Window parameter determination method and device, computer equipment and storage medium

Abstract

The invention relates to a window parameter determination method and device, computer equipment and a storage medium. The method comprises the steps that a three-dimensional target modal image is obtained, the target modal image comprises a plurality of continuous two-dimensional images, a gray level distribution vector corresponding to the target modal image is obtained according to the gray level value of each pixel point of each two-dimensional image in the target modal image, and the target modal image is obtained according to the gray level distribution vector and a machine learning model; and determining a window parameter corresponding to the target modal image. By adopting the method, the window parameters of the three-dimensional image can be determined.

Description

Technical Field

[0001] The present application relates to the field of medical technology, and in particular to a method, apparatus, computer equipment and storage medium for determining window parameters. Background Art

[0002] With the development of medical data visualization, more and more application scenarios require viewing rendered three-dimensional images, so as to more intuitively observe various tissue parts of the patient to analyze the disease, and appropriate window parameters can better display the rendered three-dimensional images. Among them, the window parameters include window level and / or window width.

[0003] In some related technologies, users often need to manually adjust window parameters based on experience, which is cumbersome. Moreover, compared with the grayscale value range of computed tomography (CT) images, the grayscale value distribution range of magnetic resonance imaging (MRI) images is wider, so it takes a lot of time to manually adjust the window parameters to the appropriate ones.

[0004] At present, some related technologies have also emerged to determine the window parameters of two-dimensional images. For example, the window position can be determined by statistically analyzing the grayscale histogram of the medical image and finding the peak, and then the window width can be determined along the cumulative distribution on both sides of the peak. This method works better for CT images, but for MR images, due to the wider grayscale distribution and more noise grayscale values, the effect of using grayscale histogram to determine the window parameters is not good.

[0005] In summary, the current method of determining window parameters is for two-dimensional images and is not applicable to three-dimensional images. Therefore, how to propose a method for determining window parameters suitable for three-dimensional images is a key research topic for researchers in this field. Summary of the invention

[0006] Based on this, it is necessary to provide a method, device, computer equipment and storage medium for determining window parameters that can be adapted to three-dimensional images in order to solve the above technical problems.

[0007] In a first aspect, the present application provides a method for determining window parameters, comprising:

[0008] Acquire a three-dimensional target modality image; the target modality image includes a plurality of continuous two-dimensional images;

[0009] According to the grayscale value of each pixel point of each two-dimensional image in the target modality image, a grayscale distribution vector corresponding to the target modality image is obtained;

[0010] According to the grayscale distribution vector and the machine learning model, the window parameters corresponding to the target modality image are determined.

[0011] In one embodiment, obtaining the grayscale distribution vector corresponding to the target modality image according to the grayscale value of each pixel point of each two-dimensional image in the target modality image includes:

[0012] According to the gray value of each pixel in each two-dimensional image, the gray distribution information corresponding to the target modality image is obtained;

[0013] According to the grayscale distribution information corresponding to the target modality image, the grayscale distribution vector is determined.

[0014] In one embodiment, the grayscale distribution information corresponding to the target modality image is obtained according to the grayscale value of each pixel in each two-dimensional image, including:

[0015] Determine the mapping grayscale value corresponding to each pixel point in each two-dimensional image according to the grayscale value of each pixel point in each two-dimensional image and the type of the target modality image;

[0016] The mapping grayscale value corresponding to each pixel in each two-dimensional image is counted to obtain the grayscale distribution information.

[0017] In one embodiment, determining the mapping grayscale value corresponding to each pixel in each two-dimensional image according to the grayscale value of each pixel in each two-dimensional image and the type of the target modality image includes:

[0018] If the type is a preset type, the grayscale value of each pixel in the two-dimensional image is used as the mapping grayscale value corresponding to each pixel in the two-dimensional image;

[0019] If the type is not a preset type, a mapping grayscale value corresponding to each pixel in the two-dimensional image is determined according to the mapping relationship and the grayscale value of each pixel in the two-dimensional image.

[0020] In one embodiment, the machine learning model includes a recognition sub-model and a prediction sub-model; determining the window parameters corresponding to the target modality image according to the grayscale distribution vector and the machine learning model includes:

[0021] According to the grayscale distribution vector and the recognition sub-model, a tissue probability vector corresponding to the target modality image is obtained; the tissue probability vector is used to indicate the probability that the tissue in the target modality image belongs to each preset tissue;

[0022] The window parameters corresponding to the target modality image are determined according to the tissue probability vector and the prediction sub-model.

[0023] In one embodiment, the window parameters are obtained according to the tissue probability vector and the prediction sub-model, including:

[0024] The window parameters are obtained according to the tissue probability vector, the grayscale distribution vector and the prediction sub-model in the machine learning model.

[0025] In one embodiment, the window parameters are obtained according to the tissue probability vector, the grayscale distribution vector and the prediction sub-model, including:

[0026] The tissue probability vector and the grayscale distribution vector are concatenated to obtain a concatenated vector;

[0027] The window parameters are obtained according to the concatenated vector and the prediction sub-model.

[0028] In one embodiment, the window parameters are obtained according to the splicing vector and the prediction sub-model, including:

[0029] Input the concatenated vector into the prediction sub-model to obtain the initial window parameters;

[0030] When the type of the target modality image is a preset type, the initial window parameters are used as window parameters; or, when the type of the target modality image is not a preset type, the window parameters are determined according to the initial window parameters and the mapping relationship.

[0031] In one embodiment, the target modality image is a reconstructed computed tomography image or a magnetic resonance image.

[0032] In one embodiment, the method further comprises:

[0033] Acquire a plurality of three-dimensional sample modal images; the sample modal images include a plurality of continuous two-dimensional images;

[0034] According to the grayscale value of each pixel point of each two-dimensional image in each sample modality image, a grayscale distribution vector sample corresponding to each sample modality image is obtained;

[0035] According to each grayscale distribution vector sample and the window parameter label corresponding to each grayscale distribution vector sample, machine learning training is performed on the initial model to obtain a machine learning model.

[0036] In one embodiment, according to each grayscale distribution vector sample and the window parameter label corresponding to each grayscale distribution vector sample, the initial model is subjected to machine learning training to obtain a machine learning model, including:

[0037] Determine a first loss according to each grayscale distribution vector sample, a tissue probability vector label corresponding to each grayscale distribution vector sample, and a first sub-model in the initial model;

[0038] Determine a second loss according to each grayscale distribution vector sample, a tissue probability vector label corresponding to each grayscale distribution vector sample, a window parameter label corresponding to each grayscale distribution vector sample, and a second sub-model in the initial model;

[0039] According to the first loss and the second loss, the initial model is trained by machine learning to obtain a machine learning model.

[0040] In one embodiment, the method further comprises:

[0041] Performing three-dimensional rendering on the target modality image to obtain a rendering result of the target modality image;

[0042] Displays rendering results based on window parameters.

[0043] In one embodiment, the method further comprises:

[0044] In response to an adjustment operation based on the rendering result, the window parameters are adjusted.

[0045] In a second aspect, the present application further provides a window parameter determination device, comprising:

[0046] A first acquisition module, used for acquiring a three-dimensional target modality image;

[0047] A first determination module is used to obtain a grayscale distribution vector corresponding to the target modality image according to the grayscale value of each pixel point of each two-dimensional image in the target modality image;

[0048] The second determination module is used to determine the window parameters corresponding to the target modality image according to the grayscale distribution vector and the machine learning model.

[0049] In a third aspect, the present application further provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any of the above methods when executing the computer program.

[0050] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of any of the above methods when executed by a processor.

[0051] In a fifth aspect, the present application also provides a computer program product, including a computer program, which implements the steps of any of the above methods when executed by a processor.

