KR20200021156A - High Quality Four Dimensional Cone Beam Computerized Tomography System Using Prior Image - Google Patents

Abstract

Embodiments of the present invention provide a high quality four-dimensional cone beam computerized tomography system capable of reconfiguring a high quality image using the small number of projection images by classifying a plurality of pieces of projection data into a plurality of groups according to a breathing cycle of a subject and generating a restored image by applying a weight relaxation map to partial projection data belonging to a classified group and reconstructing the partial projection data.

Description

High Quality Four Dimensional Cone Beam Computerized Tomography System Using Prior Image}

TECHNICAL FIELD The present invention relates to a four-dimensional cone beam computed tomography system.

The contents described in this section merely provide background information on the present embodiment and do not constitute a prior art.

Four Dimensional Cone Beam Computed Tomography (4D CBCT) provides anatomical volumetric images over the respiratory cycle. 4D CBCT images are images that are acquired continuously during breathing prior to radiation therapy and are used for the purpose of correcting the patient's position or evaluating whether the target (or tumor) is within the planned physical range in the same manner as the treatment plan.

FIG. 1 is a diagram illustrating an operation in which 4D CBCT classifies projection data and reconstructs an image. The 4D CBCT classifies the x-ray projection into multiple respiratory cycle bins according to the projection image or respiration signal extracted from an external device, and then reconstructs the 3D CBCT images independently in each respiratory cycle bin using a reconstruction algorithm such as FDK.

2 is a diagram illustrating an image in which 4D CBCT reconstructs non-uniformly sampled projection data. Since 4D CBCT is photographed with large intervals between slices in order to acquire several CBCT images within a certain time, there is little information and motion artifact occurs due to continuous photographing during respiration. In 4D CBCT, the gantry rotates at low speed, increasing scan time and increasing radiation dose to the patient.

Therefore, the appropriate number of projections in each respiratory bin is required. In particular, a larger number of projections are needed because each breathing cycle bin contains adjacent projection data containing duplicate information that captured the same subject at slightly different time points. Inadequate number of projections for each phase bin results in a low quality 4D CBCT image with clear artifacts.

Korea Patent Publication No. 10-1767069 (2017.08.04.)

Embodiments of the present invention classify a plurality of projection data into a plurality of groups according to the respiratory cycle of the subject, apply a weighted relaxation map to the partial projection data belonging to the classified group, and reconstruct the partial projection data to generate a reconstructed image Therefore, the main purpose of the invention is to improve the quality of the reconstructed image reconstructed by a small number of projection images by the four-dimensional cone beam computerized tomography system.

Still other objects of the present invention may be additionally considered within the scope that can be easily inferred from the following detailed description and effects thereof.

According to an aspect of the present embodiment, in a four-dimensional cone beam computerized tomography system, a gantry formed to be rotatable in at least one direction, a radiation head connected to the gantry and irradiating a radiation beam to an object to be inspected, connected to the gantry A panel detector that detects a radiation beam that has passed through the subject and obtains a plurality of projection data, and classifies the plurality of projection data into a plurality of groups according to the breathing cycle of the subject, and partial projection belonging to the classified group Provided is a four-dimensional cone beam computed tomography system including a processor for applying a weighted relaxation map to data and reconstructing the partial projection data to generate a reconstructed image.

According to another aspect of the present invention, in the image reconstruction method by a four-dimensional cone beam computed tomography system, a gantry formed to be rotatable in at least one direction, a radiation head connected to the gantry and irradiating a radiation beam to the object under test, A panel detector connected to the gantry and detecting a radiation beam passing through the test object to obtain a plurality of projection data, and the processing unit classifies the plurality of projection data acquired by the panel detector into a plurality of groups according to the breathing cycle of the test object. And a processing unit applying a weighted relaxation map to partial projection data belonging to the classified group, and reconstructing the partial projection data to generate a reconstructed image.

As described above, according to the exemplary embodiments of the present invention, the plurality of projection data are classified into a plurality of groups according to the breathing period of the subject, a weighted relaxation map is applied to the partial projection data belonging to the classified group, and the partial projection is performed. By reconstructing the data to generate a reconstructed image, the 4D cone beam computed tomography system has an effect of improving the quality of the reconstructed image reconstructed into a small number of projection images.

