KR20180003400A - Apparatus and method for processing medical image - Google Patents

Abstract

An objective of the present invention is to provide an apparatus and a method for processing a medical image which accurately indicate a motion relationship between adjacent points of time. According to an embodiment of the present invention, the apparatus for processing a medical image comprises a processing unit which acquires raw data in a first phase section, generates first motion information by using at least one partial angle reconstruction (PAR) image pair including two PAR images acquired in two phase sections facing each other in the first phase section, sums up a plurality of PAR images acquired in different phases in the first phase section by using the first motion information to generate a sum image, updates the first motion information to minimize an image metric indicating a motion artifact when the image metric is calculated from the sum image to generate second motion information, and applies the second motion information to the raw data to generate a reconstruction image.

Description

[0001] Apparatus and method for processing medical image [0002]

The disclosed embodiments relate to a medical image processing apparatus, a medical image processing method, and a computer-readable recording medium storing computer program code for performing the medical image processing method.

X-ray computed tomography (CT) image reconstruction is a method of acquiring raw data representing the difference in degree of X-ray attenuation of each tissue of the human body by transmitting X-rays in various directions from outside the human body, And reconstruct the internal image of the human body. X-ray CT reconstruction is a technique that can acquire high-resolution images in a short time compared to other human internal image acquisition systems such as positron emission tomography (PET), magnetic resonance imaging (MRI), and single photon emission computed tomography . However, when a moving object is photographed using an X-ray CT system, motion artifacts are generated and image quality deteriorates.

The disclosed embodiments are intended to provide a medical image processing apparatus and method that more accurately indicate motion relationships of adjacent times.

In addition, the disclosed embodiments provide a medical image processing apparatus and method capable of shortening a scan time and increasing a pitch to acquire a 3D image with reduced inter cyclic motion artifact in one cycle .

According to an aspect of an embodiment, there is provided a method for acquiring raw data in a first phase interval and generating two PAR partial images obtained in two mutually opposing phase intervals in the first phase interval, And generates a sum image by summing a plurality of PAR images obtained at different phases in the first phase section using the first motion information, And generating second motion information by updating the first motion information so that the image metric becomes minimum when calculating an image metric representing motion artifacts from the summed image, And a processing unit for generating a reconstructed image by applying the reconstructed image to the data.

The processor may generate the first motion information using the at least one PAR image pair having a 180 degree phase difference when generating the first motion information.

The processing unit may update the first motion information so that a difference of motion information between spatially adjacent control points becomes smaller.

The processing unit may update the first motion information so that the motion information is zero in the reference phase.

The processor may warp a ray of the raw data based on the second motion information when generating the reconstructed image.

The first phase interval may be a phase interval greater than 180 degrees.

The two opposing phase intervals may be greater than zero and less than 180 degrees.

Wherein the processing unit generates a PAR stack by summing a plurality of PAR images belonging to the same phase on the basis of a position at which the source projects the first PAR information and generates a PAR stack, The first motion information may be generated using the pair.

Wherein the processor applies the first motion information to a plurality of PAR images obtained in the different phases to generate a sum image, and compensates the motion by summing the plurality of motion compensated PAR images, Images can be generated.

According to another aspect of one embodiment,

Obtaining raw data in a first phase interval;

Generating first motion information using two partial angle reconstruction (PAR) images obtained in two phase sections opposed to each other in the first phase section;

A plurality of PAR images obtained in different phases in the first phase section are summed to generate a summed image and an image metric representing a motion artifact is calculated from the summed image so that the image metric is minimized Updating the first motion information to generate second motion information; And

And applying the second motion information to the RO data to generate a reconstructed image.

According to another aspect of an embodiment,

A computer readable recording medium storing computer program codes for performing a medical image processing method when read and executed by a processor,

Obtaining raw data in a first phase interval;

Generating first motion information using two partial angle reconstruction (PAR) images obtained in two phase sections opposed to each other in the first phase section;

Wherein when the image metric representing motion artifacts is calculated from the summed image by summing a plurality of PAR images obtained in different phases within the first phase section to generate a summed image, 1 motion information to generate second motion information; And

And applying the second motion information to the RO data to generate a reconstructed image.

According to the disclosed embodiments, it is possible to provide a medical image processing apparatus and method that more accurately indicate a motion relationship between adjacent times.

Also, according to the disclosed embodiments, there is provided a medical image processing apparatus and method capable of shortening a scan time and increasing a pitch, thereby acquiring a 3D image with reduced inter cyclic motion artifacts There is an effect that can be done.

1 is a diagram illustrating the structure of a CT system 100 according to an embodiment disclosed herein.
2 is a diagram illustrating a medical imaging apparatus 100a and external apparatuses according to an embodiment.
3 is a flowchart illustrating a medical image processing method according to an exemplary embodiment.
4 is a diagram illustrating a process of generating a first MVF according to an embodiment of the present invention.
5 is a diagram for explaining a method for acquiring motion information according to the disclosed embodiment.
FIG. 6 is a diagram illustrating a process of calculating an MVF from a PAR image according to an embodiment of the present invention.
7 is a diagram illustrating a process of generating a motion compensated reconstructed image according to an embodiment.
8 is a diagram illustrating a PAR sequence according to an embodiment.

The present specification discloses the principles of the present invention and discloses embodiments of the present invention so that those skilled in the art can carry out the present invention without departing from the scope of the present invention. The disclosed embodiments may be implemented in various forms.

Like reference numerals refer to like elements throughout the specification. The present specification does not describe all elements of the embodiments, and redundant description between general contents or embodiments in the technical field of the present invention will be omitted. As used herein, the term " part " may be embodied in software or hardware, and may be embodied as a unit, element, or section, Quot; element " includes a plurality of elements. Hereinafter, the working principle and embodiments of the present invention will be described with reference to the accompanying drawings.