[0052] The above-mentioned window parameter determination method, device, computer equipment and storage medium first obtain a three-dimensional target modality image, which includes multiple continuous two-dimensional images. Since the grayscale distribution vector corresponding to the three-dimensional target modality image can be obtained according to the grayscale value of each pixel point of each two-dimensional image in the target modality image, the grayscale distribution vector takes into account each grayscale value of the continuous two-dimensional image in the target modality image. In this way, the window parameters corresponding to the target modality image can be efficiently determined according to the grayscale distribution vector and the machine learning model. On the one hand, there is no need to manually determine the window parameters based on experience in the whole process. On the other hand, the method provided by this embodiment can be applied not only to the target modality image of a single two-dimensional image, but also to the three-dimensional target modality image composed of multiple two-dimensional images, so as to accurately and efficiently determine the window parameters applicable to the three-dimensional target modality image. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the drawings required for use in the embodiments or the related technical descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0054] FIG1 is a diagram showing an application environment of a method for determining window parameters in an embodiment of the present application;

[0055] FIG2 is a flow chart of a method for determining window parameters in an embodiment of the present application;

[0056] Figure 3 A schematic diagram of a process for determining a grayscale distribution vector in an embodiment of the present application;

[0057] Figure 4 This is a schematic diagram of a process for obtaining grayscale distribution information in an embodiment of the present application;

[0058] Figure 5 This is a schematic diagram of a process for determining a mapping grayscale value in an embodiment of the present application;

[0059] Figure 6 A schematic diagram of a process for determining window parameters in an embodiment of the present application;

[0060] Figure 7 A schematic diagram of an organization probability vector in an embodiment of the present application;

[0061] Figure 8 This is a schematic diagram of another process for determining window parameters in an embodiment of the present application;

[0062] Fig. 9 This is a schematic diagram of the architecture of a prediction sub-model in an embodiment of the present application;

[0063] Fig.10 This is a schematic diagram of another process for determining window parameters in an embodiment of the present application;

[0064] Fig.11 A schematic diagram of a process for obtaining a machine learning model in an embodiment of the present application;

[0065] Fig.12 This is another flow chart of obtaining a machine learning model in an embodiment of the present application;

[0066] Fig.13 A schematic diagram of an organization probability vector label in an embodiment of the present application;

[0067] Fig.14 This is a schematic diagram of a process for displaying rendering results in an embodiment of the present application;

[0068] Fig.15 A schematic diagram of a process of determining a window parameter in an embodiment of the present application;

[0069] Fig.16 This is a workflow diagram of a method for determining window parameters in an embodiment of the present application;

[0070] Fig.17 This is a data flow diagram of a method for determining window parameters in an embodiment of the present application;

[0071] Fig.18 4 is a structural block diagram of a device for determining window parameters in an embodiment of the present application. DETAILED DESCRIPTION

[0072] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0073] In the process of medical image visualization, since the grayscale value range of medical images is large, and the grayscale value range that ordinary displays can display is limited, it is difficult to display all the data details in the medical images. Therefore, window parameters are usually used to define the grayscale value area of ​​interest, so as to map the grayscale value area of ​​interest to the grayscale value range that the display can display. Based on this problem, it is necessary to propose a window parameter determination method suitable for three-dimensional images. The following will introduce this method.

[0074] FIG. 1 is an application environment diagram of the window parameter determination method in the embodiment of the present application. In an exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be shown in FIG. 1. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store relevant data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for determining window parameters is implemented.

[0075] This embodiment uses the method applied to a server as an example for illustration. It is understandable that the method can also be applied to a terminal, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. The terminal can be, but is not limited to, various personal computers, laptops, smart phones, and tablet computers. The server can be implemented as an independent server or a server cluster consisting of multiple servers.

[0076] Those skilled in the art will understand that the structure shown in FIG. 1 is merely a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0077] FIG2 is a flow chart of a method for determining window parameters in an embodiment of the present application. In an exemplary embodiment, as shown in FIG2 , a method for determining window parameters is provided, which is described by taking the method applied to the computer device in FIG1 as an example, and includes the following S201 to S203.

[0078] S201, acquiring a three-dimensional target modality image; the target modality image includes a plurality of continuous two-dimensional images.

[0079] In this embodiment, the target modality image may include but is not limited to CT images, MR images, positron emission computed tomography (PET) images, etc. The target modality image is a three-dimensional image, and the target modality image includes a plurality of continuous two-dimensional images.

[0080] Each two-dimensional image in the target modality image may belong to the same sequence, for example, Digital Imaging and Communications in Medicine (DICOM) sequence data under the same sequence. That is, each two-dimensional image in the target modality image may be a continuous image obtained using the same scanning protocol.

[0081] Optionally, the computer device may obtain the three-dimensional target modality image from a picture archiving and communication system (PACS), or may obtain the three-dimensional target modality image in response to a user input, but the present embodiment is not limited thereto.

[0082] In an exemplary embodiment, optionally, the target modality image is a reconstructed computed tomography image or magnetic resonance image. In other words, the target modality image is a 3D reconstructed image.

[0083] S202, obtaining a grayscale distribution vector corresponding to the target modality image according to the grayscale value of each pixel point of each two-dimensional image in the target modality image.

[0084] In this embodiment, each two-dimensional image includes a plurality of pixels, and each pixel corresponds to a grayscale value. Therefore, after acquiring the target modality image, the computer device can obtain the grayscale distribution vector corresponding to the target modality image according to the grayscale value of each pixel of each two-dimensional image in the target modality image. It can be understood that each pixel of each two-dimensional image is also each voxel in the three-dimensional target modality image.

[0085] Further optionally, the sizes of the two-dimensional images in the target modality image may be the same, that is, each two-dimensional image may include the same number of pixels.

[0086] The grayscale vector distribution may include N grayscale value elements, each grayscale value element corresponds to a ratio of pixels corresponding to the grayscale value to all pixels, and the sum of the ratios may be equal to 1. N is an integer greater than 1, and N may be determined according to the range of grayscale values ​​in the target modality image. For example, the range of grayscale values ​​in a CT image is 0 to 4095, that is, the grayscale values ​​in the CT image are 4096, and N may be equal to 4096.

[0087] For example, assuming that the target modality image includes two-dimensional image A, two-dimensional image B, and two-dimensional image C, and each of the above two-dimensional images includes 100 pixels, the computer device can count the number of pixels A1 corresponding to the grayscale value 1 in two-dimensional image A, the number of pixels B1 corresponding to the grayscale value 1 in two-dimensional image B, and the number C1 of pixels corresponding to the grayscale value 1 in two-dimensional image C, and use (A1+B1+C1) / 300 as the ratio of pixels with a grayscale value of 1, that is, as the grayscale value element corresponding to the grayscale value 1. By analogy, the computer device can obtain the grayscale distribution vector corresponding to the target modality image.

[0088] In some embodiments, the computer device may also perform filtering and other processing on the target modality image, and then obtain a grayscale distribution vector based on the grayscale value of each pixel point of each two-dimensional image in the processed target modality image.

[0089] S203, determining the window parameters corresponding to the target modality image according to the grayscale distribution vector and the machine learning model.

[0090] The window parameters include at least one of a window level and a window width. The computer device may also determine the window parameters corresponding to the target modality image according to the grayscale distribution vector and the machine learning model.

[0091] Among them, the machine learning model can be a deep learning model. Specifically, it can include supervised learning models, semi-supervised learning models, unsupervised learning models, etc. For example, the machine learning model can include but is not limited to a convolutional neural network (CNN) model, a recurrent neural network (RNN), a fully convolutional neural network (FCN) model, a generative adversarial network (GAN) model, a back-propagation (BP) machine learning model, a radial basis function (RBF) model, a deep belief network (DBN) model, an Elman model, or at least one of their combined models.

