Three-dimensional heart segmentation image screening method, device, equipment and storage medium

Abstract

The present invention relates to the field of image processing technology, and in particular to a three-dimensional cardiac segmentation image screening method, device, equipment and storage medium. The method comprises: performing mask processing on a fixed image to obtain a floating image with a gold standard image, wherein the fixed image is a sample cardiac image; performing deformation processing on the floating image to obtain a registration image, and performing deformation processing on the gold standard image in the floating image to obtain a pseudo gold standard image; obtaining mutual information based on the fixed image and the registration image, and obtaining a similarity coefficient based on the fixed image and the pseudo gold standard image; screening cardiac images to be screened based on the mutual information and the similarity coefficient to obtain a target cardiac image; since the present invention uses the mutual information and similarity coefficient obtained based on the fixed image and the floating image as standards to screen cardiac images to be screened, compared with existing screening by manual methods, the present invention can improve screening efficiency and reliability.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a three-dimensional heart segmentation image screening method, device, equipment and storage medium. Background Art

[0002] The heart is the most important organ in the human body, managing blood throughout the body and transporting it to various parts of the body. Heart disease directly affects a person's life and death. Therefore, conducting new technology research on early diagnosis and treatment of heart disease has very important social significance and practical value.

[0003] At present, in order to facilitate heart research, a three-dimensional active shape model is usually used to segment heart images to obtain three-dimensional grid segmentation contour images of the left and right ventricles of the heart. Researchers then study the segmented contour images.

[0004] However, prior to the existing research, due to factors such as different imaging quality of cardiac images and different anatomical conditions, the segmentation contour images obtained by the three-dimensional active shape model were of poor quality. Researchers had to manually screen out the segmentation contour images of poor quality before conducting research, which resulted in low screening efficiency.

[0005] The above content is only used to assist in understanding the technical solution of the present invention and does not constitute an admission that the above content is prior art. Summary of the Invention

[0006] The main purpose of the present invention is to provide a three-dimensional cardiac segmentation image screening method, device, equipment and storage medium, aiming to solve the technical problem of low efficiency in screening segmentation contour images in the prior art.

[0007] To achieve the above object, the present invention provides a three-dimensional heart segmentation image screening method, which includes the following steps:

[0008] performing mask processing on the fixed image to obtain a floating image with a gold standard image, wherein the fixed image is a sample heart image;

[0009] Performing deformation processing on the floating image to obtain a registration image, and performing deformation processing on the gold standard image in the floating image to obtain a pseudo gold standard image;

[0010] Obtaining mutual information based on the fixed image and the registered image, and obtaining a similarity coefficient based on the fixed image and the pseudo gold standard image;

[0011] The cardiac images to be screened are screened based on the mutual information and the similarity coefficient to obtain a target cardiac image.

[0012] Optionally, the step of screening the cardiac images to be screened based on the mutual information and the similarity coefficient to obtain a target cardiac image includes:

[0013] Performing a weighted summation on the mutual information and the similarity coefficient, and obtaining a target pseudo gold standard image according to the summation result;

[0014] obtaining a target similarity coefficient according to the target pseudo-gold standard image and the heart image to be screened;

[0015] The heart image to be screened is screened according to the target similarity coefficient to obtain a target heart image.

[0016] Optionally, before the step of obtaining a target similarity coefficient based on the target pseudo-gold standard image and the heart image to be screened, the method further includes:

[0017] Obtain an intermediate layer image of the heart image to be screened through the three-dimensional active model;

[0018] Accordingly, the step of obtaining a target similarity coefficient based on the target pseudo-gold standard image and the heart image to be screened includes:

[0019] A target similarity coefficient is obtained according to the target pseudo gold standard image and the intermediate layer image.

[0020] Optionally, before the step of obtaining mutual information according to the fixed image and the registered image, and obtaining a similarity coefficient according to the fixed image and the pseudo-gold standard image, the method further includes:

[0021] Segmenting the fixed image using a preset convolutional neural network model to obtain a segmented image;

[0022] Accordingly, the step of obtaining mutual information based on the fixed image and the registered image, and obtaining a similarity coefficient based on the fixed image and the pseudo gold standard image includes:

[0023] Mutual information is obtained based on the fixed image and the registered image, and a similarity coefficient is obtained based on the pseudo gold standard image and the segmented image.

