JP2019113474A - Image processing device and image processing method - Google Patents

Abstract

To provide a device and method with which it is possible to create a highly efficient noise-rejected tomographic image on the basis of list data collected by a radiation tomographic device.SOLUTION: An image processing device 10 includes a processing target image creation unit 11, a guidance image creation unit 12, and a filter processing unit 13. The processing target image creation unit 11 divides list data into a plurality of frames in order in which the data was collected and creates a processing target image Pof each frame using a data group, out of the list data, that is included in each frame. The guidance image creation unit 12 creates a guidance image Ithat corresponds to the processing target image Pusing a data group having a greater number of data than the data group used in creating the processing target image Pof each frame. The filter processing unit 13 performs a noise-rejection process by a guided filer using the corresponding processing target image (input image) Pand guidance image Iof each frame.SELECTED DRAWING: Figure 2

Description

  The present invention relates to an apparatus and method for creating a tomographic image after noise removal processing based on list data collected by a radiation tomography apparatus.

  Examples of a radiation tomography apparatus capable of acquiring a tomographic image of a subject (a living body) include a PET (Positron Emission Tomography) apparatus and a SPECT (Single Photon Emission Computed Tomography) apparatus.

  The PET apparatus includes a detection unit having a large number of small-sized radiation detectors arranged around the measurement space in which the subject is placed. The PET apparatus simultaneously detects the 511 keV photon pairs generated by the annihilation of electrons and positrons in the subject to which the positron emitting isotope (RI source) has been introduced by the coincidence detection method. Gather information Then, based on the collected large amount of coincidence information, a tomographic image representing the spatial distribution of the occurrence frequency of photon pairs in the measurement space (that is, the spatial distribution of the RI source) can be reconstructed. At this time, list data in which coincidence counting information collected by a PET apparatus is arranged in time series is divided into a plurality of frames in the collection order, and image reconstruction processing is performed using a data group included in each frame among the list data. By doing this, it is possible to obtain a dynamic PET image consisting of tomographic images of a plurality of frames. The PET apparatus plays an important role in the field of nuclear medicine and the like, and can be used to study, for example, biological functions and higher-order functions of the brain.

  Since the tomographic image reconstructed in this manner contains a large amount of noise, it is necessary to perform noise removal processing using an image filter. Image filters used for noise removal include Gaussian filters and guided filters. Conventionally, Gaussian filters have been used. On the other hand, the guided filter, which was developed in recent years, is characterized in that it can well preserve the boundary of light and shade in the image as compared with the Gaussian filter.

  Patent Document 1 and Non-patent Document 1 describe a technique for removing noise of a dynamic PET image by a guided filter. The techniques described in Patent Document 1 and Non-Patent Document 1 use, as a guidance image, an image obtained by integrating dynamic PET images composed of tomographic images of a plurality of frames in noise removal processing using a guided filter.

Chinese Patent Application Publication No. 103955899

Lijun Lu ,, et al. "Dynamic PET Denoising Incorporating a Composite Image Guided Filter," IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS / MIC), 2014.

  The noise removal processing technique for dynamic PET images with a guided filter described in Patent Document 1 and Non-Patent Document 1 has superior noise removal performance as compared to the case of using Gaussian filters. However, further improvement in noise removal performance is desired for PET images and SPECT images.

  The present invention has been made to solve the above problems, and provides an apparatus and method capable of creating a high-performance noise-removed tomographic image based on list data collected by a radiation tomography apparatus. Intended to be provided.

  An image processing apparatus according to a first aspect of the present invention is an apparatus for creating a tomographic image after noise removal processing based on list data collected by a radiation tomography apparatus, comprising: (1) a plurality of list data in the order of collection; A processing target image creation unit that divides a frame and creates a reconstructed image as a processing target image of the frame using a data group included in each frame of the list data; (2) a processing target image of the list data Guidance image creation for creating a reconstructed image as a guidance image corresponding to the processing target image of the frame using a data group having a larger number of data than the data group used when creating the processing target image of each frame in the creating unit Part, and (3) processing target image of each frame created in the processing subject image creating part, and processing of the frame in the guidance image creating part Using the guidance image produced in response to elephant image, it performs noise removal processing by Gaidido filter, and a filter processing section for creating a tomographic image after the noise removal processing of the frame.