In this specification, the image includes a medical image obtained by a medical imaging apparatus such as a computer tomography (CT) apparatus, a magnetic resonance imaging (MRI) apparatus, an ultrasound imaging apparatus, or an X- .

As used herein, the term " object " may include a person, an animal, or a part thereof as an object of photographing. For example, the object may comprise a part of the body (organ or organ) or a phantom.

As used herein, the term "CT system" or "CT apparatus" refers to a system or apparatus that rotates about at least one axis with respect to a target and irradiates X-rays, and detects X-rays to photograph the target.

In the present specification, the term 'CT image' means an image composed of raw data obtained by detecting X-rays and irradiating the object by rotating around at least one axis with respect to the object.

According to embodiments of the present disclosure, a 3D CT image with reduced motion artifacts can be obtained. First, consecutive PAR images are created using the obtained sinogram. Initial motion estimation is performed by minimizing the error of each PAR image pair. Thereafter, the generated PAR stack is subjected to IP maximization to estimate a refinement term, and then motion compensated image reconstruction is performed using the updated MVF.

1 is a diagram illustrating the structure of a CT system 100 according to an embodiment disclosed herein.

The CT system 100 according to an embodiment includes a gantry 110, a table 105, a controller 130, a storage unit 140, an image processing unit 150, an input unit 160, a display unit 170, And a communication unit 180.

The gantry 110 may include a rotating frame 111, an x-ray generating unit 112, an x-ray detecting unit 113, a rotation driving unit 114, and a lead-out unit 115.

The rotation frame 111 receives the drive signal from the rotation drive unit 114 and can rotate around the rotation axis RA.

The scatter prevention grid 116 is disposed between the object and the X-ray detecting unit 113 so that the main radiation can be mostly transmitted and the scattering radiation can be attenuated. The object is placed on the table 105 and the table 105 can be moved, tilted, and rotated while performing a CT scan.

The X-ray generating unit 112 receives voltage and current from a high voltage generator (HVG) to generate and emit X-rays.

The x-ray generating unit 112 may be implemented by a single source method in which one x-ray generating unit 112 and one x-ray detecting unit 113 are provided, or a dual source method in which two x-ray generating units 112 and x-

The X-ray detector 113 detects the radiation that has passed through the object. The X-ray detector 113 can detect radiation using, for example, a scintillator, a photon counting detector, or the like.

The driving method of the X-ray generating unit 112 and the X-ray detecting unit 113 may vary according to the scanning method for the object. The scan method includes an axial scan method, a helical scan method, and the like according to the movement path of the X-ray detector 113. In addition, the scan method includes a prospective mode, a retrospective mode, and the like depending on a time period during which the X-ray is irradiated.

The control unit 130 may control the operation of each component of the CT system 100. [ The control unit 130 may include a memory for storing program codes or data for performing a predetermined function, a program code, and a processor for processing data. The control unit 130 may be implemented in various combinations of one or more memories and one or more processors. The processor can create and delete program modules according to the operating state of the CT system 100, and can process operations of the program module.

The lead-out unit 115 receives the detection signal generated by the X-ray detector 113 and outputs the detection signal to the image processor 150. The lead-out unit 115 may include a data acquisition circuit (Data Acquisition System) 115-1 and a data transmission unit 115-2. The DAS 115-1 amplifies the signal output from the X-ray detecting unit 113 using at least one amplifying circuit, and outputs the amplified signal to the data transmitting unit 115-2. The data transmission unit 115-2 outputs a signal amplified by the DAS 115-1 to the image processing unit 150 using a circuit such as a multiplexer (MUX). Only a part of the data collected from the x-ray detector 113 may be provided to the image processor 150 or the image processor 150 may select only some data depending on the slice thickness or the number of slices.

The image processing unit 150 acquires the tomographic data from the signal obtained from the lead-out unit 115 (for example, pure data before machining). The image processing unit 150 may perform preprocessing on the obtained signal, conversion processing into single layer data, post-processing on the single layer data, and the like. The image processing unit 150 performs some or all of the processes exemplified in the present disclosure, and the type and order of processes performed in the image processing unit 150 may vary depending on the embodiment.

The image processing unit 150 preprocesses signals obtained from the lead-out unit 115, such as a process for correcting sensitivity nonuniformity among channels, a process for abruptly reducing and correcting signal intensity, and a process for correcting loss of signals due to an X- Can be performed.

The image processing unit 150 performs some or all of the reconstruction processing to the tomographic image to generate the tomographic data according to the embodiments. According to an embodiment, the monolayer data may have the form of back-projected data, or a tomographic image. According to embodiments, additional processing for the tomographic data may be performed by an external device such as a server, a medical device, a handheld device, or the like.

Data may be a collection of data values corresponding to the X-ray intensity passed through the object and may include projection data or a sinogram. The data that is projected backward is the data that reversely traces the data by using the angle information of the X-ray. And the tomographic image is obtained by applying the reconstruction imaging techniques including the step of backprojecting the data.

The storage unit 140 is a storage medium for storing control related data, image data, and the like, and may include a volatile or nonvolatile storage medium.

The input unit 160 receives a control signal, data, and the like from a user. The display unit 170 may display information indicating the operation state of the CT system 100, medical information, medical image data, and the like.

The CT system 100 includes a communication unit 180 and can be connected to an external device (for example, a server, a medical device, a portable device (smart phone, tablet PC, wearable device, etc.) .

The communication unit 180 may include at least one of a short-range communication module, a wired communication module, and a wireless communication module, for example, at least one component that enables communication with an external device.

The communication unit 180 may receive the control signal and data from the external device and transmit the received control signal to the control unit 130 so that the control unit 130 controls the CT system 100 according to the received control signal It is possible.

Alternatively, the control unit 130 may transmit a control signal to the external device via the communication unit 180, thereby controlling the external device in accordance with the control signal of the control unit.