[0024] Optionally, the step of obtaining mutual information based on the fixed image and the registered image includes:

[0025] Obtaining fixed information entropy according to the fixed image, obtaining registration information entropy according to the registered image, and obtaining joint information entropy according to the fixed image and the registered image;

[0026] Corresponding mutual information is obtained according to the fixed information entropy, the registration information entropy and the joint information entropy.

[0027] Optionally, before the steps of deforming the floating image to obtain a registered image and deforming the gold standard image in the floating image to obtain a pseudo gold standard image, the method further includes:

[0028] Training the floating image and the fixed image through a registration network model to obtain a target deformation field;

[0029] Accordingly, the steps of deforming the floating image to obtain a registered image, and deforming the gold standard image in the floating image to obtain a pseudo gold standard image include:

[0030] Performing deformation processing on the floating image based on the target deformation field to obtain a registered image;

[0031] The gold standard image in the floating image is deformed based on the target deformation field to obtain a pseudo gold standard image.

[0032] Optionally, before the step of performing weighted summation of the mutual information and the similarity coefficient and obtaining a target pseudo gold standard image according to the summation result, the method further includes:

[0033] Obtaining a weight value of the similarity coefficient, a weight value of the mutual information, and a preset weighting function;

[0034] Obtaining a preset weighting formula according to the weight value of the similarity coefficient, the weight value of the mutual information and a preset weighting function;

[0035] Accordingly, the step of performing weighted summation on the mutual information and the similarity coefficient and obtaining a target pseudo gold standard image according to the summation result includes:

[0036] Performing weighted summation on the mutual information and the similarity coefficient using the preset weighting formula, and obtaining a target pseudo gold standard image according to the summation result;

[0037] Wherein, the preset weighting formula is:

[0038] PG=argmax i {λ1*DICE(P i ,CF)+λ2*MI i (F,MF i )},

[0039] Where PG is the summation result, argmax i is the preset weighting function, λ1 is the weight value of the similarity coefficient, DICE(P i ,CF) is the similarity coefficient, λ2 is the weight value of the mutual information, MI i (F,MF i) is the mutual information.

[0040] In addition, to achieve the above-mentioned purpose, the present invention also provides a three-dimensional heart segmentation image screening device, the device comprising:

[0041] a mask processing module, configured to perform mask processing on a fixed image to obtain a floating image with a gold standard image, wherein the fixed image is a sample heart image;

[0042] a deformation processing module, configured to perform deformation processing on the floating image to obtain a registration image, and perform deformation processing on the gold standard image in the floating image to obtain a pseudo gold standard image;

[0043] an information acquisition module, configured to obtain mutual information based on the fixed image and the registered image, and obtain a similarity coefficient based on the fixed image and the pseudo-gold standard image;

[0044] The image screening module is used to screen the cardiac images to be screened based on the mutual information and the similarity coefficient to obtain a target cardiac image.

[0045] In addition, to achieve the above-mentioned purpose, the present invention also proposes a three-dimensional cardiac segmentation image screening device, which includes: a memory, a processor, and a three-dimensional cardiac segmentation image screening program stored on the memory and runnable on the processor, and the three-dimensional cardiac segmentation image screening program is configured to implement the steps of the three-dimensional cardiac segmentation image screening method described above.

[0046] In addition, to achieve the above-mentioned purpose, the present invention also proposes a storage medium, on which a three-dimensional heart segmentation image screening program is stored. When the three-dimensional heart segmentation image screening program is executed by a processor, the steps of the three-dimensional heart segmentation image screening method described above are implemented.

[0047] The present invention performs masking on a fixed image to obtain a floating image with a gold standard image, wherein the fixed image is a sample heart image; deforms the floating image to obtain a registration image, and deforms the gold standard image within the floating image to obtain a pseudo-gold standard image; obtains mutual information based on the fixed image and the registration image, and obtains a similarity coefficient based on the fixed image and the pseudo-gold standard image; and screens the heart images to be screened based on the mutual information and the similarity coefficient to obtain a target heart image. Because the present invention first obtains a floating image with a gold standard image using a fixed image, deforms the floating image and the gold standard image within the floating image to obtain a registration image and a pseudo-gold standard image, then obtains mutual information using the fixed image and the registration image, and obtains a similarity coefficient using the fixed image and the pseudo-gold standard image, and finally uses the mutual information and similarity coefficient as criteria to screen the heart images to be screened, the present invention can improve screening efficiency compared to existing manual screening methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A schematic diagram of the structure of a three-dimensional cardiac segmentation image screening device in the hardware operating environment involved in an embodiment of the present invention;