  An image processing apparatus according to a second aspect of the present invention is an apparatus for creating a tomographic image after noise removal processing based on list data collected by a radiation tomography apparatus, comprising: (1) a plurality of list data in the order of collection; A processing target image creation unit that creates a sinogram as a processing target image of the frame by dividing it into frames and using a data group included in each frame of the list data; (2) a processing target image creation unit of the list data A guidance image creation unit that creates a sinogram as a guidance image corresponding to the processing target image of the frame using a data group having a larger number of data than the data group used when creating the processing target image of each frame in 3) The processing target image of each frame created in the processing target image creation unit, and the guidance image creation unit A filter processing unit that performs noise removal processing by a guided filter using a guidance image created corresponding to the image to be processed, and creates a sinogram after the noise removal processing of the frame; And an image reconstructor configured to reconstruct a tomographic image of the frame using the created sinogram of each frame.

  An image processing apparatus according to a third aspect of the present invention is an apparatus for creating a tomographic image after noise removal processing based on list data collected by a radiation tomography apparatus, comprising: (1) a plurality of list data in the order of collection; A sinogram creation unit that creates a sinogram of the frame using the data group that is divided into frames and is included in each frame of the list data, (2) an initial image creation unit that creates an initial image, (3) multiple A forward projection calculation unit that performs forward projection calculation based on an input image for each of the frames in order to create a sinogram; and (4) a sinogram created by the forward projection calculation unit for each of a plurality of frames and a sinogram creation unit A back projection calculation unit that performs back projection calculation based on comparison with a sinogram; (5) a back projection calculation unit for each of a plurality of frames A processing target image creation unit that creates an image after the image update calculation as a processing target image by performing an image update calculation based on the back projection calculation result, and (6) the sinogram creation unit of the list data A guidance image creation unit that creates a reconstructed image as a guidance image corresponding to the processing target image of the frame using a data group having a larger number of data than the data group used when creating the sinogram; (7) a processing target The noise removal processing is performed by the guided filter using the processing target image of each frame created in the image creating unit and the guidance image created corresponding to the processing target image of the frame in the guidance image creating unit, And a filter processing unit that generates an image after the noise removal processing of the frame. Then, the image processing apparatus repeatedly performs forward projection calculation by the forward projection calculation unit, back projection calculation by the back projection calculation unit, image update calculation by the processing target image creation unit, and noise removal processing by the filter processing unit. The section sets the initial image as the input image at the first time of the repetitive processing, and thereafter sets the image after the noise removal processing by the filter processing section as the input image.

  In the image processing apparatus according to the present invention, the guidance image generation unit normalizes the guidance image corresponding to the processing target image of each frame so that the maximum pixel value of the guidance image becomes equal to the maximum pixel value of the processing target image. It is preferable to create. When creating a guidance image corresponding to the processing target image of each frame, the guidance image creating unit uses a data group including a data group used when creating the processing target image of the frame in the processing target image creating unit. Is preferred. Further, when the guidance image creation unit creates a guidance image corresponding to the processing target image of each frame, data before and after the data group used when creating the processing target image of the frame in the processing target image creation unit It is preferred to use groups.

  A radiation tomography system according to the present invention comprises: a radiation tomography apparatus for collecting list data for reconstructing a tomographic image of an object; and a tomography after noise removal processing based on the list data acquired by the radiation tomography apparatus And the above-mentioned image processing apparatus of the present invention for creating an image.

  An image processing method according to a first aspect of the present invention is a method for creating a tomographic image after noise removal processing based on list data collected by a radiation tomography apparatus, comprising: A process target image creating step of creating a reconstructed image as a process target image of the frame by dividing into frames and using a data group included in each frame among the list data; (2) a process target image of the list data Guidance image creation for creating a reconstructed image as a guidance image corresponding to the processing target image of the frame using a data group having a larger number of data than the data group used when creating the processing target image of each frame in the creating step And (3) processing target images of each frame created in the processing target image creating step, and guidance image creation A filter processing step of performing noise removal processing with a guided filter using a guidance image created corresponding to the processing target image of the frame in step b, and creating a tomographic image after the noise removal processing of the frame; Equipped with

  An image processing method according to a second aspect of the present invention is a method for creating a tomographic image after noise removal processing based on list data collected by a radiation tomography apparatus, comprising: A processing target image creation step of creating a sinogram as a processing target image of the frame by dividing it into frames and using a data group included in each frame among the list data; (2) processing target image creation step of the list data A guidance image creation step of creating a sinogram as a guidance image corresponding to the processing target image of the frame using a data group having a larger number of data than the data group used when creating the processing target image of each frame in 3) Process target image of each frame created in the process target image creation step, and guidance image creation (4) performing a noise removal process by the guided filter using a guidance image created corresponding to the processing target image of the frame in the step, and creating a sinogram after the noise removal process of the frame; An image reconstruction step of reconstructing a tomographic image of the frame using the sinogram of each frame created in the filtering step.