[0092] Optionally, the computer device may directly input the grayscale distribution vector into the machine learning model to obtain the window parameters output by the machine learning model. The computer device may also preprocess the grayscale distribution vector and input the preprocessed grayscale distribution vector into the machine learning model to obtain the window parameters output by the machine learning model. This embodiment does not limit this. Preprocessing may include but is not limited to outlier removal processing, fixed length processing, etc.

[0093] It is understandable that, since the machine learning model has the ability to determine the window parameters corresponding to the target modality image according to the grayscale distribution vector, it is usually necessary to train the machine learning model. Exemplarily, the computer device can obtain multiple three-dimensional sample modality images, and obtain the grayscale distribution vector samples corresponding to each sample modality image, and then input the grayscale distribution vector samples into the initial model to obtain the window parameters predicted by the initial model, and stop training to obtain the machine learning model when the accuracy between the window parameter labels corresponding to each grayscale distribution vector sample and the window parameters predicted by the initial model reaches a preset accuracy.

[0094] In the above-mentioned window parameter determination method, a three-dimensional target modality image is first obtained, and the target modality image includes multiple continuous two-dimensional images. Since the grayscale distribution vector corresponding to the three-dimensional target modality image can be obtained according to the grayscale value of each pixel point of each two-dimensional image in the target modality image, the grayscale distribution vector takes into account each grayscale value of the continuous two-dimensional image in the target modality image. In this way, the window parameters corresponding to the target modality image can be efficiently determined according to the grayscale distribution vector and the machine learning model. On the one hand, there is no need to manually determine the window parameters based on experience in the whole process. On the other hand, the method provided by this embodiment can be applied not only to the target modality image of a single two-dimensional image, but also to the three-dimensional target modality image composed of multiple two-dimensional images, so as to accurately and efficiently determine the window parameters applicable to the three-dimensional target modality image.

[0095] Figure 3 FIG. 1 is a flow chart of determining a grayscale distribution vector in an embodiment of the present application. In an exemplary embodiment, Figure 3 As shown, S202 includes S301 to S302.

[0096] S301, obtaining grayscale distribution information corresponding to the target modality image according to the grayscale value of each pixel in each two-dimensional image.

[0097] In this embodiment, the grayscale distribution information is used to indicate the distribution of each grayscale value in the target modality image. The grayscale distribution information includes but is not limited to a grayscale histogram, which can be used to indicate the number of pixels corresponding to different grayscale values ​​in the target modality image.

[0098] Optionally, the computer device may directly count the number of pixels corresponding to each grayscale value in all two-dimensional images of the target modality image to obtain grayscale distribution information corresponding to the target modality image.

[0099] S302, determining a grayscale distribution vector according to grayscale distribution information corresponding to the target modality image.

[0100] Optionally, the computer device may perform normalization processing on the grayscale distribution information to obtain a grayscale distribution vector. In some embodiments, the computer device may also perform normalization processing on the grayscale distribution information after performing outlier elimination processing to obtain a grayscale distribution vector.

[0101] In the above embodiment, since the grayscale distribution information corresponding to the target modality image can be obtained according to the grayscale value of each pixel in each two-dimensional image, and the grayscale distribution vector can be determined according to the grayscale distribution information corresponding to the target modality image, the grayscale distribution vector takes into account the grayscale distribution information of each pixel in each two-dimensional image of the target modality image, thereby improving the accuracy of the grayscale distribution vector.

[0102] Figure 4 FIG. 1 is a flow chart of obtaining grayscale distribution information in an embodiment of the present application. In an exemplary embodiment, Figure 4 As shown, S301 includes S401 to S402.

[0103] S401 , determining a mapping grayscale value corresponding to each pixel in each two-dimensional image according to the grayscale value of each pixel in each two-dimensional image and the type of the target modality image.

[0104] In this embodiment, the type of the target modality image may be used to indicate the modality type of the target modality image. Exemplarily, the type of the target modality image may include but is not limited to a CT image type, an MR image type, a PET image type, and the like.

[0105] Since different types of target modality images correspond to different ranges of grayscale values, and in order to make the window parameter determination method applicable to various types of target modality images and unify the grayscale distribution vector, this embodiment determines the mapping grayscale value corresponding to each pixel in each two-dimensional image according to the grayscale value of each pixel in each two-dimensional image and the type of the target modality image.

[0106] The mapped grayscale value is the grayscale value obtained by mapping the grayscale value of each pixel in the two-dimensional image within a certain preset range. For example, the computer device can determine the mapping relationship corresponding to each type according to the preset range and the range of grayscale values ​​corresponding to the type of each target modality image, and then determine the mapped grayscale value according to the mapping relationship corresponding to each type and the grayscale value of each pixel in the corresponding two-dimensional image.

[0107] It should be noted that the above mapping relationship may be a linear or non-linear mapping relationship.

[0108] S402, counting the mapping grayscale value corresponding to each pixel point in each two-dimensional image to obtain grayscale distribution information.

[0109] Optionally, the computer device can count the number of pixels corresponding to each mapped grayscale value in all two-dimensional images to obtain grayscale distribution information corresponding to the target modality image. In some embodiments, the computer device can also determine the grayscale distribution information corresponding to the target modality image by machine learning.

[0110] Furthermore, each element in the grayscale distribution vector can be used to indicate the proportion of pixels corresponding to each mapped grayscale value to all pixels. , To map grayscale values The corresponding pixel points account for the proportion of all pixel points, and so on. To map grayscale values The corresponding pixel points account for the proportion of all pixels. .

[0111] In the above embodiment, since it is possible to determine the mapping grayscale value corresponding to each pixel in each two-dimensional image according to the grayscale value of each pixel in each two-dimensional image and the type of the target modality image, and to count the mapping grayscale value corresponding to each pixel in each two-dimensional image, the grayscale distribution information is obtained. Therefore, it is possible to unify target modality images with different grayscale value ranges, that is, to unify the range of mapping grayscale values ​​of target modality images under different modalities, which is conducive to unifying the format of grayscale distribution vectors. Furthermore, in the case of using a machine learning model, it is possible to unify the input of the machine learning model, and also enable the machine learning model to be applicable to target modality images under various modalities.

[0112] Figure 5 FIG. 1 is a flow chart of determining a mapping gray value in an embodiment of the present application. In an exemplary embodiment, Figure 5 As shown, S401 includes S501 to S502.

[0113] S501: If the type is a preset type, the grayscale value of each pixel in the two-dimensional image is used as the mapping grayscale value corresponding to each pixel in the two-dimensional image.

[0114] In this embodiment, the preset type may be one of the modality types. Taking the preset type as the CT image type as an example, if the type of the target modality image is the CT image type, that is, the target modality image is the CT image, then the computer device may directly use the grayscale value of each pixel in the two-dimensional image as the mapping grayscale value corresponding to each pixel in the two-dimensional image.

[0115] For example, assuming that the grayscale value of pixel 1 of the two-dimensional image A is grayscale value 1, the mapped grayscale value of pixel 1 is also grayscale value 1, and the same is true for other pixels, which will not be repeated here.

[0116] S502: If the type is not a preset type, determining a mapping grayscale value corresponding to each pixel in the two-dimensional image according to the mapping relationship and the grayscale value of each pixel in the two-dimensional image.

