KR20140130786A - Super-resolution Apparatus and Method using LOR reconstruction based cone-beam in PET image - Google Patents

Abstract

The present invention provides an image based super-resolution method using cone-beam based line of response (LOR) in a positron emission tomography (PET) image, namely a super-resolution PET image reconstruction apparatus using cone-beam LOR reconstruction and a method thereof. To this end, the present invention provides a PET image reconstruction apparatus comprising: a blur modelling part generating each PSF kernel according to each position of a PET system and models a blur; an LOR acquiring part acquiring each LOR and sinogram according to each position caused by wobbling of the PET system; a cone-beam sinogram converting part format-converting the LOR based sinogram, acquired by the LOR acquiring part, into a cone-beam type sinogram; a sinogram modelling part modelling the cone-beam type sinogram format-converted by the cone-beam sinogram converting part; and a high resolution image extracting part extracts an high resolution image using the sinogram modelled by the sinogram modelling part.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus and a method for reconstructing an ultra-high resolution PET image using a cone-

The present invention is an image-based super-resolution technique using cone-beam based LOR reconstruction in positron emission tomography (PET) imaging, and is based on cone-beam based response line reconstruction Resolution PET image reconstruction apparatus and method therefor.

Positron Emission Tomography (PET) is widely used as a nuclear medicine testing method that can display physiological, chemical, and functional images of the human body in three dimensions using radiopharmaceuticals that emit positron.

This positron emission tomography (PET) is mainly used to diagnose various types of cancer, and provides useful results for differential diagnosis of cancer, staging, recurrence evaluation, and determination of therapeutic effect. In addition, positron emission tomography (PET) can be used to obtain receptor or metabolic images for heart disease, brain disease, and brain function assessment.

The positron emitted from the radioactive isotope consumes its own kinetic energy for a very short period of time after its release and combines with the neighboring electrons, and it extinguishes. At this time, it emits two extinction rays (gamma rays) at an angle of 180 °.

Cylindrical positron emission tomography (PET) scanners are capable of detecting two extinction beams emitted at the same time. When the image is reconstructed using the detected radiation, a 3D tomographic image of how much radioactive medicines are gathered in the body can be shown.

(PET) image reconstruction algorithms based on EM (Expectation Maximization) -based iterative image reconstruction techniques such as maximum likelihood expectation maximization (MLEM), ordered-subset expectation maximization (OSEM) Priori Expectation Maximization).

However, since the reconstructed PET algorithm has a low resolution, it is possible to apply a super-resolution algorithm by making a plurality of sets of singoram by wobbling the PET system to different sampling positions have. Therefore, it takes a lot of computation time and time because there are several reconfiguration processes. That is, there is a problem that the complexity of the image reconstruction is increased due to a large amount of calculation and time.

Generally, data in a PET system, for example, a line of response (LOR), can be obtained from a PET detector as shown in FIG. 1 (a). 1 is an explanatory view showing a preprocessing process of image reconstruction according to the related art.

In the prior art, in order to simplify the image reconstruction algorithm, a preprocess such as single slice rebinning (SSRB) and arc correction is applied to a sinogram as shown in FIGS. 1 (b) and (c) .

However, since a blur phenomenon occurs in the image reconstruction preprocessing process as in the prior art, if the image reconstruction is performed based on such a sinogram, the image quality of the PET image is deteriorated.

In addition, although the image reconstruction algorithm is simplified by the single slice revision (SSRB) and the arc correction, the accuracy of the conventional image reconstruction algorithm is lowered, and the forward projection and the inverse projection can not be operated in parallel. In addition, such prior art LOR reconstruction methods are not suitable for parallel processing.

Therefore, in order to acquire a PET image with improved resolution, an image reconstruction algorithm which can perform fast without processing a sinogram is required.

In addition, a technique for reducing the image reconstruction complexity is proposed by proposing a projection method and a cyanogram format suitable for parallel processing in an image reconstruction algorithm.

Also, there is a need for a technique for obtaining ultrahigh-resolution PET images in one-step by integrating existing PET image reconstruction algorithms and super-resolution algorithms (super-resolution).

