Model training method, target element determination method, apparatus and device

Abstract

The application provides a model training method, a target element determination method, device and equipment, relates to physics and computer field, and particularly relates to application of physics and computer technology in the field of radiotherapy. The model training method comprises the following steps: updating element density information of a plurality of original materials based on a target element, and obtaining a plurality of sample materials; determining a plurality of sample CT image pairs associated with the sample materials based on X characteristic parameters of a target CT device and the element density information of the sample materials; wherein the sample CT image pair comprises a first sample CT image containing the target element and a second sample CT image not containing the target element, and X is a positive integer; training an initial model according to the plurality of sample CT image pairs to obtain a target model corresponding to the target CT device, wherein the target model is used to determine the target element content in an original CT image scanned by the target CT device and a target CT image after the target element is separated.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the fields of physics and computer, in particular to the application of physics and computer technology in the field of radiotherapy, and more particularly to a model training method, a target element determination method, an apparatus and a device. BACKGROUND

[0002] Computed Tomography (CT) imaging technology has a wide application in the field of medical diagnosis. The internal structure of the detected object can be presented by CT imaging technology and image reconstruction technology. The imaging clarity of the CT image of the to-be-detected substance can be effectively improved by injecting a specific element into the to-be-detected substance and then performing CT imaging, but the CT image obtained after injecting the specific element cannot be used for subsequent calculation. SUMMARY

[0003] The present disclosure provides a model training method, a target element determination method, an apparatus and a device.

[0004] According to a first aspect of the present disclosure, a model training method is provided, comprising: updating element density information of a plurality of original materials based on a target element respectively to obtain a plurality of sample materials; determining a plurality of sample CT image pairs associated with the sample materials based on X characteristic parameters of a target CT device and the element density information of the sample materials; wherein the sample CT image pair comprises a first sample CT image containing the target element and a second sample CT image not containing the target element, and X is a positive integer; training an initial model according to the plurality of sample CT image pairs to obtain a target model corresponding to the target CT device, wherein the target model is used to determine the content of the target element in an original CT image scanned by the target CT device and a target CT image after separating the target element.

[0005] According to an embodiment of the present disclosure, the element density information of the plurality of sample materials is obtained by updating the element density information of the plurality of original materials based on the target element respectively, comprising: for each original material, the following operations are performed: in a predetermined range, the target element is randomly added to the original material to obtain Y sample materials corresponding to the original material, wherein the mass proportion of the target element in different sample materials is different, and Y is a positive integer; determining the element density information of the Y sample materials based on the mass proportion of the target element. According to an embodiment of the present disclosure, the plurality of sample CT image pairs associated with the sample materials are determined based on the X characteristic parameters and the element density information of the plurality of sample materials respectively, comprising: determining X*Y first sample CT images based on the X characteristic parameters and the element density information of the Y sample materials respectively; removing the content of the target element in the element density information of the Y sample materials respectively, and determining X*Y second sample CT images based on the element density information after removing the content of the target element.

[0006] According to an embodiment of the present application, determining a plurality of sample CT image pairs associated with sample materials based on X characteristic parameters and element density information of the plurality of sample materials respectively comprises: for each sample material, calculating a CT value of each voxel in the sample material based on the characteristic parameters and the element density information of the sample material; and determining a sample CT image corresponding to the sample material based on the CT value of each voxel.

[0007] According to an embodiment of the present application, training an initial model based on the plurality of sample CT image pairs to obtain a target model corresponding to a target CT device comprises: adding noise information to a first sample CT image associated with each sample material to generate an enhanced sample CT image pair associated with each sample material, wherein the number of the enhanced sample CT image pairs is greater than the number of the sample CT image pairs; and training the initial model based on the enhanced sample CT image pairs to obtain the target model corresponding to the target CT device.

[0008] According to an embodiment of the present application, training an initial model based on the plurality of sample CT image pairs to obtain a target model corresponding to a target CT device comprises: determining a first sample CT image in the enhanced sample CT image pair as input data of the initial model, and determining a target element content in the sample material and a second sample CT image in the enhanced sample CT image pair as output data of the initial model; and training the initial model based on the input data and the output data to obtain the target model.

[0009] According to a second aspect of the present application, a target element content determination method is provided, comprising: respectively scanning a to-be-measured substance based on X scanning energy spectrums of a target CT device to obtain X CT images; and inputting the X CT images into an element analysis model corresponding to the target CT device to determine a target element content in the to-be-measured substance and a CT image not containing the target element based on the X CT images by the element analysis model; wherein the element analysis model is a target model trained by the above model training method.

[0010] According to a third aspect of the present application, a model training apparatus is provided, comprising: an obtaining module configured to update element density information of a plurality of original materials based on a target element to obtain a plurality of sample materials; a first determining module configured to determine a plurality of sample CT image pairs associated with the sample materials based on X characteristic parameters of a target CT device and element density information of the plurality of sample materials; wherein the sample CT image pair comprises a first sample CT image containing the target element and a second sample CT image not containing the target element; and a training module configured to train an initial model based on the plurality of sample CT image pairs to obtain a target model corresponding to the target CT device, wherein the target model is used to determine a target element content in an original CT image scanned by the target CT device and a target CT image after separating the target element.