[0049] Figure 2 Schematic diagram of the process of the first embodiment of the three-dimensional heart segmentation image screening method of the present invention;

[0050] Figure 3 2. It is a flowchart of a second embodiment of the three-dimensional heart segmentation image screening method of the present invention;

[0051] Figure 4 2. It is a flowchart of a third embodiment of the three-dimensional heart segmentation image screening method of the present invention;

[0052] Figure 5 This is a structural block diagram of the first embodiment of the 3D heart segmentation image screening device of the present invention.

[0053] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0054] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0055] Reference Figure 1 , Figure 1 This is a schematic diagram of the structure of a three-dimensional cardiac segmentation image screening device in the hardware operating environment involved in an embodiment of the present invention.

[0056] like Figure 1As shown, the three-dimensional cardiac segmentation image screening device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to implement communication between these components. The user interface 1003 may include a display screen and an input unit such as a keyboard. Optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a wireless fidelity (Wi-Fi) interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk storage device. The memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0057] Those skilled in the art will understand that Figure 1 The structure shown in the figure does not constitute a limitation on the three-dimensional cardiac segmentation image screening device, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0058] like Figure 1 As shown, the memory 1005 as a storage medium may include an operating system, a network communication module, a user interface module, and a three-dimensional heart segmentation image screening program.

[0059] exist Figure 1 In the three-dimensional heart segmentation image screening device shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the three-dimensional heart segmentation image screening device of the present invention can be set in the three-dimensional heart segmentation image screening device, and the three-dimensional heart segmentation image screening device calls the three-dimensional heart segmentation image screening program stored in the memory 1005 through the processor 1001, and executes the three-dimensional heart segmentation image screening method provided by the embodiment of the present invention.

[0060] The embodiment of the present invention provides a three-dimensional heart segmentation image screening method, referring to Figure 2 , Figure 2 FIG. 4 is a flow chart of a first embodiment of a three-dimensional heart segmentation image screening method according to the present invention.

[0061] In this embodiment, the three-dimensional heart segmentation image screening method includes the following steps:

[0062] Step S10: performing mask processing on the fixed image to obtain a floating image with a gold standard image, wherein the fixed image is a sample heart image.

[0063] It should be noted that the method of this embodiment can be applied in scenarios where the quality of segmented contour images obtained from a three-dimensional active shape model is screened, or in other scenarios where image quality screening is required. The execution subject of this embodiment can be a three-dimensional cardiac segmentation image screening device with data processing, network communication, and program execution functions, such as a computer, mobile terminal, or other device capable of performing the same or similar functions. This embodiment and the following embodiments are specifically described using the above-mentioned three-dimensional cardiac segmentation image screening device (hereinafter referred to as the device).

[0064] It is understandable that the above sample heart image can be a three-dimensional image of the patient's heart obtained by a computer X-ray tomography camera, a three-dimensional image obtained by a magnetic resonance imaging machine, or a three-dimensional image of the patient's heart obtained by other equipment.

[0065] It should be understood that in order to make subsequent screening more accurate, in this embodiment, the above-mentioned sample heart images can be set in multiple numbers, and the specific number can be set according to actual conditions. The sample heart images can be randomly selected by the user from the database, and this embodiment does not limit this.

[0066] It should be emphasized that the above-mentioned mask processing can be a processing process of controlling the image processing area by blocking the sample heart image with a selected graphic or image. In this embodiment, since the main purpose is to obtain the contour images of the left ventricle and the right ventricle, the left and right ventricle marking labels in the sample heart image and the areas unrelated to the left and right ventricles can be masked to obtain several floating images. In addition, it should be noted that the above-mentioned labels used to mark the left and right ventricles can be called gold standards, and the label images corresponding to the floating images can be called gold standard images.

[0067] Meanwhile, for the convenience of subsequent explanation, the fixed image can be recorded as F, and the floating image can be recorded as S_M={M1, M2, ..., M n}, where S_M can represent a set of several floating images, n can represent the number of images, M can represent a floating image, and M i It can represent the i-th floating image, where 1≤i≤n; the gold standard image corresponding to each of the above floating images can be recorded as S_G={G1,G2,...,G n}, where S_G can represent the set of gold standard images corresponding to the floating image, G can represent the gold standard image, G i It can represent the gold standard image corresponding to the i-th floating image.

