Multi-substance decomposition method and device based on neural network, equipment and medium

Abstract

The embodiment of the invention discloses a multi-substance decomposition method and device based on a neural network, equipment and a medium, and the method comprises the steps: obtaining a dual-energy CT image of a target object; inputting the dual-energy CT image into a pre-trained base material composition judgment model to obtain a base material composition judgment result of the target object; and inputting the dual-energy CT image into a pre-trained base material absorption coefficient prediction model to obtain an equivalent absorption coefficient of each base material component in the target object, and obtaining a base material density distribution image of the target object based on the base material composition judgment result and the equivalent absorption coefficient. According to the technical scheme provided by the embodiment of the invention, the problem that the noise of the decomposed base substance image is increased or the accuracy of the decomposition result is low due to the introduction of new imaging noise in the current base substance decomposition algorithm is solved, the judgment process of the base substance combination can be modeled, the judgment precision of the base substance combination is improved, and the multi-substance decomposition precision is improved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of image processing technology, and in particular to a multi-substance decomposition method, device, equipment and medium based on a neural network. Background Art

[0002] The multi-substance decomposition algorithm based on the dual-energy CT image domain counts all the three basic material combinations of the scanned object and establishes a three-basic material combination table. The combination table can be used to classify the three basic material compositions of the pixels of the dual-energy CT image, and combined with the three-basic material decomposition equation of dual-energy CT imaging, the purpose of dual-energy multi-substance decomposition is achieved.

[0003] However, this method requires cyclically solving the combination of three base substances during the multi-substance decomposition process. This process may lead to the problem of incorrect selection of the three base substances, thereby introducing new imaging noise, resulting in increased image noise of the decomposed base substances, loss of details and decomposition errors, affecting the image effect of the base substances. Summary of the invention

[0004] The embodiments of the present invention provide a multi-substance decomposition method, device, equipment and medium based on a neural network, which can model the basic substance combination judgment process, improve the judgment accuracy of the basic substance combination, and improve the multi-substance decomposition accuracy.

[0005] In a first aspect, an embodiment of the present invention provides a multi-substance decomposition method based on a neural network, the method comprising:

[0006] Acquire a dual-energy CT image of a target object;

[0007] Inputting the dual-energy CT image into a pre-trained basic material composition judgment model to obtain a basic material composition judgment result of the target object; and

[0008] Inputting the dual-energy CT image into a pre-trained base material absorption coefficient prediction model to obtain an equivalent absorption coefficient of each base material component in the target object;

[0009] Based on the base material composition determination result and the equivalent absorption coefficient, a base material density distribution image of the target object is obtained.

[0010] Optionally, obtaining a base material density distribution image of the target object based on the base material composition judgment result and the equivalent absorption coefficient includes:

[0011] Constructing a target multi-substance decomposition equation based on the base substance composition judgment result and the equivalent absorption coefficient;

[0012] Analyze the target multi-substance decomposition equation to determine the basic substance components of the target object and the volume fraction of each basic substance component;

[0013] A base material density distribution image of the target object is obtained according to each base material component of the target object and the volume fraction of each base material component.

[0014] Optionally, the target multi-substance decomposition equation is constructed based on the base substance composition judgment result and the equivalent absorption coefficient;

[0015] Establishing target multi-substance decomposition equations respectively for the mask image corresponding to each base substance combination in the base substance composition judgment result and the base substance absorption coefficient matrix corresponding to the equivalent absorption coefficient;

[0016] Among them, each quantity of the target multi-substance decomposition equation corresponds to each quantity of the basic substance combination.

[0017] Optionally, the base material composition judgment model includes four down-sampling modules, four up-sampling modules and an output module;

[0018] Among them, each of the downsampling modules consists of two convolutional layers and a maximum pooling layer; each of the upsampling modules includes two convolutional layers and one deconvolutional layer; and the output module includes three convolutional layers.