[0011] According to a fourth aspect of the present application, a target element content determination apparatus is provided, comprising: a scanning module configured to scan a to-be-tested substance based on X scan energy spectrums of a target CT device respectively, to obtain X CT images; and a second determination module configured to input the X CT images into an element analysis model corresponding to the target CT device, to determine, by the element analysis model, a target element content in the to-be-tested substance and a CT image not containing the target element based on the X CT images; wherein the element analysis model is a target model trained by the model training apparatus.

[0012] According to a fifth aspect of the present application, an electronic device is provided, comprising: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the above method. BRIEF DESCRIPTION OF DRAWINGS

[0013] The above and other objects, features and advantages of the present application will become more apparent from the following description of embodiments of the present application, taken in conjunction with the accompanying drawings, in which:

[0014] Figure 1 A flowchart of a model training method according to an embodiment of the present application is schematically shown;

[0015] Figure 2 A flowchart of updating element density information of a plurality of original materials based on a target element respectively to obtain a plurality of sample materials according to an embodiment of the present application is schematically shown;

[0016] Figure 3 A flowchart of determining a plurality of sample CT image pairs associated with sample materials based on X characteristic parameters and element density information of the plurality of sample materials respectively according to an embodiment of the present application is schematically shown;

[0017] Figure 4 A flowchart of training an initial model based on the plurality of sample CT image pairs to obtain a target model corresponding to a target CT device according to an embodiment of the present application is schematically shown;

[0018] Figure 5 A flowchart of a target element content determination method according to an embodiment of the present application is schematically shown;

[0019] Figure 6 A structural block diagram of a model training apparatus according to an embodiment of the present application is schematically shown;

[0020] Figure 7 A structural block diagram of a target element content determination apparatus according to an embodiment of the present application is schematically shown;

[0021] Figure 8 A block diagram of an electronic device according to an embodiment of a method of the present application is schematically shown. DETAILED DESCRIPTION

[0022] In order to make the objects, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the embodiments and the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, and not all the embodiments. Based on the embodiments of the present application, all the other embodiments obtained by those skilled in the art without creative work fall within the scope of the present application.

[0023] The terms used herein are only used to describe specific embodiments, and are not intended to limit the present application. The terms "include", "contain" and the like used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0024] In the present application, unless otherwise explicitly specified and limited, the terms "mount", "connect", "connect", "fix" and the like should be understood in a broad sense, for example, can be fixedly connected, or can be detachably connected, or can be integrated; can be mechanically connected, or can be electrically connected or can communicate with each other; can be directly connected, or can be indirectly connected through an intermediate medium, can be the internal communication of two elements or the interaction relationship between two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0025] In the description of the present application, it should be understood that the terms "longitudinal", "length", "circumferential", "front", "back", "left", "right", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only used to facilitate the description of the present application and simplify the description, and do not indicate or imply that the subsystems or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application.

[0026] Throughout the drawings, the same elements are denoted by the same or similar reference numerals. When it may cause confusion in understanding the present application, the conventional structure or configuration will be omitted. And the shape, size, positional relationship of each component in the drawing does not reflect the true size, proportion and actual positional relationship. In addition, any reference symbol located between parentheses should not be constructed as a limitation.

[0027] Similarly, to the extent that the foregoing description contains object-oriented terminology, such terminology is used by way of explanation rather than limitation. It is contemplated that the inventive subject matter can be implemented in a wide variety of environments and settings, and that the scope of the invention is not limited to the specific examples described herein. In addition, although the subject matter has been described in language specific to structural features, it is to be understood that the inventive subject matter can not be limited to the specific features described. Rather, the specific features of the inventive subject matter can be described in the application as a whole, and claimed using the terminology "means for" in the claims set forth below.

[0028] In addition, the terms "first", "second", etc. are used herein only to describe various features, and do not imply a relative importance or a specific order of advantage according to the features. Therefore, the features with "first", "second" can explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "a plurality of" is at least two, for example, two, three, etc., unless otherwise specifically limited.

[0029] The embodiments of the present disclosure provide a model training method and a target element determination method and device, and before introducing the technical solutions provided by the embodiments of the present disclosure, the related technologies involved in the present disclosure are described.

[0030] In some CT imaging methods, trace elements in the human body, such as iron, copper, iodine, etc. with high atomic number, can be identified by dual-energy CT. Among them, dual-energy CT refers to an energy spectrum with two or more energy components different from each other.