[0117] Continuing with the example of the preset type being the CT image type, if the type of the target modality image is not the CT image type, that is, the target modality image is not a CT image, the computer device will determine the mapping grayscale value corresponding to each pixel in each two-dimensional image according to the mapping relationship and the grayscale value of each pixel in the two-dimensional image. Optionally, the computer device can use the product or quotient between the mapping relationship and the grayscale value of each pixel in the two-dimensional image as the mapping grayscale value corresponding to each pixel in the two-dimensional image.

[0118] Among them, the mapping relationship can be determined according to the range of grayscale values ​​corresponding to the preset type and the range of grayscale values ​​corresponding to the target modality image. For example, the quotient between the range of grayscale values ​​corresponding to the preset type and the range of grayscale values ​​corresponding to the target modality image can be used as the mapping relationship.

[0119] The mapping relationship may include but is not limited to a proportional coefficient. For example, assuming that the two-dimensional image A is an MR image, and the grayscale value of pixel 1 of the two-dimensional image A is grayscale value 1, the mapped grayscale value of pixel 1 may be equal to the product obtained by multiplying the proportional coefficient by the grayscale value 1, and the same is true for other pixels, which will not be described here. In this way, the range of grayscale values ​​corresponding to the target modality image is linearly mapped to the range of grayscale values ​​corresponding to the preset type.

[0120] It should be noted that the above takes linear mapping as an example. In some embodiments, nonlinear mapping and other methods may also be used to obtain the mapping grayscale value corresponding to each pixel point in the two-dimensional image.

[0121] In some embodiments, the computer device may also learn the above mapping process through a machine learning model. For example, the model is trained according to the modality label information and the mapping grayscale value label of the sample modality image, so that after acquiring the target modality image, the mapping grayscale value corresponding to each pixel in the two-dimensional image of the target modality image can be obtained according to the target modality image and the model.

[0122] In the above embodiment, when the type of the target modality image is a preset type, the grayscale value of each pixel in the two-dimensional image is used as the mapping grayscale value corresponding to each pixel in the two-dimensional image, and when the type of the target modality image is not a preset type, the mapping grayscale value corresponding to each pixel in each two-dimensional image is determined according to the mapping relationship and the grayscale value of each pixel in the two-dimensional image. Therefore, the mapping grayscale value corresponding to each pixel in each two-dimensional image can be determined according to the grayscale value of each pixel in each two-dimensional image and the type of the target modality image, so that the machine learning model can be applicable to target modality images under various modalities.

[0123] Figure 6 FIG. 1 is a flow chart of determining window parameters in an embodiment of the present application. In an exemplary embodiment, as shown in FIG. Figure 6 As shown, S203 includes S601 to S602.

[0124] S601, obtaining a tissue probability vector corresponding to the target modality image according to the grayscale distribution vector and the recognition sub-model; the tissue probability vector is used to indicate the probability that the tissue in the target modality image belongs to each preset tissue.

[0125] In this embodiment, the machine learning model includes an identification sub-model and a prediction sub-model. The identification sub-model is used to determine the tissue probability vector corresponding to the target modality image according to the grayscale distribution vector, and the prediction sub-model is used to determine the window parameters according to the tissue probability vector. The types of the identification sub-model and the prediction sub-model can be different. For example, the identification sub-model can be a residual network (Residual Network, ResNet) model for classification, and the prediction sub-model can be a CNN model.

[0126] Optionally, the computer device may directly input the grayscale distribution vector into the recognition sub-model to obtain the tissue probability vector output by the prediction sub-model. The computer device may also pre-process the grayscale distribution vector and input the pre-processed grayscale distribution vector into the recognition sub-model to obtain the tissue probability vector output by the prediction sub-model.

[0127] Among them, the organizational probability vector It is used to indicate the probability that the tissue in the target modality image belongs to each preset tissue. K is the number of each preset organization, which is an integer greater than 1, that is, each preset organization includes preset organization 1 to preset organization K. Indicates the probability that the tissue in the target modality image belongs to the preset tissue 1, and so on. Indicates the probability that the tissue in the target modality image belongs to the preset tissue K. It can be understood that the higher the probability, the higher the possibility that the tissue in the target modality image belongs to the preset tissue.

[0128] Figure 7 is a schematic diagram of an organization probability vector in an embodiment of the present application, such as Figure 7 As shown, K=9, =0.1, indicating that the probability that the tissue in the target modality image belongs to the heart is 0.1; =0.7, indicating that the probability that the tissue in the target modality image belongs to the lung is 0.7; =0.2, indicating that the probability that the tissue in the target modality image belongs to bone is 0.2, and the remaining elements in the tissue probability vector are 0.

[0129] S602, obtaining window parameters according to the tissue probability vector and the prediction sub-model.

[0130] In this embodiment, the computer device can input the tissue probability vector into the prediction sub-model, and the prediction sub-model outputs the window parameters corresponding to the target modality image, but this embodiment is not limited thereto. The prediction sub-model can be trained based on the tissue probability vector labels and window parameter labels corresponding to different sample modality images.

[0131] In an exemplary embodiment, the optional above-mentioned S603 may be implemented in the following manner:

[0132] The window parameters are obtained according to the tissue probability vector, the grayscale distribution vector and the prediction sub-model in the machine learning model.

[0133] Optionally, the computer device may input both the tissue probability vector and the grayscale distribution vector into the prediction sub-model to obtain the window parameters output by the prediction sub-model.

[0134] Further optionally, the computer device can train the recognition sub-model and the prediction sub-model separately to obtain a trained machine learning model. Exemplarily, when training the recognition sub-model, the computer device can train the first sub-model according to the grayscale distribution vector samples corresponding to each sample modality image and the tissue probability vector labels corresponding to the grayscale distribution vector samples, until the accuracy between the tissue probability vector predicted by the first sub-model and the tissue probability vector label reaches a preset accuracy, and then stop training to obtain the recognition sub-model. When training the prediction sub-model, the computer device can train the second sub-model according to the grayscale distribution vector samples corresponding to each sample modality image, the tissue probability vector labels corresponding to the grayscale distribution vector samples, and the corresponding window parameter labels, until the accuracy between the window parameters predicted by the second sub-model and the window parameter labels reaches a preset accuracy, and then stop training to obtain the prediction sub-model.

[0135] In the above embodiment, since the tissue probability vector corresponding to the target modality image can be obtained based on the grayscale distribution vector and the recognition sub-model, and the tissue probability vector is used to indicate the probability that the tissue in the target modality image belongs to each preset tissue, therefore, according to the tissue probability vector, the grayscale distribution vector and the prediction sub-model in the machine learning model, the window parameters suitable for the three-dimensional image can be obtained efficiently and accurately.

[0136] Figure 8 FIG. 1 is another flow chart of determining window parameters in an embodiment of the present application. In an exemplary embodiment, as shown in FIG. Figure 8 As shown, S603 includes S801 to S802.

[0137] S801, concatenate the tissue probability vector and the grayscale distribution vector to obtain a concatenated vector.

[0138] Continuing with the above example, the computer device may organize the probability vector and grayscale distribution vector Perform splicing to get the splicing vector That is, the concatenated vector includes M elements, M=K+N.

[0139] The above is organized into probability vectors In front, the grayscale distribution vector Taking the following splicing order as an example, in some embodiments, the tissue probability vector After that, the grayscale distribution vector The previous splicing order.

[0140] S802, obtaining window parameters according to the concatenated vector and the prediction sub-model.

[0141] Optionally, the computer device may directly input the concatenated vector into the prediction sub-model to obtain the window parameters output by the prediction sub-model. In some embodiments, the computer device may also perform feature extraction processing on the concatenated vector and input the result after the feature extraction processing into the prediction sub-model to obtain the window parameters output by the prediction sub-model.