[0019] Optionally, in the basic material composition judgment model, the activation function of the last convolutional layer is sigmoid, and the activation function of the convolutional layers other than the last convolutional layer is tanh.

[0020] Optionally, the output module includes a plurality of output channels, wherein each of the output channels corresponds to a base material combination and outputs a black-and-white mask image corresponding to the base material combination.

[0021] Optionally, the method according to any one of claims 1-3 is characterized in that the base material absorption coefficient prediction model includes 9 downsampling modules, wherein each of the downsampling modules includes two convolutional layers and one maximum pooling layer.

[0022] In a second aspect, an embodiment of the present invention provides a multi-substance decomposition device based on a neural network, the device comprising:

[0023] An image acquisition module, used for acquiring a dual-energy CT image of a target object;

[0024] A first image analysis module is used to input the dual-energy CT image into a pre-trained basic material composition judgment model to obtain a basic material composition judgment result of the target object; and

[0025] A second image analysis module is used to input the dual-energy CT image into a pre-trained base material absorption coefficient prediction model to obtain an equivalent absorption coefficient of each base material component in the target object;

[0026] The base material determination module is used to obtain a base material density distribution image of the target object based on the base material composition judgment result and the equivalent absorption coefficient.

[0027] In a third aspect, an embodiment of the present invention provides a computer device, the computer device comprising:

[0028] one or more processors;

[0029] A memory for storing one or more programs;

[0030] When the one or more programs are executed by the one or more processors, the one or more processors implement the neural network-based multi-substance decomposition method described in any embodiment.

[0031] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the multi-substance decomposition method based on a neural network as described in any embodiment.

[0032] The technical solution provided by the embodiment of the present invention obtains a dual-energy CT image of the target object; inputs the dual-energy CT image into a pre-trained base material component judgment model to obtain a base material component judgment result of the target object; and inputs the dual-energy CT image into a pre-trained base material absorption coefficient prediction model to obtain an equivalent absorption coefficient of each base material component in the target object; based on the base material component judgment result and the equivalent absorption coefficient, a base material density distribution image of the target object is obtained. The technical solution of the embodiment of the present invention solves the problem that the current base material decomposition algorithm will introduce new imaging noise, resulting in an increase in the noise of the decomposed base material image or a low accuracy of the decomposition result. The base material combination judgment process can be modeled to improve the judgment accuracy of the base material combination and the multi-material decomposition accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 is a flow chart of a multi-substance decomposition method based on a neural network provided by an embodiment of the present invention;

[0034] Figure 2 It is a structural diagram of a basic material composition judgment model provided by an embodiment of the present invention;

[0035] Figure 3 It is a structural diagram of a base material absorption coefficient prediction model provided by an embodiment of the present invention;

[0036] Figure 4 is a flow chart of a multi-substance decomposition method based on a neural network provided by an embodiment of the present invention;

[0037] Figure 5 This is a multi-substance decomposition effect diagram of a dual-energy CT image provided by an embodiment of the present invention;

[0038] Figure 6 is a schematic structural diagram of a multi-substance decomposition device based on a neural network provided by an embodiment of the present invention;

[0039] Figure 7 It is a structural schematic diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] Embodiment 1

[0042] Figure 1 It is a flow chart of a multi-substance decomposition method based on a neural network provided by an embodiment of the present invention. The embodiment of the present invention can be applied to the scenario of multi-substance decomposition of dual-energy CT images. The method can be executed by a multi-substance decomposition device based on a neural network, and the device can be implemented by software and / or hardware and integrated in a computer device with application development function.

[0043] like Figure 1 As shown, the multi-material decomposition method based on neural network includes the following steps:

[0044] S110, acquiring a dual-energy CT image of the target object;

[0045] Among them, the target object can be a body part of a person or an animal, or an imaging area of ​​interest in a certain part. The dual-energy CT image is an image obtained by performing CT imaging of the target object through two X-rays with different energies. Different energy values ​​are set according to actual needs based on different target objects. Among them, the image obtained based on the X-ray with higher energy value is a high-energy CT image, and the image obtained based on the X-ray with lower energy value is a low-energy CT image.