Even if the effects are not explicitly mentioned herein, the effects described in the following specification and the tentative effects expected by the technical features of the present invention are treated as described in the specification of the present invention.

1 is a diagram illustrating an operation of classifying projection data and reconstructing an image by a 4D cone beam computed tomography system.
FIG. 2 is a diagram illustrating an image reconstructed by non-uniformly sampled projection data by a 4D cone beam computed tomography system.
Figure 3 is a block diagram illustrating a four-dimensional cone beam computerized tomography system according to an embodiment of the present invention.
4 is a diagram illustrating an operation of reconstructing an image using a prior image and a weighted relaxation map by a 4D cone beam computed tomography system according to an exemplary embodiment of the present invention.
FIG. 5 is a diagram illustrating a beam driving method for improving a quality of a prior image by a 4D cone beam computed tomography system according to an exemplary embodiment of the present invention.
FIG. 6 is a diagram illustrating a divided preliminary image and a divided target image by the 4D cone beam computed tomography system according to an exemplary embodiment.
FIG. 7 is a diagram illustrating an operation of reconstructing an image according to a constraint optimization based repetitive reconstruction algorithm according to an embodiment of the present invention.
8 is a flowchart illustrating an image reconstruction method according to another embodiment of the present invention.
9 is a diagram illustrating an image reconstructed by embodiments of the present invention.

Hereinafter, in the following description of the present invention, if it is determined that the subject matter of the present invention may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted and some embodiments of the present invention will be omitted. It will be described in detail through an exemplary drawing.

The four-dimensional cone beam computed tomography system is a device for irradiating conical divergent X-rays and restoring to a tomography image from a projection image acquired according to the patient's breathing cycle.

3 is a block diagram illustrating a four-dimensional cone beam computed tomography system.

As shown in FIG. 3, the cone beam computerized tomography system 100 includes a gantry 110, a radiation head 120, a panel detector 130, and a processor 140. do. The cone beam computed tomography system 100 may omit some of the various components illustrated in FIG. 3 or further include other components.

The gantry 110 is connected to the main body and is formed to be rotatable in at least one direction. The gantry 110 rotates 360 degrees around the inspected object located between the radiation head 120 and the panel detector 130. The gantry 120 may further include an inclinometer sensor.

The radiation head 120 is connected to the gantry 110 and irradiates a beam of radiation to the object under test. The radiation head 120 operates during the scan time for photographing the subject.

The panel detector 130 is connected to the gantry 110 and detects a radiation beam that has passed through the object to obtain a plurality of projection data.

The processor 260 generates a reconstructed image by applying a reconstruction algorithm to the plurality of projection data. The processor 260 classifies the plurality of projection data into a plurality of groups according to the breathing cycle of the subject. The processor 260 generates a reconstructed image by applying a weighted relaxation map to the partial projection data belonging to the classified group and reconstructing the partial projection data.

4 is a diagram illustrating an operation of reconstructing an image using a prior image and a weighted relaxation map by a 4D cone beam computed tomography system.

The processor applies the partial projection data as the pre-image, applies the weighted relaxation map as the target image, and minimizes the first objective function defined as the difference image comparing the pre-image with the target image. A reconstructed image is generated according to an iterative reconstruction algorithm. A reconstructed image is generated by reconstructing the projection volume, a reprojection is generated by reprojecting the reconstructed image, a weighted relaxation map is applied to the reprojected image, and an iterative reconstruction algorithm is applied again.

The weighted relaxation map is calculated by detecting an inconsistency region between images to be compared and distance converting intensity information of pixels or voxels of the inconsistency region. A pixel or voxel included in the weighted relaxation map represents a relaxation coefficient representing a distribution characteristic of a mismatched region. The relaxation coefficient is a coefficient for adjusting the number of iterations of the optimization function, and the number of iterations of the iterative reconstruction algorithm may be changed. The relaxation factor of the weighted relaxation map is assigned a higher weight as the degree of inconsistency is higher, and a lower weight is assigned as the degree of inconsistency is lower.

The first objective function is expressed by Equation 1 in consideration of the difference image μ _d between the pre-image and the target image in the compressed function. Preliminary images may be taken on the first day of radiation therapy treatment and reconstructed using full projection.

