section type, g″ other is the final probability value that the ultrasound image to be identified belongs to other section types.

[0046] Optionally, the ultrasonic cardiographic section recognition method further includes:

[0047] If the identification result of the section type of the to-be-identified ultrasound cardiogram is other section types, determining that the section corresponding to the to-be-identified ultrasound cardiogram is a non-standard section;

[0048] If the recognition result of the section type of the ultrasonic cardiac image to be recognized is one of the preset section types, executing the following steps;

[0049] Acquiring inherent structure information of a second section corresponding to the ultrasonic cardiogram to be identified according to a detection result of the inherent structure of the section of the ultrasonic cardiogram to be identified;

[0050] Inputting the inherent structural information of the second section corresponding to the ultrasound cardiogram to be identified into a pre-trained section quality control model to obtain a probability value of a section quality standard and a probability value of a section quality non-standard for each preset section type; and

[0051] The section quality control score of the ultrasonic cardiac image to be identified is calculated according to the probability value of the section quality standard of the preset section type corresponding to the identification result of the section type of the ultrasonic cardiac image to be identified and the probability value of the section quality non-standard.

[0052] Optionally, the ultrasonic cardiographic section recognition method further includes:

[0053] If the section quality control score of the to-be-identified ultrasonic cardiology image is greater than a first preset threshold, the section corresponding to the to-be-identified ultrasonic cardiology image is determined to be a standard section.

[0054] Optionally, acquiring the inherent structure information of a second section corresponding to the ultrasonic cardiogram to be identified according to the detection result of the inherent structure of the section of the ultrasonic cardiogram to be identified includes:

[0055] For each inherent structure, the second detection frame with the highest confidence among all the second detection frames corresponding to the inherent structure in the detection results is used as the second target detection frame of the inherent structure, and the second section inherent structure information corresponding to the ultrasonic cardiac image to be identified is obtained according to the position information of each second target detection frame and the confidence of the second target detection frame.

[0056] To achieve the above object, the present invention further provides an ultrasonic cardiology video slice quality control method, the ultrasonic cardiology video slice quality control method comprising:

[0057] For each frame of the ultrasonic cardiogram in the ultrasonic cardiogram video, the ultrasonic cardiogram section recognition method described above is used to recognize the frame of the ultrasonic cardiogram, so as to obtain a recognition result of the section type of the frame of the ultrasonic cardiogram;

[0058] For each frame of the ultrasonic cardiogram in the ultrasonic cardiogram video, if the recognition result of the section type of the frame of the ultrasonic cardiogram is other section types, it is determined that the section corresponding to the frame of the ultrasonic cardiogram is a non-standard section; if the recognition result of the section type of the frame of the ultrasonic cardiogram is one of the preset section types, the following steps are performed;

[0059] Inputting the inherent structural information of the section corresponding to the frame of ultrasound cardiology image into a pre-trained section quality control model to obtain a probability value of a section quality standard and a probability value of a section quality non-standard for each preset section type;

[0060] Calculating a slice quality control score for the frame of ultrasound cardiology image according to a probability value of a slice quality standard for a preset slice type and a probability value of a slice quality non-standard for the slice type corresponding to the recognition result of the slice type for the frame of ultrasound cardiology image; and

[0061] If the section quality control score of the frame of ultrasound cardiology image is greater than a preset threshold, the section corresponding to the frame of ultrasound cardiology image is determined to be a standard section;

[0062] The number of frames of the ultrasonic cardiogram whose sections are standard sections in the ultrasonic cardiogram video is counted, and the section quality control score of the ultrasonic cardiogram video is calculated according to the ratio between the number of frames of the ultrasonic cardiogram whose sections are standard sections and the total number of frames of the ultrasonic cardiogram in the ultrasonic cardiogram video.

[0063] Optionally, the ultrasound cardiology video section quality control method further includes:

[0064] It is determined whether the section quality control score of the ultrasound cardiology video is greater than a second preset threshold value. If so, it is determined that the section quality of the ultrasound cardiology video is qualified.

[0065] To achieve the above object, the present invention further provides an electronic device, comprising a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, the ultrasonic cardiac image section recognition method described above is implemented.

[0066] To achieve the above object, the present invention further provides a readable storage medium, wherein the readable storage medium stores a computer program, and when the computer program is executed by a processor, the ultrasonic cardiac image section recognition method described above is implemented.

[0067] Compared with the prior art, the ultrasonic cardiology image section recognition method, ultrasonic cardiology video section quality control method, electronic device and readable storage medium provided by the present invention have the following advantages:

[0068] The ultrasonic cardiac image section recognition method provided by the present invention first uses a trained target detection model to detect the inherent structure of the section of the acquired ultrasonic cardiac image to be recognized, so as to obtain the first section inherent structure information corresponding to the ultrasonic cardiac image to be recognized; then, according to the first section inherent structure information and the multivariate Gaussian distribution models corresponding to the various preset section types acquired in advance, respectively calculate the initial probability value of the ultrasonic cardiac image to be recognized belonging to each of the preset section types; then, respectively normalize the initial probability value of the ultrasonic cardiac image to be recognized belonging to each of the preset section types, so as to obtain the normalized probability value of the ultrasonic cardiac image to be recognized belonging to each of the preset section types; then, according to the normalized probability value of the ultrasonic cardiac image to be recognized belonging to each of the preset section types, respectively calculate the probability value of the ultrasonic cardiac image to be recognized belonging to each of the preset section types. type and the final probability values ​​of other section types, wherein the other section types are section types other than all the preset section types; finally, the maximum final probability value is determined from all the final probability values, and the section type corresponding to the maximum final probability value is used as the recognition result of the section type of the ultrasonic cardiac image to be identified. Therefore, the ultrasonic cardiac image section recognition method provided by the present invention makes full use of the knowledge in the field of ultrasonic clinical examination to give full application to the inherent structural information in the standard section of the ultrasonic cardiac image, so that the ultrasonic cardiac image section recognition method provided by the present invention is more inclined to the thinking mode of professional doctors, thereby effectively improving the accuracy of ultrasonic cardiac image section type recognition, and further can better assist doctors in section type recognition, so as to reduce the difficulty in the process of ultrasonic cardiac image section type recognition, and further can better assist the standardized scanning of ultrasonic cardiograms. In addition, by normalizing the initial probability values ​​of the ultrasound image to be identified belonging to each preset section type, the initial probability values ​​of the ultrasound image to be identified belonging to each preset section type can be normalized to between 0 and 1, thereby making it easier to subsequently calculate the final probability values ​​of the ultrasound image to be identified belonging to each of the preset section types and other section types according to the results after normalization, and finally determine the section type of the ultrasound image to be identified according to the final probability values ​​of the ultrasound image to be identified belonging to each of the preset section types and other section types.