[0046] S120, inputting the dual-energy CT image into a pre-trained basic material composition judgment model to obtain a basic material composition judgment result of the target object.

[0047] Among them, the basic material composition judgment model is a neural network model, which is based on the dual-energy CT image with known basic material composition as a sample, and the basic material combination type in the basic material combination table corresponding to the sample is used as the sample label for model training. The basic material composition judgment model can be used to judge the basic material composition of each pixel in the dual-energy CT image. The basic material composition type can be a non-repeating triplet obtained according to mathematical permutations and combinations. The input of the basic material composition judgment model is the dual-energy CT image, and the output is a black and white mask image. The number of channels of the output of the black and white mask image is consistent with the number of corresponding basic material combination categories, and each black and white mask image corresponds to a basic material combination.

[0048] In this embodiment, in the process of constructing the model training sample, specifically, Figure 2 As shown, when the target object is the abdomen of a human body, an iodine contrast agent is injected into the abdomen of the human body and then scanned to obtain a dual-energy CT image of the abdomen of the human body. The dual-energy CT image of the abdomen is input into the basic substance composition judgment model, and a black-and-white mask image of three possible basic substance combinations of "iodine-blood-air", "fat-muscle-air" and "fat-bone-air" can be obtained. Among them, the six basic substances contained in the dual-energy CT image of the abdomen of the human body are iodine-fat-muscle-bone-blood-air.

[0049] In an optional implementation, the base material composition determination model includes four down-sampling modules, four up-sampling modules and an output module;

[0050] Among them, each downsampling module consists of two convolutional layers and a maximum pooling layer; each upsampling module includes two convolutional layers and one deconvolution layer; and the output module includes three convolutional layers.

[0051] Specifically, the layer structure of each downsampling module is convolution layer-convolution layer-pooling layer, which is used to downsample the dual-energy CT image. The layer structure of each upsampling module is convolution layer-convolution layer-deconvolution layer, and the layer structure of the output module is convolution layer-convolution layer-convolution layer.

[0052] like Figure 2 As shown in FIG. 1 , the output feature map of the downsampling module and the input feature map of the upsampling module are connected using a Concatenate operation, which can speed up network training and improve image accuracy.

[0053] In an optional implementation, in the base material composition judgment model, the activation function of the last convolutional layer is sigmoid, and the activation function of the convolutional layers other than the last convolutional layer is tanh.

[0054] Using the tanh function as the activation function of the convolutional layer except the last convolutional layer converges faster, making the training process easier to converge and preventing gradient explosion and gradient disappearance; using the sigmoid function as the activation function of the last convolutional layer is smoother and can enhance the expressiveness of the model.

[0055] In an optional embodiment, the output module includes a plurality of output channels, wherein each output channel corresponds to a base material combination and outputs a black-and-white mask image corresponding to the base material combination.

[0056] The network input of the basic material composition judgment model is a dual-energy CT image with an image size of 512x512, and the output is a black and white mask image with three channels with an image size of 512x512. The black value of the black and white mask image is 0, and the white value is 1. At the same pixel position of the three-channel mask image, at most one channel has a value of 1, where the three basic material distributions corresponding to the channel with a pixel value of 1 are the three basic material distributions of the pixel.

[0057] S130, inputting the dual-energy CT image into a pre-trained base material absorption coefficient prediction model to obtain an equivalent absorption coefficient of each base material component in the target object.

[0058] The base material absorption coefficient prediction model is a neural network model, which is used to predict the equivalent absorption coefficients of different base materials under low-energy X-rays and high-energy X-rays. Its input is a dual-energy CT image, and its output is a base material equivalent absorption coefficient matrix.