는 재구성 단면 영상 (l, m, n) 좌표에서 이산 기울기 변환 함수를 통해 얻어진 기울기를 의미하며, 수학식 2와 같이 표현된다.

Denotes the slope obtained through the discrete gradient transformation function in the reconstructed cross-sectional image (l, m, n) coordinates, and is represented by Equation 2 below.

The processor applies a constraint optimization technique that alternately minimizes the data fit function and the image fit function to minimize the objective function.

The processor adds an anisotropic coefficient (ω _j ) that separates the boundary region and the noise region to the first objective function to calculate differentially according to the contrast with the surrounding pixels. The first objective function to which the anisotropy coefficient is added is expressed as in Equation 3, and the anisotropic coefficient is expressed in Equation 4. Borders showing a relatively high contrast compared to the surrounding pixels [mu] _m are preserved, and noise voxels showing a low contrast are suppressed.

The retention parameter δ for separating the boundary area and the noise area may be calculated by calculating a gradient value of the brightness value from each cross-sectional image and determining an optimal retention parameter through a cumulative cumulative distribution function (CDF) histogram. .

FIG. 5 is a diagram illustrating a beam driving method for improving a quality of a previous image by a 4D cone beam computed tomography system.

The processor applies a ray driven backprojection method to improve the quality of the previous image. Beam-driven reverse projection is a technique that reflects the actual X-ray geometric projection trajectory. To model, we need to sample the position (x1, y1, z1) of the X-ray source along all angles.

The processing unit calculates the position of the pixel corresponding to the panel sensor from eight corners of the voxel of the prior image, in order to shorten the calculation time. The processor calculates the length crossing the voxel while passing through the voxel starting the ray from the center position of each channel of the panel detector only for the pixels inside the convex hull including the corresponding pixel. The processor may update the voxel by applying a weight equal to the length that is crossed. The flux-driven backprojection softens the noisy pixels while maintaining prominent boundaries.

The processor may generate a segmented dictionary image by using a semi-automatic segmentation tool on the high-definition dictionary image obtained by using a beam-driven reverse projection method, and use the same to update the weighted relaxation map. The divided pre-image is an image divided by comparing with a threshold based on the brightness value of the pre-image.

The processor may generate a split target image by applying a clustering algorithm to the reconstructed cross-sectional image updated during each iteration while performing an iterative reconstruction algorithm, and then update the weighted relaxation map through comparison with the split prior image.

FIG. 6 (a) is a segmented pre-image obtained by applying a semi-automatic segmentation tool to a high-definition pre-image in a 4D cone beam computed tomography system, and FIG. 6 (b) is updated during every iteration while an iterative reconstruction algorithm is performed. The segmented target image, which is a clustering algorithm applied to the reconstructed target image, is visualized in a different color for each grouped voxel.

When the iterative reconstruction algorithm is applied, the processor divides the region of the target image updated during each iteration. The processor groups the target images by optimizing the second objective function defined as the brightness values of the target images based on the possibility that the brightness values of the target images belong to a specific group.

The processor classifies the brightness value of the prior image by setting a vector that is close to the center vector of the specific group to have a high weight and a vector that is far from the center vector of the specific group to have a low weight. The processor recalculates the center vectors of all groups and repeatedly performs the process of assigning weights until the calculated center vectors do not change. The processing unit classifies groups having the same number.

The processor determines the number of clusters c (2 = c ≦ = n) and selects the exponential weight m (1 <m <∞). When initializing the initial membership function U ⁽⁰⁾ and expressing the number of iterations of the algorithm as r (r = 0, 1, 2, ...), the membership function U has a value between 0 and 1, The sum can always be set to one. Degree of belonging is expressed as in Equation 5.

The second objective function is expressed as in Equation 6, and is repeatedly performed in order to minimize the value of the second objective function.

Where u _ik is a value between 0 and 1, _indicating the degree of belonging to the k-th data of x _k belonging to the first cluster. v _i is the i th cluster center vector and j (j = 1, 2, 3, ..., L) is a variable in a specific space.

The processor calculates the center {v _i | i = 1, 2, ..., c} of the fuzzy cluster using Equation 7.

The new membership function U ^{(r + 1)} is calculated through equation (8).