[0142] Further optionally, when training the prediction sub-model, the computer device can splice the grayscale distribution vector samples and the tissue probability vector labels to obtain spliced ​​vector samples, and train the second sub-model based on the spliced ​​vector samples and the corresponding window parameter labels until the accuracy between the window parameters predicted by the second sub-model and the window parameter labels reaches a preset accuracy, and then stop training to obtain the prediction sub-model.

[0143] Fig. 9 Schematic diagram of the architecture of a prediction sub-model in an embodiment of the present application, such as Fig. 9 As shown, the prediction sub-model may include four convolutional layers and three fully connected layers, and the length of the splicing vector is K+N, with K=9. Taking N=4096 as an example, after the splicing vector is input into the prediction sub-model, the convolutional layers connected in sequence perform dimensionality reduction processing, firstly, the first convolutional layer processes the splicing vector to obtain feature A with a dimension of 4096, then the second convolutional layer processes feature A to obtain feature B with a dimension of 2048, then the third convolutional layer processes feature B to obtain feature C with a dimension of 2048, and finally the fourth convolutional layer processes feature C to obtain feature D with a dimension of 1024.

[0144] The fourth convolutional layer is connected to the first of the three fully connected layers. Each of the three fully connected layers is connected to the previous fully connected layer to further extract features. Furthermore, the first fully connected layer outputs feature E with a dimension of 4096 based on feature D, the second fully connected layer outputs feature F with a dimension of 2048 based on feature E, and the third fully connected layer outputs a window parameter with a final dimension of 2 based on feature F.

[0145] In the above embodiment, since the tissue probability vector and the grayscale distribution vector are spliced ​​to obtain a splicing vector, and the window parameters are obtained based on the splicing vector and the prediction sub-model, the process of first classifying the target modality image into tissue categories and then predicting parameters based on the grayscale distribution vector and the tissue probability vector obtained after tissue identification is implemented.

[0146] Fig.10 FIG. 1 is another flow chart of determining window parameters in an embodiment of the present application. In an exemplary embodiment, as shown in FIG. Fig.10 As shown, S802 includes S1001 to S1003.

[0147] S1001, input the concatenated vector into the prediction sub-model to obtain initial window parameters.

[0148] In this embodiment, the computer device inputs the splicing vector into the prediction sub-model to obtain the initial window parameters. It can be understood that the initial window parameters also include at least one of the window level and the window width.

[0149] S1002: When the type of the target modality image is a preset type, use the initial window parameters as window parameters.

[0150] Furthermore, since the grayscale value of each pixel in the two-dimensional image is used as the mapping grayscale value corresponding to each pixel in the two-dimensional image when the type of the target modality image is a preset type, the initial window parameters do not need to be reverse mapped, and therefore, the initial window parameters can be directly used as window parameters. Exemplarily, when the target modality image is a CT image, the computer device inputs the splicing vector corresponding to the target modality image into the prediction sub-model, and the window parameters of the target modality image can be directly obtained.

[0151] S1003: When the type of the target modality image is not a preset type, determine the window parameters according to the initial window parameters and the mapping relationship.

[0152] Since when the type of the target modality image is not the preset type, the mapping grayscale value corresponding to each pixel in each two-dimensional image will be determined according to the mapping relationship and the grayscale value of each pixel in the two-dimensional image, therefore, when the type of the target modality image is not the preset type, the window parameters will still be determined according to the initial window parameters and the mapping relationship, so as to reversely map the initial window parameters to obtain the window parameters.

[0153] For example, taking the target modality image as an MR image, assuming that the mapping grayscale value corresponding to the MR image is determined according to the product of the mapping relationship and the grayscale value of each pixel in the two-dimensional image, then after the splicing vector corresponding to the target modality image is input into the prediction sub-model to obtain the initial window parameters, the initial window parameters are divided by the mapping relationship to obtain the window parameters of the MR image.

[0154] In the above embodiment, the initial window parameters are obtained by inputting the splicing vector into the prediction sub-model, and when the type of the target modality image is the preset type, the initial window parameters are used as the window parameters, and when the type of the target modality image is not the preset type, the window parameters are determined according to the initial window parameters and the mapping relationship. In this way, accurate window parameters corresponding to the target modality image can be obtained.

[0155] The above introduces the use process of the window parameter determination method. It can be understood that before determining the window parameters, it is necessary to train a machine learning model. The following will introduce the training process of the machine learning model. Fig.11 FIG. 1 is a flow chart of obtaining a machine learning model in an embodiment of the present application. In an exemplary embodiment, Fig.11 As shown, the above-mentioned training method of the machine learning model also includes S1101 to S1103.

[0156] S1101, acquiring a plurality of three-dimensional sample modality images; the sample modality images include a plurality of continuous two-dimensional images.

[0157] In this embodiment, the sample modality image is a three-dimensional image. Similarly, the sample modality image includes a plurality of continuous two-dimensional images. Among them, the two-dimensional images in the sample modality image can be images in the same sequence. The principle of the sample modality image is similar to that of the target modality image, and will not be repeated here.

[0158] Sample modality images should include images under multiple modalities. For example, sample modality images include CT sample images, MR sample images, etc. In order to improve the generalization performance of the machine learning model, the coverage of sample modality images should be large enough. For example, sample modality images can include images under multiple parts, and can also include images of patients of various age groups, etc. The two-dimensional sample sequence corresponding to the sample modality image includes but is not limited to DICOM sequence data.

[0159] Optionally, the computer device can obtain multiple three-dimensional sample modality images from PACS. In some embodiments, the computer device can also simulate and generate sample modality images.

[0160] S1102 , obtaining a grayscale distribution vector sample corresponding to each sample modality image according to the grayscale value of each pixel point of each two-dimensional image in each sample modality image.

[0161] Furthermore, the same principle as S202, according to the gray value of each pixel point of each two-dimensional image in each sample modality image, the gray distribution vector sample corresponding to each sample modality image can be obtained, which will not be repeated here. The gray vector distribution sample can also include N gray value elements, each gray value element corresponds to the proportion of the pixel points corresponding to the gray value to all the pixel points, and the sum of each element can be equal to 1.

[0162] S1103, performing machine learning training on the initial model according to each grayscale distribution vector sample and the window parameter label corresponding to each grayscale distribution vector sample to obtain a machine learning model.

[0163] In this embodiment, the window parameter labels corresponding to the grayscale distribution vector samples may be manually labeled or automatically determined by a computer device, but this embodiment is not limited thereto.

[0164] Optionally, the computer device may input each grayscale distribution vector sample into the initial model to obtain the window parameters predicted by the initial model, and update the model parameters of the initial model according to the evaluation parameters between the window parameters predicted by the initial model and the corresponding window parameter labels, until the evaluation parameters between the window parameters predicted by the initial model and the corresponding window parameter labels meet the preset conditions, stop training the initial model, and use the initial model at this time as the machine learning model. The evaluation parameters may include but are not limited to at least one of accuracy, missed detection rate, false detection rate, cross entropy, etc.

[0165] In the above embodiment, since multiple three-dimensional sample modal images are obtained, and the grayscale distribution vector samples corresponding to each sample modal image are obtained according to the grayscale values ​​of each pixel point of each two-dimensional image in each sample modal image, and then according to each grayscale distribution vector sample and the window parameter label corresponding to each grayscale distribution vector sample, the initial model can be trained by machine learning to obtain a machine learning model. Therefore, after obtaining the machine learning model, the window parameters corresponding to the target modal image can be determined according to the grayscale distribution vector and the machine learning model.

[0166] Fig.12 FIG. 1 is another flow chart of obtaining a machine learning model in an embodiment of the present application. In an exemplary embodiment, Fig.12 As shown, S1103 includes S1201 to S1203.