[0059] Specifically, the base material components may be different types of base materials, and the equivalent absorption coefficient may be the absorption coefficient of various base materials under two X-rays of different energies, which may include a low-energy absorption coefficient and a high-energy absorption coefficient. The low-energy absorption coefficient is the absorption coefficient of the base material for lower-energy X-rays, and the high-energy absorption coefficient is the absorption coefficient of the base material for higher-energy X-rays. Since the same base material has different absorption capabilities for two different energy X-rays, the low-energy absorption coefficient and high-energy absorption coefficient of the same base material are also different. The base material absorption coefficient prediction model may be a model obtained through training based on dual-energy CT images with known absorption coefficients of various base materials as model training samples.

[0060] The equivalent absorption coefficient matrix is ​​composed of the equivalent absorption coefficients of each base material. The number of rows in the matrix represents the number of X-rays with different energies, and the number of columns represents the number of base material types, such as Figure 3 As shown in the figure, taking the input of the dual-energy CT image of the abdomen of the human body injected with iodine contrast agent as an example, the number of rows is 2, which means that there are two types of X-rays, namely high-energy and low-energy X-rays, and the number of columns is 6, which means that there are 6 types of base substances, namely iodine-fat-muscle-bone-blood-air. The equivalent absorption coefficient matrix of the base substances is:

[0061]

[0062] in, Represents the high energy absorption coefficient of a certain base substance, Represents the low energy absorption coefficient of a certain base substance, x represents the type of substance, and can also be represented by a number, for example This represents the high energy absorption coefficient of the base substance iodine, and so on.

[0063] In an optional embodiment, the base material absorption coefficient prediction model includes 9 downsampling modules, wherein each downsampling module includes two convolutional layers and one maximum pooling layer.

[0064] like Figure 3 As shown in the figure, the layer structure of the downsampling module is convolution layer-convolution layer-maximum pooling layer. The convolution kernel size of the convolution layer can be 3x3, the step size can be 1x1, and the activation function can be Leaky ReLU. The step size of the maximum pooling layer is 2x2. After each downsampling module, the image size of the dual-energy CT image is reduced to half of the original size. For example, for a dual-energy CT image with an input image size of 512x512, after 9 downsamplings, its size is compressed to a 1x1 array, and the array length is twice the number of base material types. The array is transformed through the reshape function in Python, and the output is an equivalent absorption coefficient matrix.

[0065] S140, obtaining a base material density distribution image of the target object based on the base material composition judgment result and the equivalent absorption coefficient.

[0066] Specifically, the base material composition judgment result output by the base material composition judgment model is combined with the equivalent absorption coefficient matrix output by the base material absorption coefficient prediction model to obtain a base material density distribution image.

[0067] The technical solution provided by the embodiment of the present invention obtains a dual-energy CT image of the target object; inputs the dual-energy CT image into a pre-trained base material component judgment model to obtain a base material component judgment result of the target object; and inputs the dual-energy CT image into a pre-trained base material absorption coefficient prediction model to obtain an equivalent absorption coefficient of each base material component in the target object; based on the base material component judgment result and the equivalent absorption coefficient, a base material density distribution image of the target object is obtained. The technical solution of the embodiment of the present invention solves the problem that the current base material decomposition algorithm will introduce new imaging noise, resulting in an increase in the noise of the decomposed base material image or a low accuracy of the decomposition result. The base material combination judgment process can be modeled to improve the judgment accuracy of the base material combination and the multi-material decomposition accuracy.

[0068] Embodiment 2

[0069] Figure 4 This is a flowchart of a multi-substance decomposition method based on a neural network provided by an embodiment of the present invention. Based on the above disclosed embodiments, the present disclosed embodiment describes in detail the basic substance composition judgment model. The explanations of the terms that are the same or corresponding to the above disclosed embodiments are not repeated here. The method can be performed by a multi-substance decomposition device based on a neural network, which can be implemented by software and / or hardware and integrated into a computer device with application development functions.

[0070] like Figure 4 As shown, the multi-material decomposition method based on neural network includes the following steps:

[0071] S210: Acquire a dual-energy CT image of the target object.