Also, there is a demand for a technique for obtaining a more accurate ultrahigh resolution PET image by using the blur modeling of each point according to the position of the PET system.

Accordingly, the present invention has been proposed in order to solve the above-mentioned problems and to meet the above-mentioned problems, and cone-beam-based LOR reconstruction in a positron emission tomography (PET) Resolution PET image reconstruction using a super-resolution technique using cone-beam based response line reconstruction, and a method therefor.

The objects of the present invention are not limited to the above-mentioned objects, and other objects and advantages of the present invention which are not mentioned can be understood by the following description, and will be more clearly understood by the embodiments of the present invention. It will also be readily apparent that the objects and advantages of the invention may be realized and attained by means of the instrumentalities and combinations particularly pointed out in the appended claims.

According to an aspect of the present invention, there is provided an apparatus for reconstructing an image of positron emission tomography (PET), comprising: a blur modeling unit for generating blur models by generating respective PSF kernels corresponding to respective positions of a PET system; part; A reaction line acquiring unit for acquiring respective reaction lines and a sinogram according to respective positions by wobbling the PET system; A cone-beam intersymbol transformer for transforming the reaction-line-based sinogram obtained by the reaction-line acquiring unit into a cone-beam type sinogram format; A sinogram modeling unit for modeling a sinogram format-converted into a cone-beam shape in the cone-beam cosine transform unit; And a high-resolution image deriving unit for deriving a high-resolution image using the sinogram modeled by the sinogram modeling unit.

The present invention also provides a method of reconstructing an image of positron emission tomography (PET), comprising: generating a PSF kernel corresponding to each position of a PET system to model a blur; Obtaining each reaction line and a sinogram according to each position by wobbling the PET system; Converting the obtained response line based sinogram to a cone-beam type sinogram format; Modeling the sinogram formatted in the cone-beam format; And deriving a high-resolution image using the modeled sinogram.

According to the present invention as described above, it is possible to quickly perform the image reconstruction process without preprocessing the sinogram, and it is possible to acquire an ultrahigh resolution PET image with improved resolution compared to the existing image reconstruction algorithm.

In addition, the present invention proposes a projection technique and a sinogram format suitable for parallel processing in an image reconstruction algorithm, thereby reducing image reconstruction complexity, reducing computational complexity, and speeding up processing.

In addition, the present invention has an effect of obtaining an ultrahigh-resolution PET image by merely performing an algorithm processing by integrating the EM-based repetitive image reconstruction and the super-resolution algorithm.

Further, the present invention has the effect of obtaining a more accurate ultrahigh resolution PET image by using the blur modeling of each point according to the position of the PET system.

FIG. 1 is an explanatory view showing a preprocessing process of image reconstruction according to the prior art.
FIG. 2 is an explanatory view showing a change of a point spread function (PSF) according to a position in a PET system used in the present invention. FIG.
FIG. 3 is an explanatory view showing a sampling method according to a wobble motion of a PET system used in the present invention. FIG.
FIG. 4 is a flowchart illustrating an ultrahigh-resolution PET image reconstruction method using cone-beam-based response line reconstruction according to the present invention.
FIG. 5 is an explanatory view showing a cone-beam based reaction line (LOR) reconstruction according to the present invention; FIG.
6 is a block diagram of an apparatus for reconstructing an ultrahigh resolution PET image using cone-beam-based response line reconstruction according to an embodiment of the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS The above and other objects, features and advantages of the present invention will become more apparent from the following detailed description of the present invention when taken in conjunction with the accompanying drawings, It can be easily carried out. In the following description, well-known functions or constructions are not described in detail since they would obscure the invention in unnecessary detail. Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

First, referring to FIG. 2, PSF kernel-based blur modeling used in the present invention and wobbling-based high-resolution sampling will be described with reference to FIG.

FIG. 2 is an explanatory view showing a change of a point spread function (PSF) according to a position in a PET system used in the present invention. FIG. 3 is a graph showing a sampling method according to a wobble motion of a PET system used in the present invention Fig.