[0068] In a specific implementation, the above device can perform mask processing on a certain number of sample cardiac images and use the processed images as floating images.

[0069] Step S20: performing deformation processing on the floating image to obtain a registration image, and performing deformation processing on the gold standard image in the floating image to obtain a pseudo gold standard image.

[0070] It should be noted that the above deformation process may be such that the floating image M i After deformation, it is as similar as possible to the fixed image F, so that the corresponding gold standard image G i After deformation, it is as similar as possible to the label image of the fixed image F. The above registered image is the above floating image M i The image after deformation processing, the pseudo gold standard image is the gold standard image G i For the sake of understanding, we can record the registered image as MF after deformation processing. i , the pseudo gold standard image is recorded as P i .

[0071] Furthermore, in order to obtain the floating image M more accurately i The corresponding registration image MF i and the gold standard image G i The corresponding pseudo gold standard image P i Before the above step S20, it also includes: training the floating image and the fixed image through a registration network model to obtain a target deformation field; accordingly, the above step S20 includes: deforming the floating image based on the target deformation field to obtain a registered image; and deforming the gold standard image in the floating image based on the target deformation field to obtain a pseudo gold standard image.

[0072] It should be noted that the above registration network model can be used to obtain the registration image MF i and the pseudo gold standard image P i The registration network model can be downloaded from the Internet and adjusted according to the actual situation. The target deformation field can be the displacement field predicted by the registration network model, and the floating image M can be obtained through the displacement field. i The corresponding registration image MF i and the gold standard image G i The corresponding pseudo gold standard image P i , each floating image M i The corresponding target deformation fields can be different, and the corresponding registration images MF i and the pseudo gold standard image P i may also be different, so in order to facilitate distinction, the above target deformation field can be recorded as Through the target deformation field Get the corresponding registration image MF i We can record the process as Through the target deformation field Get the corresponding pseudo gold standard image P i We can record the process as

[0073] In a specific implementation, the above-mentioned device will train the floating image and the fixed image through a registration network model to obtain a target deformation field, then deform the above-mentioned floating image based on the target deformation field to obtain a registered image, and deform the above-mentioned gold standard image based on the target deformation field to obtain a pseudo gold standard image.

[0074] Step S30: obtaining mutual information according to the fixed image and the registered image, and obtaining a similarity coefficient according to the fixed image and the pseudo gold standard image.

[0075] It should be noted that the above mutual information can refer to the registered image MF i The information content of the fixed image F is included in the similarity coefficient, which can be used to compare the fixed image F and the pseudo gold standard image P. i The coefficient of the similarity metric.

[0076] In a specific implementation, the above-mentioned device obtains the amount of information of the fixed image contained in the registered image as mutual information, and obtains a similarity measurement index between the fixed image and the pseudo gold standard image as a similarity coefficient.

[0077] Step S40: screening the cardiac images to be screened based on the mutual information and the similarity coefficient to obtain a target cardiac image.

[0078] It is understandable that after the three-dimensional image of the patient's heart is segmented by the three-dimensional active shape model, several slice images can be obtained. Since the intermediate layer image can represent the quality of the entire segmented contour image obtained by the three-dimensional active shape model, the device can select the intermediate layer image from the slice image as the heart image to be screened, and the image that meets the user's needs after screening will be used as the target heart image.

[0079] In a specific implementation, the above-mentioned device will screen the intermediate layer images in the slice image based on the mutual information and similarity coefficient, and use the image that meets the user's needs as the target heart image.

[0080] The above-mentioned device in this embodiment can perform mask processing on a certain number of sample cardiac images, and use the processed images as floating images; train the floating images and fixed images through a registration network model to obtain a target deformation field, and then deform the above-mentioned floating image based on the target deformation field to obtain a registered image, and deform the above-mentioned gold standard image based on the target deformation field to obtain a pseudo-gold standard image; then obtain the amount of information contained in the fixed image in the registered image as mutual information, and obtain a similarity measurement index between the fixed image and the pseudo-gold standard image as a similarity coefficient, and screen the intermediate layer images in the slice image based on the mutual information and the similarity coefficient, and use the image that meets the user's needs as the target cardiac image; since this embodiment first uses the mutual information and the similarity coefficient as standards to screen the cardiac images to be screened, compared with the existing manual screening, the screening efficiency can be improved. At the same time, manual screening may have errors, thereby improving the accuracy and reliability of screening.