[0167] S1201, determining a first loss according to each grayscale distribution vector sample, a tissue probability vector label corresponding to each grayscale distribution vector sample, and a first sub-model in the initial model.

[0168] In this embodiment, the tissue probability vector label corresponding to each gray distribution vector sample is used to indicate which preset tissue the tissue in the corresponding sample modality image belongs to. In the tissue probability vector label, only one element is equal to 1, and the other elements are all equal to 0. Fig.13 Schematic diagram of an organization probability vector label in an embodiment of the present application, such as Fig.13 As shown, the organization probability vector label .by Fig.13 For example, , indicating that the tissue in the corresponding sample modality image belongs to the lung.

[0169] In some embodiments, in order to improve the accuracy of the tissue probability vector label, the computer device may filter sample modality images that only include one tissue.

[0170] Furthermore, the computer device determines the first loss based on each grayscale distribution vector sample, the tissue probability vector label corresponding to each grayscale distribution vector sample, and the first sub-model in the initial model.

[0171] Optionally, the computer device can input each grayscale distribution vector sample into the first sub-model in the initial model to obtain the tissue probability vector predicted by the first sub-model, and determine the first loss based on the tissue probability vector predicted by the first sub-model and the tissue probability vector label corresponding to each grayscale distribution vector sample.

[0172] Further optionally, the computer device may calculate the cross entropy between the tissue probability vector predicted by the first sub-model and the corresponding tissue probability vector label as the first loss.

[0173] Exemplarily, the computer device may determine the first loss according to the following formula (1): In formula (1), m is a positive integer greater than or equal to 1 and less than or equal to M, and M is the number of sample modality images in a single training process. k is a positive integer greater than or equal to 1 and less than or equal to K, The number of organizational label types. Whether the sample modality image m is a tissue label ,Right now When the sample modality image m is the tissue label .

[0174] is a softmax function used to calculate the sample modality image m is classified as a tissue label The probability of It can be determined according to the following formula (2). In formula (2), Between 0 and 1, represents the tissue probability vector of the sample modality image m predicted by the first sub-model, represents the probability that the predicted sample modality image m belongs to the tissue label k. It should be noted that in formula (1) To make the first loss Normalization to improve the first loss accuracy.

[0175] (1)

[0176] (2)

[0177] S1202, determining a second loss according to each grayscale distribution vector sample, a tissue probability vector label corresponding to each grayscale distribution vector sample, a window parameter label corresponding to each grayscale distribution vector sample, and a second sub-model in the initial model.

[0178] In this embodiment, the computer device can splice each grayscale distribution vector sample and the tissue probability vector label corresponding to each grayscale distribution vector sample to obtain a spliced ​​vector sample, and input the spliced ​​vector sample to the second sub-model to obtain the window parameter predicted by the second sub-model, and then determine the second loss according to the window parameter predicted by the second sub-model and the window parameter label corresponding to the grayscale distribution vector sample. It can be understood that the window parameter label corresponding to the grayscale distribution vector sample is also the window parameter label of the sample modality image corresponding to the grayscale distribution vector sample.

[0179] Exemplarily, the computer device may determine the second loss according to the following formula (3): In formula (3), represents the window width of the sample modality image m predicted by the second sub-model, represents the window width in the window parameter label of the sample modality image m; represents the window position of the sample modality image m predicted by the second sub-model, represents the window position in the window parameter label of the sample modality image m. It should be noted that in formula (3) To make the second loss Normalization to improve the second loss accuracy.

[0180] (3)

[0181] S1203: Perform machine learning training on the initial model according to the first loss and the second loss to obtain a machine learning model.

[0182] In this embodiment, the computer device may determine a first product between the first coefficient and the first loss, and determine a second product between the second coefficient and the second loss, and then perform machine learning training on the initial model according to the sum of the first product and the second product to obtain a machine learning model. The sum of the first coefficient and the second coefficient may be equal to one.

[0183] Exemplarily, the computer device may determine the third loss according to the following formula (4), and perform machine learning training on the initial model according to the third loss L to obtain the machine learning model. For example, the computer device may stop training the initial model when the third loss L is less than a preset loss, and use the initial model at this time as the machine learning model. The preset loss may be a number set according to demand, such as a number close to 0. In other words, the training goal of the entire machine learning model is to make the third loss L as small as possible, that is, the identified tissue and the predicted window parameters are as close as possible to the corresponding tissue probability vector label and window parameter label, so as to achieve accurate prediction and provide better three-dimensional rendering effect.

[0184] (4)

[0185] In the above embodiment, the first loss is determined according to each grayscale distribution vector sample, the tissue probability vector label corresponding to each grayscale distribution vector sample, and the first sub-model in the initial model, and the second loss is determined according to each grayscale distribution vector sample, the tissue probability vector label corresponding to each grayscale distribution vector sample, the window parameter label corresponding to each grayscale distribution vector sample, and the second sub-model in the initial model. In this way, after the initial model is trained for machine learning according to the first loss and the second loss, a machine learning model including a recognition sub-model and a prediction sub-module can be obtained. Then, during use, the tissue probability vector corresponding to the target modality image can be obtained according to the grayscale distribution vector and the recognition sub-model, and the window parameters can be obtained according to the tissue probability vector, the grayscale distribution vector and the prediction sub-model in the machine learning model.

[0186] Fig.14 FIG. 1 is a flow chart showing a rendering result in an embodiment of the present application. In an exemplary embodiment, Fig.14 As shown, the above-mentioned window parameter determination method also includes S1401 to S1402.

[0187] S1401, performing three-dimensional rendering on the target modality image to obtain a rendering result of the target modality image.

[0188] In this embodiment, if the target modality image is a three-dimensional image, the computer device will perform three-dimensional rendering on the target modality image to display the rendered three-dimensional image, and obtain the rendering result of the target modality image. It can be understood that the rendering result is also the rendered target modality image.

[0189] S1402: Display rendering results based on window parameters.

[0190] Afterwards, the computer device can display the rendering results based on the window parameters obtained by the machine learning model. This embodiment does not limit the form of displaying the rendering results.

[0191] In the above embodiment, since the target modality image is three-dimensionally rendered to obtain the rendering result of the target modality image, and the window parameters are parameters applicable to the three-dimensional target modality image, after displaying the rendering result based on the window parameters, the phenomena of incomplete tissue displayed in the rendering result, display of irrelevant fine tissue, display of a large amount of air, etc. can be reduced, thereby improving the display effect and facilitating the user to observe the real tissue situation.

[0192] In an exemplary embodiment, the above window parameter determination method may further include the following steps:

[0193] In response to an adjustment operation based on the rendering result, the window parameters are adjusted.

[0194] In this embodiment, the adjustment operation may include but is not limited to at least one operation based on a mouse, a keyboard, a gesture, and a voice. That is, if the user is still not satisfied with the rendering result displayed in S1402, the user may initiate an adjustment operation, and the computer device may respond to the adjustment operation to adjust the window parameters.

[0195] In the above embodiment, since the window parameters can be adjusted in response to adjustment operations based on the rendering results, the window parameters determined by the machine learning model can be used as an initial reference when displaying the rendering results, and the user can subsequently make further fine adjustments to the global or local window parameters.

[0196] In order to more clearly introduce the window parameter determination method in this application, Figure 15~Figure 17 Provide explanation. Fig.15 FIG. 1 is a process diagram of a method for determining window parameters in an embodiment of the present application. Fig.15 As shown, the computer device can execute the window parameter determination method according to the following steps.

[0197] S1501, acquiring multiple three-dimensional sample modality images.

[0198] S1502 , obtaining a grayscale distribution vector sample corresponding to each sample modality image according to the grayscale value of each pixel point of each two-dimensional image in each sample modality image.