  An image processing method according to a third aspect of the present invention is a method of creating a tomographic image after noise removal processing based on list data collected by a radiation tomography apparatus, comprising: (1) a plurality of list data in the order of collection; A sinogram creation step of creating a sinogram of the frame by dividing it into frames and using a data group included in each frame of the list data, (2) an initial image creating step of creating an initial image, (3) multiple Forward projection calculation step of performing forward projection calculation based on the input image for each of the frames to create a sinogram, and (4) a sinogram created in the forward projection calculation step for each of a plurality of frames and a sinogram created step Backprojection calculation step of performing backprojection calculation based on comparison with the sinogram; (5) A process target image creation step of performing image update calculation for each of the frames based on the back projection calculation result in the back projection calculation step to create an image after the image update calculation as a processing target image; Among them, using a data group having a larger number of data than the data group used when creating the sinogram of each frame in the sinogram creation step, creating a guidance image that creates a reconstructed image as a guidance image corresponding to the processing target image of the frame. And (7) using the guidance target image of each frame created in the processing target image creation step and the guidance image created corresponding to the processing target image of the frame in the guidance image creation step. The noise is removed by the filter and the noise of the frame is Comprises a filtering step of creating an image after removal process, the. Then, this image processing method repeatedly performs the forward projection calculation step, the back projection calculation step, the processing target image creation step and the filter processing step, and in the forward projection calculation step, the initial image is used as the input image at the first time of the repeated processing. In the following, the image after the noise removal processing in the filtering step is used as the input image.

  In the image processing method of the present invention, the guidance image creating step normalizes the guidance image corresponding to the processing target image of each frame so that the maximum pixel value of the guidance image becomes equal to the maximum pixel value of the processing target image. It is preferable to create. The guidance image creating step uses a data group including a data group used when creating a processing target image of the frame in the processing target image creating step when creating a guidance image corresponding to the processing target image of each frame. Is preferred. Also, when the guidance image creation step creates a guidance image corresponding to the processing target image of each frame, data before and after the data group used when creating the processing target image of the frame in the processing target image creation step It is preferred to use groups.

  According to the present invention, it is possible to create a high performance noise-removed tomographic image based on the list data collected by the radiation tomography apparatus.