[0081] refer to Figure 3 , Figure 3 FIG. 4 is a flow chart of a second embodiment of a three-dimensional heart segmentation image screening method according to the present invention.

[0082] In order to obtain a more accurate similarity coefficient, based on the first embodiment, in this embodiment, before step S30, the following steps are further included:

[0083] Step S301: Segment the fixed image using a preset convolutional neural network model to obtain a segmented image. Accordingly, the above step S30 includes:

[0084] Step S30 ′: obtaining mutual information according to the fixed image and the registered image, and obtaining a similarity coefficient according to the pseudo gold standard image and the segmented image.

[0085] It should be noted that the above-mentioned preset convolutional neural network model can be used for image segmentation, and the preset convolutional neural network model can be downloaded from the Internet. The above-mentioned device can segment different structures in the fixed image F through the preset convolutional neural network model to obtain a segmented image, which can be denoted as CF here.

[0086] It is understandable that the similarity coefficient can be recorded as DICE (P i , CF), the calculation formula is:

[0087]

[0088] Then the similarity coefficient DICE (P i , CF).

[0089] Furthermore, the step of obtaining mutual information based on the fixed image and the registered image in the above-mentioned step S30' includes: obtaining fixed information entropy based on the fixed image, obtaining registration information entropy based on the registered image, and obtaining joint information entropy based on the fixed image and the registered image; and obtaining corresponding mutual information based on the fixed information entropy, the registration information entropy and the joint information entropy.

[0090] It should be understood that the above information entropy can eliminate the concept of uncertainty and further improve the accuracy of the screening results. It is to calculate the fuzzy information concept to obtain an accurate information entropy value. Among them, the above fixed information entropy can be recorded as H(F), and the calculation formula is:

[0091]

[0092] Where f is a pixel in a fixed image F, and p(f) is the probability of belonging to pixel f;

[0093] The above registration information entropy can be recorded as H(MF i ), the calculation formula is:

[0094]

[0095] In the formula, mf i For the i-th registration image MF i The pixels in p(mf i ) belongs to pixel mf i probability;

[0096] The above joint information entropy can be recorded as H(F,MF i ), the calculation formula is:

[0097]

[0098] Where, p(f, mf i ) belongs to pixel f and belongs to pixel mf i The probability of p(f|mf i ) is the pixel mf i The probability of belonging to pixel f under the condition of ;

[0099] The above mutual information can be recorded as MI i (F,MF i ), the calculation formula is:

[0100] MI i (F,MF i )=H(F)+H(MF i )-H(F,MF i ).

[0101] In a specific implementation, the above-mentioned device first segments a fixed image through a preset neural network model to obtain a segmented image, and then obtains fixed information entropy, registration information entropy and joint information entropy respectively through formulas, obtains mutual information through fixed information entropy, registration information entropy and joint information entropy, and obtains a similarity coefficient based on the pseudo gold standard image and the segmented image.

[0102] In this embodiment, the fixed image is segmented in advance to remove interference factors, thereby making the obtained similarity coefficient more accurate. At the same time, the concept of uncertainty can be eliminated through information entropy, further improving the accuracy of the screening results.

[0103] refer to Figure 4 , Figure 4 FIG. 4 is a flow chart of a fourth embodiment of a three-dimensional heart segmentation image screening method according to the present invention.

[0104] Furthermore, in order to screen more accurately, Figure 4 Based on the first and second embodiments, step S40 includes:

[0105] Step S41: performing weighted summation on the mutual information and the similarity coefficient, and obtaining a target pseudo gold standard image according to the summation result.

[0106] It should be noted that, before the above step S41, the method further includes: obtaining the weight value of the similarity coefficient, the weight value of the mutual information, and a preset weighting function; obtaining a preset weighting formula according to the weight value of the similarity coefficient, the weight value of the mutual information, and the preset weighting function;

[0107] Accordingly, the above step S41 includes: performing weighted summation on the mutual information and the similarity coefficient using the preset weighting formula, and obtaining a target pseudo gold standard image according to the summation result;

[0108] Wherein, the preset weighting formula is:

[0109] PG=argmax i {λ1*DICE(P i ,CF)+λ2*MI i (F,MF i )},

[0110] Where PG is the summation result, argmax i is the preset weighting function, λ1 is the weight value of the similarity coefficient, and λ2 is the weight value of the mutual information.