[0031] By injecting a specific element into the material to be tested, the imaging clarity of the CT image is improved, thereby better assisting in delineating the target area and critical organs. However, the CT image obtained after injecting the specific element cannot be used for dose calculation. On the one hand, the specific element significantly increases the stopping power of the specific material, affecting the range calculation of the proton / heavy ion beam. On the other hand, the specific element has a fast metabolism time in the human body. From the CT image acquisition to the start of radiotherapy, usually several days are needed. When the radiotherapy starts, the specific element is no longer contained in the material to be tested. At this time, the material element density distribution in the target area is different from the CT image containing the specific element. The dose calculation based on the CT image containing the specific element will cause calculation errors. Therefore, it is necessary to separate the specific element in the CT image to obtain a pseudo-regular CT image that does not contain the specific element, and to perform subsequent calculations based on the pseudo-regular CT image. The existing specific element separation uses a material decomposition method, that is, it is assumed that there are only iodine and water in the material, and the material is approximated as water, which does not conform to the actual situation, which is easy to cause errors. Moreover, this method is easily affected by image noise and artifacts, and may produce deviations.

[0032] Figure 1 A flowchart of a model training method according to an embodiment of the present application is schematically shown.

[0033] As shown in Figure 1 The model training method of this embodiment includes operations S110-S130.

[0034] In operation S110, the element density information of a plurality of original materials is updated based on a target element, respectively, to obtain a plurality of sample materials.

[0035] In some embodiments, the element density information of the original material can be prior information obtained from a database in advance. The element density information of different types of original materials is different. The element density information of each original material can include the types of elements contained in the tissue material, the density of each element, and the distribution of multiple elements in the tissue material. The sample material is used as a basic database for subsequent determination of a plurality of sample CT image pairs.

[0036] In some embodiments, for each original material, a mass proportion of the target element is randomly assigned within a reasonable range to obtain a plurality of sample materials corresponding to the original material, the element density information of the sample material being different from the element density information of the original material. Each original material A can correspond to a plurality of sample materials B. The number of sample materials corresponding to each original material is the same as the number of mass proportions of the target element assigned to the original material. For example, if the random proportion of the target element corresponding to each original material is 100, then the number of sample materials corresponding to the original material is also 100. Taking the original material as bone tissue material and the target element as iodine element as an example, the mass proportion of the iodine element is randomly assigned to the bone tissue material within a predetermined range to obtain 100 bone tissue materials with different mass proportions of iodine element. Taking 71 different tissue materials as original materials as an example, by randomly assigning a mass proportion of the target element to each of the original materials, 7100 sample materials can be obtained.

[0037] The element density information of each sample material is obtained by updating the element density of the original material based on the mass proportion of the target element after the mass proportion of the target element is randomly assigned.

[0038] It should be noted that due to the characteristics of each tissue material, the target element is not present in all voxels. Therefore, when constructing sample materials, a sample material with a target element mass proportion of 0% should be provided for each original material to simulate the local characteristics of the target element in real CT, thereby improving the accuracy of subsequent model training and improving the prediction accuracy of the model.

[0039] In operation S120, a plurality of sample CT image pairs associated with the sample materials are determined based on the X characteristic parameters of the target CT device and the element density information of the plurality of sample materials. The sample CT image pair includes a first sample CT image containing the target element and a second sample CT image not containing the target element.

[0040] In some embodiments, the characteristic parameters are associated with a scanning energy spectrum used by the target CT device for scanning. Different scanning energy spectrums correspond to different characteristic parameters. The characteristic parameters can be used to describe the energy spectrum characteristics of the target CT device. The reference phantom can be scanned using the scanning energy spectrum of the target CT device to determine the characteristic parameters corresponding to the scanning energy spectrum, and the sample CT image pairs associated with the sample materials are determined according to the plurality of characteristic parameters, wherein the sample CT image pair includes a first sample CT image containing the target element and a second sample CT image not containing the target element.

[0041] In some embodiments, the tissue material is scanned by a spectrum to obtain a CT image of the tissue material, the CT image including a distribution of CT values of the tissue material, the CT values reflecting interactions between matter and photons. The CT image includes a distribution of CT values of the tissue material, the CT values reflecting interactions between matter and photons. The target CT device can be a CT device capable of transmitting multiple energy spectra, and the target device scans the tissue material based on energy spectra different in energy composition, and the obtained CT images can be different, effectively avoiding the case that the CT values of different element composition tissue materials are the same.

[0042] Taking a dual-energy CT device as an example, the dual-energy CT device has two scanning energy spectra. When multiple tissue materials are scanned based on a single ray source, the element compositions of the multiple tissue materials are different but the CT values under a single energy spectrum can be the same. In this case, the two tissue materials cannot be further distinguished in the CT image, which leads to deviation of the diagnostic result. Therefore, by using a dual-energy CT, two CT images of the two tissue materials scanned based on two energy spectra can be compared. When the CT values of the two tissue materials are the same under a certain energy spectrum, the CT values under another energy spectrum are often different, thereby further distinguishing the two tissue materials.

[0043] For example, a virtual scanning environment of the target CT device is simulated by a characteristic parameter, and a virtual tissue corresponding to the sample material is simulated based on the element density information of the sample material, and a sample CT image obtained by scanning the virtual tissue in the virtual scanning environment is calculated.

[0044] In some embodiments, the first sample CT image and the second sample CT image are calculated based on the characteristic parameters of the element density distribution and the scanning energy spectrum. Based on the characteristic parameters, the scanning environment of the CT device can be simulated, and the photon energy spectrum emitted by the target CT device can be determined. Based on the element density information, the element density at different positions in the sample material can be simulated, thereby calculating the absorption capacity of the photon energy at different positions in the sample material, and further calculating the CT values at different positions in the sample material based on the absorption capacity, to obtain the first sample CT image and the second sample CT image.