[0072] S220, inputting the dual-energy CT image into a pre-trained basic material composition judgment model to obtain a basic material composition judgment result of the target object.

[0073] S230, inputting the dual-energy CT image into a pre-trained base material absorption coefficient prediction model to obtain an equivalent absorption coefficient of each base material component in the target object.

[0074] S240. Construct a target multi-substance decomposition equation based on the basic substance composition judgment result and the equivalent absorption coefficient.

[0075] Specifically, the target multi-substance decomposition equation is constructed using the basic substance composition judgment results and the equivalent absorption coefficient output by the basic substance composition judgment model.

[0076] Taking the target object as the human abdomen as an example, the target multi-material decomposition equation can be:

[0077]

[0078] Among them, u LE 、u HE are the absorption coefficients of the target object, i.e., the abdomen of the human body, under low-energy X-rays LE and high-energy X-rays HE, respectively. LE is the pixel value of low-energy CT image, u HE is the pixel value of high-energy CT image, is the absorption coefficient of base substance i under low-energy and high-energy X-rays. For example, the type of base substance iodine can be marked as 1, and its low-energy absorption coefficient can be expressed as The high energy absorption coefficient can be expressed as α i is the volume fraction of base substance i, for example, the volume fraction of base substance iodine can be expressed as α1.

[0079] Optionally, a target multi-substance decomposition equation is constructed based on the basic substance composition judgment result and the equivalent absorption coefficient, including:

[0080] The target multi-material decomposition equations are established respectively for the mask image corresponding to each base material combination in the base material composition judgment result and the base material absorption coefficient matrix corresponding to the equivalent absorption coefficient.

[0081] Among them, each quantity of the target multi-substance decomposition equation corresponds to each quantity of the basic substance combination.

[0082] Specifically, the base substance composition judgment model outputs a black-and-white mask image with a distribution value of 0-1, and the number of channels of the mask image is consistent with the number of three base substance combinations in the three base substance combination table of the decomposed target object. For example, taking a dual-energy CT image of the abdomen of a human body injected with iodine contrast agent as an example, its three base substance combination table contains three base substance combinations: "iodine-blood-air", "fat-muscle-air", and "fat-bone-air". According to the combination, the number of channels of the black-and-white mask image output by the base substance composition judgment model is 3, and each channel corresponds to one of the three base substance combinations.

[0083] Substitute each combination of three base substances into the target multi-substance decomposition equation and solve it.

[0084] For example, when the combination "iodine-blood-air" is substituted into the target multi-substance decomposition equation, the types i of iodine, blood, and air are marked as 1, 2, and 3 respectively, then the volume fractions of iodine, blood, and air are α1, α2, and α3 respectively, and the low-energy absorption coefficients are as well as The high energy absorption coefficient is as well as Substitute the above-mentioned base material volume fraction, base material absorption coefficient, and target object absorption coefficient into equation (1).

[0085] S250, analyzing the target multi-substance decomposition equation to determine the basic substance components of the target object and the volume fraction of each basic substance component.

[0086] Specifically, solve equation (1) to determine whether 0≦α i ≦1, the basic material composition and volume fraction of the dual-energy CT image can be obtained.

[0087] S260 , obtaining a base material density distribution image of the target object according to each base material component of the target object and the volume fraction of each base material component.

[0088] Specifically, by combining the black and white mask image output by the base material composition judgment model with the equivalent absorption coefficient matrix output by the base material absorption coefficient prediction model, the three base material compositions of each dual-energy CT image pixel and the high and low energy absorption coefficients of the corresponding base materials can be obtained. By jointly establishing the target multi-material decomposition equation, the density distribution image of the base material can be output.