FIG. 1 is a graph showing a time change of acquisition frequency of coincidence counting information in a subject into which an RI radiation source is input. FIG. 2 is a view showing the arrangement of a radiation tomography system 1A according to the first embodiment. FIG. 3 is a flowchart illustrating the image processing method of the first embodiment. Figure 4 is a diagram illustrating a list data, frame, the processing target image P _m and guidance image I _m. FIG. 5 is a diagram for explaining a simulation method. FIG. 5 (a) shows a numerical phantom. FIG. 5 (b) shows a sinogram. FIG. 5 (c) shows a noise addition sinogram. FIG. 5 (d) shows a reconstructed image. FIG. 6 is a graph showing a model of temporal change in activity of each of a tumor part (Tumor), a white matter part (WM) and a gray matter part (GM) used in the simulation. FIG. 7 is a view showing a reconstructed image obtained in the simulation. FIG. 8 is a view showing an image obtained by performing the noise removal processing of Comparative Example 1 on the reconstructed image of FIG. 7. FIG. 9 is a view showing an image obtained by performing the noise removal processing of Comparative Example 2 on the reconstructed image of FIG. 7. FIG. 10 is a view showing an image obtained by performing the noise removal processing of the present embodiment on the reconstructed image of FIG. 7. FIG. 11 is a graph showing temporal changes in MSE of reconstructed images in Comparative Example 1 and Comparative Example 2 obtained in the simulation and in the present embodiment. FIG. 12 is a graph showing a time change of PSNR of reconstructed images in Comparative Example 1, Comparative Example 2 and this embodiment obtained in the simulation. FIG. 13 is a graph showing temporal changes in SSIM of reconstructed images in Comparative Example 1, Comparative Example 2 and the present embodiment obtained in the simulation. FIG. 14 is a graph showing the time change of the activity of the tumorous part (Tumor) in each of Comparative Example 1 and Comparative Example 2 obtained in the simulation and the present embodiment. FIG. 15 is a graph showing the time change of the activity of the white matter portion (WM) in Comparative Example 1, Comparative Example 2 and the present embodiment obtained in the simulation. FIG. 16 is a graph showing the time change of the activity of the gray matter portion (GM) in Comparative Example 1, Comparative Example 2 and the present embodiment obtained in the simulation. FIG. 17 is a graph showing the MAE of each of the tumorous part, the white matter part and the gray matter part for each of Comparative Example 1 and Comparative Example 2 obtained in the simulation and the present embodiment. FIG. 18 is a graph showing the RMSE of each of the tumorous part, the white matter part and the gray matter part for each of Comparative Example 1 and Comparative Example 2 obtained in the simulation and the present embodiment. FIG. 19 is a view showing an MRI image obtained in the example. FIG. 20 is an enlarged view of a part of the MRI image of FIG. FIG. 21 is a view showing a PET reconstructed image obtained in the example. FIG. 22 is a view showing an image obtained by performing the noise removal processing of Comparative Example 1 on the reconstructed image of FIG. FIG. 23 is a view showing an image obtained by performing the noise removal processing of Comparative Example 2 on the reconstructed image of FIG. FIG. 24 is a view showing an image obtained by performing the noise removal processing of the present embodiment on the reconstructed image of FIG. FIG. 25 is a diagram showing the configuration of a radiation tomography system 1B according to the second embodiment. FIG. 26 is a flowchart for explaining the image processing method of the second embodiment. FIG. 27 is a graph showing a time change of MSE of reconstructed images in Comparative Example 1 and Comparative Example 2 obtained in the simulation and in the present embodiment. FIG. 28 is a graph showing a time change of PSNR of reconstructed images in Comparative Example 1, Comparative Example 2 and this embodiment obtained in the simulation. FIG. 29 is a graph showing temporal changes in SSIM of reconstructed images in Comparative Example 1, Comparative Example 2 and the present embodiment obtained in simulation. FIG. 30 is a diagram showing a configuration of a radiation tomography system 1C of the third embodiment. FIG. 31 is a flow chart for explaining the image processing method of the third embodiment. FIG. 32 is a view showing an image obtained by the noise removal processing of the third embodiment. FIG. 33 is a graph showing time change of MSE of reconstructed images in Comparative Example 1, Comparative Example 2, First Embodiment, and Third Embodiment obtained in the simulation. FIG. 34 is a graph showing a time change of PSNR of reconstructed images in Comparative Example 1, Comparative Example 2, First Embodiment, and Third Embodiment obtained in the simulation. FIG. 35 is a graph showing temporal changes in SSIM of reconstructed images in Comparative Example 1, Comparative Example 2, and the first and third embodiments obtained in the simulation.

  Hereinafter, with reference to the accompanying drawings, modes for carrying out the present invention will be described in detail. In the description of the drawings, the same elements will be denoted by the same reference symbols, without redundant description. The present invention is not limited to these exemplifications, is shown by the claims, and is intended to include all modifications within the scope and meaning equivalent to the claims.

First, an overview of the guided filter will be described. The guided filter performs noise removal processing using the input image P and the guidance image I on the assumption of a linear conversion model represented by the following equation (1), and creates an output image Q after the noise removal processing. Q _i is the pixel value of pixel i in the output image Q. I _i is the pixel value of pixel i in the guidance image I. ω _k represents a small area (patch area) composed of a plurality of pixels centered on the pixel k in the input image P. a _k and b _k are coefficients.

The error function E (a _k , b _k ) is expressed by the following equation (2). The coefficients a _k and b _k that minimize the error function E (a _k , b _k ) are _obtained . ε is a regularization parameter with a _k as a penalty, and is expressed by the following equation (3). σ is an estimated standard deviation of noise in the entire input image P. α is a constant.

If this equation (2) is solved by the regularized least squares method (linear ridge regression), the following equations (4) and (5) are obtained. <P _k > is an average pixel value in the small area ω _k of the input image P. mu _k is the average pixel value in the small area omega _k of the guidance image I. sigma _k is the variance of pixel values in the small regions omega _k of the guidance image I. | ω _k | is the number of pixels in the small area ω _k .

In order to reduce the influence of noise, as expressed by the following equation (6) and (7), the average value <a _k coefficients _a k, _{b k} in small region ω _{i>, <b k>} Ask for Then, the <a _k>, using <b _k>, by the following equation (8), it is possible to obtain an output image Q after the noise removal processing.

In general, in a guided filter, the guidance image I may be the same as the input image P. That is, assuming that the dynamic PET image consists of tomographic images of M frames, the tomographic image of the mth frame of these M frames is used as both the input image P _m and the guidance image I _m by the guided filter. it may generate the output image Q _m after the noise removal processing of the m-th frame.