[0111] It is understandable that the weight value λ1 of the similarity coefficient and the weight value λ2 of the mutual information can be set according to actual conditions.

[0112] It should be emphasized that, because 1≤i≤n, there will be n results of weighted summation. The above device will select the summation with the largest value from the n weighted summation results and record it as PG. Then, the value of i at this time is determined according to PG, and the corresponding pseudo gold standard image P is determined at this time. i , which further represents the pseudo gold standard image P at this time i It can be used as the best standard image to judge the quality of the segmented contour image.

[0113] In a specific implementation, the above-mentioned device will obtain the weight value of the similarity coefficient, the weight value of the mutual information and the preset weighting function, and calculate the similarity coefficient and mutual information obtained previously through a preset weighting formula to obtain the summation result, and select the pseudo-gold standard image corresponding to the summation result with the largest value as the target pseudo-gold standard image.

[0114] Step S42: obtaining a target similarity coefficient according to the target pseudo-gold standard image and the heart image to be screened.

[0115] It should be noted that, since the intermediate layer image can represent the quality of the entire segmented contour image obtained through the three-dimensional active shape model, before the above-mentioned step S42, it also includes: obtaining the intermediate layer image of the heart image to be screened through the three-dimensional active model; accordingly, the above-mentioned step S42: includes: obtaining the target similarity coefficient based on the target pseudo-gold standard image and the intermediate layer image.

[0116] It should be understood that the target similarity coefficient can be recorded as DICE(PG, X), and the calculation formula is:

[0117]

[0118] Wherein, X is the heart image to be screened, and the target similarity coefficient DICE(PG, X) can be obtained by the above calculation formula.

[0119] In a specific implementation, the above-mentioned device can obtain the target similarity coefficient between the target pseudo-gold standard image and the heart image to be screened through the above-mentioned calculation formula.

[0120] Step S43: screening the cardiac images to be screened according to the target similarity coefficient to obtain a target cardiac image.

[0121] It should be emphasized that the screening criterion may be that when the target similarity coefficient DICE(PG, X) is greater than a preset threshold, it can be determined that the cardiac image X to be screened corresponding to the target similarity coefficient DICE(PG, X) meets the user's requirements, and the cardiac image X to be screened that meets the requirements is used as the target cardiac image; when the target similarity coefficient is less than the preset threshold, it can be determined that the cardiac image X to be screened corresponding to the target similarity coefficient DICE(PG, X) does not meet the user's requirements, and the cardiac image X to be screened that does not meet the requirements can be deleted.

[0122] It should be noted that, in this embodiment, the preset threshold may be 0.7 or other values.

[0123] In a specific implementation, the above-mentioned device can determine whether the obtained target similarity coefficient is greater than a preset threshold. If it is greater, it can be determined that the heart image to be screened corresponding to the target similarity coefficient meets the requirements, and the heart image to be screened that meets the requirements can be used as the target heart image.

[0124] In this embodiment, the above-mentioned device will obtain the weight value of the similarity coefficient, the weight value of the mutual information and the preset weighting function, and calculate the similarity coefficient and mutual information obtained previously through a preset weighting formula to obtain a summation result, and select the pseudo-gold standard image corresponding to the summation result with the largest value as the target pseudo-gold standard image; obtain the target similarity coefficient between the target pseudo-gold standard image and the heart image to be screened through the above-mentioned calculation formula; judge whether it is greater than the preset threshold value according to the obtained target similarity coefficient. If it is greater than, it can be determined that the heart image to be screened corresponding to the target similarity coefficient meets the requirements, and the heart image to be screened that meets the requirements is used as the target heart image; this embodiment can select the best image for judging the quality of the segmented contour image from each pseudo-gold standard image, thereby improving the accuracy of screening, and at the same time obtain the target similarity coefficient through the intermediate layer image in the heart image to be screened for screening, further ensuring the accuracy of screening.