[0045] For example, a plurality of first sample CT images associated with the sample material are calculated based on the characteristic parameters of different scanning energy spectra and the element density information of the sample material, and it should be noted that for the sample material with a target element content of 0%, the sample CT image is also marked as a first sample CT image.

[0046] And the target element is removed from the element density information of each sample material, and a plurality of second sample CT images associated with the sample material are calculated based on the element density information after the target element is removed.

[0047] It should be noted that the first sample CT image and the second sample CT image mentioned in the embodiments of the present disclosure are obtained based on the characteristic parameters of at least two scanning energy spectrums, that is, the first sample CT image and the second sample CT image are determined by using the characteristic parameters of two scanning energy spectrums, and the first sample CT image refers to a plurality of CT images containing a target element obtained based on different energy spectrum characteristic parameters, and the second sample CT image refers to a plurality of CT images not containing a target element obtained based on different energy spectrum characteristic parameters.

[0048] Taking a target CT device as a dual-energy CT device and the characteristic parameters as K 1,H and K 1,L , for example, the first sample CT image in each sample CT image pair refers to a dual-energy first sample CT image, and the second sample CT image refers to a dual-energy second sample CT image. Wherein, K 1,H is a characteristic parameter of a high-energy CT spectrum, and K 1,L is a characteristic parameter of a low-energy CT spectrum. Here, the high-energy spectrum and the low-energy spectrum are relative concepts. The dual-energy CT device simultaneously generates two groups of scanning energy spectrums, and the high-energy spectrum is the one with high energy and the low-energy spectrum is the one with low energy.

[0049] In operation S130, an initial model is trained according to the plurality of sample CT image pairs to obtain a target model corresponding to the target CT device, wherein the target model is used to determine the content of the target element in the original CT image scanned by the target CT device and the target CT image after the target element is separated.

[0050] In some embodiments, the initial model can be a fully connected neural network. By training the fully connected neural network using the plurality of sample CT image pairs, the correlation between the CT value of the first sample CT image and the content of the target element and the mapping relationship between the first sample CT image and the second sample CT image can be extracted by using the fully connected operation, so that the target model obtained by training can determine the content of the target element based on the original CT image scanned by the CT device, and generate a CT image not containing the target element (i.e., the target CT image) based on the original CT image. Wherein, the original CT image refers to a CT image scanned after the target element is injected into the material to be tested. The content of the target element refers to the density mass ratio of the target element in the material to be tested.

[0051] It should be noted that the energy spectrum parameters of different CT devices can be different. For each CT device, the above method needs to be used. Figure 1The model training method shown trains the target model corresponding to the CT device. The model training method provided in the embodiments of the present disclosure introduces known tissue materials as prior information, randomly increases the mass proportion of the target element in the tissue materials, obtains a plurality of sample materials, calculates the CT images corresponding to different energy spectrums of the sample materials in the case of containing the target element and not containing the target element, constructs a sample CT image pair for training the model, uses a machine learning method, takes the CT image containing the target element as input data, takes the CT image not containing the target element and the mass proportion of the target element as output data, and trains the target model, so as to realize the separation of the content of the target element in the CT image through the target model, and facilitate to improve the accuracy of subsequent operations.

[0052] Figure 2 A flowchart for updating element density information of a plurality of original materials based on a target element respectively and obtaining a plurality of sample materials according to the embodiments of the present disclosure is schematically shown.

[0053] As Figure 2 shown, the embodiment of updating the element density information of the plurality of original materials based on the target element respectively and obtaining the plurality of sample materials includes operations S210-S220, and operations S210-S220 are performed for each original material.

[0054] In operation S210, the target element is randomly added in the original material in a preset range, and Y sample materials corresponding to the original material are obtained, wherein the mass proportion of the target element in different sample materials is different, and Y is a positive integer.

[0055] In operation S220, the element density information of each of the Y sample materials is determined based on the mass proportion of the target element.

[0056] In some embodiments, the original material and the element density information thereof can be obtained from a database in advance. The original material includes a plurality of tissue materials, such as bone tissue material, soft tissue material, etc., and the element density information of different original materials is different.

[0057] For each tissue material, the mass proportion of the target element is randomly assigned in a preset range, and Y sample materials corresponding to the original material are obtained, wherein the mass proportion of the target element in different sample materials is different.

[0058] Taking the target element as iodine element and Y = 100 as an example, for each tissue material, the mass proportion of iodine element is randomly assigned in a preset range, and each tissue material can correspond to 100 random proportions of iodine element, so as to obtain 100 sample materials corresponding to the original material. The preset range may be 0% to 3%, for example. It should be noted that for each tissue material, the sample material with a mass proportion of 0% of iodine element is also included.

[0059] In some embodiments, based on the mass proportion of the target element in the sample material, the element density information corresponding to each sample material is updated, for example, the mass proportion of each element in the normalized element density information is updated, so that the total mass proportion of each element in the sample material is 100%.