[0069] Since the ultrasonic video slice quality control method, electronic device and readable storage medium provided by the present invention belong to the same inventive concept as the ultrasonic image slice recognition method provided by the present invention, the ultrasonic video slice quality control method, electronic device and readable storage medium provided by the present invention have all the advantages of the ultrasonic image slice recognition method provided by the present invention. In addition, the ultrasonic video slice quality control method provided by the present invention judges whether the slice quality is standard for each frame of ultrasonic image in the video according to the inherent structural information of the second slice corresponding to the frame of ultrasonic image. The judgment integrates the judgment habits of doctors in the ultrasound department on whether the slice quality of the image is standard, thereby providing a basis for realizing the control of the quality of the ultrasonic video slice. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 A flowchart of a method for identifying sections of an ultrasonic cardiac image provided by one embodiment of the present invention;

[0071] Figure 2a This is an echocardiographic training image of the PLAX-LV section with intrinsic structures marked;

[0072] Figure 2b This is an echocardiographic training image of the PSAX-AV section with the intrinsic structures marked;

[0073] Figure 2c PSAX-Mitral valve orifice short-axis echocardiography training image with inherent structures marked;

[0074] Figure 2d This is an echocardiography training image of the PSAX-papillary muscle horizontal section with intrinsic structures marked;

[0075] Figure 2e PSAX-apical echocardiography training image with intrinsic structures marked;

[0076] Figure 2f This is an echocardiography training image of the A4C section with the intrinsic structures marked;

[0077] Figure 2g An echocardiographic training image of the A2C section with intrinsic structures marked;

[0078] Figure 2h This is an echocardiography training image of the A3C-apical long axis section with the intrinsic structures marked;

[0079] Figure 3 A flow chart of a method for quality control of an ultrasound cardiology video slice provided in one embodiment of the present invention;

[0080] Figure 4A schematic diagram of the block structure of an electronic device provided in one embodiment of the present invention. DETAILED DESCRIPTION

[0081] The ultrasonic image section recognition method, ultrasonic video section quality control method, electronic device and storage medium proposed in the present invention are further described in detail below in combination with the accompanying drawings and specific embodiments. According to the following description, the advantages and features of the present invention will be clearer. It should be noted that the drawings are in a very simplified form and use non-precise proportions, which are only used to conveniently and clearly assist in explaining the purpose provided by the present invention. In order to make the purposes, features and advantages of the present invention more obvious and easy to understand, please refer to the accompanying drawings. It should be noted that the structures, proportions, sizes, etc. illustrated in the drawings of this specification are only used to match the contents disclosed in the specification for people familiar with this technology to understand and read, and are not used to limit the limiting conditions for the implementation of the present invention. Any structural modification, change in proportional relationship or adjustment of size, when the effects and purposes that can be achieved by the present invention are the same or similar, should still fall within the scope of the technical content disclosed by the present invention.

[0082] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0083] In addition, in the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0084] The core idea of ​​the present invention is to provide an ultrasonic cardiogram section recognition method, an ultrasonic cardiogram video section quality control method, an electronic device and a storage medium, which can effectively improve the accuracy of ultrasonic cardiogram section type recognition by making full use of the inherent structural information in the standard section of the ultrasonic cardiogram. It should be noted that, as can be understood by those skilled in the art, the section recognition referred to in the present invention refers to identifying the type of cardiac section corresponding to the ultrasonic cardiogram acquisition, and the section inherent structure referred to in the present invention refers to the anatomical structure contained in the human organ under each section, for example, the heart under the PLAX-LV (parasternal long axis-left ventricle) section includes five anatomical structures of the right ventricle, mitral valve, aorta, left ventricle and left atrium, and the heart under the A2C (apical two-chamber heart) section includes three anatomical structures of the left ventricle, left atrium and mitral valve; the section inherent structure information referred to in the present invention refers to some parameter information that can characterize the section inherent structure, such as coordinate parameters and size parameters of the section inherent structure. It should also be noted that, as can be understood by those skilled in the art, the ultrasonic cardiology image section recognition method and the ultrasonic cardiology video section quality control method provided by the present invention can be applied to the electronic device provided by the present invention, wherein the electronic device can be a personal computer, a mobile terminal, etc., and the mobile terminal can be a mobile phone, a tablet computer, and other hardware devices with various operating systems.

[0085] To realize the above idea, the present invention provides a method for identifying sections of an ultrasonic cardiac image. Figure 1 , which schematically shows a flow chart of an ultrasonic cardiac image section recognition method provided by an embodiment of the present invention. Figure 1 As shown, the ultrasonic cardiac image section recognition method provided by the present invention comprises the following steps:

[0086] Step S110: Use the trained target detection model to detect the inherent structure of the section of the acquired ultrasound cardiogram to be identified, so as to obtain the inherent structure information of the first section corresponding to the ultrasound cardiogram to be identified.

[0087] Step S120, calculating the initial probability value of the ultrasonic cardiac image to be identified belonging to each of the preset section types according to the inherent structural information of the first section and the multivariate Gaussian distribution models corresponding to the various preset section types acquired in advance.

[0088] Step S130 , respectively normalizing the initial probability values ​​of the ultrasonic cardiac image to be identified belonging to each of the preset section types to obtain normalized probability values ​​of the ultrasonic cardiac image to be identified belonging to each of the preset section types.

[0089] Step S140, according to the normalized probability value of the ultrasound cardiac image to be identified belonging to each of the preset section types, respectively calculate the final probability value of the ultrasound cardiac image to be identified belonging to each of the preset section types and the final probability values ​​of other section types, wherein the other section types are section types other than all the preset section types.

[0090] Step S150: determine the maximum final probability value from all the final probability values, and use the section type corresponding to the maximum final probability value as the recognition result of the section type of the ultrasound cardiac image to be recognized.

[0091] Therefore, the ultrasonic section recognition method provided by the present invention makes full use of the knowledge in the field of ultrasonic clinical examination to give full application to the inherent structural information in the standard section of the ultrasonic cardiac image, so that the ultrasonic section recognition method provided by the present invention is more inclined to the way of thinking of professional doctors, thereby effectively improving the accuracy of ultrasonic cardiac image section type recognition, and thus can better assist doctors in section type recognition, so as to reduce the difficulty in the process of ultrasonic cardiac image section type recognition, and thus can better assist the standardized scanning of ultrasonic cardiograms. By normalizing the initial probability value of the ultrasonic image to be identified belonging to each preset section type, the initial probability value of the ultrasonic image to be identified belonging to each preset section type can be normalized to between 0 and 1, so that it is more convenient to calculate the final probability value of the ultrasonic image to be identified belonging to each preset section type and other section types according to the results after normalization, and finally determine the section type of the ultrasonic image to be identified according to the final probability value of the ultrasonic image to be identified belonging to each preset section type and other section types. It should be noted that, as can be understood by those skilled in the art, the sum of the final probability values ​​of the to-be-identified ultrasound image belonging to the various preset section types and the final probability values ​​of the to-be-identified ultrasound image belonging to other section types is equal to 1.