[0089] Specifically, Figure 5 As shown, the dual-energy CT image of the pig leg containing iodine solution is subjected to multi-substance (bone-iodine-water-muscle-fat-air) decomposition, and the dual-energy CT image is input into a pre-trained basic substance composition judgment model, and the basic substance composition judgment results of the pig leg containing iodine solution are obtained as "bone-fat-air", "iodine-water-air" and "muscle-fat-air", and the basic substance density distribution images obtained based on the basic substance composition judgment results and the equivalent absorption coefficient are the density distribution images of bone-iodine-water-muscle-fat, respectively.

[0090] The technical solution provided by the embodiment of the present invention obtains a dual-energy CT image of the target object; inputs the dual-energy CT image into a pre-trained basic material composition judgment model to obtain a basic material composition judgment result of the target object; the basic material composition judgment model includes four downsampling modules, four upsampling modules and an output module; wherein each downsampling module is composed of two convolutional layers and a maximum pooling layer; each upsampling module includes two convolutional layers and a deconvolution layer; the output module includes three convolutional layers. And, the dual-energy CT image is input into a pre-trained basic material absorption coefficient prediction model to obtain the equivalent absorption coefficient of each basic material component in the target object; based on the basic material composition judgment result and the equivalent absorption coefficient, the basic material density distribution image of the target object is obtained. The technical solution of the embodiment of the present invention solves the problem that the current basic material decomposition algorithm will introduce new imaging noise, resulting in an increase in the noise of the decomposed basic material image or a low accuracy of the decomposition result, and can model the basic material combination judgment process, improve the judgment accuracy of the basic material combination, and improve the decomposition accuracy of multiple materials.

[0091] Embodiment 3

[0092] Figure 6 It is a structural schematic diagram of a multi-substance decomposition device based on a neural network provided in an embodiment of the present invention. The embodiment of the present invention can be applied to the scenario of multi-substance decomposition of dual-energy CT images. The device can be implemented by software and / or hardware and integrated into a computer device with application development function.

[0093] like Figure 6 As shown, the multi-substance decomposition device based on neural network includes: an image acquisition module 310, a first image analysis module 320, a second image analysis module 330 and a base substance determination module 340.

[0094] The image acquisition module 310 is used to acquire a dual-energy CT image of the target object.

[0095] The first image analysis module 320 is used to input the dual-energy CT image into a pre-trained basic material composition judgment model to obtain a basic material composition judgment result of the target object.

[0096] The second image analysis module 330 is used to input the dual-energy CT image into a pre-trained base material absorption coefficient prediction model to obtain the equivalent absorption coefficient of each base material component in the target object.

[0097] The base material determination module 340 is used to obtain a base material density distribution image of the target object based on the base material composition judgment result and the equivalent absorption coefficient.

[0098] The technical solution provided by the embodiment of the present invention obtains a dual-energy CT image of the target object; inputs the dual-energy CT image into a pre-trained base material component judgment model to obtain a base material component judgment result of the target object; and inputs the dual-energy CT image into a pre-trained base material absorption coefficient prediction model to obtain an equivalent absorption coefficient of each base material component in the target object; based on the base material component judgment result and the equivalent absorption coefficient, a base material density distribution image of the target object is obtained. The technical solution of the embodiment of the present invention solves the problem that the current base material decomposition algorithm will introduce new imaging noise, resulting in an increase in the noise of the decomposed base material image or a low accuracy of the decomposition result. The base material combination judgment process can be modeled to improve the judgment accuracy of the base material combination and the multi-material decomposition accuracy.

[0099] Optionally, the base material determination module 340 includes:

[0100] An equation building unit is used to build a target multi-substance decomposition equation based on the base substance composition judgment result and the equivalent absorption coefficient.

[0101] The equation analysis unit is used to analyze the target multi-substance decomposition equation to determine the basic substance components of the target object and the volume fraction of each basic substance component.

[0102] The distribution image generating unit obtains a base material density distribution image of the target object according to each base material component of the target object and the volume fraction of each base material component.

