KR102590461B1 - Method and apparatus for decomposing multi-material based on dual-energy technique - Google Patents

Abstract

The dual energy technique-based multi-substance separation method and device according to a preferred embodiment of the present invention separates and images not only two substances but three or more substances from the dual energy image of the object, thereby providing additional information for diagnosis and testing. Accuracy can be improved, and more accurate material separation can be performed through the process of correcting image deterioration and removing the influence of the densest material.

Description

{Method and apparatus for decomposing multi-material based on dual-energy technique}

The present invention relates to a method and device for separating multiple materials based on dual energy technology, and more specifically, to a method and device for separating materials.

In the case of general digital radiography (DR) imaging, the attenuated radiation information passing through the inside of a three-dimensional object is detected by a two-dimensional detector and imaged. The property of showing different degrees of attenuation is used to capture the inside of the object. An image with organs and substances separated appears. At this time, complex attenuation occurs as various materials are located in the path through which the radiation passes, and materials with small attenuation (e.g., soft tissue, etc.) have the disadvantage of being difficult to distinguish from other objects due to low contrast. To overcome this problem, researchers attempted to create images in which only soft tissue was separated, using the difference in attenuation of radiation with two different energies to attenuate the signals of high-density and low-density materials in the image. We are currently researching a dual energy technique that increases the ability to distinguish between materials in an image by amplifying or removing them.

When using the dual energy technique, it can be seen that the ability to distinguish substances is improved compared to existing digital radiography (DR) images. In particular, in the case of lesions or foreign substances that are difficult to identify hidden behind high-density materials (e.g., bone, etc.), diagnostic accuracy can be confirmed to be improved by increasing material contrast using dual energy techniques.

However, in the case of the existing dual energy method, the increase in patient radiation dose due to excessive dose irradiation for double imaging and quantum noise compensation and image distortion due to motion blur generated during dual imaging are major problems.

Recently, the above problems can be overcome through the popularization of photon counting detector-based imaging devices such as dual energy x-ray absorptiometry (DEXA), and the separation of two existing materials in 3D images Beyond this, research is underway to separate three or more substances.

The purpose of the present invention is to separate multiple substances from the dual energy image of the object based on the dual energy technique, through a process of correcting image deterioration and removing the influence of the densest substance, to produce images for each substance. The aim is to provide a dual energy technique-based multi-material separation method and device.

Other unspecified objects of the present invention can be additionally considered within the scope that can be easily inferred from the following detailed description and its effects.

A dual energy technique-based multiple material separation method according to a preferred embodiment of the present invention to achieve the above object includes the steps of acquiring a dual energy image of an object; Obtaining a restored dual energy image by correcting image deterioration of the dual energy image of the object; Obtain a high-density material image corresponding to the densest material from the reconstructed dual energy image of the object based on a thickness-intensity look-up table corresponding to the densest material among the plurality of materials to be separated. steps; Obtaining a weight map image corresponding to the high-density material image using a pre-learned image transformation model; acquiring a high-density material-removed dual-energy image based on the reconstructed dual-energy image of the object and the weight map image; and removal of the high-density material based on a thickness-strength look-up table for each material obtained through a calibration phantom corresponding to the remaining materials excluding the highest density material among the plurality of materials to be separated. and obtaining a material image for each of the remaining materials from the energy image.

Here, the thickness-intensity lookup table is obtained based on the restored dual energy image obtained by correcting image deterioration of the dual energy image of the calibration phantom and the thickness of the material corresponding to the calibration phantom, The same image deterioration correction method may be used to correct image deterioration of a dual energy image corresponding to a phantom and to correct image deterioration of a dual energy image corresponding to the object.

Here, in the step of acquiring the weight map image, the high-density material image is input to the pre-learned image conversion model, and a first weight map image and a second energy for the first energy output from the pre-learned image conversion model are obtained. The weight map image corresponding to the high-density material image may be obtained through a second weight map image for .

Here, the method may further include learning the image conversion model using learning data including computed tomography (CT) images of actual patients.

Here, in the image conversion model learning step, for each computed tomography (CT) image included in the learning data, a first radiation image (digitally) for the first energy is generated using the computed tomography (CT) image. Reconstructed Radiography (DRR) and a second radiation image (DRR) for the second energy are acquired, and the first radiation image (DRR) and the second radiation image (DRR) are obtained based on a thickness-intensity lookup table corresponding to the densest material. A high-density material learning image is acquired using a second radiation image (DRR), and a first high-density material removal radiation for the first energy from which the densest material is removed using the computed tomography (CT) image. Obtaining an image and a second high-density material removal radiation image for the second energy, and using the first radiation image (DRR) and the first high-density material removal radiation image to obtain a first high-density material removal radiation image for the first energy Obtain a second high-density material image for the second energy using the second radiation image (DRR) and the second high-density material removal radiation image, and obtain the high-density material learning image and the first high-density material. Obtaining a first weight map learning image for the first energy using an image, and obtaining a second weight map learning image for the second energy using the high-density material learning image and the second high-density material image; , Based on the high-density material learning image, the first weight map learning image, and the second weight map learning image obtained for each of the computed tomography (CT) images included in the learning data, the high-density material learning image may be used as input data, and the image transformation model may be learned using the first weight map learning image and the second weight map learning image as correct answer labels.