[0092] Specifically, the type and number of inherent structures of each preset section type are predetermined based on existing medical knowledge. The preset section types include but are not limited to eight sections: PLAX-LV (parasternal long axis-left ventricle) section, PSAX-AV (parasternal short axis-greater vessel level) section, PSAX (parasternal short axis)-mitral valve orifice short axis section, PSAX (parasternal short axis)-papillary muscle level section, PSAX (parasternal short axis)-apical horizontal section, A4C (apical four-chamber heart) section, A2C (apical two-chamber heart) section, and A3C (apical three-chamber heart)-apical long axis section. Among them, the PLAX-LV (parasternal long axis-left ventricle) section includes the five intrinsic structures of the right ventricle, mitral valve, aorta, left ventricle and left atrium; the PSAX-AV (parasternal short axis-greater vessel level) section includes the five intrinsic structures of the aortic valve, pulmonary valve, left atrium, right atrium and tricuspid valve; the PSAX (parasternal short axis)-mitral valve orifice short axis section includes the four intrinsic structures of the right ventricle, anterior leaflet of the mitral valve, posterior leaflet of the mitral valve and left ventricle; the PSAX (parasternal short axis)-papillary muscle level section includes the right ventricle, anterior ventricular leaflet of the mitral valve, posterior leaflet of the mitral valve and left ventricle. The four intrinsic structures are the papillary muscle, posterior papillary muscle, and left ventricle; PSAX (parasternal short axis) - apical horizontal section includes the two intrinsic structures of the left ventricular apex and left ventricle; A4C (apical four-chamber heart) section includes the six intrinsic structures of the left ventricle, left atrium, right ventricle, right atrium, tricuspid valve, and mitral valve; A2C (apical two-chamber heart) section includes the three intrinsic structures of the left ventricle, left atrium, and mitral valve; A3C (apical three-chamber heart) - apical long axis section includes the five intrinsic structures of the left atrium, left ventricle, aortic valve, mitral valve, and aorta.

[0093] In an exemplary embodiment, the target detection model is trained by the following steps:

[0094] Acquire a first training data set, wherein the first training data set includes a plurality of first training samples, wherein the first training samples include ultrasound cardiology training images of preset section types and corresponding section intrinsic structure labels, wherein the section intrinsic structure labels include intrinsic structure category labels and intrinsic structure position labels;

[0095] The pre-built target detection model is trained using the first training data set to obtain a trained target detection model.

[0096] Specifically, the ultrasound training images in the first training data set are derived from ultrasound video frames of different preset section types (e.g., the eight standard section types listed above). By annotating these ultrasound training images, the corresponding section intrinsic structure labels can be obtained. Furthermore, for each intrinsic structure, the intrinsic structure can be annotated with a corresponding rectangular frame. For details, please refer to Figure 2a to Figure 2h ,in Figure 2a This is an echocardiographic training image of the PLAX-LV section with intrinsic structures marked; Figure 2b This is an echocardiographic training image of the PSAX-AV section with the intrinsic structures marked; Figure 2c PSAX-Mitral valve orifice short-axis echocardiography training image with inherent structures marked; Figure 2d This is an echocardiography training image of the PSAX-papillary muscle horizontal section with intrinsic structures marked; Figure 2e PSAX-apical echocardiography training image with intrinsic structures marked; Figure 2f This is an echocardiography training image of the A4C section with the intrinsic structures marked; Figure 2g An echocardiographic training image of the A2C section with intrinsic structures marked; Figure 2h The A3C-apical long axis section of the ultrasound training image is annotated with the inherent structure. It should be noted that, as can be understood by those skilled in the art, for each ultrasound training image, the inherent structure category label corresponding to the ultrasound training image can be set in the form of section type + inherent structure type. For example, the inherent structure category label of the left ventricle in the ultrasound training image of the PLAX-LV section is PLAX-LV-left ventricle. Therefore, the total number of inherent structures that can be identified by the target detection model is Where N is the total number of preset section types, n i is the number of inherent structures corresponding to the i-th preset section type. Taking the preset section types including PLAX-LV section, PSAX-AV section, PSAX-mitral valve orifice short axis section, PSAX-papillary muscle horizontal section, PSAX-apical horizontal section, A4C section, A2C section, A3C-apical long axis section as an example, the total number of inherent structures that can be identified by the target detection model is 34. In addition, it should be noted that, as those skilled in the art can understand, the present invention does not limit the specific network structure of the target detection model, and the target detection model can be a CNN (convolutional neural network) model in the prior art. It should also be noted that, as those skilled in the art can understand, more content about how to use the first training data set to train the pre-built target detection model can refer to the relevant content of the training process of the neural network model in the prior art, which will not be repeated here.

[0097] In an exemplary embodiment, the multivariate Gaussian distribution model corresponding to each of the preset section types is obtained by the following steps:

[0098] For each of the preset slice types:

[0099] Selecting a first training sample belonging to the preset section type from the first training data set as a target sample;

[0100] For each of the target samples belonging to the preset section type, the trained target detection model is used to detect the section inherent structure of the ultrasound cardiology training image of the target sample to obtain a first detection result of the corresponding section inherent structure, and for each inherent structure corresponding to the preset section type, the first detection frame with the highest confidence among all the first detection frames corresponding to the inherent structure in the first detection result is used as the first target detection frame of the inherent structure, and the first section inherent structure information of the preset section type corresponding to the target sample is obtained according to the position information of each of the first target detection frames corresponding to the target sample;

[0101] The first section inherent structural information of the preset section type corresponding to each of the target samples belonging to the preset section type is combined to obtain a first section inherent structural information set of the preset section type, and a multivariate Gaussian distribution model corresponding to the preset section type is constructed based on the first section inherent structural information set of the preset section type.

[0102] Therefore, for each preset section type, the trained target detection model is used to detect the ultrasonic cardiac training image of each target sample belonging to the preset section type, and for each target sample belonging to the preset section type, the first detection frame with the highest confidence in the first detection frame corresponding to each inherent structure of the preset section type in the first detection result of the ultrasonic cardiac training image of the target sample is used as the corresponding first target detection frame, thereby obtaining the first section inherent structure information of the preset section type corresponding to the target sample, which can lay a good foundation for the subsequent acquisition of an accurate multivariate Gaussian distribution model, and help to further improve the accuracy of the ultrasonic cardiac image section recognition method provided by the present invention. It should be noted that, as can be understood by those skilled in the art, when the preset section types include eight standard sections, namely, PLAX-LV section, PSAX-AV section, PSAX-mitral valve orifice short axis section, PSAX-papillary muscle horizontal section, PSAX-apical horizontal section, A4C section, A2C section, and A3C-apical long axis section, for each ultrasonic cardiac training image, the first detection result of the ultrasonic cardiac training image output by the target detection model includes the recognition results of 34 inherent structures (corresponding to eight standard sections), wherein each inherent structure corresponds to a plurality of first detection frames with different confidences, so that for each inherent structure of the preset section type corresponding to the ultrasonic cardiac training image, the first detection frame with the highest confidence is selected from the plurality of first detection frames with different confidences corresponding to each of the inherent structures as the first target detection frame of the inherent structure.

[0103] In an exemplary embodiment, the position information of the first object detection frame includes a horizontal coordinate and a vertical coordinate of a center point of the first object detection frame and a width and a length of the first object detection frame;

[0104] The step of acquiring the first section intrinsic structure information of the preset section type corresponding to the ultrasonic cardiology training image of the target sample according to the position information of each of the first target detection frames corresponding to the target sample includes:

[0105] Acquire a first position information matrix according to the position information of each of the first target detection frames corresponding to the target sample;

[0106] The first position information matrix is ​​reconstructed into a corresponding first position information vector to obtain the first section intrinsic structure information of the preset section type corresponding to the target sample.