The processor calculates Equation (9), and if Δ> ε, sets r = r + 1, recalculates the fuzzy cluster center, and repeats the above steps. If Δ≤ε, the algorithm ends. ε means the threshold value.

FIG. 7 is a diagram illustrating an operation of reconstructing an image according to a constraint optimization based repetitive reconstruction algorithm, using a 4D cone beam computed tomography system.

The processor minimizes data suitability of the first objective function and minimizes image suitability of the first objective function in order to minimize the first objective function.

The processing unit may apply an iterative reconstruction algorithm based on a weight relaxation map to minimize data suitability. The weighted relaxation coefficient may be applied to the SART (Simultaneous Algebraic Reconstruction Technique).

The processing unit may apply adaptive gradient descent to minimize image fit. By applying adaptive gradient descent, the optimal number of iterations of the optimal gradient descent that softens the noise voxels while minimizing the first objective function can be determined through fine tuning. Fine tuning updates parameters by entering additional data into an existing model. Minimize the canonical function with respect to global variation.

The processor may filter the reconstructed image by applying an adaptive gradient descent method to the third objective function defined as the brightness value of the reconstructed image and repeating the process of detecting a parameter that optimizes the third objective function. .

The third objective function to which the Total Variation Denoising (TVD) technique is applied is expressed by Equation 10.

P ( _{u, v} ) is the brightness value at the (u, v) coordinate of the X-ray image, and G (P ( _{u, v} )) is the discrete gradient conversion function calculated at (u, v) of the X-ray image. It means the slope value obtained through. Adaptive Gradient Descent is applied to minimize the third objective function, which is expressed by Equation 11.

[lambda] is an adaptation parameter that controls the degree of smoothness as the repetition step proceeds, and the updated value in each gradient descent step is controlled to be set to a smaller value as the number of repetitions increases. t means the number of repetitions.

▽ R (P _j ) means the slope at the j-th indexed position of the X-ray image of the third objective function calculated at each gradient descent step and the sum of the square roots of the slopes calculated at all positions _j ) | is required for the normalized slope calculation. This slope is expressed by Equation 12 and Equation 13.

The algorithm for the adaptive gradient descent can be expressed as shown in Table 1.

Although components included in the 4D cone beam computed tomography system are separately illustrated in FIG. 3, the plurality of components may be combined with each other to be implemented as at least one module. The components are connected to a communication path connecting a software module or a hardware module inside the device and operate organically with each other. These components communicate using one or more communication buses or signal lines.

The four-dimensional cone beam computed tomography system may be implemented in a logic circuit by hardware, firmware, software, or a combination thereof, or may be implemented using a general purpose or special purpose computer. The device may be implemented using a hardwired device, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), or the like. The device may also be implemented as a System on Chip (SoC) including one or more processors and controllers.

The four-dimensional cone beam computerized tomography system may be implemented in the form of software, hardware, or a combination thereof in a computing device provided with hardware elements. The computing device includes various or all communication devices such as a communication modem for performing communication with various devices or wired and wireless communication networks, a memory for storing data for executing a program, and a microprocessor for executing and operating a program. It can mean a device.

8 is a flowchart illustrating an image reconstruction method according to another embodiment of the present invention. The image reconstruction method may be performed by a 4D cone beam computed tomography system, and a detailed description of the operations performed by the 4D cone beam computed tomography system and a duplicate description thereof will be omitted.

In step S810, the 4D cone beam computed tomography system classifies the obtained plurality of projection data into a plurality of groups according to the breathing cycle of the subject.

In operation S820, the 4D cone-beam computed tomography system applies a weighted relaxation map to the partial projection data belonging to the classified group, and reconstructs the partial projection data to generate a reconstructed image.

The generating of the reconstructed image (S820) may include applying partial projection data as a pre-image, applying a weighted relaxation map as a target image, and minimizing a first objective function defined as a difference image comparing the pre-image with the target image. A reconstructed image is generated according to an Iterative Reconstruction algorithm based on a Constraint Optimization.

The weighted relaxation map is calculated by detecting an inconsistency region between compared images and distance converting intensity information of pixels or voxels of the inconsistency region, and the pixels or voxels included in the weighted relaxation map represent relaxation coefficients representing distribution characteristics of the inconsistency region. In addition, the number of iterations of the iterative reconstruction algorithm may be changed according to the relaxation coefficient. The relaxation coefficient of the weighted relaxation map may be assigned a higher weight as the degree of inconsistency is higher, and a lower weight may be assigned as the degree of inconsistency is lower.