Here, the step of acquiring the high-density material removal dual energy image includes using the first weight map image from the first reconstructed dual energy image for the first energy among the reconstructed dual energy images of the object. 1 Obtain a high-density material-removed dual energy image, and obtain a second high-density map image for the second energy using the second weight map image from the second reconstructed dual energy image for the second energy among the reconstructed dual energy images of the object. Material removal can be achieved by acquiring dual energy images.

Here, the step of acquiring a material image for each of the remaining materials is based on a thickness-intensity lookup table for each material corresponding to the remaining materials, and the first high-density material removal dual energy image and the second high-density material removal dual energy image. It can be achieved by obtaining a material image for each of the remaining materials from the high-density material removal dual energy image.

Here, the step of acquiring a material image for each of the remaining materials includes a thickness-intensity lookup corresponding to the target material in the order of materials from high density to low density of the remaining materials to be separated from the high-density material removal dual energy image. Based on the table, a target material separation image is obtained from the high-density material removal dual energy image, the target material separation image is removed from the high-density material removal dual energy image, and the high-density material removal dual energy image from which the target material separation image is removed is obtained. The process of performing the next material separation using the energy image may be repeatedly performed to obtain a material image for each of the remaining materials from the high-density material removal dual energy image.

Here, when the number of materials to be separated is n, the calibration phantom is n-1 sub-phantoms composed of a combination of a first material, which is the lowest density material among the n materials, and a second material, which is one of the remaining materials. The sub-phantom may include a first sub-phantom made of the first material in a stepped form and a second sub-phantom located on one side of the first sub-phantom and made of the second material in a stepped form.

A computer program according to a preferred embodiment of the present invention for achieving the above technical problem is stored in a computer-readable storage medium and executes any one of the above-described dual energy technique-based multi-material separation methods on a computer.

The dual energy technique-based multi-material separation device according to a preferred embodiment of the present invention for achieving the above object is a multi-material separation device that separates multiple substances based on the dual energy technique. memory for storing one or more programs to separate; and one or more processors that perform an operation to separate multiple substances based on a dual energy technique according to the one or more programs stored in the memory, wherein the processor acquires a dual energy image of the object, and A restored dual energy image is acquired by correcting image deterioration of the dual energy image of the object, and the object is separated based on a thickness-intensity look-up table corresponding to the material with the highest density among the plurality of materials to be separated. Obtain a high-density material image corresponding to the densest material from the reconstructed dual energy image, obtain a weight map image corresponding to the high-density material image using a pre-learned image transformation model, and reconstruct the dual energy of the object. Based on the image and the weight map image, a dual energy image of high-density material removal is obtained, and the thickness of each material is obtained through a calibration phantom corresponding to the remaining materials excluding the highest density material among the plurality of materials to be separated- Based on an intensity look-up table, a material image for each of the remaining materials is obtained from the high-density material removal dual energy image.

Here, the thickness-intensity lookup table is obtained based on the restored dual energy image obtained by correcting image deterioration of the dual energy image of the calibration phantom and the thickness of the material corresponding to the calibration phantom, The same image deterioration correction method may be used to correct image deterioration of a dual energy image corresponding to a phantom and to correct image deterioration of a dual energy image corresponding to the object.

Here, the processor inputs the high-density material image into the pre-learned image transformation model, and outputs a first weight map image for the first energy and a second weight map image for the second energy output from the pre-learned image transformation model. Through the weight map image, the weight map image corresponding to the high-density material image can be obtained.

According to the dual energy technique-based multi-substance separation method and device according to a preferred embodiment of the present invention, diagnosis is provided by providing additional information by separating and imaging not only two substances but three or more substances from the dual energy image of the object. And the accuracy of inspection can be improved.

In addition, more accurate material separation can be performed through a process of correcting image deterioration and removing the influence of the densest material.

In addition, the present invention can be expected to increase patient convenience with a faster imaging time compared to existing 3D image-based methods, and can reduce the radiation dose received by patients.

In addition, the present invention relates to various radiological imaging devices such as X-ray panoramic imaging equipment, digital tomosynthesis, single-photon emission computed tomography, and positron emission tomography. It can be applied to video equipment.

The effects of the present invention are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description below.

Figure 1 is a block diagram illustrating a dual energy technique-based multi-material separation device according to a preferred embodiment of the present invention.
Figure 2 is a flowchart illustrating a dual energy technique-based multi-material separation method according to a preferred embodiment of the present invention.
Figure 3 is a diagram for explaining an example of a calibration phantom according to a preferred embodiment of the present invention.
Figure 4 is a diagram for explaining the process of obtaining a thickness-strength lookup table according to a preferred embodiment of the present invention.
FIG. 5 is a diagram for explaining an example of the acquisition process of the thickness-strength lookup table shown in FIG. 4.
Figure 6 is a diagram for explaining the learning process of an image conversion model according to a preferred embodiment of the present invention.
FIG. 7 is a diagram for explaining an example of the acquisition process of the high-density material learning image shown in FIG. 6.
FIG. 8 is a diagram for explaining an example of the acquisition process of the weight map learning image shown in FIG. 6.
Figure 9 is a diagram illustrating an example of a dual energy technique-based multi-material separation process according to a preferred embodiment of the present invention.
FIG. 10 is a diagram illustrating an example of a dual energy image of an object according to a preferred embodiment of the present invention. FIG. 10(a) represents simulation-based mathematical object information, and FIG. 10(b) represents the image of FIG. 10. Shows a dual energy image of the object according to (a).
FIG. 11 is a diagram showing a weight map image corresponding to the dual energy image of the object shown in FIG. 10.
FIG. 12 is a diagram showing the results of performing multi-material separation on the dual energy image of the object shown in FIG. 10.

Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings. The advantages and features of the present invention and methods for achieving them will become clear by referring to the embodiments described in detail below along with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below and may be implemented in various different forms. The present embodiments are merely provided to ensure that the disclosure of the present invention is complete and to provide a general understanding of the technical field to which the present invention pertains. It is provided to fully inform those with knowledge of the scope of the invention, and the present invention is only defined by the scope of the claims. Like reference numerals refer to like elements throughout the specification.

Unless otherwise defined, all terms (including technical and scientific terms) used in this specification may be used with meanings that can be commonly understood by those skilled in the art to which the present invention pertains. Additionally, terms defined in commonly used dictionaries are not interpreted ideally or excessively unless clearly specifically defined.

In this specification, terms such as “first” and “second” are used to distinguish one component from another component, and the scope of rights should not be limited by these terms. For example, a first component may be named a second component, and similarly, the second component may also be named a first component.

In this specification, identification codes (e.g., a, b, c, etc.) for each step are used for convenience of explanation. The identification codes do not describe the order of each step, and each step is clearly ordered in a specific order in context. Unless specified, it may occur differently from the specified order. That is, each step may occur in the same order as specified, may be performed substantially simultaneously, or may be performed in the opposite order.

In this specification, expressions such as “have,” “may have,” “includes,” or “may include” indicate the presence of the corresponding feature (e.g., a numerical value, function, operation, or component such as a part). indicates, does not rule out the presence of additional features.

Hereinafter, preferred embodiments of the dual energy technique-based multi-material separation method and device according to the present invention will be described in detail with reference to the accompanying drawings.

First, a multi-material separation device based on dual energy technology according to a preferred embodiment of the present invention will be described with reference to FIG. 1.

Figure 1 is a block diagram illustrating a dual energy technique-based multi-material separation device according to a preferred embodiment of the present invention.

Referring to FIG. 1, a dual energy technique-based multi-material separation device (hereinafter referred to as 'multi-material separation device') 100 according to a preferred embodiment of the present invention includes a process of correcting image deterioration and separation of the highest density material. Through the process of removing the influence, multiple substances can be separated from the dual energy image of the object based on the dual energy technique, and images for each substance can be obtained.

Here, the dual energy image refers to a low-energy image obtained by irradiating low-energy radiation of 30 keV to 60 keV to the object and a high-energy image obtained by irradiating high-energy radiation of 80 keV to 144 keV to the object.

That is, the present invention is a digital radiography (DR) and dual energy x-ray absorptiometry (DEXA) imaging system using X-rays and gamma rays, by separating multiple substances based on a dual energy technique. , the accuracy of diagnosis and testing can be improved by obtaining more material information under the same conditions as the existing dual energy technique.

To this end, the multi-material separation device 100 may include one or more processors 110, a computer-readable storage medium 130, and a communication bus 150.

The processor 110 may control the multi-material separation device 100 to operate. For example, the processor 110 may execute one or more programs 131 stored in the computer-readable storage medium 130. One or more programs 131 may include one or more computer-executable instructions, which, when executed by the processor 110, cause the multi-material separation device 100 to separate multiple materials based on a dual energy technique. It may be configured to perform an operation to separate.

The computer-readable storage medium 130 is configured to store computer-executable instructions, program code, program data, and/or other suitable forms of information for separating multiple substances based on dual energy techniques. The program 131 stored in the computer-readable storage medium 130 includes a set of instructions executable by the processor 110. In one embodiment, computer-readable storage medium 130 includes memory (volatile memory, such as random access memory, non-volatile memory, or an appropriate combination thereof), one or more magnetic disk storage devices, optical disk storage devices, flash It may be memory devices, another form of storage medium that can be accessed by the multi-material separation device 100 and store desired information, or a suitable combination thereof.

Communication bus 150 interconnects various other components of multi-material separation device 100, including processor 110 and computer-readable storage medium 130.

The multi-material separation device 100 may also include one or more input/output interfaces 170 and one or more communication interfaces 190 that provide an interface for one or more input/output devices. The input/output interface 170 and communication interface 190 are connected to the communication bus 150. Input/output devices (not shown) may be connected to other components of the multi-material separation device 100 via input/output interface 170.

Then, a multi-material separation method based on dual energy technique according to a preferred embodiment of the present invention will be described with reference to FIGS. 2 to 8.

FIG. 2 is a flowchart illustrating a multi-material separation method based on dual energy technique according to a preferred embodiment of the present invention, FIG. 3 is a diagram illustrating an example of a calibration phantom according to a preferred embodiment of the present invention, and FIG. 4 is a diagram for explaining the acquisition process of the thickness-strength lookup table according to a preferred embodiment of the present invention, and Figure 5 is a diagram for explaining an example of the acquisition process for the thickness-strength lookup table shown in Figure 4. 6 is a diagram illustrating the learning process of the image conversion model according to a preferred embodiment of the present invention, FIG. 7 is a diagram illustrating an example of the acquisition process of the high-density material learning image shown in FIG. 6, and FIG. 8 is This is a diagram to explain an example of the acquisition process of the weight map learning image shown in FIG. 6.