As shown in FIG. 2, the point spread function (PSF) of a sample measured in a detector of a general PET system differs depending on the position. Therefore, the PSF of the reconstructed PET image also varies depending on the location. Using these properties, the PSF according to the position of the PET image is reconstructed, and then the PSF kernel (PSF kernel) of the hyper-high resolution algorithm is applied to the PSF according to the position when the super-resolution algorithm is applied If you use it, you can get better images.

In FIG. 3, the PET system is wobbled such as wobbling the detector or wobbling a bed in which the person in the detector is worn, classifying the measured sinogram according to each position, and reconstructing the sinogram You can create a translated image set. Using these PET images, an ultra-high resolution algorithm can be applied.

The present invention will now be described in detail with reference to the drawings.

FIG. 4 is a flowchart illustrating a method of reconstructing an ultrahigh resolution PET image using a cone-beam-based response line reconstruction according to the present invention. FIG. 6 is a block diagram of an apparatus for reconstructing an ultrahigh-resolution PET image using a cone-beam-based response line reconstruction according to an embodiment of the present invention.

A blur modeling DB (410) is created by generating each PSF kernel (PSF kernel) according to each position of the PET system using Monte-Carlo simulation. That is, as shown in FIG. 2, a PSF kernel is generated according to a Monte Carlo simulation based on a point source to create a blur modeling DB.

The PSF kernel-based blur modeling of the present invention will be described as follows.

PET systems can experience blur even without motion due to physical phenomena such as positron range, non-collinearity, detector crystal width, block effect, and the like. Blur may occur differently depending on the structure of the detector. That is, each PET system can have its own blur. Therefore, it is important to recognize the unique blur kernel of the PET system to effectively apply ultra-high resolution algorithms.

In the prior art, it is assumed that the blur kernel of the PET system is recognized in advance. That is, it can be predicted using information on a predetermined blur kernel. However, since the blur kernel may be different for each PET system, it is necessary to accurately estimate the blur kernel information of the PET system during image reconstruction, but the performance of the ultra high resolution algorithm can be improved.

Accordingly, in the present invention, each PSF kernel corresponding to each position of the PET system is generated using the Monte-Carlo simulation, and a blur modeling DB is created to apply it to the super-high resolution algorithm in image reconstruction.

Then, as shown in FIG. 3, the PET system is wobbled to obtain respective reaction lines and sinograms according to respective positions (420). The obtained sinogram is made based on the reaction line.

The relationship between the wobbling of the present invention and an ultra high resolution algorithm will be described below. A sinogram set can be obtained using a sinogram obtained at each location with wobbling.

Then, the entire reaction line based sinogram obtained in the PET system is converted into a sinogram format in the form of a cone beam (aka quadrangular pyramid) as shown in FIG. 5 (430). Using this cone-beam based response line reconstruction, the present invention is to perform image reconstruction in parallel using a GPU or the like. In order to understand the present invention, the cone-beam-based response line reconfiguration of the present invention will be described later in more detail.

Thereafter, a sinogram corresponding to format conversion into the cone-beam shape is modeled using Equation (1) below (440). In the present invention, a high-resolution image is warped by warping as shown in Equation (1), and the projected data is subjected to blurring and downsampling and then modeled as a sinogram. The modeling of a sinogram using Equation (1) is defined as an ultra-high resolution algorithm in the present invention. That is, in the present invention, a sinogram that is format converted into a cone-beam format is converted into a high-resolution sinogram to be reconstructed into a high-resolution image.

Where y _k is a k-th low-resolution sinogram, A is a system projection matrix, D is a down sampling matrix, B is a blurring matrix, R is a wobbling matrix, and X is a high-resolution image.

Particularly, in modeling a sinogram that is format converted into a cone-beam shape, a modeling method in which an image is projected and then blurred can be considered. However, this method requires a large amount of computation because the projection data must be blurred, thereby increasing the complexity and slowing the processing speed.

Accordingly, in the present invention, as shown in Equations (1) and (2), after the image is blurred, a sinogram-based projection that is format-converted into a cone-beam shape is performed to reduce the calculation amount and reduce the complexity, The image reconstruction processing speed can be guaranteed.