Generating the reconstructed image (S820) may include minimizing data suitability of the first objective function, minimizing image suitability of the first objective function, and minimizing data suitability in order to minimize the first objective function. Applying an iterative reconstruction algorithm based on this, and minimizing the image fit can apply an adaptive gradient descent method. An anisotropic coefficient that separates the boundary region and the noise region may be added to the first objective function to calculate differentially according to the degree of contrast with the surrounding pixels.

Generating the reconstructed image (S820) calculates the position of the pixel corresponding to the panel detector from the eight corners of the voxel and for each channel of the panel detector only for the pixels inside the convex hull containing the corresponding pixel. The length of the crossing in the voxel may be calculated while passing through the voxel from the center position of the ray, and the voxel may be updated by applying a weight equal to the length of the crossing.

The generating of the reconstructed image (S820) may be performed by using the segmentation target image obtained by applying the clustering algorithm to the reconstructed target image updated during every iteration of the pre-partitioned image and the repetitive reconstruction algorithm using the semi-automatic segmentation tool. Can be updated. In the process of dividing an area of the target image, the target image may be grouped by optimizing a second objective function defined as the brightness value of the target image based on the possibility that the brightness value of the target image belongs to a specific group. In the process of dividing the region of the target image, the brightness vector of the target image is classified by setting a vector close to the center vector of the specific group to have a high weight and a vector far from the center vector of the specific group to have a low weight. The process of repeating the process of assigning the weights until the calculated center vector does not change again can be performed.

The generating of the reconstructed image (S820) may include applying an adaptive gradient descent method to a third objective function defined as a brightness value of the reconstructed image and detecting a parameter for optimizing the third objective function. The reconstructed image may be filtered repeatedly.

9 (a) is an image reconstructed using the FDK reconstruction algorithm, FIG. 9 (b) is an image reconstructed using the CS reconstruction algorithm, and FIG. 9 (c) is reconstructed using the PICCS reconstruction algorithm. 9 (d) is an image reconstructed using the reconstruction algorithm according to the present embodiment. It can be seen that the 4D cone-beam computed tomography system according to the present embodiment can obtain a high quality reconstructed image using a small number of projection data.

Although each process is described as being sequentially executed in FIG. 8, this is merely an example, and a person skilled in the art may change the order described in FIG. 8 without departing from the essential characteristics of the exemplary embodiment of the present invention. It may be possible to apply various modifications and variations, or to execute one or more processes in parallel or to add other processes.

The operations according to the embodiments may be implemented in the form of program instructions that may be executed by various computer means, and may be recorded in a computer readable medium. Computer-readable media refers to any medium that participates in providing instructions to a processor for execution. Computer-readable media can include program instructions, data files, data structures, or a combination thereof. For example, there may be a magnetic medium, an optical recording medium, a memory, and the like. The computer program may be distributed over networked computer systems so that the computer readable code is stored and executed in a distributed fashion. Functional programs, codes, and code segments for implementing the present embodiment may be easily inferred by programmers in the art to which the present embodiment belongs.

The present embodiments are for explaining the technical idea of the present embodiment, and the scope of the technical idea of the present embodiment is not limited by these embodiments. The scope of protection of the present embodiment should be interpreted by the following claims, and all technical ideas within the scope equivalent thereto should be construed as being included in the scope of the present embodiment.

100: four-dimensional cone beam computed tomography system
110: gantry
120: radiation head
130: panel detector
140: processing unit

Claims (17)