Referring to FIG. 2, the processor 110 of the multi-material separation device 100 may obtain a thickness-strength look-up table for each material using a calibration phantom (S110).

Here, the thickness-intensity lookup table can be obtained based on the restored dual energy image obtained by correcting image deterioration of the dual energy image of the calibration phantom and the thickness of the material corresponding to the calibration phantom.

In addition, the calibration phantom may be formed in a stepped form made of the material to be separated (or a material having a linear attenuation coefficient similar to the material to be separated). That is, the calibration phantom may be formed in a stepped shape so that it is made up of a combination of areas with different thicknesses.

At this time, when the number of materials to be separated is n, the calibration phantom will include n-1 sub-phantoms composed of a combination of a first material, which is the lowest density material among the n materials, and a second material, which is one of the remaining materials. You can.

At this time, as shown in FIG. 3, one sub-phantom may include a first sub-phantom made of a first material in a stepped form and a second sub-phantom located on one side of the first sub-phantom and made of a second material in a stepped form. You can. At this time, the first sub-phantom and the second sub-phantom are divided into the step direction of the first sub-phantom (i.e., the direction in which the thickness of the first material is different) and the step direction of the second sub-phantom (i.e., the direction in which the thickness of the second material is different). Different directions) may be arranged to be orthogonal to each other. By irradiating radiation to such a sub-phantom, a dual energy image (a first dual energy image and a second dual energy image) can be obtained.

In more detail, as shown in FIG. 4, the processor 110 may obtain a thickness-strength lookup table through each sub-phantom included in the calibration phantom corresponding to the plurality of materials to be separated.

That is, the processor 110 can acquire a dual energy image for the sub-phantom. Here, the dual energy image may include a first dual energy image for a first energy (low energy of 30 keV to 60 keV) and a second dual energy image for a second energy (high energy of 80 keV to 144 keV). You can.

Additionally, the processor 110 may obtain a restored dual energy image by correcting image deterioration of the dual energy image (the first dual energy image and the second dual energy image). Here, the restored dual energy image is a first restored dual energy image obtained by correcting image deterioration of the first dual energy image for the first energy and a restored dual energy image obtained by correcting the image deterioration of the second dual energy image for the second energy. It may include a second reconstructed dual energy image.

In general, radiation images suffer from image deterioration such as scattering and noise, and can be modeled as D = I + S + N. Here, D represents the acquired dual energy image, I represents the ideal dual energy image, S represents the image including the scattered ray component, and N represents the image including the noise component. In the case of scattered lines, hardware methods (Grid technique, Air-gap technique, Narrow-detector technique, etc.), software methods (Hybrid-cassette technique, De-convolution technique, Prior-information technique, etc.), or hardware and software methods. A mixed method (see T. Kawamura et al, Improvement in image quality and workflow of x-ray examinations using a new image processing method, "virtual grid technology", FUJIFLIM RESEARCH & DEVELOPMENT, 60, 2015, 21-27 ) can be used to remove the influence of scattered rays. In the case of noise, noise can be reduced through software methods (Filtering technique using mean filter, average filter, wiener filter, moving average filter, median filter, etc., Prior-information technique using Domain transformation technique, machine learning, deep learning, etc.). The impact can be eliminated. Image deterioration of a dual energy image according to the present invention can be performed through various conventional techniques and is not limited to a specific technique. However, the same image deterioration correction method (image deterioration correction technique and related parameters) can be used for image deterioration correction of the dual energy image corresponding to the calibration phantom according to the present invention and image deterioration correction of the dual energy image corresponding to the object. there is.

Then, the processor 110 obtains a thickness-intensity lookup table for the corresponding material based on the reconstructed dual energy image (the first reconstructed dual energy image and the second reconstructed dual energy image) and the thickness of the material corresponding to the sub-phantom. can do.

For example, the material separation algorithm may be an equation representing the relationship between the thickness of the material to be separated and the dual energy image, as shown in [Equation 1] below.

Here, t _d represents the thickness of the first material to be separated. t _s represents the thickness of the second material to be separated. P _L represents the first dual energy image for the first energy (low energy) among the dual energy images. P _H represents the second dual energy image for the second energy (high energy) among the dual energy images.

In other words, an equation such as [Equation 1] above can be derived using the linear attenuation coefficient and the low-energy and high-energy intensity calculation equation according to thickness, such as [Equation 2] below.

And, the constant values of the material separation algorithm as in [Equation 1] above (a ₀ , a ₁ , a ₂ , a ₃ , a ₄ , a ₅ , b ₀ , b ₁ , b ₂ , b ₃ , b ₄ , b ₅ ) can be calculated in advance to obtain a thickness-strength lookup table for the corresponding material. At this time, the processor 110 may obtain the constant value of the material separation algorithm through various analysis methods such as deep learning techniques such as Singular Value Decomposition (SVD) and neural network, and least squares method.

The processor 100 may perform the above process for each sub-phantom included in the calibration phantom to obtain a thickness-intensity lookup table corresponding to each of the plurality of materials to be separated.

For example, when separating three substances (bone, muscle, and fat), a first sub-phantom in a stepped form with the substance with the lowest density among the three substances (fat) as the first substance and the other substance (bone) are separated. Sub-phantom 1 was produced by arranging the second sub-phantom in a stepped form made of two materials as shown on the left side of FIG. 5 (a), and a dual energy image (Low kV _p (bone) as shown on the left side of FIG. 5 (b) -fat) and High kV _p (bone-fat)) can be obtained.