[0103] Optionally, the equation building unit is specifically used for:

[0104] Establishing target multi-substance decomposition equations respectively for the mask image corresponding to each base substance combination in the base substance composition judgment result and the base substance absorption coefficient matrix corresponding to the equivalent absorption system;

[0105] Among them, each quantity of the target multi-substance decomposition equation corresponds to each quantity of the basic substance combination.

[0106] Optionally, the base material composition judgment model includes four down-sampling modules, four up-sampling modules and an output module;

[0107] Among them, each of the downsampling modules consists of two convolutional layers and a maximum pooling layer; each of the upsampling modules includes two convolutional layers and one deconvolutional layer; and the output module includes three convolutional layers.

[0108] Optionally, in the basic material composition judgment model, the activation function of the last convolutional layer is sigmoid, and the activation function of the convolutional layers other than the last convolutional layer is tanh.

[0109] Optionally, the output module includes a plurality of output channels, wherein each of the output channels corresponds to a base material combination and outputs a black-and-white mask image corresponding to the base material combination.

[0110] Optionally, the base material absorption coefficient prediction model includes 9 downsampling modules, wherein each of the downsampling modules includes two convolutional layers and one maximum pooling layer.

[0111] The neural network-based multi-substance decomposition device provided in the embodiment of the present invention can execute the neural network-based multi-substance decomposition method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0112] Embodiment 4

[0113] Figure 7 A schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Figure 7 A block diagram of an exemplary computer device 12 suitable for use in implementing embodiments of the present invention is shown. Figure 4 The computer device 12 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention. The computer device 12 can be any terminal device with computing capability and can be configured in the angiography image display device.

[0114] like Figure 7As shown, the computer device 12 is in the form of a general-purpose computing device. The components of the computer device 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 that connects various system components (including the system memory 28 and the processing unit 16).

[0115] The bus 18 may be one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor or a local bus using any of a variety of bus architectures. For example, these architectures include, but are not limited to, an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MAC) bus, an Enhanced ISA bus, a Video Electronics Standards Association (VESA) local bus, and a Peripheral Component Interconnect (PCI) bus.

[0116] The computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the computer device 12, including volatile and non-volatile media, removable and non-removable media.

[0117] The system memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache 32. The computer device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system 34 may be used to read and write non-removable, non-volatile magnetic media ( Figure 7 not shown, usually called a "hard drive"). Although Figure 7 Not shown in the figure, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, a DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to the bus 18 via one or more data medium interfaces. The system memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of the various embodiments of the present invention.

[0118] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in system memory 28, such program modules 42 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment. Program modules 42 generally perform the functions and / or methods of the embodiments described herein.

[0119] The computer device 12 may also communicate with one or more external devices 14 (e.g., keyboards, pointing devices, displays 24, etc.), one or more devices that enable a user to interact with the computer device 12, and / or any device that enables the computer device 12 to communicate with one or more other computing devices (e.g., network cards, modems, etc.). Such communication may be performed via an input / output (I / O) interface 22. Furthermore, the computer device 12 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 20. As shown, the network adapter 20 communicates with the other modules of the computer device 12 via the bus 18. It should be understood that although Figure 7 Not shown, other hardware and / or software modules may be used in conjunction with computer device 12, including but not limited to microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0120] The processing unit 16 executes various functional applications and data processing by running the program stored in the system memory 28, for example, implementing the multi-substance decomposition method based on a neural network provided in the embodiment of the present invention, the method comprising:

[0121] Acquire a dual-energy CT image of a target object;

[0122] Inputting the dual-energy CT image into a pre-trained basic material composition judgment model to obtain a basic material composition judgment result of the target object; and

[0123] Inputting the dual-energy CT image into a pre-trained base material absorption coefficient prediction model to obtain an equivalent absorption coefficient of each base material component in the target object;

[0124] Based on the base material composition determination result and the equivalent absorption coefficient, a base material density distribution image of the target object is obtained.