[0107] Specifically, it is assumed that the sample set consisting of each target sample belonging to the i-th preset section type is expressed as Where i is a positive integer, and i∈[1,N], N is the total number of preset section types, M i is the number of target samples belonging to the i-th preset section type, h j Represents the sample set H j The jth target sample in the sample set H is traversed using the trained target detection model. j The original detection results of the ultrasound training images of all target samples in can be expressed as where h' j represents the original detection result of the ultrasonic training image of the jth target sample (i.e., the first detection result, the original detection result also includes the category of the inherent structure of the section corresponding to each first detection frame), j∈[1,M], h' j is a matrix of size A×6, where A represents the number of first detection boxes output by the target detection model. Each first detection box is represented by a six-dimensional vector, which is represented by the abscissa (abscissa of the center point), ordinate (ordinate of the center point), width, length, confidence, and corresponding intrinsic structure category (section type + intrinsic structure type, such as PLAX-LV-left ventricle) of the first detection box. For the jth target sample h j , extract the n corresponding to the i-th preset section type i The first target detection frames corresponding to the inherent structures are n i ×4 first position information matrix, by i×4 first position information matrix is ​​reconstructed to obtain the corresponding first position information vector (whose dimension is n i ×4), thereby obtaining the jth target sample h j The first slice inherent structure information of the corresponding i-th preset slice type Where S i Indicates the i-th preset section type. Taking the preset section type as PLAX-LV section as an example, it has five inherent structures of right ventricle, mitral valve, aorta, left ventricle and left atrium (that is, when the i-th preset section type (that is, the preset section type S i ) is the PLAX-LV section, n i =5), therefore, the first target detection frames corresponding to the five inherent structures of the PLAX-LV section right ventricle, PLAX-LV section mitral valve, PLAX-LV section aorta, PLAX-LV section left ventricle, and PLAX-LV section left atrium are extracted from the original detection results. The position information of these five first target detection frames can form a 5×4 first position information matrix. Thus, for each preset section type, the first section inherent structure information of the preset section type corresponding to each target sample in the corresponding sample set is combined to obtain the first section inherent structure information set of the preset section type.

[0108] In an exemplary embodiment, constructing a multivariate Gaussian distribution model corresponding to the preset section type according to the first section intrinsic structure information set of the preset section type includes:

[0109] Calculate the slice inherent structure mean and the slice inherent structure covariance matrix corresponding to the preset slice type according to the first slice inherent structure information set of the preset slice type;

[0110] According to the mean and covariance matrix corresponding to the preset section type, a multivariate Gaussian distribution model corresponding to the preset section type is constructed.

[0111] Specifically, for the relevant contents on how to calculate the mean and the covariance matrix, reference may be made to the prior art, which will not be elaborated here.

[0112] Furthermore, constructing a multivariate Gaussian distribution model corresponding to the preset section type according to the mean and covariance matrix corresponding to the preset section type includes:

[0113] The multivariate Gaussian distribution model is constructed according to the following formula:

[0114]

[0115] Among them, f i(x) is the probability density function of the multivariate Gaussian distribution corresponding to the i-th preset section type, Σ i is the inherent structural covariance matrix of the slice corresponding to the i-th preset slice type, k i is the inherent structural covariance matrix Σ of the slice corresponding to the i-th preset slice type i rank, μ i is the mean of the inherent structure of the slice corresponding to the i-th preset slice type, n i is the number of inherent structures corresponding to the i-th preset section type, and x is an independent variable related to the first section inherent structure information of the i-th preset section type.

[0116] Therefore, by introducing the dimension compensation factor into the multivariate Gaussian distribution model This can make the probabilities between different slice space embedding dimensions comparable, which is more helpful for subsequent comparison of the final probability values ​​of the to-be-identified ultrasound cardiac image belonging to each of the preset slice types and other slice types.

[0117] In an exemplary embodiment, the step of using a trained target detection model to detect the inherent structure of a section of the acquired ultrasound image to be identified, so as to obtain the inherent structure information of a first section corresponding to the ultrasound image to be identified, includes:

[0118] Using the trained target detection model to detect the inherent structure of the section of the acquired ultrasound cardiac image to be identified, so as to obtain a corresponding second detection result;

[0119] For each inherent structure of each preset section type, the second detection frame with the highest confidence among all the second detection frames corresponding to the inherent structure of the preset section type in the second detection result is used as the second target detection frame of the inherent structure of the preset section type, and the first section inherent structure information of the preset section type corresponding to the ultrasonic cardiac image to be identified is obtained according to the position information of each second target detection frame corresponding to the preset section type.

[0120] Therefore, for each inherent structure in each preset section type, the second detection frame with the highest confidence in the second detection frame corresponding to the inherent structure in the second detection result corresponding to the ultrasonic cardiac image to be identified is used as the second target detection frame corresponding to the inherent structure, which can effectively ensure the accuracy of the first section inherent structure information of the preset section type.

[0121] Furthermore, the position information of the second object detection frame includes a horizontal coordinate, a vertical coordinate of a center point of the second object detection frame, and a width and a length of the second object detection frame.

[0122] The step of acquiring the first section intrinsic structure information of the preset section type corresponding to the ultrasonic cardiac image to be identified according to the position information of each second target detection frame corresponding to the preset section type includes:

[0123] Acquire a second position information matrix according to the position information of each of the second object detection frames;

[0124] The second position information matrix is ​​reconstructed into a corresponding second position information vector to obtain the first section intrinsic structure information of the preset section type corresponding to the ultrasonic cardiac image to be identified.

[0125] Specifically, for more details on how to obtain the inherent structural information of the first slice of each preset slice type corresponding to the to-be-identified ultrasonic cardiac image, reference may be made to the above related description, which will not be elaborated herein.

[0126] It should be noted that, as can be understood by those skilled in the art, after obtaining the first section inherent structure information (i.e., the second position information vector) of each preset section type corresponding to the ultrasound image to be identified, for each preset section type, the second position information vector corresponding to the preset section type is substituted as an independent variable x into the multivariate Gaussian distribution model corresponding to the preset section type, so as to obtain the probability value g of the ultrasound image to be identified belonging to the preset section type. i .

[0127] In an exemplary embodiment, normalizing the initial probability values ​​of the to-be-identified ultrasound image belonging to each preset section type respectively includes:

[0128] The following formula (1) is used to normalize the initial probability value of the ultrasound image to be identified belonging to each preset section type:

[0129]

[0130] Among them, g i is the initial probability value of the ultrasound image to be identified belonging to the i-th preset section type, g′ i g i The normalized result of , i is a positive integer.

[0131] Therefore, according to formula (1), the initial probability values ​​of the ultrasound image to be identified belonging to each preset section type are normalized respectively, so that the initial probability values ​​of the ultrasound image to be identified belonging to each preset section type can be effectively normalized to between 0 and 1.