[0199] S1503, determining a first loss according to each grayscale distribution vector sample, a tissue probability vector label corresponding to each grayscale distribution vector sample, and a first sub-model in the initial model.

[0200] S1504, determining a second loss according to each grayscale distribution vector sample, a tissue probability vector label corresponding to each grayscale distribution vector sample, a window parameter label corresponding to each grayscale distribution vector sample, and a second sub-model in the initial model.

[0201] S1505: Perform machine learning training on the initial model according to the first loss and the second loss to obtain a machine learning model.

[0202] S1506, acquiring a three-dimensional target modality image.

[0203] S1507: If the type is a preset type, the grayscale value of each pixel in the two-dimensional image is used as the mapping grayscale value corresponding to each pixel in the two-dimensional image.

[0204] S1508: If the type is not a preset type, determine the mapping grayscale value corresponding to each pixel in the two-dimensional image according to the mapping relationship and the grayscale value of each pixel in the two-dimensional image.

[0205] S1509, counting the mapping grayscale value corresponding to each pixel point in each two-dimensional image to obtain grayscale distribution information.

[0206] S1510, determining a grayscale distribution vector according to grayscale distribution information corresponding to the target modality image.

[0207] S1511, obtaining a tissue probability vector corresponding to the target modality image according to the grayscale distribution vector and the recognition sub-model, wherein the tissue probability vector is used to indicate the probability that the tissue in the target modality image belongs to each preset tissue.

[0208] S1512, concatenating the tissue probability vector and the grayscale distribution vector to obtain a concatenated vector.

[0209] S1513, input the concatenated vector into the prediction sub-model to obtain initial window parameters.

[0210] S1514: When the type of the target modality image is a preset type, the initial window parameters are used as window parameters corresponding to the target modality image.

[0211] S1515: When the type of the target modality image is not a preset type, determine the window parameters corresponding to the target modality image according to the initial window parameters and the mapping relationship.

[0212] S1516, performing three-dimensional rendering on the target modality image to obtain a rendering result of the target modality image.

[0213] S1517, display the rendering result based on the window parameters.

[0214] S1518, adjusting window parameters in response to an adjustment operation based on the rendering result.

[0215] Among them, S1501~S1505 is the training process, and S1506~S1518 is the training process. The steps of S1501~S1518 can refer to the above embodiment and will not be repeated here. It can be seen that due to the particularity of medical images, although the range of grayscale values ​​of medical images of different modalities is different, the range of grayscale values ​​corresponding to different tissues of the human body under a single modality is relatively fixed. Utilizing this characteristic, the present embodiment maps the statistical range of grayscale distribution information of target modality images of different modalities to be consistent, thereby unifying the input, intermediate results and output format of the preset network, and training the window parameters corresponding to target modality images of different modalities based on the machine learning model.

[0216] Fig.16 is a flowchart of a method for determining window parameters in an embodiment of the present application, Fig.17 This is a data flow diagram of a method for determining window parameters in an embodiment of the present application. Fig.16 and Fig.17 , the user can input the target modality image into the computer device, and then the computer device will obtain the grayscale distribution vector corresponding to the target modality image. The computer device can first obtain the grayscale distribution information based on the target modality image, and then obtain the grayscale distribution vector based on the grayscale distribution information.

[0217] Next, the tissue probability vector can be obtained based on the recognition sub-model. For example, the grayscale distribution vector is input into the recognition sub-model to obtain the tissue probability vector output by the recognition sub-model. After that, the computer device obtains the window parameters corresponding to the target modality image based on the prediction sub-model. Among them, the computer device will obtain the window parameters based on the tissue probability vector, the grayscale probability vector and the preset sub-model. Finally, the computer device can display the rendering results based on the window parameters. After viewing the rendering effect, the user can continue to adjust the window parameters.

[0218] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0219] Based on the same inventive concept, the embodiment of the present application also provides a window parameter determination device for implementing the above-mentioned window parameter determination method. The implementation solution provided by the device to solve the problem is similar to the implementation solution recorded in the above-mentioned method, so the specific limitations in one or more embodiments of the window parameter determination device provided below can refer to the limitations of the window parameter determination method above, and will not be repeated here.

[0220] Fig.18 is a structural block diagram of a window parameter determination device in an embodiment of the present application. In an exemplary embodiment, Fig.18 As shown, a window parameter determination device 1800 is provided, comprising: a first acquisition module 1801, a first determination module 1802 and a second determination module 1803, wherein:

[0221] The first acquisition module 1801 is used to acquire a three-dimensional target modality image; the target modality image includes a plurality of continuous two-dimensional images.

[0222] The first determining module 1802 is used to obtain a grayscale distribution vector corresponding to the target modality image according to the grayscale value of each pixel point of each two-dimensional image in the target modality image.

[0223] The second determination module 1803 is used to determine the window parameters corresponding to the target modality image according to the grayscale distribution vector and the machine learning model.

[0224] In the above-mentioned window parameter determination device, a three-dimensional target modality image is first obtained, and the target modality image includes multiple continuous two-dimensional images. Since the grayscale distribution vector corresponding to the three-dimensional target modality image can be obtained according to the grayscale value of each pixel point of each two-dimensional image in the target modality image, the grayscale distribution vector takes into account each grayscale value of the continuous two-dimensional image in the target modality image. In this way, the window parameters corresponding to the target modality image can be efficiently determined according to the grayscale distribution vector and the machine learning model. On the one hand, there is no need to manually determine the window parameters based on experience in the whole process. On the other hand, the method provided by this embodiment can be applied not only to the target modality image of a single two-dimensional image, but also to the three-dimensional target modality image composed of multiple two-dimensional images, so as to accurately and efficiently determine the window parameters applicable to the three-dimensional target modality image.

[0225] Optionally, the first determining module 1802 includes:

[0226] The first determining unit is used to obtain grayscale distribution information corresponding to the target modality image according to the grayscale value of each pixel in each two-dimensional image.

[0227] The second determining unit is used to determine the grayscale distribution vector according to the grayscale distribution information corresponding to the target modality image.

[0228] Optionally, the first determining unit includes:

[0229] The first determination subunit is used to determine the mapping grayscale value corresponding to each pixel point in each two-dimensional image according to the grayscale value of each pixel point in each two-dimensional image and the type of the target modality image.

[0230] The statistical subunit is used to count the mapping grayscale values ​​corresponding to each pixel point in each two-dimensional image to obtain grayscale distribution information.

[0231] Optionally, the first determination subunit is used to use the grayscale value of each pixel in the two-dimensional image as the mapping grayscale value corresponding to each pixel in the two-dimensional image if the type is a preset type; if the type is not a preset type, determine the mapping grayscale value corresponding to each pixel in the two-dimensional image based on the mapping relationship and the grayscale value of each pixel in the two-dimensional image.

[0232] Optionally, the machine learning model includes an identification sub-model and a prediction sub-model; the second determination module 1803 includes:

[0233] A third determination unit is used to obtain a tissue probability vector corresponding to the target modality image according to the grayscale distribution vector and the recognition sub-model; the tissue probability vector is used to indicate the probability that the tissue in the target modality image belongs to each preset tissue;

[0234] The fourth determination unit is used to obtain the window parameter according to the tissue probability vector and the prediction sub-model.

[0235] Optionally, a fourth determination unit is used to obtain window parameters based on the tissue probability vector, the grayscale distribution vector and the prediction sub-model in the machine learning model.

[0236] Optionally, the fourth determining unit includes:

[0237] The splicing subunit is used to splice the tissue probability vector and the grayscale distribution vector to obtain a splicing vector.