The noise removal processing technique for dynamic PET images with a guided filter described in Patent Document 1 and Non-patent Document 1 uses the sum of tomographic images (input images P _m ) of M frames of dynamic PET images as a guidance image I Used (the following (9) formula). The guidance image I has noise reduced from each input image P _m . Therefore, in Patent Document 1 and Non-patent Document 1, by using this guidance image I, it is considered that noise removal processing with higher performance can be performed on each input image P _m .

  In general, the frequency of electron-positron pair annihilation (acquisition frequency of coincidence counting information) in a subject into which an RI radiation source is injected is, as shown in FIG. It becomes higher and eventually reaches a peak at a certain time Tp, and after that peak is gradually lowered gradually. FIG. 1 is a graph showing a time change of acquisition frequency of coincidence counting information in a subject into which an RI radiation source is input.

  From this, the period of each frame is set relatively short in a period in which the frequency of acquisition of coincidence count information is high immediately after the RI source is turned on, and the frequency of acquisition of coincidence count information thereafter gradually decreases. In the period, the period of each frame is set relatively long. In addition, tomographic images are significantly different between the period (before time Tp) until the frequency of acquisition of coincidence count information reaches a peak immediately after RI source injection and the period after time Tp (after time Tp). It is known that

The noise removal processing techniques described in Patent Document 1 and Non-patent Document 1 do not take into account that the period of each frame is not constant, nor do they take into account that the tomographic images before and after the peak are largely different. , A simple sum of tomographic images (input image P _m ) of M frames is used as a guidance image I. From this, there is a limit to the performance of noise removal for the tomographic image of each frame.

  The radiation tomography system, the image processing apparatus, and the image processing method of the embodiments described below can create a high-performance noise-removed tomographic image based on the list data collected by the radiation tomography apparatus.

First Embodiment
FIG. 2 is a view showing the arrangement of a radiation tomography system 1A according to the first embodiment. The radiation tomography system 1A includes a radiation tomography apparatus 2 and an image processing apparatus 10. The image processing apparatus 10 includes a processing target image creation unit 11, a guidance image creation unit 12, a filter processing unit 13, and a storage unit 15. A computer is used as the image processing apparatus 10.

  The radiation tomography apparatus 2 is an apparatus for collecting list data for reconstructing a tomographic image of a subject. Examples of the radiation tomography apparatus 2 include a PET apparatus and a SPECT apparatus. In the following, it is assumed that the radiation tomography apparatus 2 is a PET apparatus.

  The radiation tomography apparatus 2 includes a detection unit having a large number of small-sized radiation detectors arranged around a measurement space in which a subject is placed. The radiation tomography apparatus 2 detects, by the coincidence counting method, a photon pair having an energy of 511 keV generated with the annihilation of the electron-positron pair in the subject into which the positron-emitting isotope (RI source) is introduced, This coincidence counting information is accumulated. Then, the radiation tomography apparatus 2 outputs, to the image processing apparatus 10, an arrangement of the accumulated large number of simultaneous counting information in time series as list data.

  The list data includes identification information and detection time information of a pair of radiation detectors that simultaneously count photon pairs. The list data may further include energy information of photons detected by each radiation detector and detection time difference information of a pair of radiation detectors.

  The image processing apparatus 10 reconstructs a tomographic image based on the list data. As a technique for reconstructing a tomographic image based on list data, a ML-EM (maximum likelihood expectation maximization) method and a successive approximation image reconstruction technique by a block iteration method which is an improvement thereof are known. Also, as successive approximation image reconstruction technology by block iteration method, OSEM (ordered subset ML-EM) method, RAMLA (row-action maximum likelihood algorithm) method, DRAMA (dynamic RAMLA) method, etc. are known. The image processing apparatus 10 also performs noise removal processing using a guided filter to create a tomographic image after the noise removal processing.

The processing target image creation unit 11 divides the list data into a plurality of frames in the order of collection, and creates a reconstructed image as the processing target image P _m of the frame using a data group included in each frame among the list data. Do. The guidance image creation unit 12 processes the frame using a data group having a larger number of data than the data group used when creating the processing target image P _m of each frame in the processing target image creation unit 11 among the list data. creating a reconstructed image as guidance image I _m corresponding to the target image P _m. The processing target image creating unit 11 and the guidance image creating unit 12 may be common or may be provided separately.