In addition, a first sub-phantom in a stepped form with the least dense material (fat) among the three materials as the first material and a second sub-phantom in a stepped form with another material (muscle) as the second material are shown in FIG. 5. Create sub-phantom 2 by placing it as shown on the right side of (a), and acquire dual energy images (Low kV _p (muscle-fat) and High kV _p (muscle-fat)) as shown on the right side of (b) of Figure 5. can do.

Then, the image degradation of the dual energy image (Low kV _p (bone-fat) and High kV _p (bone-fat)) for sub-phantom 1 is corrected to obtain a restored dual energy image (corrected Low kV _p (bone-fat) ) and corrected High kV _p (bone-fat)) can be obtained.

In addition, the image degradation of the dual energy image (Low kV _p (muscle-fat) and High kV _p (muscle-fat)) for Sub-Phantom 2 was corrected and the restored dual energy image (corrected Low kV _p (muscle-fat)) was restored. and corrected High kV _p (muscle-fat)) can be obtained.

Then, the thickness of the material (bone-fat) corresponding to sub-phantom 1, the reconstructed dual energy image corresponding to sub-phantom 1 (corrected Low kV _p (bone-fat) and corrected High kV _p (bone-fat) ), the thickness of the material (muscle-fat) corresponding to sub-phantom 2, and the reconstructed dual-energy image (corrected Low kV _p (muscle-fat) and corrected High kV _p (muscle-fat)) corresponding to sub-phantom 2. Using the thickness-strength lookup table for bone t _bone (P _L , P _H ), the thickness-strength lookup table for fat t _fat (P _L , P _H ) and the thickness-strength lookup table t _muscle (P _L , P _H ) can be obtained.

Additionally, the processor 110 may learn an image transformation model including a neural network using the training data (S120).

Here, the learning data may include computed tomography (CT) images of actual patients.

In more detail, as shown in FIG. 6, the processor 110 may perform a training data acquisition process and a weight map prediction training process. The processor 110 performs a digitally reconstructed radiography (DRR) image acquisition process, a high-density material learning image acquisition process, a high-density material removal radiation image acquisition process, and a high-density material image for each computed tomography (CT) image included in the learning data. The training data acquisition process can be performed by performing the acquisition process and the weight map learning image acquisition process. And, the processor 110 is based on the high-density material learning image and the weight map learning image (the first weight map learning image and the second weight map learning image) obtained for each computed tomography (CT) image included in the learning data. In this way, an image transformation model can be learned by using the high-density material learning image as input data and the weight map learning image (the first weight map learning image and the second weight map learning image) as the correct answer label. For example, the processor 110 inputs a high-density material learning image to a machine learning/deep learning-based image conversion model, and the weight map prediction image (a first weight map prediction image and a second weight map prediction image) output from the image conversion model. ) and the correct answer label, the weight map learning image (the first weight map learning image and the second weight map learning image), calculate the error, and feed back the calculated error to modify the parameters of the image conversion model. The weight map prediction image (the first weight map prediction image and the second weight map prediction image), which is the output of the transformation model, and the weight map training image (the first weight map training image and the second weight map training image) which are the answer labels are substantially different from each other. You can learn an image conversion model by performing it repeatedly until it is the same.

That is, the processor 110 may acquire a first radiation image (DRR) for the first energy and a second radiation image (DRR) for the second energy using a computed tomography (CT) image.

In addition, the processor 110 can acquire a high-density material learning image using the first radiation image (DRR) and the second radiation image (DRR) based on a thickness-intensity lookup table corresponding to the material with the highest density. .

In addition, the processor 110 uses computed tomography (CT) images to create a first high-density material-removed radiation image for the first energy from which the densest material has been removed and a second high-density material-removed radiation image for the second energy. can be obtained.

Then, the processor 110 acquires a first high-density material image for the first energy using the first radiation image (DRR) and the first high-density material removal radiation image, and uses the second radiation image (DRR) and the second high-density material removal radiation image. A second high-density material image for the second energy can be obtained using a material-removed radiation image.

Then, the processor 110 acquires a first weight map learning image for the first energy using the high-density material learning image and the first high-density material image, and obtains a second weight map learning image using the high-density material learning image and the second high-density material image. A second weight map learning image for energy can be obtained.

For example, as shown in FIG. 7, a first radiology image (DDR image for low energy (L)) and a second radiology image (for high energy (H)) are obtained through a computed tomography (CT) image of an actual patient. DDR video) can be acquired. Then, based on the thickness-intensity lookup table t _bone (P _L , P _H ) for the bone, the first radiological image (DDR image for low energy (L) P _L ) A high-density material learning image can be obtained using the DDR image P _H ) for ).