Then, a high-resolution image X in the above-mentioned sinogram modeling is derived 450 by a GPU parallel processing operation using an iterative reconstruction algorithm. MLEM (Maximum Likelihood Expectation Maximization) is a typical iterative reconfiguration algorithm. However, it is not possible to use the EM (Expectation Maximization) -based iterative reconstruction algorithm such as ordered-subset expectation maximization (OSEM) It can be used in the invention.

That is, a high-resolution image X is derived using MLEM as shown in the following equation (2).

It can be seen that a number of forward projections "A" and back-projections "A ^T " are required to perform the iterative reconstruction algorithm to derive the high resolution image X as in Equation (2) have. Hereinafter, the cone-beam-based response line reconfiguration of the present invention will be described in more detail.

Parallel operation processing becomes difficult when modeling based on a general reaction line (LOR) in performing a plurality of forward projections "A" and back-projection "A ^T " operations.

In other words, both the forward projection and the reverse projection can be performed in a line-driven and voxel-driven manner, using different operators as shown in the following Table 1 do. If the scatter operator is executed in parallel, data loss occurs. You can use the atomic operator to prevent this loss, but because of the high cost, parallel processing performance is greatly degraded. Therefore, you should use the gather operator to maximize parallel processing performance.

In other words, when MLEM is processed in parallel, it is optimal to perform the forward projection line-driven and the reverse projection voxel-driven. However, it is not easy to perform reverse projection in a voxel-driven manner in a general parallel beam type sinogram format. For voxel-driven, the combination of each voxel and neighboring reaction lines (LORs) must be obtained. In the above-mentioned common sinogram format, the process of calculating the combination is complicated.

Therefore, in order to solve this problem, in the present invention, the entire reaction line obtained in the PET system is converted into a sinogram format in the form of a cone beam (aka quadrangular pyramid) cone-beam based re-line reconstruction.

Referring again to FIGS. 5 and 430, when the reaction line is reconstructed on the basis of a cone-beam, the combination of the reaction lines is calculated using a cone having a crystal crystal vertex as a vertex, So that the voxel-driven based reverse projection can be performed by the gather operator. Even if the sinogram format is changed in cone-beam format, the forward-projection can continue to use line-driven, so that the entire MLEM process can be performed by a gather operator suitable for the GPU.

For example, the reaction lines may be converted to a cone-shaped sinogram format with the vertex of the crystal of the first detector in the first region. The reaction lines converted into the cone-shaped sinogram format are spread out in a cone shape in a second region having a second detector as a vertex of the crystal of the first detector in the first region. At this time, the second detector (second region) may be located on a circular shape facing the first detector (first region).

The following Table 1 shows scatter and gather operations in MLEM.

Forward projection Backprojection Line-driven Gather Scatter Voxel-driven Scatter Gather

The parallel processing of the forward projection and the back projection in the image reconstruction of the present invention will be described as follows. Parallel processing is a method of dividing work of the same kind in a plurality of arithmetic cores and simultaneously calculating them. A CPU or GPU with multicore can be used for parallel processing. GPUs typically have dozens or hundreds of cores to perform large-scale parallel processing, as well as graphics operations as well as mathematical operations.

6 is a block diagram of an apparatus for reconstructing an ultrahigh-resolution PET image using cone-beam-based response line reconstruction according to an embodiment of the present invention.

6, an apparatus for reconstructing an ultrahigh resolution PET image using cone-beam based response line reconstruction according to the present invention includes a blur modeling unit 610, a reaction line acquiring unit 620, a cone- A conversion unit 630, a sinogram modeling unit 640, and a high resolution image deriving unit 650. [

The blur modeling unit 610 generates a blur modeling database (DB) by generating a PSF kernel for each position of the PET system using a Monte-Carlo simulation.

The reaction line acquisition unit 620 acquires each reaction line and a sino graph according to each position by wobbling the PET system as shown in FIG.

The cone-beam sinogram transformer 630 transforms the entire reaction line-based sinogram obtained in the PET system into a sinogram format in the form of a cone beam (aka quadrangular pyramid) as shown in Fig.

The sinogram modeling unit 640 models a sinogram that has undergone format conversion into the cone-beam format using Equation (1).