[0132] In an exemplary embodiment, the final probability values ​​of the ultrasound image to be identified belonging to each of the preset section types and other section types are calculated respectively according to the normalized result, including:

[0133] The following formula (2) is used to calculate the final probability value of the ultrasound image to be identified belonging to each of the preset section types:

[0134]

[0135] The final probability value of the ultrasound image to be identified belonging to other section types is calculated using the following formula (3):

[0136]

[0137] Among them, g″ i is the final probability value of the ultrasound image to be identified belonging to the i-th preset section type, N is the total number of preset section types, g′ i is the normalized probability value that the ultrasound image to be identified belongs to the i-th preset section type, g' h is the normalized probability value that the ultrasound image to be identified belongs to the hth preset section type, g' j is the normalized probability value that the ultrasound image to be identified belongs to the jth preset section type, g' m is the initial probability value of the ultrasound image to be identified belonging to the mth preset section type, g' k is the normalized probability value that the ultrasound image to be identified belongs to the kth preset section type, g″ other is the final probability value that the ultrasound image to be identified belongs to other section types.

[0138] Therefore, the above formulas (2) and (3) can effectively normalize the sum of the final probability values ​​of the ultrasound image to be identified belonging to each preset section type and other section types to 1. It should be noted that, as can be understood by those skilled in the art, after the final probability value of the ultrasound image to be identified belonging to each of the preset section types is calculated according to formula (2), the final probability value of the ultrasound image to be identified belonging to each of the preset section types can be directly subtracted from 1 as the sum of the final probability values ​​of the ultrasound image to be identified belonging to each of the preset section types as the final probability value of the ultrasound image to be identified belonging to other section types. Therefore, when calculating g″ 1 , g″ 2 ,……,g" N and g″ other After that, the one with the highest value is found as the maximum final probability value, and the section type corresponding to the maximum final probability value is the section type of the ultrasound cardiac image to be identified.

[0139] In an exemplary embodiment, the ultrasound cardiac image section recognition method provided by the present invention further includes:

[0140] If the identification result of the section type of the to-be-identified ultrasound cardiogram is other section types, determining that the section corresponding to the to-be-identified ultrasound cardiogram is a non-standard section;

[0141] If the recognition result of the section type of the ultrasonic cardiac image to be recognized is one of the preset section types, executing the following steps;

[0142] Acquiring inherent structure information of a second section corresponding to the ultrasonic cardiogram to be identified according to a detection result of the inherent structure of the section of the ultrasonic cardiogram to be identified;

[0143] Inputting the inherent structural information of the second section corresponding to the ultrasound cardiogram to be identified into a pre-trained section quality control model to obtain a probability value of a section quality standard and a probability value of a section quality non-standard for each preset section type; and

[0144] The section quality control score of the ultrasonic cardiac image to be identified is calculated according to the probability value of the section quality standard of the preset section type corresponding to the identification result of the section type of the ultrasonic cardiac image to be identified and the probability value of the section quality non-standard.

[0145] Therefore, the present invention judges whether the section quality of the ultrasound image to be identified is standard according to the inherent structural information of the second section corresponding to the ultrasound image to be identified, and integrates the judgment habit of ultrasound doctors on whether the section quality of the image is standard (ultrasound doctors generally judge whether the section quality is standard according to whether the inherent structure in the image is clear and whether the inherent structure is complete). It should be noted that, as can be understood by those skilled in the art, the sum of the probability values ​​of the section quality standard and the probability values ​​of the section quality non-standard for all preset section types is 1. For example, if the total number of the preset section types is 8, after the inherent structural information of the second section corresponding to the ultrasound image to be identified is input into the pre-trained section quality control model, the section quality control model will output the probability values ​​of the section quality standard and the probability values ​​of the section quality non-standard for these 8 section types, that is, output 16 probability values ​​(the sum of these 16 probability values ​​is 1), that is, for each preset section type, the probability value of the section quality standard and the probability value of the section quality non-standard will be output.

[0146] Specifically, the section quality control score of the ultrasound cardiac image to be identified can be calculated according to the following formula (4):

[0147]

[0148] Among them, q Standard-S is the probability value of the slice quality standard of the preset slice type corresponding to the recognition result of the slice type of the ultrasonic cardiac image to be recognized, q Non-standard-S It is a probability value of non-standard section quality of a preset section type corresponding to the recognition result of the section type of the ultrasonic cardiac image to be recognized.

[0149] In an exemplary embodiment, the step of acquiring the second section intrinsic structure information corresponding to the ultrasonic cardiogram to be identified according to the detection result of the section intrinsic structure of the ultrasonic cardiogram to be identified includes:

[0150] For each inherent structure, the second detection frame with the highest confidence among all the second detection frames corresponding to the inherent structure in the detection results is used as the second target detection frame of the inherent structure, and the second section inherent structure information corresponding to the ultrasonic cardiac image to be identified is obtained according to the position information of each second target detection frame and the confidence of the second target detection frame.

[0151] Specifically, the slice intrinsic structure matrix can be first obtained according to the position information of each second target detection frame and the confidence of the second target detection frame, and then the slice intrinsic structure matrix can be reconstructed into the corresponding slice intrinsic structure vector, so as to obtain the second slice intrinsic structure information corresponding to the ultrasonic cardiac image to be identified. It should be noted that, as can be understood by those skilled in the art, for each inherent structure on the ultrasonic cardiac image to be identified, the position information of the second target detection frame corresponding to the inherent structure can reflect whether the inherent structure is complete, and the confidence of the second target detection frame corresponding to the inherent structure can reflect whether the inherent structure is clear.

[0152] In an exemplary embodiment, the ultrasound cardiac image section recognition method provided by the present invention further includes:

[0153] If the section quality control score of the to-be-identified ultrasonic cardiology image is greater than a first preset threshold, the section corresponding to the to-be-identified ultrasonic cardiology image is determined to be a standard section.

[0154] It should be noted that, as those skilled in the art can understand, the first preset threshold can be set according to specific circumstances, and the present invention is not limited to this.

[0155] In an exemplary embodiment, the training process of the section quality control model includes the following steps:

[0156] Acquire a second training data set, where the second training data set includes a plurality of second training samples, where the second training samples include section intrinsic structure training data and corresponding labels indicating whether a preset section type is standard or not;

[0157] The pre-built section quality control model is trained using the second training data set to obtain a trained section quality control model.

[0158] Specifically, the target detection model described above can be used to detect the inherent structure of the section for each of the ultrasound training images in the first training data set, so as to obtain the second inherent structure information of the section corresponding to each of the ultrasound training images. For each of the ultrasound training images, the corresponding inherent structure matrix of the section can be obtained according to the position information of each of the first target detection frames corresponding to the ultrasound training image and the confidence of the first target detection frame, and then the inherent structure matrix of the section can be reconstructed into the corresponding inherent structure vector of the section, so as to obtain the second inherent structure information of the section corresponding to the ultrasound training image, and the second inherent structure information of the section corresponding to the ultrasound training image is the inherent structure training data of the section. Furthermore, for each of the ultrasound training images, a senior ultrasound doctor can judge whether the ultrasound training image meets the preset section type standard, and the judgment result is used as a label of whether the preset section type standard is met and the second inherent structure information of the section corresponding to the ultrasound training image (i.e., the inherent structure training data of the section) constitutes a second training sample. For example, if the result of determining whether a preset section type standard of a certain ultrasound cardiology training image is a PLAX section standard, the PLAX section standard is used as a label and the second section inherent structure information corresponding to the ultrasound cardiology training image (i.e., section inherent structure training data) is used to form a second training sample. It should be noted that, as can be understood by those skilled in the art, the present invention does not limit the specific network structure of the section quality control model, and the network structure of the section quality model can adopt the network structure of any classification model in the prior art.