[0125] In addition, an embodiment of the present invention further proposes a storage medium storing a three-dimensional heart segmentation image screening program. When the three-dimensional heart segmentation image screening program is executed by a processor, the steps of the three-dimensional heart segmentation image screening method described above are implemented.

[0126] In addition, refer to Figure 5 , Figure 5 This is a structural block diagram of a first embodiment of a three-dimensional heart segmentation image screening device according to the present invention. This embodiment of the present invention further provides a three-dimensional heart segmentation image screening device, comprising:

[0127] a mask processing module 501 for performing mask processing on a fixed image to obtain a floating image with a gold standard image, wherein the fixed image is a sample heart image;

[0128] a deformation processing module 502 for performing deformation processing on the floating image to obtain a registration image, and performing deformation processing on the gold standard image in the floating image to obtain a pseudo gold standard image;

[0129] An information acquisition module 503 is configured to obtain mutual information based on the fixed image and the registered image, and obtain a similarity coefficient based on the fixed image and the pseudo-gold standard image;

[0130] The image screening module 504 is configured to screen the cardiac images to be screened based on the mutual information and the similarity coefficient to obtain a target cardiac image.

[0131] The above-mentioned device in this embodiment can perform mask processing on a certain number of sample cardiac images, and use the processed images as floating images; train the floating images and fixed images through a registration network model to obtain a target deformation field, and then deform the above-mentioned floating image based on the target deformation field to obtain a registered image, and deform the above-mentioned gold standard image based on the target deformation field to obtain a pseudo-gold standard image; then obtain the amount of information contained in the fixed image in the registered image as mutual information, and obtain a similarity measurement index between the fixed image and the pseudo-gold standard image as a similarity coefficient, and screen the intermediate layer images in the slice image based on the mutual information and the similarity coefficient, and use the image that meets the user's needs as the target cardiac image; since this embodiment first uses the mutual information and the similarity coefficient as standards to screen the cardiac images to be screened, compared with the existing manual screening, the screening efficiency can be improved. At the same time, manual screening may have errors, thereby improving the accuracy and reliability of screening.

[0132] Other embodiments or specific implementations of the three-dimensional heart segmentation image screening device of the present invention can refer to the above-mentioned method embodiments and will not be described in detail here.

[0133] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or system comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or system. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or system comprising the element.

[0134] The serial numbers of the above embodiments of the present invention are for description only and do not represent the advantages or disadvantages of the embodiments.

[0135] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better embodiment. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as read-only memory / random access memory, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present invention.

[0136] The above are only preferred embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A three-dimensional heart segmentation image screening method, characterized in that: The method comprises the following steps: Performing mask processing on the fixed image to obtain a floating image with a gold standard image, wherein the fixed image is a sample heart image, and the floating images are multiple, each of which corresponds to one gold standard image; Performing deformation processing on the floating image to obtain a registration image, and performing deformation processing on the gold standard image in the floating image to obtain a pseudo gold standard image; Obtaining mutual information based on the fixed image and the registered image, and obtaining a similarity coefficient based on the fixed image and the pseudo gold standard image; screening the cardiac images to be screened based on the mutual information and the similarity coefficient to obtain a target cardiac image; The step of screening the cardiac images to be screened based on the mutual information and the similarity coefficient to obtain a target cardiac image comprises: Performing a weighted summation on the mutual information and the similarity coefficient, and obtaining a target pseudo-gold standard image according to the summation result, wherein the target pseudo-gold standard image is the pseudo-gold standard image corresponding to the summation result with the largest value; obtaining a target similarity coefficient according to the target pseudo-gold standard image and the heart image to be screened; The heart image to be screened is screened according to the target similarity coefficient to obtain a target heart image.

2. The three-dimensional heart segmentation image screening method according to claim 1, wherein: Before the step of obtaining a target similarity coefficient based on the target pseudo gold standard image and the heart image to be screened, the method further includes: Obtain an intermediate layer image of the heart image to be screened through the three-dimensional active model; Accordingly, the step of obtaining a target similarity coefficient based on the target pseudo-gold standard image and the heart image to be screened includes: A target similarity coefficient is obtained according to the target pseudo gold standard image and the intermediate layer image.