For example, an external device can process data of an external device according to a control signal of a control unit received through a communication unit.

The external device may be provided with a program capable of controlling the CT system 100, and the program may include an instruction to perform a part or all of the operation of the control unit 130. [

The program may be installed in an external device in advance, or a user of the external device may download and install the program from a server that provides the application. The server providing the application may include a recording medium storing the program.

The CT system 100 according to the disclosed embodiments may or may not use a contrast agent at the time of CT imaging according to an embodiment, or may be implemented in the form of a device associated with other devices.

The medical imaging apparatus according to the disclosed embodiments is an apparatus that acquires data obtained by imaging a target object and generates a reconstructed CT image from the data. The medical imaging device according to one embodiment may be implemented with the CT system 100 of FIG.

When an object moving with X-ray CT, such as a heart that repeats rapid contraction and relaxation, is synchronized with an electrocardiogram (ECG), it is possible to obtain a CT image with reduced motion artifacts. For example, X-ray CT can be used for evaluation of coronary artery stenosis, evaluation of graft vessels after coronary artery bypass grafting, and evaluation of restenosis after coronary stenting, and is also useful for the differentiation of acute chest pain in the emergency room And can be used as a screening test for coronary artery stenosis in patients with high risk of coronary artery disease.

However, since the time resolution of X-ray CT is still insufficient for imaging the heart, it causes image blur or motion artifacts caused by cardiac movements during cardiac imaging in cases of very rapid beating or arrhythmia, The coronary artery can not be distinguished from the shape and thickness of the coronary artery.

As a hardware approach to increase the time resolution, a method of increasing the physical rotation speed of the gantry, a method of using a dual-source, or the like can be used. A dual source system can have twice the time resolution of a single source system. However, there are disadvantages such as increasing the rotation speed of the hardware or installing additional facilities in the system, there is a physical limitation of the improvement, and the cost is increased due to the increase of the hardware complexity.

Alternatively, motion information may be extracted from the image, or motion information may be extracted from data such as a cineogram to compensate for motion. Embodiments of the present disclosure extract motion information extracted from PAR images and include motion models in the image reconstruction process to improve image quality degradation.

In the 4D cardio motion estimation and compensation method based on partial angle reconstruction (PAR) image, the disclosed embodiments estimate and compensate for motion so that the difference between a plurality of pairs of PAR image pairs is small. The pair of PAR image pairs means RO image data obtained from angular intervals having mutually conjugate relations, for example, PAR image pairs reconstructed from the RO data obtained respectively in two angular intervals having a phase difference of 180 degrees. Furthermore, the disclosed embodiments disclose a method and apparatus for reconstructing a tomographic image that can further correct the motion model based on the information potential (IP), thereby accurately indicating the motion relationship of adjacent times.

In addition, the disclosed embodiments increase the pitch at which the object is scanned so that the image can be acquired in a short time. You can use a low pitch helical scan or a step and shoot method to capture a moving object, for example, a heart. However, according to the low-pitch scan method or the step-and-shoot method, there is a disadvantage in that the scan time and the dose are excessively increased in consideration of the motion period of the moving object and the like. There is a disadvantage that inter-cycle movement can occur. According to the disclosed embodiments, there is provided an apparatus and method for obtaining a 3D image with reduced motion artifacts in the context of a standard pitch helical scan. The medical image processing apparatus according to the disclosed embodiments can acquire the data and the reconstructed image in a short time by increasing the pitch of the gantry. Also, according to the disclosed embodiments, it is possible to prevent the generation of motion artifacts between cycles that occurs when scanning the entire object, for example, the heart, at a low pitch and imaging. Also, according to the disclosed embodiments, movement can be estimated and compensated for the whole image region in a standard pitch spiral scan environment.

Furthermore, the disclosed embodiments can compensate for motion of the heart even when the motion phases of the heart are not matched, because the image reconstruction is possible with only a single cycle of raw data when high pitch is used, Maintenance and so on.

As an approach to increase the time resolution of X-ray CT, there is a method of reconstructing an ECG synchronized image as described above. In addition, there is a method of increasing the performance of basic hardware or adding the number of sources. However, the disclosed embodiments provide an apparatus and a method for algorithmically reducing motion artifacts, since improvements in hardware have disadvantages such as physical limitations and increased cost due to increased complexity.

Also, according to one embodiment of the present disclosure, a motion metric can be estimated based on an optimization method such that a motion artifact of a motion compensated reconstruction image becomes small after defining a metric for measuring the degree of motion artifacts in the image. This embodiment is useful for estimating a motion model for a region of interest with intra-cardiac movements, such as coronary arteries, myocardial muscles, and valves.

As an example of motion compensation, it is also possible to perform motion estimation and motion compensation using data such as a sinogram. Since the sinogram is a set of projection images acquired at a single moment, it generally has better temporal resolution than the reconstructed image. Therefore, if the motion of the object is extracted from the sinogram, the motion can be estimated with the motion artifacts and the like being significantly reduced compared with the reconstructed image. However, since the algorithm using the sinogram estimates the motion based on the image projected in one direction, the motion information in the direction in which the ray is projected accumulates in the linear integration model, It is difficult to estimate information. Using a dual source system, two projected images can be obtained simultaneously from two sources at a distance of about 90 degrees, so that they can complement each other in the two images. However, in a general single-source system, the accuracy of estimated motion information may be relatively lower than that of the ray projection direction in the vertical direction of the ray projection direction, and an optimization technique for extracting motion information from all the sinograms is used. There are many disadvantages.

According to the disclosed embodiments there is provided a method and apparatus for estimating and compensating 4D movement of the heart in a low pitch spiral scan situation and a method for reconstructing a single 3D image with reduced overall heart motion artifacts in a standard pitch spiral scan situation And an apparatus are provided.

2 is a diagram illustrating a medical imaging apparatus 100a and external apparatuses according to an embodiment.