[0060] By randomly assigning the content of the target element to the original material, the difference in uptake of the target element by different tissue materials in a real scenario can be simulated, ensuring that the training data can cover the full range from no target element to high concentration of target element, effectively improving data diversity and thus improving the generalization ability of the trained model. Moreover, the original material is similar to the real tissue material, and the sample material obtained based on this method can more accurately reflect the complex composition in the actual tissue material, thereby improving the accuracy of the target element separation model.

[0061] Figure 3 A flowchart for determining a plurality of sample CT image pairs associated with sample materials based on X feature parameters and element density information of the sample materials, according to an embodiment of the present disclosure, is shown schematically.

[0062] As shown in Figure 3 The embodiment of determining a plurality of sample CT image pairs associated with sample materials based on X feature parameters and element density information of the sample materials includes operations S310 to S320.

[0063] In operation S310, X*Y first sample CT images are determined based on X feature parameters and element density information of Y sample materials.

[0064] In operation S320, the content of the target element in the element density information of the Y sample materials is removed, and X*Y second sample CT images are determined based on the element density information after removing the content of the target element.

[0065] In some embodiments, the sample CT image of each sample material can be divided into a plurality of voxels, and a voxel is the smallest volume unit in a CT image. The CT value corresponding to each voxel is used to represent the interaction between the substance in the volume unit and the photon. If the substance in the voxel has strong absorption ability to photon energy, the CT value of the voxel will also be high.

[0066] In some embodiments, the following operations are performed on each sample material: based on the characteristic parameter and the element density information of the sample material, the CT value of each voxel in the sample material is calculated; based on the CT value of each voxel, the sample CT image corresponding to the sample material is determined. For example, based on the element density information of the sample material and the characteristic parameter, the attenuation coefficient μ of each voxel in the sample material can be calculated, and the CT value of each voxel can be calculated based on the attenuation coefficient μ, and the sample CT image of the sample material can be obtained based on the CT value of each voxel of the sample material.

[0067] The characteristic parameter is k 1,H and k 2,L For example, k 1,H is the characteristic parameter corresponding to the high-scan energy spectrum, k 2,L is the characteristic parameter corresponding to the low-scan energy spectrum, and based on k 1,H and k 2,L and the element density information of each sample material, the high-energy and low-energy first sample CT images of each sample material are determined, which are denoted as the dual-energy first sample CT images. It should be noted that for the sample material with a target element content of 0%, the sample CT image is also denoted as the dual-energy first sample CT image.

[0068] After obtaining the dual-energy first sample CT images of each sample material, the target element content in the element density information of each sample material is removed, and based on the element density information after removing the target element content and the characteristic parameter, the high-energy and low-energy second sample CT materials of each sample material are determined, which are denoted as the dual-energy second sample CT images, and the sample CT image pair associated with each sample material is obtained, each sample CT image pair contains the dual-energy first sample CT image and the dual-energy second sample CT image of the sample material.

[0069] For example, the characteristic parameter corresponding to the scan energy spectrum of the target CT device can be determined by a known reference phantom. For example, the X scan data of the reference material is obtained by scanning the reference material with the X scan energy spectrum distribution of the CT device; and the X characteristic parameters of the X scan energy spectrum are obtained by comparing the X scan data with the reference data of the reference material obtained in advance.

[0070] The reference material can be a known standard phantom. For example, the element density distribution, the total density of the material, and the CT image obtained based on the known scan energy spectrum of the reference material are known.

[0071] For example, the reference material can be a Gammex phantom, a CIRS phantom, or the like. The Gammex phantom is scanned by the target CT device based on M scan energy spectrums, and X CT images of the Gammex phantom are obtained as scan data. Based on the known CT images of the Gammex phantom, the element density distribution, and the total density of the material, the absorption degree of each voxel in the Gammex phantom to a specific photon energy can be determined. Based on the absorption degree and the scan data, the photon energy emitted by the emission source of the CT device is calculated, and thus the characteristic energy spectrum of the CT device is determined.

[0072] For example, the absorption ability of a voxel with a certain material in the Gammex phantom to a ray with a specific energy can be considered as fixed. Based on the known CT images of the Gammex phantom, the correspondence between the CT value and the photon energy of the energy spectrum is determined. Based on this correspondence, the photon energy of the energy spectrum is calculated based on the CT value in the scan data, and thus the characteristic parameters of the X scan energy spectrums can be determined based on the X CT images in the scan data.

[0073] Figure 4 A flowchart of training an initial model according to a plurality of sample CT images to obtain a target model corresponding to a target CT device is schematically shown.

[0074] As shown in Figure 4 The embodiment of training an initial model according to a plurality of sample CT images to obtain a target model corresponding to a target CT device includes operations S410-S420.

[0075] In operation S410, noise information is added to each sample material associated sample CT image pair to generate an enhanced sample CT image pair associated with each sample material, and the number of the enhanced sample CT image pairs is greater than the number of the sample CT image pairs.

[0076] In operation S420, the initial model is trained based on the enhanced sample CT image pairs to obtain a target model corresponding to the target CT device.