The filter processing unit 13 performs noise removal processing by the guided filter using the processing target image (input image) P _m and the guidance image I _m corresponding to each frame, and the tomographic image of the frame after the noise removal processing is create. The storage unit 15 stores list data, and stores the processing target image P _m and the guidance image I _m corresponding to each frame. In addition, the storage unit 15 stores the tomographic image after the noise removal processing of each frame.

  FIG. 3 is a flowchart illustrating the image processing method of the first embodiment. The image processing method according to the present embodiment includes a list data acquisition step S10 for acquiring list data collected by the radiation tomography apparatus 2, a processing target image generation step S11 performed by the processing target image generation unit 11, and a guidance image generation unit A guidance image creation step S12 performed by 12 and a filter processing step S13 performed by the filter processing unit 13 are provided.

In the list data acquisition step S10, the list data acquired by the radiation tomography apparatus 2 is acquired. In the processing target image creation step S11, the list data is divided into a plurality of frames in the order of collection, and a reconstructed image is created as the processing target image P _m of the frame using a data group included in each frame in the list data. Do. In the guidance image producing step S12, by using a large number of data of the data group from the data group used in preparing the processing target image P _m for each frame in the processed image producing step S11, the processing target image P _m of the frame creating a reconstructed image as guidance image I _m corresponding to. The processing target image creation step S11 and the guidance image creation step S12 may be performed in any order, or may be performed in parallel. In the filter processing step S13, noise removal processing is performed by the guided filter using the corresponding processing target image (input image) P _m and guidance image I _m of each frame, and the tomographic image of the frame after the noise removal processing is create.

Figure 4 is a diagram illustrating a list data, frame, the processing target image P _m and guidance image I _m. In this figure, the horizontal axis represents the time when list data was collected, and its time axis is divided into a plurality of frames. The processing target image P _m of the m-th frame is a tomographic image reconstructed using a data group included in the m-th frame of the list data. The guidance image I _m corresponding to the processing target image P _m of the m-th frame is a tomographic image reconstructed using a data group having a larger number of data than the data group included in the m-th frame in the list data.

The guidance image I _m corresponding to the processing target image P _m of the m-th frame is preferably created using a data group having twice or more the number of data of the data group of the m-th frame. Guidance image I _m corresponding to the processing target image P _m of the m-th frame may be created using a data group including a group of data of the m frame, the data groups before and after the data group of the m frame It may be created using. When the m-th frame is time Tp previously (flow-region) in FIG. 1, the guidance image I _m corresponding to the processing target image P _m of the m-th frame, the time Tp previous data group among the list data It is preferable to create using. When the m-th frame is the time Tp after (PET ligand kinetics dependent area) in FIG. 1, the guidance image I _m corresponding to the processing target image P _m of the m-th frame, the time Tp after the data of the list data It is preferred that they be created using groups. Guidance image I _m corresponding to the processing target image P _m of the m-th frame may be created by using the whole of the list data.

In the guidance image generation unit 12 (the guidance image creating step S12), the guidance image I _m corresponding to the processing target image P _m of the m-th frame, the maximum pixel value of the maximum pixel value of the guidance image I _m is the processing target image P _m It is preferable that normalization is performed to be equal to. That is, it is preferable to use a normalized guidance image I ' _m represented by the following equation (10). MaxP _m is the maximum pixel value of the processing target image P _m of the m-th frame. MaxI _m is the maximum pixel value of the guidance image I _m corresponding to the processing target image P _m of the m-th frame.

Next, simulation results will be described. FIG. 5 is a diagram for explaining a simulation method. First, a numerical phantom (FIG. 5 (a)) was prepared. This numerical phantom simulates a tomographic image of the monkey brain into which ¹⁸ F-FDG (fluorodeoxyglucose) has been added. The numerical phantom includes a white matter part (WM) and a gray matter part (GM) of the brain as well as a tumor part (Tumor). FIG. 6 is a graph showing a model of temporal change in activity of each of a tumor part (Tumor), a white matter part (WM) and a gray matter part (GM) used in the simulation. Time-Activity Curve (TAC) of activity generally increases gradually with time immediately after turning on the RI source, reaches a peak at a certain time, and then gradually decreases after that peak. It will be. The total count number (the number of simultaneous count information) was set to 15,000,000.

  A sinogram (FIG. 5 (b)) of each frame was created based on this numerical phantom. The sinogram is a histogram of coincidence counting information for each pair of radiation detectors of the radiation tomography apparatus 2. Each period of the first to sixth frames is 10 seconds (1 minute subtotal), each period of the 7th to 8th frames is 30 seconds (1 minute subtotal), and each period of the 9th to 16th frames Is 60 seconds (a total of 8 minutes), each period of the 17th to 26th frames is 300 seconds (a total of 50 minutes), and each period of the 27th to 32nd frames is 600 seconds (a total of 60 minutes). The collection time of the entire list data was 120 minutes.