The medical imaging device 100a according to one embodiment includes at least one processing unit 210. [

According to one embodiment, the processing unit 210 may acquire the acquired data by photographing the object from the scanner 220. [ The scanner 220 may include an x-ray generating unit 112, an x-ray detecting unit 113, and a lead-out unit 115. The scanner 220 may be controlled by a control unit 130 separate from the processing unit 210 of the medical image processing apparatus 100a or may be controlled by the processing unit 210 of the medical image processing apparatus 100a to scan the object, As shown in FIG. According to one embodiment, the scanner 220 is provided in the medical image processing apparatus 100a. According to another embodiment, the scanner 220 is provided as a separate device from the medical image processing apparatus 100a, and can transmit data to the processing unit 210 through a wired or wireless I / O device or a communication device.

According to another embodiment, the processing unit 210 may obtain data from the external device 230 connected via the network 240. [ The external device 230 may be, for example, a server that stores medical image data, another medical imaging device, a user terminal, an external storage device, or the like. The network 230 includes various types of wired and wireless networks, including, for example, a wired and wireless local area network (LAN), a mobile communication network, the Internet, and the like.

For example, is a sinogram or projection data including phase information.

3 is a flowchart illustrating a medical image processing method according to an exemplary embodiment. The operation of the processing unit 210 will be described with reference to the flowchart of FIG. The medical image processing method of the present disclosure can be performed by various electronic devices including at least one memory and at least one processing unit. In the present disclosure, an embodiment in which a medical image processing method is performed by the medical imaging apparatus 100a shown in FIG. 2 will be described.

The processing unit 210 scans the object and acquires the acquired data (S302). Data can be obtained in various manners such as being obtained from a scanner of the medical image processing apparatus 100a, received from an external apparatus, or the like. The processing unit 210 acquires the data of the first phase section. The first phase interval may be set to a phase interval larger than 180 degrees, for example, 360 degrees, 720 degrees, and the like.

Next, the processing unit 210 reconstructs at least one partial angle reconstruction (PAR) image pair from the RO data, and generates first motion information using the reconstructed two PAR images (S304). The first motion information may be represented by, for example, a motion vector field (MVF). In the present specification, the motion information is represented by the MVF, but the motion information may be expressed in various forms other than the MVF format according to the embodiment. The PAR image pair includes two PAR images reconstructed using data of an angular section that is greater than 0 and smaller than 180 degrees and corresponding to two angular sections facing each other. The two angular intervals facing each other may be, for example, an angular interval having a phase difference of 180 degrees. The processing unit 210 may generate the first MVF using the pair of PAR images and may generate the first MVF using the plurality of PAR image pairs.

In addition, the processing unit 210 generates a summation image by summing a plurality of PAR images having different phases reconstructed by applying the first MVF (S306). According to one embodiment, the processing unit 210 generates a PAR sequence including a plurality of PAR images covering a phase interval greater than 180 degrees from the RO data, and generates a PAR sequence using the PAR images included in the PAR sequence, Generates a first MVF, and generates a sum image. For example, the PAR sequence may cover a phase section of 360 degrees or cover a phase section of 360 degrees or more.

Next, the processing unit 210 calculates an image metric representing the motion artifact using the sum image (S307). For example, the processing unit 210 may calculate the image metric based on the histogram of the sum image. The image metric is a value quantitatively representing motion artifacts of the data. The processing unit 210 may iteratively perform the process of updating the first MVF and calculating the summed image and the image metric with the updated first MVF so that the image metric is reduced.

Next, when the first MVF is converged, the processing unit 210 determines the updated first MVF as the second MVF (S308). Here, the convergence of the first MVF means that the image metric converges.

When the second MVF is determined, the processing unit 210 applies the second motion information, i.e., the second MVF, to data such as a cineogram or projection data to generate a reconstructed image (S310). For example, the processing unit 210 may warp the ray of the sinogram in the reconstruction process based on the second MVF to generate a motion compensated reconstruction image.

According to one embodiment, the first MVF and the second MVF are three-dimensional MVFs. Also, according to one embodiment, the first MVF and the second MVF may have phase-dependent motion vector components.

In accordance with one embodiment of the present disclosure, a 4D motion estimation and compensation algorithm for a low pitch spiral scan is described.

4 is a diagram illustrating a process of generating a first MVF according to an embodiment of the present invention.

According to the 4D cardiac motion estimation and compensation algorithm according to the embodiment, the processing unit 210 generates the PAR stack with reference to the acquired sinogram and the ECG signal prior to motion estimation. The processing unit 210 performs initial motion estimation by minimizing errors between the pair of PARs. Thereafter, the processing unit 210 estimates a refinement term so as to be a complete image when the PAR stack of consecutive viewpoints is added based on the first MVF. Finally, the processing unit 210 updates the first MVF and performs the motion-compensated image reconstruction using the estimated second MVF.

First, the process of generating a PAR image will be described.

The processing unit 210 generates a PAR sequence as shown in FIG. 4 using the projection data of a short scan period for motion estimation. For example, the X-ray generating unit 112 rotates while sequentially passing through the angular sections 401 to 413 in FIG. 4, irradiates the X-ray to the object 410, and outputs data from the X-ray detected by the X- Is generated. According to one embodiment, the angle range for performing the scan is greater than 180 degrees. For example, the angle range for performing the scan may be 360 degrees or more. The processing unit 210 generates 421 to 433 PAR images corresponding to the 401 to 413 angular intervals from the raw data. Among 401 to 413 PAR images, PAR images having a phase difference of 180 degrees from each other become a pair of PAR images. For example, a 421 PAR image and a 427 PAR image can form a pair of PAR images, and a 422 PAR image and a 428 PAR image can form a pair of PAR images.