[0077] In some embodiments, considering the noise (such as device error, environmental interference, etc.) in the actual CT scanning process, a basic data set expansion can be performed based on the sample CT image pairs to obtain a model training database, wherein the model training database includes a plurality of enhanced sample CT image pairs, the enhanced sample CT image pairs are obtained by adding noise to the first sample CT image, and the model trained by the model training data set can adapt to the uncertainty of the real CT scanning scene, and the robustness of the model is improved.

[0078] Exemplarily, assuming that the measurement error of each first sample CT image obeys a Gaussian distribution, independent noise is added to the high-energy CT value and the low-energy CT value of each original first sample CT image respectively to generate a new perturbed data (i.e., a first sample CT image after noise), and an enhanced sample CT image pair is constructed based on the first sample CT image after noise to complete the expansion of the sample CT image pair. Exemplarily, the second sample CT image in the enhanced sample CT image pair and the target element mass fraction remain unchanged, and the enhanced sample CT image pair is obtained by establishing a corresponding relationship between the first sample CT image after noise, the second sample CT image and the target element mass fraction.

[0079] Based on the enhanced sample CT image pair obtained by adding noise, the real scene can be better simulated, and based on the enhanced sample CT image pair, the model training can make the model contact more noise data in the training, learn to ignore random disturbance, and improve the accuracy and robustness of the model prediction. Moreover, data increase can expand the scale of the training set and avoid model overfitting.

[0080] In some embodiments, the first sample CT image in the enhanced sample CT image pair is determined as the input data of the initial model, the target element content in the sample material and the second sample CT image in the enhanced sample CT image pair are determined as the output data of the initial model, and the initial model is trained based on the input data and the output data to obtain a target model.

[0081] In some embodiments, the first sample CT image is taken as the input data of the model, and the target element content and the second sample CT image are taken as the output data of the model, so that the model can calculate the output target element content and the CT image without the target element based on the first sample CT image.

[0082] Exemplarily, the model can make independent prediction based on voxels, calculate the target element density mass fraction of each voxel and the second CT value of the voxel based on the first CT value of the voxel. The target element content corresponding to the sample material is obtained by combining the target element content of each voxel in the sample material, and the second sample CT image of the sample material is obtained by combining the second CT values of each voxel in the sample material according to the spatial position. Wherein, the first CT value can be the CT value of the voxel in the first sample CT image, and the second CT value can be the CT value of the voxel in the second sample CT image.

[0083] Continuing to take the target element as the iodine element as an example, the dual-energy iodine-containing CT value of the sample object is taken as the input data of the model, and the dual-energy non-iodine-containing CT value of the sample object and the iodine element density mass fraction are taken as the output data, so that the model can determine the iodine element density mass fraction and the dual-energy non-iodine-containing CT value based on the dual-energy iodine-containing CT value of the sample object. Wherein, the sample object is a voxel in the sample material.

[0084] Exemplarily, in the training process, the model establishes a mapping from the dual-energy iodine-containing CT value to the iodine element density mass ratio and a mapping from the dual-energy iodine-containing CT value to the dual-energy non-iodine element CT value. The model learns the attenuation characteristics of the iodine element under different energies and different tissue materials through the dual-energy iodine-containing CT value, decomposes the CT value containing iodine into a combination of "non-iodine element tissue" and "iodine element", and thereby outputs the non-iodine CT value and the iodine element density mass ratio.

[0085] Figure 5 FIG. 1 is a flowchart of a target element content determination method according to an embodiment of the present disclosure.

[0086] As shown in FIG. 1, the target element content determination method according to the embodiment includes operations S510-S520. Figure 5

[0087] In operation S510, the target substance is scanned based on X scan energy spectrums of a target CT device respectively, and X CT images are obtained.

[0088] In operation S520, the X CT images are input to an element analysis model corresponding to the target CT device, and the element analysis model determines the target element content in the target substance and a CT image not containing the target element based on the X CT images.

[0089] In the embodiment of the present disclosure, the element analysis model is a target model trained by the model training method provided in the embodiment of the present disclosure. Figures 1 to 4

[0090] For example, the X CT images containing a specific element are input to the element analysis model, and the element analysis model can output the content of the specific element and a conventional CT image not containing the specific element.

[0091] In the embodiment of the present disclosure, the process of determining the element density distribution of the target substance based on the M CT images through the element analysis model is similar to the training process described above, and is not described herein for simplicity.

[0092] Figure 6 FIG. 6 is a structural block diagram of a model training apparatus according to an embodiment of the present disclosure.

[0093] As shown in FIG. 6, the model training apparatus 600 according to the embodiment includes an obtaining module 610, a first determining module 620, and a training module 630. Figure 6

[0094] ​​​The obtaining module 610 is used to update the elemental density information of multiple original materials based on the target element, thereby obtaining multiple sample materials. In one embodiment, the obtaining module 610 can be used to perform the operation S110 described above, which will not be repeated here.