[0159] Based on the same inventive concept, the present invention also provides an ultrasound cardiology video section quality control method, please refer to Figure 3 , which schematically shows a flow chart of an ultrasound cardiology video slice quality control method provided by an embodiment of the present invention. Figure 3 As shown, the ultrasound cardiology video section quality control method provided by the present invention comprises the following steps:

[0160] Step S210: for each frame of the ultrasonic cardiogram in the ultrasonic cardiogram video, the ultrasonic cardiogram section recognition method described above is used to recognize the frame of the ultrasonic cardiogram, so as to obtain a recognition result of the section type of the frame of the ultrasonic cardiogram;

[0161] Step S220: for each frame of the ultrasonic cardiogram in the ultrasonic cardiogram video, if the recognition result of the section type of the frame of the ultrasonic cardiogram is other section types, it is determined that the section corresponding to the frame of the ultrasonic cardiogram is a non-standard section; if the recognition result of the section type of the frame of the ultrasonic cardiogram is one of the preset section types, the following steps are performed;

[0162] Inputting the inherent structural information of the section corresponding to the frame of ultrasound cardiology image into a pre-trained section quality control model to obtain a probability value of a section quality standard and a probability value of a section quality non-standard for each preset section type;

[0163] Calculating a slice quality control score for the frame of ultrasound cardiology image according to a probability value of a slice quality standard for a preset slice type and a probability value of a slice quality non-standard for the slice type corresponding to the recognition result of the slice type for the frame of ultrasound cardiology image; and

[0164] If the section quality control score of the frame of ultrasound cardiology image is greater than a preset threshold, the section corresponding to the frame of ultrasound cardiology image is determined to be a standard section;

[0165] Step S230, counting the number of frames of the ultrasonic cardiogram whose sections are standard sections in the ultrasonic cardiogram video, and calculating the section quality control score of the ultrasonic cardiogram video according to the ratio between the number of frames of the ultrasonic cardiogram whose sections are standard sections and the total number of frames of the ultrasonic cardiogram in the ultrasonic cardiogram video.

[0166] Since the ultrasonic video section quality control method provided by the present invention and the ultrasonic image section recognition method provided by the present invention belong to the same inventive concept, the ultrasonic video section quality control method provided by the present invention has all the advantages of the ultrasonic image section recognition method provided by the present invention. In addition, the ultrasonic video section quality control method provided by the present invention judges whether the section quality is standard for each frame of ultrasonic image in the video according to the inherent structural information of the second section corresponding to the frame of ultrasonic image. The judgment integrates the judgment habits of doctors in the ultrasound department on whether the section quality of the image is standard, thereby providing a basis for realizing the control of the quality of ultrasonic video sections.

[0167] Furthermore, the ultrasound cardiology video section quality control method provided by the present invention also includes:

[0168] It is determined whether the section quality control score of the ultrasound cardiology video is greater than a second preset threshold value. If so, it is determined that the section quality of the ultrasound cardiology video is qualified.

[0169] It should be noted that, as those skilled in the art can understand, the second preset threshold can be set according to actual conditions, and the present invention is not limited to this.

[0170] Based on the same inventive concept, the present invention also provides an electronic device, please refer to Figure 4 , which schematically shows a block diagram of an electronic device provided by an embodiment of the present invention. Figure 4 As shown, the electronic device includes a processor 101 and a memory 103, and a computer program is stored in the memory 103. When the computer program is executed by the processor 101, the ultrasonic cardiac image section recognition method or the ultrasonic cardiac video section quality control method described above is implemented. Since the electronic device provided by the present invention and the ultrasonic cardiac image section recognition method provided by the present invention or the ultrasonic cardiac video section quality control method provided by the present invention belong to the same inventive concept, the electronic device provided by the present invention has all the advantages of the ultrasonic cardiac image section recognition method or the ultrasonic cardiac video section quality control method provided by the present invention. For details, please refer to the relevant description above, which will not be repeated here.

[0171] like Figure 4 As shown, the electronic device also includes a communication interface 102 and a communication bus 104, wherein the processor 101, the communication interface 102, and the memory 103 communicate with each other through the communication bus 104. The communication bus 104 can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus 104 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus. The communication interface 102 is used for communication between the above-mentioned electronic device and other devices.

[0172] The processor 101 referred to in the present invention may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor 101 is the control center of the electronic device, and various interfaces and lines are used to connect various parts of the entire electronic device.

[0173] The memory 103 may be used to store the computer program. The processor 101 implements various functions of the electronic device by running or executing the computer program stored in the memory 103 and calling the data stored in the memory 103 .

[0174] The memory 103 may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0175] The present invention also provides a readable storage medium, wherein a computer program is stored in the readable storage medium, and when the computer program is executed by a processor, the ultrasonic cardiac image section recognition method or the ultrasonic cardiac video section quality control method described above can be implemented. Since the readable storage medium provided by the present invention and the ultrasonic cardiac image section recognition method or the ultrasonic cardiac video section quality control method provided by the present invention belong to the same inventive concept, the readable storage medium provided by the present invention has all the advantages of the ultrasonic cardiac image section recognition method or the ultrasonic cardiac video section quality control method provided by the present invention. For details, please refer to the relevant description above, which will not be repeated here.

[0176] The readable storage medium provided by the present invention can adopt any combination of one or more computer-readable media. The readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, a system, device or component of electricity, magnetism, light, electromagnetic, infrared or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer hard disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this article, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, a device or a device or used in combination with it.

[0177] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program codes. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, device, or device. The program code contained on the computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0178] In summary, compared with the prior art, the ultrasonic cardiology image section recognition method, ultrasonic cardiology video section quality control method, electronic device and readable storage medium provided by the present invention have the following advantages:

[0179] The ultrasonic cardiac image section recognition method provided by the present invention first uses a trained target detection model to detect the inherent structure of the section of the acquired ultrasonic cardiac image to be recognized, so as to obtain the first section inherent structure information corresponding to the ultrasonic cardiac image to be recognized; then, according to the first section inherent structure information and the multivariate Gaussian distribution models corresponding to the various preset section types acquired in advance, respectively calculate the initial probability value of the ultrasonic cardiac image to be recognized belonging to each of the preset section types; then, respectively normalize the initial probability value of the ultrasonic cardiac image to be recognized belonging to each of the preset section types, so as to obtain the normalized probability value of the ultrasonic cardiac image to be recognized belonging to each of the preset section types; then, according to the normalized probability value of the ultrasonic cardiac image to be recognized belonging to each of the preset section types, respectively calculate the probability value of the ultrasonic cardiac image to be recognized belonging to each of the preset section types. type and the final probability values ​​of other section types, wherein the other section types are section types other than all the preset section types; finally, the maximum final probability value is determined from all the final probability values, and the section type corresponding to the maximum final probability value is used as the recognition result of the section type of the ultrasonic cardiac image to be identified. Therefore, the ultrasonic cardiac image section recognition method provided by the present invention makes full use of the field knowledge of ultrasonic clinical examination to give full application to the inherent structural information in the standard section of the ultrasonic cardiac image, so that the ultrasonic cardiac image section recognition method provided by the present invention is more inclined to the thinking mode of professional doctors, thereby effectively improving the accuracy of ultrasonic cardiac image section type recognition, and further can better assist doctors in section type recognition, so as to reduce the difficulty in the process of ultrasonic cardiac image section type recognition, and further can better assist the standardized scanning of ultrasonic cardiograms. In addition, by normalizing the initial probability values ​​of the ultrasound image to be identified belonging to each preset section type, the initial probability values ​​of the ultrasound image to be identified belonging to each preset section type can be normalized to between 0 and 1, thereby making it easier to subsequently calculate the final probability values ​​of the ultrasound image to be identified belonging to each of the preset section types and other section types according to the results after normalization, and finally determine the section type of the ultrasound image to be identified according to the final probability values ​​of the ultrasound image to be identified belonging to each of the preset section types and other section types.