In addition, as shown in FIG. 8, a first radiological image (DDR image for low energy (L)) and a second radiological image (DDR image for high energy (H)) are obtained using computed tomography (CT) images. ) is acquired, and a first high-density material removal radiology image from which bone information is removed (DDR image for low energy (L) from which bone information is removed) and a second high-density material removal radiation image from which bone information is removed using computed tomography (CT) images. An image (DDR image for high energy (H) with bone information removed) can be acquired. Then, the first radiological image (DDR image for low energy (L)) and the first high-density material-depleted radiological image (DDR image for low energy (L) with bone information removed) are used to create a first high-density material image. (Bone image for low energy (L)) is acquired, and a second radiological image (DDR image for high energy (H)) and a second high-density material-removed radiological image (high energy (H) with bone information removed) are acquired. A second high-density material image (bone image for high energy (H)) can be acquired using the DDR image for high energy (H). Then, the first weight map learning image (weight map image for low energy (L)) is acquired using the high-density material learning image and the first high-density material image (bone image for low energy (L)), and the high-density material image (bone image for low energy (L)) is used to obtain the first weight map learning image (weight map image for low energy (L)) A second weight map learning image (weight map image for high energy (H)) can be obtained using the material learning image and the second high-density material image (bone image for high energy (H)).

Afterwards, the processor 110 may acquire a dual energy image of the object (S130).

That is, the processor 110 can acquire a dual energy image by irradiating radiation to the object.

Here, the dual energy image may include a first dual energy image for a first energy (low energy of 30 keV to 60 keV) and a second dual energy image for a second energy (high energy of 80 keV to 144 keV). You can.

Then, the processor 110 may correct image deterioration of the dual energy image of the object and obtain a restored dual energy image (S140).

That is, the processor 110 may obtain a restored dual energy image by correcting image deterioration of the dual energy image (the first dual energy image and the second dual energy image) of the object.

Here, the restored dual energy image is a first restored dual energy image obtained by correcting image deterioration of the first dual energy image for the first energy and a restored dual energy image obtained by correcting the image deterioration of the second dual energy image for the second energy. It may include a second reconstructed dual energy image.

At this time, the same image deterioration correction method (image deterioration correction technique and related parameters) can be used to correct image deterioration of the dual energy image corresponding to the calibration phantom and to correct image deterioration of the dual energy image corresponding to the object.

Then, the processor 110 may obtain a high-density material image corresponding to the material with the highest density from the reconstructed dual energy image of the object (S150).

That is, the processor 110 generates a restored dual energy image of the object (a first restored dual energy image for the first energy and a second restored dual energy image for the first energy) based on a thickness-intensity lookup table corresponding to the material with the highest density among the plurality of materials to be separated. A high-density material image corresponding to the material with the highest density can be obtained from the second restored dual energy image for 2 energies.

For example, a first reconstructed dual energy image (reconstructed dual energy image P _L for low energy (L)) and a second reconstructed dual energy image based on the thickness-intensity lookup table t _bone (P _L , P _H ) for bone. A high-density material image can be obtained using (restored dual energy image P _H for high energy (H)).

Then, the processor 110 may obtain a weight map image corresponding to the high-density material image using a pre-learned image transformation model (S160).

That is, the processor 110 inputs a high-density material image into a pre-learned image conversion model, and generates a first weight map image for the first energy and a second weight map for the second energy output from the pre-learned image conversion model. Through the image, a weight map image corresponding to the high-density material image can be obtained.

Then, the processor 110 may acquire a high-density material-removed dual energy image based on the reconstructed dual energy image and the weight map image of the object (S170).

That is, the processor 110 acquires the first high-density material removal dual energy image for the first energy using the first weight map image from the first reconstructed dual energy image for the first energy among the reconstructed dual energy images of the object. You can.

And, the processor 110 acquires a second high-density material removal dual energy image for the second energy using the second weight map image from the second reconstructed dual energy image for the second energy among the reconstructed dual energy images of the object. You can.

Finally, the processor 110 may acquire a material image for each of the materials remaining except for the material with the highest density among the plurality of materials to be separated from the high-density material removal dual energy image (S180).

That is, the processor 110 removes the high-density material using a dual energy image based on the thickness-intensity lookup table for each material obtained through the calibration phantom corresponding to the remaining materials excluding the material with the highest density among the plurality of materials to be separated. It is possible to obtain material images for each of the remaining materials.

In more detail, the processor 110 uses the first high-density material removal dual energy image and the second high-density material removal dual energy image to remove the high-density material based on a thickness-intensity lookup table for each material corresponding to the remaining materials. From the energy image, material images for each of the remaining materials can be obtained.

That is, the processor 110 determines the density of the remaining material to be separated from the high-density material removal dual energy image (the first high-density material removal dual energy image for the first energy and the second high-density material removal dual energy image for the second energy). In order from high to low materials, high-density material removal dual energy image (first high-density material removal dual energy image for the first energy and second energy for the second energy) based on the thickness-intensity lookup table corresponding to the target material. A target material separation image can be obtained from the second high-density material removal dual energy image.

Then, the processor 110 removes the target material separation image from the high-density material removal dual energy image (the first high-density material removal dual energy image for the first energy and the second high-density material removal dual energy image for the second energy). You can.

Then, the processor 110 converts the target material separation image into a removed high-density material removal dual energy image (a first high-density material removal dual energy image for the first energy and a second high-density material removal dual energy image for the second energy). The process of performing the following material separation is performed iteratively to remove the remaining dense material from the dual energy image (the first dense material removed dual energy image for the first energy and the second dense material removed dual energy image for the second energy). Material images can be obtained for each material.

Then, an example of a dual energy technique-based multi-material separation process according to a preferred embodiment of the present invention will be described with reference to FIGS. 9 to 12.