The high-resolution image derivation unit 650 derives a high-resolution image X from the sinogram modeling by a GPU parallel processing operation using a recursive reconstruction algorithm such as Equation (2).

Meanwhile, the method of the present invention as described above can be written in a computer program. And the code and code segments constituting the program can be easily deduced by a computer programmer in the field. In addition, the created program is stored in a computer-readable recording medium (information storage medium), and is read and executed by a computer to implement the method of the present invention. And the recording medium includes all types of recording media readable by a computer.

It will be apparent to those skilled in the art that various modifications and variations can be made in the present invention without departing from the spirit or scope of the invention. The present invention is not limited to the drawings.

610: Blur modeling unit 620: Reaction line acquiring unit
630: Cone-beam sinogram conversion unit 640: Sinogram modeling unit
650: High resolution image deriving unit

Claims (10)

In an image reconstruction apparatus of positron emission tomography (PET)
A blur modeling unit for modeling blur by generating respective PSF kernels according to respective positions of the PET system;
A reaction line acquiring unit for acquiring respective reaction lines and a sinogram according to respective positions by wobbling the PET system;
A cone-beam intersymbol transformer for transforming the reaction-line-based sinogram obtained by the reaction-line acquiring unit into a cone-beam type sinogram format;
A sinogram modeling unit for modeling a sinogram format-converted into a cone-beam shape in the cone-beam cosine transform unit; And
A high-resolution image deriving unit for deriving a high-resolution image using the sinogram modeled by the sinogram modeling unit,
Resolution PET reconstruction using cone - beam - based response line reconstruction.
The method according to claim 1,
The sinogram modeling unit may include a high-resolution image, a wobbling obtained by the response-line acquiring unit, and a blur modeling unit used in the blur modeling unit to convert the format into cone-beam format in the cone-beam intersymbol conversion unit. Resolution PET image reconstruction using cone-beam-based response line reconstruction.
The method according to claim 1,
Wherein the high-resolution image derivation unit derives a high-resolution image by performing a forward projection and a reverse projection on a sinogram modeled by the sinogram modeling unit by a parallel processing operation. High resolution PET image reconstruction device.
4. The method according to any one of claims 1 to 3,
The sinogram modeling unit warps a high-resolution image by wobbling, and after the blurring and downsampling, the projected data is modeled as a sinogram. An ultrahigh-resolution PET Image reconstruction device.
5. The method of claim 4,
Wherein the blur modeling unit models the blur by generating respective PSF kernels corresponding to each position according to a Monte-Carlo simulation based on a point source, and reconstructs the ultra-high resolution PET image using the cone-beam based response line reconstruction.
5. The method of claim 4,
Wherein the cone-beam cosine transform unit transforms the sinogram into a cone shape in a second region having a second detector with the crystal of the first detector on the first region of the PET system as a vertex, High Resolution PET Image Reconstruction Using Cone - Beam - Based Reactor Reconstruction.
5. The method of claim 4,
Wherein the high resolution image derivation unit derives a high resolution image by performing parallel projection processing based on the iterative reconstruction algorithm based forward projection and the inverse projection to derive a high resolution image.
8. The method of claim 7,
Wherein the high resolution image derivation unit uses a Gather operator in a forward projection and a Scatter operator in a back projection to perform a line-driven based parallel processing operation, and reconstructs an ultra high resolution PET image using a cone-beam based response line reconstruction Device.
8. The method of claim 7,
Wherein the high-resolution image derivation unit uses a scatter operator in a forward projection and a Gather operator in a reverse projection to perform a voxel-driven based parallel processing operation, and reconstructs an ultra-high resolution PET image using a cone-beam based response line reconstruction Device.
In an image reconstruction method of positron emission tomography (PET)
Generating each PSF kernel according to each position of the PET system to model a blur;
Obtaining each reaction line and a sinogram according to each position by wobbling the PET system;
Converting the obtained response line based sinogram to a cone-beam type sinogram format;
Modeling the sinogram formatted in the cone-beam format; And
And deriving a high-resolution image using the modeled sinogram
Resolution PET image reconstruction using cone - beam - based response line reconstruction.

Images