[0180] Since the ultrasonic video slice quality control method, electronic device and readable storage medium provided by the present invention belong to the same inventive concept as the ultrasonic image slice recognition method provided by the present invention, the ultrasonic video slice quality control method, electronic device and readable storage medium provided by the present invention have all the advantages of the ultrasonic image slice recognition method provided by the present invention. In addition, the ultrasonic video slice quality control method provided by the present invention judges whether the slice quality is standard for each frame of ultrasonic image in the video according to the inherent structural information of the second slice corresponding to the frame of ultrasonic image. The judgment integrates the judgment habits of doctors in the ultrasound department on whether the slice quality of the image is standard, thereby providing a basis for realizing the control of the quality of the ultrasonic video slice.

[0181] It should be noted that the computer program code for performing the operation of the present invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages-such as Java, Smalltalk, C++, and also conventional procedural programming languages-such as "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network-including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, using an Internet service provider to connect through the Internet).

[0182] It should be noted that the devices and methods disclosed in the embodiments of this article can also be implemented in other ways. The device implementation described above is only schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of this article. In this regard, each box in the flowchart or block diagram can represent a part of a module, program or code, and the module, program segment or part of the code contains one or more executable instructions for implementing the specified logical function, and the module, program segment or part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented by a dedicated hardware-based system for performing a specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions.

[0183] In addition, the functional modules in the various embodiments of this document may be integrated together to form an independent part, or each module may exist independently, or two or more modules may be integrated to form an independent part.

[0184] The above description is only a description of the preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any changes and modifications made by a person skilled in the art in the field of the present invention based on the above disclosure are within the scope of protection of the present invention. Obviously, a person skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations fall within the scope of the present invention and its equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for identifying sections of an ultrasound cardiac image. It is characterized in that include: Using the trained target detection model to detect the inherent structure of the section of the acquired ultrasound cardiogram to be identified, so as to obtain the inherent structure information of the first section corresponding to the ultrasound cardiogram to be identified; Calculating respectively the initial probability value of the to-be-identified ultrasound cardiac image belonging to each of the preset section types according to the inherent structural information of the first section and the pre-acquired multivariate Gaussian distribution models corresponding to the various preset section types; Normalizing the initial probability values ​​of the to-be-identified ultrasound cardiac image belonging to each of the preset section types respectively, so as to obtain the normalized probability values ​​of the to-be-identified ultrasound cardiac image belonging to each of the preset section types; According to the normalized probability values ​​of the ultrasound image to be identified belonging to each of the preset section types, respectively calculating the final probability value of the ultrasound image to be identified belonging to each of the preset section types and the final probability values ​​of other section types, wherein the other section types are section types other than all the preset section types; Determine the maximum final probability value from all the final probability values, and use the section type corresponding to the maximum final probability value as the recognition result of the section type of the ultrasound cardiac image to be recognized; The following formula is used to calculate the final probability value of the to-be-identified ultrasound cardiac image belonging to each of the preset section types: The final probability value of the to-be-identified echocardiographic image belonging to other section types is calculated using the following formula: Among them, g i is the final probability value of the ultrasound image to be identified belonging to the i-th preset section type, N is the total number of preset section types, g i ' is the normalized probability value of the ultrasound image to be identified belonging to the i-th preset section type, g' h is the normalized probability value that the ultrasonic cardiac image to be identified belongs to the hth preset section type, g' j is the normalized probability value that the ultrasonic cardiac image to be identified belongs to the jth preset section type, g' m is the initial probability value of the ultrasound image to be identified belonging to the mth preset section type, g' k is the normalized probability value that the ultrasound image to be identified belongs to the kth preset section type, g'' other is the final probability value that the ultrasound cardiac image to be identified belongs to other section types.

2. The method for recognizing sections of an ultrasonic cardiac image according to claim 1, It is characterized in that The target detection model is trained by the following steps: Acquire a first training data set, wherein the first training data set includes a plurality of first training samples, wherein the first training samples include ultrasound cardiology training images of preset section types and corresponding section intrinsic structure labels, wherein the section intrinsic structure labels include intrinsic structure category labels and intrinsic structure position labels; The pre-built target detection model is trained using the first training data set to obtain a trained target detection model.

3. The method for recognizing sections of an ultrasonic cardiac image according to claim 2, It is characterized in that Obtain the multivariate Gaussian distribution model corresponding to each preset slice type by following the steps below: For each of the preset slice types: Selecting a first training sample belonging to the preset section type from the first training data set as a target sample; For each of the target samples belonging to the preset section type, the trained target detection model is used to detect the section inherent structure of the ultrasound cardiology training image of the target sample to obtain a first detection result of the corresponding section inherent structure, and for each inherent structure corresponding to the preset section type, the first detection frame with the highest confidence among all the first detection frames corresponding to the inherent structure in the first detection result is used as the first target detection frame of the inherent structure, and the first section inherent structure information of the preset section type corresponding to the target sample is obtained according to the position information of each of the first target detection frames corresponding to the target sample; The first section inherent structural information of the preset section type corresponding to each of the target samples belonging to the preset section type is combined to obtain a first section inherent structural information set of the preset section type, and a multivariate Gaussian distribution model corresponding to the preset section type is constructed based on the first section inherent structural information set of the preset section type.

4. The method for recognizing sections of an ultrasonic cardiac image according to claim 3, It is characterized in that The position information of the first object detection frame includes the horizontal coordinate and the vertical coordinate of the center point of the first object detection frame and the width and length of the first object detection frame; The acquiring, according to the position information of each of the first target detection frames corresponding to the target sample, the first section intrinsic structure information of the preset section type corresponding to the target sample comprises: Acquire a first position information matrix according to the position information of each of the first target detection frames corresponding to the target sample; The first position information matrix is ​​reconstructed into a corresponding first position information vector to obtain the first section intrinsic structure information of the preset section type corresponding to the target sample.

5. The method for recognizing sections of an ultrasonic cardiac image according to claim 3, It is characterized in that The step of constructing a multivariate Gaussian distribution model corresponding to the preset section type according to the first section intrinsic structure information set of the preset section type includes: Calculate the slice inherent structure mean and the slice inherent structure covariance matrix corresponding to the preset slice type according to the first slice inherent structure information set of the preset slice type; According to the mean and covariance matrix corresponding to the preset section type, a multivariate Gaussian distribution model corresponding to the preset section type is constructed.