Figure 9 is a diagram for explaining an example of a multi-material separation process based on dual energy technique according to a preferred embodiment of the present invention, and Figure 10 is a diagram for explaining an example of a dual energy image of an object according to a preferred embodiment of the present invention. As a diagram, Figure 10(a) shows simulation-based mathematical object information, Figure 10(b) shows a dual energy image of the object according to Figure 10(a), and Figure 11 shows the simulation-based mathematical object information shown in Figure 10. This is a diagram showing a weight map image corresponding to the dual energy image of the object, and FIG. 12 is a diagram showing the result of performing multi-material separation on the dual energy image of the object shown in FIG. 10.

Referring to FIG. 9, when three substances (bone, muscle, and fat) are separated, a first sub-phantom in a stepped form with “fat” as the first material and a stepped form with “bone” as the second material. A dual energy image of sub-phantom 1 is acquired through sub-phantom 1, which includes a second sub-phantom, and a step-shaped first sub-phantom with “fat” as the first material and “muscle” as the second material. A dual energy image of sub-phantom 2 can be obtained through sub-phantom 2, which includes a second sub-phantom in a stepped form. Then, image deterioration of the dual energy image for sub-phantom 1 and the dual energy image for sub-phantom 2 can be corrected. At this time, the de-convolution technique was used to remove the effect of scattering lines, and the domain transformation technique was used to remove the effect of noise. Then, the thickness of the material (bone-fat) corresponding to sub-phantom 1, the dual energy image with image degradation corrected for sub-phantom 1, the thickness of the material (muscle-fat) corresponding to sub-phantom 2, and the sub-phantom _Using a _dual - _energy image with _image degradation _{corrected corresponding} to ) and the thickness-strength lookup table for muscle t _muscle (P _L , P _H ) can be obtained.

Thereafter, image deterioration of the dual energy image of the object as shown in FIG. 10 may be corrected. At this time, the technique used to correct image deterioration of the dual energy image for the calibration phantom (De-convolution technique to remove the effect of scattering lines and Domain transformation technique to remove the effect of noise) and parameters were used to determine the size of the object. Image degradation of the dual energy image was corrected.

Then, a high-density material image (i.e., bone image) can be obtained based on the dual energy image with image deterioration corrected for the object and the thickness-intensity lookup table t _bone (P _L , P _H ) for bone. .

Then, using the image transformation model learned and built in advance, a weight map image corresponding to the high-density material image (i.e., bone image) can be obtained, as shown in FIG. 11.

Then, based on the dual energy image and the weight map image from which image deterioration corresponding to the object has been corrected, a bone separation dual energy image from which high-density material (i.e., bone) has been removed can be obtained.

Finally, dense material separation dual energy imaging with the dense material (i.e. bone) removed, a thickness-intensity lookup table for fat t _fat (P _L , P _H ) and a thickness-intensity lookup table for muscle t _muscle (P Based on _L , P _H ), fat separation images and muscle separation images can be obtained. Accordingly, as shown in FIG. 12, a bone separation image, a fat separation image, and a muscle separation image can be obtained through dual energy imaging.

Operations according to the present embodiments may be implemented in the form of program instructions that can be performed through various computer means and recorded on a computer-readable storage medium. A computer-readable storage medium refers to any medium that participates in providing instructions to a processor for execution. A computer-readable storage medium may include program instructions, data files, data structures, or combinations thereof. For example, there may be magnetic media, optical recording media, memory, etc. A computer program may be distributed over networked computer systems so that computer-readable code can be stored and executed in a distributed manner. Functional programs, codes, and code segments for implementing this embodiment can be easily deduced by programmers in the technical field to which this embodiment belongs.

These embodiments are intended to explain the technical idea of the present embodiment, and the scope of the technical idea of the present embodiment is not limited by these examples. The scope of protection of this embodiment should be interpreted in accordance with the claims below, and all technical ideas within the equivalent scope should be interpreted as being included in the scope of rights of this embodiment.

100: multi-material separation device,
110: processor,
130: computer-readable storage medium,
131: program,
150: communication bus,
170: input/output interface,
190: communication interface

Claims (13)