[0238] The second determination subunit is used to obtain the window parameter according to the splicing vector and the prediction submodel.

[0239] Optionally, a second determination subunit is used to input the splicing vector into the prediction submodel to obtain initial window parameters; when the type of the target modality image is a preset type, the initial window parameters are used as window parameters; when the type of the target modality image is not a preset type, the window parameters are determined based on the initial window parameters and the mapping relationship.

[0240] Optionally, the target modality image is a reconstructed computed tomography image or a magnetic resonance image.

[0241] Optionally, the window parameter determination device 1800 further includes:

[0242] The second acquisition module is used to acquire a plurality of three-dimensional sample modality images; the sample modality images include a plurality of continuous two-dimensional images.

[0243] The third determination module is used to obtain the grayscale distribution vector samples corresponding to each sample modality image according to the grayscale value of each pixel point of each two-dimensional image in each sample modality image.

[0244] The training module is used to perform machine learning training on the initial model according to each grayscale distribution vector sample and the window parameter label corresponding to each grayscale distribution vector sample to obtain a machine learning model.

[0245] Optional training modules include:

[0246] The fifth determining unit is used to determine the first loss according to each grayscale distribution vector sample, the tissue probability vector label corresponding to each grayscale distribution vector sample, and the first sub-model in the initial model.

[0247] The sixth determination unit is used to determine the second loss according to each grayscale distribution vector sample, the tissue probability vector label corresponding to each grayscale distribution vector sample, the window parameter label corresponding to each grayscale distribution vector sample, and the second sub-model in the initial model.

[0248] The training unit is used to perform machine learning training on the initial model according to the first loss and the second loss to obtain a machine learning model.

[0249] Optionally, the window parameter determination device 1800 further includes:

[0250] The rendering module is used to perform three-dimensional rendering on the target modality image to obtain the rendering result of the target modality image.

[0251] The display module is used to display the rendering results based on the window parameters.

[0252] Optionally, the window parameter determination device 1800 further includes:

[0253] The adjustment module is used to adjust the window parameters in response to the adjustment operation based on the rendering result.

[0254] Each module in the above-mentioned window parameter determination device can be implemented in whole or in part by software, hardware and a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each module above.

[0255] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0256] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0257] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0258] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0259] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0260] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for determining window parameters, characterized in that: The method comprises: Acquire a three-dimensional target modality image; the target modality image includes a plurality of continuous two-dimensional images; Obtaining a grayscale distribution vector corresponding to the target modality image according to the grayscale value of each pixel point of each two-dimensional image in the target modality image; According to the grayscale distribution vector and the machine learning model, the window parameters corresponding to the target modality image are determined.

2. The method according to claim 1, characterized in that The step of obtaining the grayscale distribution vector corresponding to the target modality image according to the grayscale value of each pixel point of each two-dimensional image in the target modality image comprises: Obtaining grayscale distribution information corresponding to the target modality image according to the grayscale value of each pixel in each of the two-dimensional images; The grayscale distribution vector is determined according to the grayscale distribution information corresponding to the target modality image.

3. The method according to claim 2, characterized in that The step of obtaining grayscale distribution information corresponding to the target modality image according to the grayscale value of each pixel in the two-dimensional image includes: Determining a mapping grayscale value corresponding to each pixel point in each of the two-dimensional images according to the grayscale value of each pixel point in each of the two-dimensional images and the type of the target modality image; The mapping grayscale value corresponding to each pixel point in each of the two-dimensional images is counted to obtain the grayscale distribution information.

4. The method according to claim 3, characterized in that Determining the mapping grayscale value corresponding to each pixel point in each of the two-dimensional images according to the grayscale value of each pixel point in each of the two-dimensional images and the type of the target modality image includes: If the type is a preset type, the grayscale value of each pixel in the two-dimensional image is used as the mapping grayscale value corresponding to each pixel in the two-dimensional image; If the type is not the preset type, a mapping grayscale value corresponding to each pixel in the two-dimensional image is determined according to the mapping relationship and the grayscale value of each pixel in the two-dimensional image.

5. The method according to any one of claims 1 to 4, characterized in that: The machine learning model includes a recognition sub-model and a prediction sub-model; and determining the window parameters corresponding to the target modality image according to the grayscale distribution vector and the machine learning model includes: According to the grayscale distribution vector and the identification sub-model, a tissue probability vector corresponding to the target modality image is obtained; the tissue probability vector is used to indicate the probability that the tissue in the target modality image belongs to each preset tissue; The window parameters are obtained according to the tissue probability vector and the prediction sub-model.

6. The method according to claim 5, characterized in that The step of obtaining the window parameter according to the tissue probability vector and the prediction sub-model comprises: The window parameters are obtained according to the tissue probability vector, the grayscale distribution vector and the prediction sub-model.

7. The method according to claim 6, characterized in that The step of obtaining the window parameters according to the tissue probability vector, the grayscale distribution vector and the prediction sub-model comprises: Splicing the tissue probability vector and the grayscale distribution vector to obtain a splicing vector; The window parameter is obtained according to the splicing vector and the prediction sub-model.

8. The method according to claim 7, characterized in that The step of obtaining the window parameter according to the splicing vector and the prediction sub-model includes: Inputting the splicing vector into the prediction sub-model to obtain initial window parameters; When the type of the target modality image is a preset type, the initial window parameters are used as the window parameters; or, when the type of the target modality image is not the preset type, the window parameters are determined according to the initial window parameters and a mapping relationship.

9. The method according to any one of claims 1 to 4, characterized in that: The target modality image is a reconstructed computed tomography image or a magnetic resonance image.

10. The method according to any one of claims 1 to 4, characterized in that: The method further comprises: Acquire a plurality of three-dimensional sample modality images; the sample modality images include a plurality of continuous two-dimensional images; Obtaining grayscale distribution vector samples corresponding to each of the sample modality images according to the grayscale values ​​of each pixel point of each two-dimensional image in each of the sample modality images; According to each of the grayscale distribution vector samples and the window parameter labels corresponding to each of the grayscale distribution vector samples, machine learning training is performed on the initial model to obtain the machine learning model.

11. The method according to claim 10, characterized in that The step of performing machine learning training on the initial model according to each of the grayscale distribution vector samples and the window parameter labels corresponding to each of the grayscale distribution vector samples to obtain the machine learning model comprises: Determine a first loss according to each of the grayscale distribution vector samples, a tissue probability vector label corresponding to each of the grayscale distribution vector samples, and a first sub-model in the initial model; Determine a second loss according to each of the grayscale distribution vector samples, a tissue probability vector label corresponding to each of the grayscale distribution vector samples, a window parameter label corresponding to each of the grayscale distribution vector samples, and a second sub-model in the initial model; According to the first loss and the second loss, machine learning training is performed on the initial model to obtain the machine learning model.

12. The method according to any one of claims 1 to 4, characterized in that: The method further comprises: Performing three-dimensional rendering on the target modality image to obtain a rendering result of the target modality image; The rendering result is displayed based on the window parameters.

13. The method according to claim 12, characterized in that The method further comprises: In response to an adjustment operation based on the rendering result, the window parameter is adjusted.

14. A window parameter determination device, characterized in that: The device comprises: A first acquisition module is used to acquire a three-dimensional target modality image; the target modality image includes a plurality of continuous two-dimensional images; A first determination module, configured to obtain a grayscale distribution vector corresponding to the target modality image according to the grayscale value of each pixel point of each two-dimensional image in the target modality image; The second determination module is used to determine the window parameters corresponding to the target modality image according to the grayscale distribution vector and the machine learning model.

15. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 13 are implemented.

16. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 13 are implemented.

17. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 13 are implemented.

Images