According to the present embodiment, since the data acquisition condition is a low pitch spiral scan, a plurality of PAR images are generated for the same phase. The processing unit 210 creates a PAR stack by stacking the plurality of PAR images belonging to the same phase in accordance with the position where the source (for example, the X-ray generating unit 112) projects them in terms of efficiency of calculation. Each PAR stack reflects the motion information of the subject at the corresponding phase in high temporal resolution, but shows only a partial structure of the subject. Therefore, it is difficult to estimate the motion by directly comparing with the adjacent phase using the PAR stack, but it is possible to estimate the motion by directly comparing the PAR stack with the PAR stack which differs in phase by 180 degrees.

5 is a diagram for explaining a method for acquiring motion information according to the disclosed embodiment.

According to the disclosed embodiment, the processing unit 210 generates motion information using at least one PAR image pair including a first PAR image and a second PAR image for a conjugate angle, Obtain MVF. According to one embodiment, the first PAR image and the second PAR image are images having an angle difference of 180 degrees on the rotation path of the X-ray generator 112. Also, the first PAR image and the second PAR image may be a tomographic image for a slice corresponding to a z-axis position within an error range of the object. The number of PAR image pairs used to generate the first MVF may vary depending on the embodiment.

Referring to FIG. 5, the first angle section 520 and the second angle section 522 may have a conjugate angle relationship, which is an angle between the first angle section 520 and the second angle section 522. The angular difference between the two angular sections in the conjugate relationship is 180 degrees. For example, the 401 angular section of FIG. 4 may be the first angular section 520, and the 407 angular section may be the second angular section 522. FIG. In addition, the 402 angular section of FIG. 4 may be the first angular section 520 and the 408 angular section may be the second angular section 522. Since the x-ray generating unit 112 moves while rotating around the subject at a predetermined speed, the first PAR image and the second PAR image reconstructed for the first angle interval 520 and the second angle interval 522, respectively, And has a time difference by an angle difference of 180 degrees.

If the first angle section 520 and the second angle section 522 have a mutually congruent relationship, the view in the first angular section 520 and the second angular section 522 is the same, The surface of the object 505 to be sensed when the object 505 is sensed when the object 505 is sensed in the second angle interval 522 and the surface of the object 505 sensed when the object 505 is sensed in the angle interval 520, ) Are the same. Accordingly, the first PAR image for the first angle section 522 and the second PAR image for the second angle section 522 represent states of the surface of the same object 505 at different times, The PAR image is compared with the second PAR image to obtain the motion information of the object during the time difference due to the 180 degree angle difference 510.

According to the disclosed embodiment, the first angular interval 520 and the second angular interval 522 are greater than 0 degrees and less than 180 degrees, and the processing unit 210 performs a partial angle reconstruction (PAR) The first PAR image and the second PAR image are reconstructed from the RO data obtained for the first angle section 520 and the second angle section 522. [ The PAR image reconstructed by the partial angle reconstruction technique may be an incomplete image or a partial image described above. According to the present embodiment, the reconstruction of an image using a relatively small angular section, compared to half reconstruction or full reconstruction, can increase temporal resolution and minimize motion artifacts. In the embodiment of the present disclosure, the amount of motion of a target object can be more accurately measured by measuring the amount of motion of the target object using the first PAR image and the second PAR image, which are partial angle images.

FIG. 6 is a diagram illustrating a process of calculating an MVF from a PAR image according to an embodiment of the present invention.

According to an embodiment, prior to motion estimation using a PAR image, band-pass filtering (BPF) is performed so that the characteristics of the image are appropriately changed for motion estimation. As shown in FIG. 6, in the PAR image, shading appears in a direction projected near an object having a high contrast ratio, and this shading affects the intensity of other nearby objects, Drop it. Since the shading is basically a low frequency component and the feature used in motion estimation is a specific high frequency component, the present embodiment increases the motion estimation accuracy by performing BPF on the PAR image. The bandpass filter may be implemented by the processing unit 210, for example.

In addition, one embodiment of the present disclosure uses a multi-resolution technique to improve the robustness of motion estimation. That is, the original PAR stack is blurred and downsampled to produce a low-resolution image to perform motion estimation. In this case, when the estimated motion vector is used as an initial value in estimating the motion at the higher resolution, it helps to improve the accuracy of the motion estimation.

The processing unit 210 compares the first PAR images 602, 604 and 606 of the multi-resolution and the second PAR images 612, 614 and 616 of the multi-resolution, Gt; MVF 620 < / RTI >

Next, the motion estimation process will be described. First, the initial motion estimation will be described.

According to one embodiment, the processing unit 210 uses a B-spline-based 4D freeform deformation (FFD) model to express the movement of the heart over time as a motion model. The 4D FFD model shows the correspondence relationship between the reference image of the specific reference time r and the image of each viewpoint, with respect to the 4D video sequence. The transform function of the 4D FFD model is shown in Equation (1).

는 i시점의 PAR 영상의 중심 뷰 각도(center view angle)에 해당한다.

는 시간 도메인(temporal domain)에서의 콘트롤 포인트(control point) 간의 포인트 간의 간격을 나타내고,

는 공간 도메인(spatial domain)에서의 콘트롤 포인트간의 간격을 나타낸다.

는 t시점에 j위치의 콘트롤 포인트의 변위 벡터(displacement vector)를 나타낸다.

는 기준 시점 r에서 각 시점으로 움직임 관계를 나타내기 위한 4D 콘트롤 포인트 세트의 변위 벡터 세트를 나타낸다. 콘트롤 포인트는 모션을 추정하고자 하는 특정 샘플 포인트를 의미한다. 처리부(210)는 이들 특정 샘플 포인트에서 추정된 모션 벡터를 보간 함으로써, 임의의 복셀(x)에서의 모션 벡터를 추정하게 된다.In Equation (1), β represents 1D cubic B-spline and B represents 3D tensor product of 1D cubic B-spline.