In a four-dimensional cone beam computed tomography system,
A gantry formed to be rotatable in at least one direction;
A radiation head connected to the gantry and irradiating a radiation beam to the object under test;
A panel detector connected to the gantry and configured to detect a radiation beam passing through the inspected object to obtain a plurality of projection data; And
A processor configured to classify the plurality of projection data into a plurality of groups according to a breathing cycle of the subject, apply a weighted relaxation map to the partial projection data belonging to the classified group, and reconstruct the partial projection data to generate a reconstructed image
Four-dimensional cone beam computerized tomography system comprising a. The method of claim 1,
The processor may apply the partial projection data as a prior image, apply the weighted relaxation map as a target image, and minimize a first objective function defined as a difference image comparing the prior image with the target image. And a reconstructed image according to an iterative reconstruction algorithm based on an optimization. The method of claim 2,
The weighted relaxation map is calculated by detecting an inconsistency region between images to be compared and distance converting intensity information of pixels or voxels of the inconsistency region, and the pixels or voxels included in the weighted relaxation map are distributed of the inconsistency region. And a repetition coefficient of the repetitive reconstruction algorithm according to the remission coefficient. The method of claim 3,
The relaxation coefficient of the weighted relaxation map is assigned a higher weight as the degree of inconsistency is higher, and a lower weight is assigned as the degree of inconsistency is lower. The method of claim 2,
In order to minimize the first objective function, the processor minimizes data suitability of the first objective function, minimizes image suitability of the first objective function, and minimizing the data suitability based on the weight relaxation map. Applying an iterative reconstruction algorithm, and minimizing the image suitability, applies an adaptive gradient descent method. The method of claim 2,
And the processor adds an anisotropy coefficient for separating a boundary region and a noise region to the first objective function to calculate differentially according to the degree of contrast with surrounding pixels. The method of claim 2,
The processing unit calculates a position of a pixel corresponding to the panel detector from eight corners of the voxel of the preliminary image, and each channel of the panel detector only for pixels in a convex hull including the corresponding pixel. Four-dimensional cone-beam computed tomography, characterized in that the line is transmitted from the center to calculate the length crossing the voxel while passing through the voxel and updating the voxel by applying the weight of the crossing length. system. The method of claim 2,
The processor divides the region of the preliminary image using a semi-automatic segmentation tool, and uses the divided target image obtained by applying a clustering algorithm to the target image updated during every iteration of the preliminary image and the iterative reconstruction algorithm. And a weighted mitigation map. The method of claim 8,
The segmentation of the region of the target image may include grouping the dictionary images by optimizing a second objective function defined as the brightness value of the prior image based on a possibility that the brightness value of the prior image belongs to a specific group. 4D cone beam computed tomography system. The method of claim 9,
The processor is configured to classify the brightness value of the target image by setting a vector close to the center vector of the specific group to have a high weight and a vector far from the center vector of the specific group to have a low weight. And calculating the weight again until the calculated center vector does not change. The method of claim 2,
The processor applies an adaptive gradient descent method to a third objective function defined as a brightness value of the reconstructed image, and repeats the process of detecting a parameter for optimizing the third objective function. 4D cone beam computed tomography system, characterized in that for filtering. In the image reconstruction method by a four-dimensional cone beam computerized tomography system comprising a gantry, a radiation head, a panel detector, and a processing unit,
The processing unit classifying the plurality of projection data acquired by the panel detector into a plurality of groups according to the breathing period of the subject; And
The processor generates a reconstructed image by applying a weighted relaxation map to the partial projection data belonging to the classified group and reconstructing the partial projection data.
Image reconstruction method comprising a. The method of claim 12,
Generating the reconstructed image,
Constraint Optimization based on applying the partial projection data as a prior image, applying the weighted relaxation map as a target image, and minimizing a first objective function defined as a difference image comparing the prior image and the target image And reconstructing the reconstructed image according to an iterative reconstruction algorithm. The method of claim 12,
Generating the reconstructed image,
In order to minimize the first objective function, minimizing data suitability of the first objective function, minimizing image suitability of the first objective function, and minimizing the data suitability are iterative reconstruction algorithms based on the weight relaxation map. And minimizing the image suitability by applying an adaptive gradient descent method. The method of claim 12,
Generating the reconstructed image,
And dividing the region of the preliminary image and the region of the target image and updating the weighted relaxation map by using the divided preliminary image and the divided target image. The method of claim 15,
The segmentation of the region of the target image may include grouping the target images by optimizing a second objective function defined as the brightness value of the target image based on a possibility that the brightness value of the target image belongs to a specific group. Image reconstruction method. The method of claim 12,
Generating the reconstructed image,
Filtering the reconstructed image by applying adaptive gradient descent to a third objective function defined as a brightness value of the reconstructed image and repeating the process of detecting a parameter that optimizes the third objective function. The image reconstruction method characterized by.

Images