Obtaining a dual energy image of an object;
Obtaining a restored dual energy image by correcting image deterioration of the dual energy image of the object;
Obtain a high-density material image corresponding to the densest material from the reconstructed dual energy image of the object based on a thickness-intensity look-up table corresponding to the densest material among the plurality of materials to be separated. steps;
Obtaining a weight map image corresponding to the high-density material image using a pre-learned image transformation model;
acquiring a high-density material-removed dual-energy image based on the reconstructed dual-energy image of the object and the weight map image; and
Dual energy to remove the high-density material based on a thickness-strength look-up table for each material obtained through a calibration phantom corresponding to the materials other than the densest material among the plurality of materials to be separated. Obtaining a material image for each of the remaining materials from the image;
Including,
The weight map image acquisition step is,
Inputting the high-density material image into the pre-learned image conversion model, through a first weight map image for first energy and a second weight map image for second energy output from the pre-learned image conversion model, Consisting of acquiring the weight map image corresponding to the high-density material image,
Multi-material separation method based on dual energy technique. In paragraph 1:
The thickness-strength lookup table is,
It is obtained based on the reconstructed dual energy image obtained by correcting image deterioration of the dual energy image of the calibration phantom and the thickness of the material corresponding to the calibration phantom,
Image deterioration correction of the dual energy image corresponding to the calibration phantom and image deterioration correction of the dual energy image corresponding to the object use the same image deterioration correction method,
Multi-material separation method based on dual energy technique. delete In paragraph 1:
Learning the image conversion model using learning data including computed tomography (CT) images of actual patients;
A dual energy technique-based multi-material separation method further comprising: In paragraph 4,
The image conversion model learning step is,
For each computed tomography (CT) image included in the learning data,
A first digitally reconstructed radiography (DRR) image for the first energy and a second radiological image (DRR) for the second energy are acquired using the computed tomography (CT) image, and the highest density is obtained. Acquire a high-density material learning image using the first radiological image (DRR) and the second radiological image (DRR) based on a thickness-intensity lookup table corresponding to the high-density material,
Using the computed tomography (CT) image, obtain a first high-density material-removed radiation image for the first energy and a second high-density material-removed radiation image for the second energy from which the densest material has been removed; ,
Obtain a first high-density material image for the first energy using the first radiation image (DRR) and the first high-density material removal radiation image, and obtain the second radiation image (DRR) and the second high-density material removal. Obtaining a second high-density material image for the second energy using a radiation image,
Obtain a first weight map learning image for the first energy using the high-density material learning image and the first high-density material image, and obtain a first weight map learning image for the first energy using the high-density material learning image and the second high-density material image. Obtain a second weight map training image for
Based on the high-density material learning image, the first weight map learning image, and the second weight map learning image obtained for each of the computed tomography (CT) images included in the learning data, the high-density material learning image is generated. Consists of learning the image transformation model using input data and the first weight map learning image and the second weight map learning image as correct answer labels,
Multi-material separation method based on dual energy technique. In paragraph 1:
The high-density material removal dual energy image acquisition step is,
Obtaining a first high-density material-removed dual energy image for the first energy using the first weight map image from the first reconstructed dual energy image for the first energy among the reconstructed dual energy images of the object,
Consisting of obtaining a second high-density material-removed dual energy image for the second energy using the second weight map image from the second reconstructed dual energy image for the second energy among the reconstructed dual energy images of the object,
Multi-material separation method based on dual energy technique. In paragraph 6:
The material image acquisition step for each of the remaining materials is,
Based on a thickness-intensity lookup table for each material corresponding to the remaining material, the remaining material is removed from the high-density material removal dual energy image using the first high-density material removal dual energy image and the second high-density material removal dual energy image. Consists of acquiring a material image for each,
Multi-material separation method based on dual energy technique. In paragraph 7:
The material image acquisition step for each of the remaining materials is,
Removal of the high-density material Based on the thickness-intensity lookup table corresponding to the target material, in the order of material from high to low density of the remaining material to be separated from the dual energy image, removal of the high-density material from the dual energy image The process of acquiring a separation image, removing the target material separation image from the high-density material removal dual energy image, and performing the next material separation with the high-density material removal dual energy image from which the target material separation image was removed, is repeated. Consisting of obtaining a material image for each of the remaining materials from the high-density material removal dual energy image,
Multi-material separation method based on dual energy technique. In paragraph 1:
The calibration phantom is,
When there are n substances to be separated, it includes n-1 sub-phantoms consisting of a combination of a first substance, which is the lowest density substance among the n substances, and a second substance, which is one of the remaining substances,
The sub-phantom is,
Comprising a first sub-phantom in a stepped form made of the first material and a second sub-phantom in a stepped form located on one side of the first sub-phantom and made of the second material,
Multi-material separation method based on dual energy technique. A computer program stored in a computer-readable storage medium for executing the dual energy technique-based multi-material separation method according to any one of claims 1, 2, 4 to 9 on a computer. A multi-material separation device that separates multiple materials based on dual energy technology,
A memory storing one or more programs for the separation of multiple substances based on dual energy techniques; and
One or more processors that perform an operation to separate multiple substances based on a dual energy technique according to the one or more programs stored in the memory;
Including,
The processor,
Acquire a dual energy image of the object,
Obtaining a restored dual energy image by correcting image deterioration of the dual energy image of the object,
Obtain a high-density material image corresponding to the densest material from the reconstructed dual energy image of the object based on a thickness-intensity look-up table corresponding to the densest material among the plurality of materials to be separated. do,
Obtaining a weight map image corresponding to the high-density material image using a pre-learned image transformation model,
Obtaining a high-density material removal dual energy image based on the reconstructed dual energy image of the object and the weight map image,
Dual energy to remove the high-density material based on a thickness-strength look-up table for each material obtained through a calibration phantom corresponding to the materials other than the densest material among the plurality of materials to be separated. Obtaining a material image for each of the remaining materials from the image,
The processor,
Inputting the high-density material image into the pre-learned image conversion model, through a first weight map image for first energy and a second weight map image for second energy output from the pre-learned image conversion model, Obtaining the weight map image corresponding to the high-density material image,
Multi-material separation device based on dual energy technique. In paragraph 11:
The thickness-strength lookup table is,
It is obtained based on the reconstructed dual energy image obtained by correcting image deterioration of the dual energy image of the calibration phantom and the thickness of the material corresponding to the calibration phantom,
Image deterioration correction of the dual energy image corresponding to the calibration phantom and image deterioration correction of the dual energy image corresponding to the object use the same image deterioration correction method,
Multi-material separation device based on dual energy technique. delete

Images