Corresponds to the center view angle of the PAR image at time i .

Represents an interval between points between control points in a temporal domain,

Represents the spacing between the control points in the spatial domain.

Represents the displacement vector of the control point at position j at time t .

Represents a set of displacement vectors of a set of 4D control points for indicating a motion relationship from the reference point r to each point in time. A control point means a specific sample point for which motion is to be estimated. The processing unit 210 estimates the motion vector in an arbitrary voxel x by interpolating the estimated motion vector at these specific sample points.

Next, the processing unit 210 calculates the difference between each pair of the PAR stacks.

)를 정의하였다.According to one embodiment, the processing unit 210 determines the parameters of the 4D motion model such that the difference between each pair of the PAR stacks is small. For example, the processing unit 210 defines the parameters of the 4D motion model so that the difference between the 401 PAR image and the 407 PAR image becomes smaller, and defines the parameters of the 4D motion model so that the difference between the 402 PAR image and the 408 PAR image becomes smaller . Since the two PAR stacks used in the motion estimation are of the same modality, the following dissimilarity metric, D (

) Were defined.

는

의 스캔 범위 동안의 PAR 스택의 개수를 나타내고, r은 기준 위상의 인덱스를 나타내며, P_i는 i번째 PAR영상을 나타낸다. Pi(x)는 i번째 PAR 영상이고, Pr(x)는 기준 위상의 PAR 영상을 의미한다. 처리부(210)는 모든 켤레 쌍의 PAR 스택들 간의 에러를 최소화 한다. here

The

R denotes the index of the reference phase, and P _i denotes the i- th PAR image. Pi (x) is the i-th PAR image, and Pr (x) is the PAR image of the reference phase. The processing unit 210 minimizes errors between the PAR pairs of all pairs of pairs.

On the other hand, since the problem of estimating motion parameters is ill-posed because the degree of freedom of the 4D FFD motion model is high, Similarly, a penalize term is used to reduce the large difference in parameter values between adjacent control points in time and space.

In Equation (3), K _j represents a set of neighbor indexes on the space of the jth control point, and K _t represents a set of neighbor indexes in time of the control point at time t .

According to one embodiment, the processing unit 210 updates the first MVF so that MVF is zero in the reference phase. Since the MVF may not be 0 in the reference phase due to the 4D B-spline interpolation characteristic, in the disclosed embodiment, a second normalization item for correcting it may be defined as Equation (4).

The processing unit 210 according to an exemplary embodiment may use a data item defined above (e.g., Equation 2) and two normalized items (e.g., Equation 3 and Equation 4) to calculate a final cost function ) Is defined as Equation (5), and a motion model parameter that minimizes it is found.

과

는 정규화 파라미터이다.here

and

Is a normalization parameter.

One embodiment disclosed to find motion parameters that minimize the cost function defined above uses the Gauss-Newton method.

According to one embodiment, the processing unit 210 performs an additional update to the first MVF to obtain a more accurate motion model.

Since the first MVF is estimated using the conjugate PAR stack, it can not accurately represent the motion relationship over a continuous time flow. According to one embodiment, a new cardiac motion model is defined by adding a refining term to the previously estimated first MVF as shown in Equation (6) to more accurately represent the motion relationship over a continuous time flow.

In Equation (6) ,? _C (t) can be defined as Equation (7).

, 집합이고, β _C는 β를 180도 주기로 반복시킨 것으로 콘트롤 포인트를 180도마다 재사용하여 MVF를 만들게 한다. Where Φ is the displacement vector of the control point,

, And β _C is a 180-degree repetition of β, which allows MVF to be created by reusing control points at 180 degrees.

Next, the processing unit 210 performs an additional update of the first MVF using the energy function.

The processing unit 210 according to an exemplary embodiment uses an information potential (IP) to estimate a refining term. IP is an image metric according to one embodiment. IP is derived by applying Parzen's windowing technique to Renyi's quadratic entropy, and can quantify motion artifacts present in the image, such as a general Shannon's entropy. IP is defined as shown in Equation (8).

는

의 표준편차를 가지는 가우시안(Gaussian) 분포를 나타내고 N은 I_MC의 전체 voxel 수이다. IP는 shannon's entropy에 비해 수식적으로 단순한 형태를 가지고, bin size나 interval 등의 파라미터를 정할 필요가 없기 때문에 최적화 과정에서 이를 다루기 쉽다. here

The

And N is the total number of voxels of I _MC . IP has a simpler form than the shannon's entropy, and it is easy to handle it in the optimization process because it does not need to specify parameters such as bin size or interval.

As described earlier, the 4D FFD model is ill-posed because of its high degrees of freedom, so the problem of estimating parameters is unknown. According to one embodiment, to solve this problem, the processor 210 may use Equation 9 so that the difference between the spatially adjacent control points as the normalization term does not increase.

Meanwhile, in this embodiment, the normalized term is used to prevent the currently estimated MVF from greatly varying from the first MVF, and to prevent the shape change due to the large MVF change. The normalization term such as Equation 10 can be expressed as making the correction term zero.

The energy function for motion model estimation can be defined by subtracting two normalized terms from IP.

The processing unit 210 obtains the improved MVF to maximize the energy function defined above and combines it with the first MVF to calculate the second MVF.

7 is a diagram illustrating a process of generating a motion compensated reconstructed image according to an embodiment.

In one embodiment, motion compensation is implemented by referring to Schafer's method based on Weighted Filtered Back Projection (WFBP) method to obtain a motion compensated image with an estimated 4D motion model. The WFBP method is a method of back projection by applying a different weight to the cone angle by rebinning the projection data acquired by the fan-beam into a parallel-beam. Schafer's method is a method of back projection by reflecting a motion vector (ray is warped) in back projection process in order to restore the object when there is motion in an object and motion vector is given. The processing unit 210 generates MVF to the viewpoint of each view at the target viewpoint and performs motion compensation by warping MVF when the view is back-projected. That is, the processing unit 210 performs motion compensation by waving a ray of a sinogram in an image reconstruction process. The processing unit 210 may perform 4D B-spline interpolation for each view as shown in FIG. 7 to obtain a 4D motion model.