[0095] The first determining module 620 is used to determine multiple pairs of sample CT images associated with sample materials based on X feature parameters of the target CT device and element density information of multiple sample materials; wherein, the pair of sample CT images includes a first sample CT image containing the target element and a second sample CT image not containing the target element. In one embodiment, the first determining module 620 can be used to perform the operation S120 described above, which will not be repeated here.

[0096] The training module 630 is used to train an initial model based on multiple sample CT images to obtain a target model corresponding to the target CT device. The target model is used to determine the distribution information content of target elements in the original CT image obtained by the target CT device and the target CT image after separating the target elements. In one embodiment, the training module 630 can be used to perform the operation S130 described above, which will not be repeated here.

[0097] According to embodiments of the present invention, any plurality of modules among the obtaining module 610, the first determining module 620, and the training module 630 may be combined into one module, or any one of these modules may be split into multiple modules. Alternatively, at least a portion of the functionality of one or more of these modules may be combined with at least a portion of the functionality of other modules and implemented in one module. According to embodiments of the present invention, at least one of the obtaining module 610, the first determining module 620, and the training module 630 may be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or any other reasonable means of integrating or packaging circuitry, or implemented in software, hardware, or firmware, or in any appropriate combination of any of these three implementation methods. Alternatively, at least one of the obtaining module 610, the first determining module 620, and the training module 630 may be at least partially implemented as a computer program module, which, when run, can perform corresponding functions.

[0098] Figure 7 A schematic block diagram of a target element content determination device according to an embodiment of the present invention is shown.

[0099] like Figure 7 As shown, the target element content determination device 700 of this embodiment includes a scanning module 710 and a second determination module 720.

[0100] The scanning module 710 is configured to scan the target material based on X scan energy spectrums of the target CT device respectively, and obtain X CT images. In an embodiment, the scanning module 710 can be configured to perform the operation S510 described above, and details are not repeated here.

[0101] The second determining module 720 is configured to input the X CT images into an element analysis model corresponding to the target CT device, and determine the target element content in the target material and the CT image not containing the target element based on the X CT images by the element analysis model. In an embodiment, the second determining module 720 can be configured to perform the operation S520 described above, and details are not repeated here.

[0102] The element analysis model is a target model trained by a device as shown in Figure 6 The element analysis model is a target model trained by a device as shown in

[0103] According to embodiments of the present application, any of the scanning module 710 and the second determining module 720 can be combined in one module, or any of them can be split into multiple modules. Alternatively, at least part of the function of one or more of these modules can be combined with at least part of the function of other modules, and implemented in one module. According to embodiments of the present application, at least one of the scanning module 710 and the second determining module 720 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on chip, a system on board, a system on package, an application specific integrated circuit (ASIC), or any other reasonable way of hardware or firmware that can be integrated or packaged, or any one of software, hardware and firmware or any appropriate combination of any of them. Alternatively, at least one of the scanning module 710 and the second determining module 720 can be at least partially implemented as a computer program module which can perform corresponding functions when running.

[0104] Figure 8 The block diagram of an electronic device according to the method of embodiments of the present application is schematically shown.

[0105] As shown in Figure 8As shown, the electronic device 800 according to an embodiment of the present application includes a processor 801 which can perform various appropriate actions and processes in accordance with a program stored in a read only memory (ROM) 802 or a program loaded from a storage section 808 into a random access memory (RAM) 803. The processor 801 can include, for example, a general purpose microprocessor (e.g., a CPU), an instruction set processor, and / or a related chip set, and / or a special purpose microprocessor (e.g., an application specific integrated circuit (ASIC)), and so on. The processor 801 can also include an on-board memory for cache use. The processor 801 can include a single processing unit or multiple processing units to perform the various actions of the method processes according to embodiments of the present application.

[0106] In the RAM 803, various programs and data required for the operation of the electronic device 800 are stored. The processor 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. The processor 801 performs various operations of the method processes according to embodiments of the present application by executing the programs in the ROM 802 and / or the RAM 803. Note that the programs can also be stored in one or more memories other than the ROM 802 and the RAM 803. The processor 801 can also perform various operations of the method processes according to embodiments of the present application by executing the programs stored in the one or more memories.

[0107] According to an embodiment of the present application, the electronic device 800 can further include an input / output (I / O) interface 805 which is also connected to the bus 804. The electronic device 800 can further include one or more of the following components connected to the input / output (I / O) interface 805: an input section 806 including a keyboard, a mouse, etc.; an output section 807 including a display such as a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 808 including a hard disk, etc.; and a communication section 809 including a network interface card such as a LAN card, a modem, etc. The communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the input / output (I / O) interface 805 as necessary. A removable recording medium 811 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc. is attached to the drive 810 as necessary, so that a computer program read out therefrom is installed in the storage section 808 as necessary.

[0108] The present disclosure also provides a computer readable storage medium, which can be included in the device / apparatus / system described in the above embodiments, or exist separately without being assembled into the device / apparatus / system. The above computer readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of the present disclosure.

[0109] According to the embodiments of the present disclosure, the computer readable storage medium can be a non-volatile computer readable storage medium. For example, it can include, but is not limited to, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In this disclosure, a computer readable storage medium can be any tangible medium that contains or stores a program, which can be used by or in connection with an instruction execution system, apparatus, or device.