[0125] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the multi-substance decomposition method based on a neural network provided in any embodiment of the present invention is implemented, including:

[0126] Acquire a dual-energy CT image of a target object;

[0127] Inputting the dual-energy CT image into a pre-trained basic material composition judgment model to obtain a basic material composition judgment result of the target object; and

[0128] Inputting the dual-energy CT image into a pre-trained base material absorption coefficient prediction model to obtain an equivalent absorption coefficient of each base material component in the target object;

[0129] Based on the base material composition determination result and the equivalent absorption coefficient, a base material density distribution image of the target object is obtained.

[0130] The computer storage medium of the embodiment of the present invention can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to: an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0131] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0132] The program code embodied on the computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0133] Computer program code for performing the operations of the present invention may be written in one or more programming languages ​​or combinations thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0134] It should be understood by those skilled in the art that the modules or steps of the present invention described above can be implemented by a general-purpose computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, optionally, they can be implemented by a program code executable by a computer device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0135] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A multi-substance decomposition method based on neural network, characterized in that: include: Acquire a dual-energy CT image of a target object; Inputting the dual-energy CT image into a pre-trained basic material composition judgment model to obtain a basic material composition judgment result of the target object; and Inputting the dual-energy CT image into a pre-trained base material absorption coefficient prediction model to obtain an equivalent absorption coefficient of each base material component in the target object; Based on the base material composition determination result and the equivalent absorption coefficient, a base material density distribution image of the target object is obtained.

2. The method according to claim 1, characterized in that Based on the base material composition determination result and the equivalent absorption coefficient, a base material density distribution image of the target object is obtained, including: Constructing a target multi-substance decomposition equation based on the base substance composition judgment result and the equivalent absorption coefficient; Analyze the target multi-substance decomposition equation to determine the basic substance components of the target object and the volume fraction of each basic substance component; A base material density distribution image of the target object is obtained according to each base material component of the target object and the volume fraction of each base material component.

3. The method according to claim 2, characterized in that The target multi-substance decomposition equation is constructed based on the base substance composition judgment result and the equivalent absorption coefficient; Establishing target multi-substance decomposition equations respectively for the mask image corresponding to each base substance combination in the base substance composition judgment result and the base substance absorption coefficient matrix corresponding to the equivalent absorption coefficient; Among them, each quantity of the target multi-substance decomposition equation corresponds to each quantity of the basic substance combination.

4. The method according to any one of claims 1 to 3, characterized in that: The basic material composition judgment model includes four down-sampling modules, four up-sampling modules and an output module; Among them, each of the downsampling modules consists of two convolutional layers and a maximum pooling layer; each of the upsampling modules includes two convolutional layers and one deconvolutional layer; and the output module includes three convolutional layers.

5. The method according to claim 4, characterized in that In the basic material composition judgment model, the activation function of the last convolutional layer is sigmoid, and the activation function of the convolutional layers other than the last convolutional layer is tanh.

6. The method according to claim 4, characterized in that The output module includes a plurality of output channels, wherein each of the output channels corresponds to a base material combination and outputs a black and white mask image corresponding to the base material combination.

7. The method according to any one of claims 1 to 3, characterized in that: The base material absorption coefficient prediction model includes 9 downsampling modules, wherein each of the downsampling modules includes two convolutional layers and one maximum pooling layer.

8. A multi-substance decomposition device based on a neural network, characterized in that: The device comprises: An image acquisition module, used for acquiring a dual-energy CT image of a target object; A first image analysis module is used to input the dual-energy CT image into a pre-trained basic material composition judgment model to obtain a basic material composition judgment result of the target object; and A second image analysis module is used to input the dual-energy CT image into a pre-trained base material absorption coefficient prediction model to obtain an equivalent absorption coefficient of each base material component in the target object; The base material determination module is used to obtain a base material density distribution image of the target object based on the base material composition judgment result and the equivalent absorption coefficient.

9. A computer device, characterized in that: The computer device comprises: one or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the neural network-based multi-substance decomposition method as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the neural network-based multi-substance decomposition method as described in any one of claims 1 to 7 is implemented.

Images