Next, in accordance with another embodiment of the present disclosure, a 3D motion estimation and compensation algorithm for a standard pitch spiral scan is described.

According to the present embodiment, a 3D image is obtained in which movement artifacts are reduced with respect to the entire heart in a standard pitch spiral scan environment. First, consecutive PAR images are created using the obtained sinogram. Then, initial motion estimation is performed by extracting the intermediate frequency and minimizing the error of each PAR image pair. The first MVF, which is the initial MVF, is used to warp all the PAR images to the virtual reference time point so that the 180 degree difference images are collected in the same virtual phase. Thereafter, the correction term is estimated by performing the IP maximization described in the embodiment for the low-pitch spiral scan on the generated PAR stack, and then the MVF is reconstructed using the estimated MVF.

8 is a diagram illustrating a PAR sequence according to an embodiment.

According to one embodiment, the processing unit 210 generates a PAR sequence as shown in FIG. 10 using projection data of a scan period shorter than a short scan for motion estimation. Since the data acquisition condition is a standard pitch helical scan, there is no target phase to cover all in the z axis, and only one PAR image exists for one phase. According to one embodiment, the processing unit 210 may calculate the MVF from a pair of PAR images corresponding to angular sections facing each other. For example, PAR images that differ by 180 degrees can be used for motion estimation because they have limited view angle artifacts of the same pattern, corresponding to overlapping regions.

Next, the initial motion estimation will be described. According to one embodiment, to estimate an initial motion model, the movement of the heart over time is shown based on a B-spline-based 4D freeform deformation (FFD) model. The conversion function is as shown in Equation (12).

Where the a _i (z) is a phase, (weight function) weighting function to create the MVF in the slice position z to zero at the position corresponding to the source intersects the z slices. This allows the reference phase to be defined differently for each z-slice. In each z-slice position, motion is estimated by using a pair of PAR images, and compensation is performed based on the temporal position of the z-slice.

As described above, the BPF is applied to improve the matching accuracy between the pairs of PAR images. Then, the motion model parameters are estimated so that the SSD becomes smaller between each pair of PAR images.

는 180도 동안의 PAR 영상 개수를 나타낸다. 또한 움직임 모델 파라미터를 추정하는 것은 ill-posed 문제이기 때문에 수학식 14와 같이, 시간 공간적으로 인접한 움직임 모델 파라미터 간의 값의 차이가 지나치게 커지는 것을 막는 정규화를 사용 하였다.here

Represents the number of PAR images for 180 degrees. Also, since the motion model parameter estimation is an ill-posed problem, normalization is used to prevent the difference in value between temporally and spatially adjacent motion model parameters from becoming too large as shown in Equation (14).

The optimization procedure for the above two equations proceeds as described above based on the Gauss-Newton method and the optimization transfer method.

Next, the processing efficiency can be increased by stacking the PAR image before IP maximization based on the initial motion estimation result. According to the present embodiment, the PAR images having the angle of 180 * n + as the center view angle are waved and stacked with the first MVF, unlike the case of collecting the same phase in advance. In the standard pitch spiral scan environment, since only one cycle is used, images of the same phase can not be collected. Instead, the images are shifted 180 degrees by using the information of the first MVF so as to make a virtual phase.

In the embodiment of the standard pitch spiral scan, as described above, a refining term is defined as shown in Equation (15) to improve the initial motion estimation result.

Next, the processing unit 210 performs motion estimation based on the local IP such as Equation (16) that can quantify the image quality of the motion compensation image using the energy function.

Here, the sum image I _MC is generated by applying MVF to the PAR stack and adding all of them to the PAR stack. In this case, the first MVF applied is the first MVF, and then the MVF which is updated (iterated) while it is applied is applied to obtain the sum image I _MC repeatedly. Further, the image metric H (?) Is repeatedly calculated from the sum image I _MC to update the first MVF so that the image metric H (?) Becomes small. The image metric H (?) According to Equation (16) is a value quantitatively representing motion artifacts based on the histogram distribution of the sum image I _MC . The final motion compensation image is obtained by applying the second MVF, which is the last MVF generated repeatedly and updated, to the data (projection data or a sinogram).

Since the 4D B-spline FFD model is ill-posed, the normalization term such as Equation 17 is used so that the values of the spatial model parameters adjacent to each other become similar.

In addition, since the processing unit 210 according to the embodiment uses IP maximization, it is similar to the initial motion compensated image when the ideal image at a certain time point is warped to the current MVF by shape preserving as in Equation (18) .

IP and two normalizations can be used to define an energy function,

IP and the first normalization are obtained by the Gauss-Newton method in the same manner as in the low-pitch spiral scan.

Similar to the case of the low-pitch spiral scan, in this embodiment, the motion compensated reconstruction is implemented by using the Schafer's method based on the WFBP reconstruction algorithm. However, in the standard pitch spiral scan algorithm, the motion model should be applied in a cascading manner as in the case of the summation method in the low pitch spiral scan algorithm.

In the cascade method, estimation must be started considering the coordinates of the final image. First, since the final image is obtained through accurate motion estimation, a refining motion model is considered as shown in Equation (20).

Where x _MC represents the coordinates of the final motion compensated image. Then, the following coordinate transformation, which generates the motion compensated PAR stack from the obtained data, is considered as shown in equation (21).

In this embodiment, the motion compensation amount to be finally generated for motion compensation is expressed by Equation (22).

The processing unit 210 performs motion compensation using the calculated motion compensation amount through bi-linear interpolation using a warping method or a pre-computed MVF set as described above.