  Poisson noise according to the count number of each frame was given to the sinogram of each frame to create a noise-added sinogram (FIG. 5 (c)). Based on the noise-added sinogram, a reconstructed image (FIG. 5 (d)) was created by the ML-EM method. The number of iterations in the ML-EM method was 40 times.

FIG. 7 is a view showing a reconstructed image obtained in the simulation. This reconstruction image shows a tomographic image of the 30th frame (period of 5400 to 6000 seconds) before the noise removal processing. FIG. 8 is a view showing an image obtained by performing the noise removal processing of Comparative Example 1 on the reconstructed image of FIG. 7. The noise removal process of Comparative Example 1 is a process using a Gaussian filter. FIG. 9 is a view showing an image obtained by performing the noise removal processing of Comparative Example 2 on the reconstructed image of FIG. 7. The noise removal process of Comparative Example 1 is the filter process described in Patent Document 1 and Non-Patent Document 1. FIG. 10 is a view showing an image obtained by performing the noise removal processing of the present embodiment on the reconstructed image of FIG. 7. Here, using normalized guidance image I _'m that was created using the entire list data. As can be seen from the comparison of these figures, compared to the noise removal processing of Comparative Examples 1 and 2, in the noise removal processing of the present embodiment, the boundary of the gradation in the image is well preserved and noise is removed well. It is done.

  FIG. 11 is a graph showing temporal changes in MSE of reconstructed images in Comparative Example 1 and Comparative Example 2 obtained in the simulation and in the present embodiment. MSE (Mean Square Error) represents an error with a true image (here, the numerical phantom image in FIG. 5A), and a lower value means closer to a true image. FIG. 12 is a graph showing a time change of PSNR of reconstructed images in Comparative Example 1, Comparative Example 2 and this embodiment obtained in the simulation. The peak signal to noise ratio (PSNR) represents the quality of an image in decibels (dB), and a higher value means better image quality. FIG. 13 is a graph showing temporal changes in SSIM of reconstructed images in Comparative Example 1, Comparative Example 2 and the present embodiment obtained in the simulation. SSIM (Structural Similarity Index) is an index that quantifies changes in brightness and contrast and structure of an image, and a higher value means better image quality. The indices MSE, PSNR and SSIM also indicate that the performance of the noise removal process of the present embodiment is superior to that of the noise removal process of Comparative Examples 1 and 2.

  FIG. 14 is a graph showing the time change of the activity of the tumorous part (Tumor) in each of Comparative Example 1 and Comparative Example 2 obtained in the simulation and the present embodiment. FIG. 15 is a graph showing the time change of the activity of the white matter portion (WM) in Comparative Example 1, Comparative Example 2 and the present embodiment obtained in the simulation. FIG. 16 is a graph showing the time change of the activity of the gray matter portion (GM) in Comparative Example 1, Comparative Example 2 and the present embodiment obtained in the simulation. Each of FIGS. 14-16 also shows the model of the time change of the activity used as the reference | standard shown in FIG.

  Based on these FIGS. 14-16, the error of the activity with respect to the reference | standard was calculated | required about each case of the comparative example 1, the comparative example 2, and this embodiment which were obtained in simulation. FIG. 17 is a graph showing the MAE of each of the tumorous part, the white matter part and the gray matter part for each of Comparative Example 1 and Comparative Example 2 obtained in the simulation and the present embodiment. FIG. 18 is a graph showing the RMSE of each of the tumorous part, the white matter part and the gray matter part for each of Comparative Example 1 and Comparative Example 2 obtained in the simulation and the present embodiment. Both the mean absolute error (MAE) and the root mean square error (RMSE) mean that the smaller the value, the better the quantitativity. Both indices of MAE and RMSE indicate that the quantitative removal is better in the case of the noise removal process of the present embodiment than the noise removal process of Comparative Examples 1 and 2.

Next, an example will be described. In the example, a monkey brain to which ¹⁸ F-FDG was added was used as a subject. An animal PET apparatus (SHR-38000) manufactured by Hamamatsu Photonics K.K. was used as the PET apparatus. Transmission measurement was performed for 30 minutes, and emission measurement was performed for 120 minutes, and absorption correction was performed on the emission measurement result based on the transmission measurement result. The list data was divided into first to thirty-second frames in the same manner as the above-described simulation. The reconstructed image of each frame was created by the ML-EM method. In addition, an MRI (Magnetic Resonance Imaging) image of the subject was also acquired.