6. The method for recognizing sections of an ultrasonic cardiac image according to claim 5, It is characterized in that The step of constructing a multivariate Gaussian distribution model corresponding to the preset section type according to the mean and covariance matrix corresponding to the preset section type includes: The multivariate Gaussian distribution model is constructed according to the following formula: Among them, f i (x) is the probability density function of the multivariate Gaussian distribution corresponding to the i-th preset section type, Σ i is the inherent structural covariance matrix of the slice corresponding to the i-th preset slice type, k i is the inherent structural covariance matrix Σ of the slice corresponding to the i-th preset slice type i rank, μ i is the mean of the inherent structure of the slice corresponding to the i-th preset slice type, n i is the number of inherent structures corresponding to the i-th preset section type, and x is an independent variable related to the first section inherent structure information of the i-th preset section type.

7. The method for identifying sections of an ultrasonic cardiac image according to claim 1, It is characterized in that The method of using the trained target detection model to detect the inherent structure of the section of the acquired ultrasound cardiogram to be identified, so as to obtain the inherent structure information of the first section corresponding to the ultrasound cardiogram to be identified, includes: Using the trained target detection model to detect the inherent structure of the section of the acquired ultrasound cardiac image to be identified, so as to obtain a corresponding second detection result; For each inherent structure of each preset section type, the second detection frame with the highest confidence among all the second detection frames corresponding to the inherent structure of the preset section type in the second detection result is used as the second target detection frame of the inherent structure of the preset section type, and the first section inherent structure information of the preset section type corresponding to the ultrasonic cardiac image to be identified is obtained according to the position information of each second target detection frame corresponding to the preset section type.

8. The method for recognizing sections of an ultrasonic cardiac image according to claim 7, It is characterized in that The step of calculating the initial probability value of the to-be-identified ultrasound cardiac image belonging to each of the preset section types according to the inherent structural information of the first section and the pre-acquired multivariate Gaussian distribution models corresponding to the various preset section types, comprises: For each preset section type, an initial probability value that the ultrasound cardiac image to be identified belongs to the preset section type is calculated according to the inherent structural information of the first section of the preset section type and the multivariate Gaussian distribution model corresponding to the preset section type.

9. The method for recognizing sections of an ultrasonic cardiac image according to claim 1, It is characterized in that The normalizing of the initial probability values ​​of the to-be-identified ultrasound cardiac image belonging to each preset section type comprises: The following formula is used to normalize the initial probability value of the ultrasonic cardiac image to be identified belonging to each preset section type: Among them, g i is the initial probability value of the ultrasound image to be identified belonging to the i-th preset section type, g' i g i The corresponding normalized probability value, i is a positive integer, g l is an initial probability value that the ultrasound cardiac image to be identified belongs to the lth preset section type, and N is the total number of preset section types.

10. The method for identifying sections of an ultrasonic cardiogram according to claim 1, It is characterized in that The method further comprises: If the identification result of the section type of the to-be-identified ultrasound cardiogram is other section types, determining that the section corresponding to the to-be-identified ultrasound cardiogram is a non-standard section; If the recognition result of the section type of the ultrasonic cardiac image to be recognized is one of the preset section types, executing the following steps; Acquiring inherent structure information of a second section corresponding to the ultrasonic cardiogram to be identified according to a detection result of the inherent structure of the section of the ultrasonic cardiogram to be identified; Inputting the inherent structural information of the second section corresponding to the ultrasound cardiogram to be identified into a pre-trained section quality control model to obtain a probability value of a section quality standard and a probability value of a section quality non-standard for each preset section type; and The section quality control score of the ultrasonic cardiac image to be identified is calculated according to the probability value of the section quality standard of the preset section type corresponding to the identification result of the section type of the ultrasonic cardiac image to be identified and the probability value of the section quality non-standard.

11. The method for identifying sections of an ultrasonic cardiogram according to claim 10, It is characterized in that The method further comprises: If the section quality control score of the to-be-identified ultrasonic cardiology image is greater than a first preset threshold, the section corresponding to the to-be-identified ultrasonic cardiology image is determined to be a standard section.

12. The method for identifying sections of an ultrasonic cardiogram according to claim 10, It is characterized in that The step of acquiring the second section intrinsic structure information corresponding to the ultrasonic cardiogram to be identified according to the detection result of the section intrinsic structure of the ultrasonic cardiogram to be identified comprises: For each inherent structure, the second detection frame with the highest confidence among all the second detection frames corresponding to the inherent structure in the detection results is used as the second target detection frame of the inherent structure, and the second section inherent structure information corresponding to the ultrasonic cardiac image to be identified is obtained according to the position information of each second target detection frame and the confidence of the second target detection frame.

13. A method for quality control of ultrasound cardiology video sections. It is characterized in that include: For each frame of the ultrasonic cardiogram in the ultrasonic cardiogram video, the ultrasonic cardiogram section recognition method according to any one of claims 1 to 9 is used to recognize the frame of the ultrasonic cardiogram, so as to obtain a recognition result of the section type of the frame of the ultrasonic cardiogram; For each frame of the ultrasonic cardiogram in the ultrasonic cardiogram video, if the recognition result of the section type of the frame of the ultrasonic cardiogram is other section types, it is determined that the section corresponding to the frame of the ultrasonic cardiogram is a non-standard section; if the recognition result of the section type of the frame of the ultrasonic cardiogram is one of the preset section types, the following steps are performed; Inputting the inherent structural information of the section corresponding to the frame of ultrasound cardiology image into a pre-trained section quality control model to obtain a probability value of a section quality standard and a probability value of a section quality non-standard for each preset section type; Calculating a slice quality control score for the frame of ultrasound cardiology image according to a probability value of a slice quality standard for a preset slice type corresponding to the recognition result of the slice type for the frame of ultrasound cardiology image and a probability value of a slice quality non-standard; as well as If the section quality control score of the frame of ultrasound cardiology image is greater than a preset threshold, the section corresponding to the frame of ultrasound cardiology image is determined to be a standard section; The number of frames of the ultrasonic cardiogram whose sections are standard sections in the ultrasonic cardiogram video is counted, and the section quality control score of the ultrasonic cardiogram video is calculated according to the ratio between the number of frames of the ultrasonic cardiogram whose sections are standard sections and the total number of frames of the ultrasonic cardiogram in the ultrasonic cardiogram video.

14. The method for quality control of ultrasound cardiology video slices according to claim 13, It is characterized in that The method further comprises: It is determined whether the section quality control score of the ultrasound cardiology video is greater than a second preset threshold value. If so, it is determined that the section quality of the ultrasound cardiology video is qualified.

15. An electronic device, It is characterized in that The method comprises a processor and a memory, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the ultrasonic cardiac image section recognition method described in any one of claims 1 to 12 or the ultrasonic cardiac video section quality control method described in any one of claims 13 to 14 is implemented.

16. A readable storage medium, It is characterized in that The readable storage medium stores a computer program, and when the computer program is executed by the processor, the ultrasonic cardiac image section recognition method described in any one of claims 1 to 12 or the ultrasonic cardiac video section quality control method described in any one of claims 13 to 14 is implemented.

Images