[0110] The above describes the embodiments of the present disclosure. However, these embodiments are only for illustrative purposes, and are not intended to limit the scope of the present disclosure. Although each embodiment is described above separately, this does not mean that the measures in each embodiment cannot be used advantageously in combination. Those skilled in the art can make various substitutions and modifications without departing from the scope of the present disclosure, and these substitutions and modifications should all be included in the protection scope of the present disclosure.

Claims

1. A model training method, characterized in that, The method comprises: updating element density information of a plurality of original materials based on a target element respectively to obtain a plurality of sample materials; determining a plurality of sample CT image pairs associated with the sample materials based on X characteristic parameters of a target CT device and element density information of the plurality of sample materials, wherein the sample CT image pair comprises a first sample CT image containing the target element and a second sample CT image not containing the target element, and X is a positive integer; training an initial model according to the plurality of sample CT image pairs to obtain a target model corresponding to the target CT device, wherein the target model is used to determine the target element content in the original CT image scanned by the target CT device and the target CT image after the target element is separated; The updating of the element density information of the plurality of original materials based on the target element respectively to obtain the plurality of sample materials comprises: for each original material, adding the target element in the original material randomly in a preset range to obtain Y sample materials corresponding to the original material, wherein the mass proportion of the target element in different sample materials is different, Y is a positive integer; and determining the element density information of the Y sample materials respectively based on the mass proportion of the target element.

2. The model training method of claim 1, wherein, The determining of the plurality of sample CT image pairs associated with the sample materials based on the X characteristic parameters and the element density information of the plurality of sample materials respectively comprises: determining X*Y first sample CT images based on the X characteristic parameters and the element density information of the Y sample materials respectively; removing the target element content in the element density information of the Y sample materials respectively, and determining X*Y second sample CT images based on the element density information after removing the target element content.

3. The model training method of claim 2, wherein, The determining of the plurality of sample CT image pairs associated with the sample materials based on the X characteristic parameters and the element density information of the plurality of sample materials respectively comprises: performing the following operations on each sample material: calculating the CT value of each voxel in the sample material based on the characteristic parameters and the element density information of the sample material; determining the sample CT image corresponding to the sample material based on the CT value of each voxel.

4. The model training method of claim 1, wherein, The training of the initial model according to the plurality of sample CT image pairs to obtain the target model corresponding to the target CT device comprises: adding noise information to the first sample CT image associated with each sample material respectively to generate an enhanced sample CT image pair associated with each sample material, wherein the number of the enhanced sample CT image pair is greater than the number of the sample CT image pair; training the initial model based on the enhanced sample CT image pair to obtain the target model corresponding to the target CT device.

5. The model training method of claim 4, wherein, The training of the initial model according to the plurality of sample CT image pairs to obtain the target model corresponding to the target CT device comprises: determining a first sample CT image in the pair of enhanced sample CT images as input data of the initial model, and determining a target element content in the sample material and a second sample CT image in the pair of enhanced sample CT images as output data of the initial model; training the initial model based on the input data and the output data to obtain a target model.

6. A method of determining the content of an element of interest, characterized by, comprising: scanning a to-be-tested substance based on X scan energy spectrums of a target CT device respectively to obtain X CT images; inputting the X CT images into an element analysis model corresponding to the target CT device to determine a target element content in the to-be-tested substance and a CT image not containing the target element based on the X CT images by the element analysis model; wherein the element analysis model is a target model trained by the method in any one of claims 1-5. 7.A model training apparatus, comprising: an obtaining module configured to update element density information of a plurality of original materials based on a target element respectively to obtain a plurality of sample materials; a first determining module configured to determine a plurality of sample CT image pairs associated with the sample materials based on X characteristic parameters of a target CT device and the element density information of the plurality of sample materials; wherein the sample CT image pair comprises a first sample CT image containing a target element and a second sample CT image not containing the target element; a training module configured to train an initial model according to the plurality of sample CT image pairs to obtain a target model corresponding to the target CT device, wherein the target model is used to determine a target element content in an original CT image scanned by the target CT device and a target CT image after the target element is separated; the updating of the element density information of the plurality of original materials based on the target element respectively to obtain the plurality of sample materials comprises: for each original material, randomly adding the target element in the original material within a preset range to obtain Y sample materials corresponding to the original material, wherein a mass proportion of the target element in different sample materials is different, and Y is a positive integer; and determining element density information of the Y sample materials respectively based on the mass proportion of the target element. 8.A target element content determination apparatus, comprising: a scanning module configured to scan a to-be-tested substance based on X scan energy spectrums of a target CT device respectively to obtain X CT images; a second determining module configured to input the X CT images into an element analysis model corresponding to the target CT device to determine a target element content in the to-be-tested substance and a CT image not containing the target element based on the X CT images by the element analysis model; wherein the element analysis model is a target model trained by the apparatus in claim 7. 9.An electronic device, comprising: at least one processor; and a memory connected with the at least one processor in communication; wherein, The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-5 and / or claim 6.

Images