3. The three-dimensional heart segmentation image screening method according to claim 2, wherein: Before the steps of obtaining mutual information based on the fixed image and the registered image, and obtaining a similarity coefficient based on the fixed image and the pseudo gold standard image, the method further includes: Segmenting the fixed image using a preset convolutional neural network model to obtain a segmented image; Accordingly, the step of obtaining mutual information based on the fixed image and the registered image, and obtaining a similarity coefficient based on the fixed image and the pseudo gold standard image includes: Mutual information is obtained based on the fixed image and the registered image, and a similarity coefficient is obtained based on the pseudo gold standard image and the segmented image.

4. The three-dimensional heart segmentation image screening method according to claim 3, wherein: The step of obtaining mutual information based on the fixed image and the registered image comprises: Obtaining fixed information entropy according to the fixed image, obtaining registration information entropy according to the registered image, and obtaining joint information entropy according to the fixed image and the registered image; Corresponding mutual information is obtained according to the fixed information entropy, the registration information entropy and the joint information entropy.

5. The three-dimensional heart segmentation image screening method according to any one of claims 1 to 4, characterized in that: Before the steps of deforming the floating image to obtain a registered image and deforming the gold standard image in the floating image to obtain a pseudo gold standard image, the method further includes: Training the floating image and the fixed image through a registration network model to obtain a target deformation field; Accordingly, the steps of deforming the floating image to obtain a registered image, and deforming the gold standard image in the floating image to obtain a pseudo gold standard image include: Performing deformation processing on the floating image based on the target deformation field to obtain a registered image; The gold standard image in the floating image is deformed based on the target deformation field to obtain a pseudo gold standard image.

6. The three-dimensional heart segmentation image screening method according to claim 5, wherein: Before the step of performing weighted summation of the mutual information and the similarity coefficient and obtaining the target pseudo gold standard image according to the summation result, the method further includes: Obtaining a weight value of the similarity coefficient, a weight value of the mutual information, and a preset weighting function; Obtaining a preset weighting formula according to the weight value of the similarity coefficient, the weight value of the mutual information and a preset weighting function; Accordingly, the step of performing weighted summation on the mutual information and the similarity coefficient and obtaining a target pseudo gold standard image according to the summation result includes: Performing weighted summation on the mutual information and the similarity coefficient using the preset weighting formula, and obtaining a target pseudo gold standard image according to the summation result; Wherein, the preset weighting formula is: PG=argmax i {λ1*DICE(P i ,CF)+λ2*MI i (F,MF i )}, Where PG is the summation result, argmax i is the preset weighting function, λ1 is the weight value of the similarity coefficient, DICE(P i ,CF) is the similarity coefficient, λ2 is the weight value of the mutual information, MI i (F,MF i ) is the mutual information, P i is the pseudo gold standard image, CF is the segmented image, F is the fixed image, MF i is the registered image.

7. A three-dimensional cardiac segmentation image screening device, characterized in that: The device comprises: a mask processing module, configured to perform mask processing on a fixed image to obtain a floating image with a gold standard image, wherein the fixed image is a sample heart image, and the floating images are multiple, each of which corresponds to one gold standard image; a deformation processing module, configured to perform deformation processing on the floating image to obtain a registration image, and perform deformation processing on the gold standard image in the floating image to obtain a pseudo gold standard image; an information acquisition module, configured to obtain mutual information based on the fixed image and the registered image, and obtain a similarity coefficient based on the fixed image and the pseudo-gold standard image; An image screening module, configured to screen the cardiac images to be screened based on the mutual information and the similarity coefficient to obtain a target cardiac image; The image screening module is further configured to perform a weighted summation on the mutual information and the similarity coefficient, and obtain a target pseudo-gold standard image based on the summation result, wherein the target pseudo-gold standard image is the pseudo-gold standard image corresponding to the summation result with the largest value; obtain a target similarity coefficient based on the target pseudo-gold standard image and the heart image to be screened; and screen the heart image to be screened based on the target similarity coefficient to obtain a target heart image.

8. A three-dimensional cardiac segmentation image screening device, characterized in that: The device includes: a memory, a processor, and a three-dimensional heart segmentation image screening program stored in the memory and executable on the processor, wherein the three-dimensional heart segmentation image screening program is configured to implement the steps of the three-dimensional heart segmentation image screening method according to any one of claims 1 to 6.

9. A storage medium, characterized in that: The storage medium stores a three-dimensional heart segmentation image screening program, which, when executed by a processor, implements the steps of the three-dimensional heart segmentation image screening method according to any one of claims 1 to 6.

Images