In the present disclosure, embodiments are disclosed in which X-ray CT is used to compensate for image quality degradation due to movement occurring during cardiac image acquisition. Although the present disclosure has been described with reference to an embodiment of the heart, the scope of rights of the present patent is not limited to the case where the object is a heart, and the object may be various body organs or parts.

Embodiments of the present disclosure provide a 4D ME / MC algorithm that can obtain improved image quality by introducing a new motion estimation method, IP maximization method, based on the PAR image based 4D ME / MC algorithm. Unlike the MAM optimization method that obtains motion that optimizes the entropy-based image metric and applies it to the image reconstruction, the IP maximization method has advantages of being able to perform motion estimation and compensation for the entire image region and further processing for ROI extraction It is advantageous that it is not necessary. The IP maximization method has the advantage that it is easier to handle in the optimization process because IP is simpler than the Shannon entropy.

Embodiments of the present disclosure also have proposed algorithms for estimating and compensating for motion in a standard pitch spiral scan environment. Embodiments of the present disclosure use the filtered PAR image to perform an initial motion estimate such that the error between the conjugate PAR images that differ by 180 degrees is smaller, and obtain the final motion estimation result by IP maximization based on the result. As described above, according to the embodiments of the present disclosure, the motion can be estimated and compensated for the entire image area, and no additional processing is required. In addition, the embodiments of the present disclosure use only a single cycle of data, so that there is no need to care about heart movement between cycles or maintenance of the same breathing state between cycles. The disclosed embodiments are expected to open a new chapter in image acquisition for cardiac diagnosis with an approach that has not previously existed.

Meanwhile, the disclosed embodiments may be embodied in the form of a computer-readable recording medium for storing instructions and data executable by a computer. The command may be stored in the form of program code, and when executed by the processor, may generate a predetermined program module to perform a predetermined operation. In addition, the instructions, when executed by a processor, may perform certain operations of the disclosed embodiments.

The embodiments disclosed with reference to the accompanying drawings have been described above. It will be understood by those of ordinary skill in the art that various changes in form and details may be made therein without departing from the spirit and scope of the present invention as defined by the following claims. The disclosed embodiments are illustrative and should not be construed as limiting.

Claims (19)

Acquiring raw data in a first phase interval and generating at least one PAR image pair including two PAR (partial angle reconstruction) images obtained respectively in two opposing phase intervals in the first phase interval, And generating a sum image by summing a plurality of PAR images obtained in different phases in the first phase section using the first motion information to generate a motion artifact from the sum image, , The first motion information is updated to generate the second motion information and the second motion information is applied to the raw data to generate a reconstructed image so that the image metric is minimized The medical image processing apparatus comprising: The method according to claim 1,
Wherein each of the at least one PAR image pair comprises two PAR images having a phase difference of 180 degrees. The method according to claim 1,
Wherein the processing unit updates the first motion information so that a difference of motion information between spatially adjacent control points becomes smaller. The method according to claim 1,
Wherein the processor updates the first motion information so that the motion information is zero in the reference phase. The method according to claim 1,
Wherein the processing unit warps a ray of the low-order data based on the second motion information when generating the reconstructed image. The method according to claim 1,
Wherein the first phase section is a phase section larger than 180 degrees. The method according to claim 1,
Wherein the two phase sections opposed to each other are phase sections greater than 0 and smaller than 180 degrees. The method according to claim 1,
Wherein the processing unit generates a PAR stack by summing a plurality of PAR images belonging to the same phase on the basis of a position at which the source projects the first PAR information and generates a PAR stack, And generates the first motion information using a pair. The method according to claim 1,
Wherein the processor applies the first motion information to a plurality of PAR images obtained in the different phases to generate a sum image, and compensates the motion by summing the plurality of motion compensated PAR images, A medical image processing apparatus for generating an image. Obtaining raw data in a first phase interval;
Generating first motion information using at least one PAR image pair including two PAR partial images obtained respectively in two opposed phase sections in the first phase section;
A plurality of PAR images obtained in different phases in the first phase section are summed to generate a summed image and an image metric representing a motion artifact is calculated from the summed image so that the image metric is minimized Updating the first motion information to generate second motion information; And
And applying the second motion information to the RO data to generate a reconstructed image. 11. The method of claim 10,
Wherein each of the at least one PAR image pair comprises two PAR images having a phase difference of 180 degrees. 11. The method of claim 10,
Further comprising updating the first motion information so that the difference of motion information between spatially adjacent control points becomes smaller. 11. The method of claim 10,
And updating the first motion information so that the motion information is zero in the reference phase. 11. The method of claim 10,
Wherein generating the reconstructed image comprises warping a ray of the raw data based on the second motion information. 11. The method of claim 10,
Wherein the first phase interval is a phase interval greater than 180 degrees. 11. The method of claim 10,
Wherein the two phase sections facing each other are phase sections greater than 0 and smaller than 180 degrees. 11. The method of claim 10,
The medical image processing method further comprises generating a PAR stack by summing a plurality of PAR images belonging to the same phase based on positions at which the source projects them,
Wherein the generating of the first motion information comprises generating the first motion information using the PAR stack pair of the two phase sections opposed to each other. 11. The method of claim 10,
The motion compensated motion compensator applies the first motion information to a plurality of PAR images obtained in the different phases to generate motion compensated motion compensated PAR images, , Medical image processing method. A computer readable recording medium storing computer program codes for performing a medical image processing method when read and executed by a processor,
Obtaining raw data in a first phase interval;
Generating first motion information using at least one PAR image pair including two PAR partial images obtained respectively in two opposed phase sections in the first phase section;
Wherein when the image metric representing motion artifacts is calculated from the summed image by summing a plurality of PAR images obtained in different phases within the first phase section to generate a summed image, 1 motion information to generate second motion information; And
And applying the second motion information to the RO data to generate a reconstructed image.

Images