  FIG. 19 is a view showing an MRI image obtained in the example. FIG. 20 is an enlarged view of a part of the MRI image of FIG. FIG. 21 is a view showing a PET reconstructed image obtained in the example. This reconstruction image shows a tomographic image of the 30th frame (period of 5400 to 6000 seconds) before the noise removal processing. FIG. 22 is a view showing an image obtained by performing the noise removal processing of Comparative Example 1 on the reconstructed image of FIG. FIG. 23 is a view showing an image obtained by performing the noise removal processing of Comparative Example 2 on the reconstructed image of FIG. FIG. 24 is a view showing an image obtained by performing the noise removal processing of the present embodiment on the reconstructed image of FIG. Here, a normalized guidance image created using the entire list data was used.

  Each of FIGS. 20 to 24 is an enlarged view of a portion corresponding to the rectangular frame portion in FIG. In each of FIG. 20 to FIG. 24, solid arrows indicate the center of the brain (cerebral peduncle), and dashed arrows indicate cerebral wrinkles (central groove). As can be seen from the comparison of these figures, compared to the noise removal processing of Comparative Examples 1 and 2, in the noise removal processing of the present embodiment, the boundary of the gradation in the image is well preserved and noise is removed well. It is a good representation of the brain structure. In addition, spike noise is reduced in the noise removal process of the present embodiment.

As described above, in the present embodiment, in the noise removal processing by the guided filter, processing of the frame is performed using a data group having a larger number of data than the data group used when creating the processing target image P _m of each frame. A guidance image I _m (or a normalized guidance image I ′ _m ) corresponding to the target image P _m is created. As a result, even when the period of each frame is not constant or the number of data of each frame is not constant, it is possible to create a noise-removed tomographic image of each frame with higher performance. Also, when noise removal processing by Gaidido filter, in particular by using a normalized guidance image I _'m, it is possible to further improve the image quality of tomographic images. In addition, when using the guidance image I _m (or the normalized guidance image I ' _m ) in consideration of the movement of the PET ligand in the subject during the noise removal processing by the guided filter, the movement PET that changes the movement rapidly The image quality of the image can be improved.

Second Embodiment
FIG. 25 is a diagram showing the configuration of a radiation tomography system 1B according to the second embodiment. The radiation tomography system 1 </ b> B includes a radiation tomography apparatus 2 and an image processing apparatus 20. The image processing apparatus 20 includes a processing target image creation unit 21, a guidance image creation unit 22, a filter processing unit 23, an image reconstruction unit 24, and a storage unit 25. A computer is used as the image processing apparatus 20.

The image processing apparatus 20 reconstructs a tomographic image based on the list data. The image processing apparatus 20 also performs noise removal processing using a guided filter to create a tomographic image after the noise removal processing. Target image creation section 21 is divided into a plurality of frames list data in order of acquisition, by using the data groups contained in each frame of the list data to create a sinogram as processing target image P _m of the frame. The guidance image creation unit 22 processes the frame using a data group having a larger number of data than the data group used when creating the processing target image P _m of each frame in the processing target image creation unit 21 among the list data. creating a sinogram as guidance image I _m corresponding to the target image P _m. The processing target image creation unit 21 and the guidance image creation unit 22 may be common or may be provided separately.

The filter processing unit 23 performs noise removal processing with a guided filter using the processing target image (input image) P _m and the guidance image I _m corresponding to each frame, and creates a sinogram after the noise removal processing of the frame. Do. The image reconstruction unit 24 reconstructs a tomographic image of the frame using the sinogram after noise removal processing of each frame created in the filter processing unit 23. Storage unit 25 stores the list data, and stores the corresponding processed image P _m and guidance image I _m for each frame. Further, the storage unit 25 stores the tomographic image after the noise removal processing of each frame.

  FIG. 26 is a flowchart for explaining the image processing method of the second embodiment. The image processing method according to the present embodiment includes a list data acquisition step S20 for acquiring list data collected by the radiation tomography apparatus 2, a processing target image generation step S21 performed by the processing target image generation unit 21, and a guidance image generation unit A guidance image generation step S22 performed by the image processing unit 22, a filter processing step S23 performed by the filter processing unit 23, and an image reconstruction step S24 performed by the image reconstruction unit 24 are provided.

In list data acquisition step S20, the list data acquired by the radiation tomography apparatus 2 is acquired. The processed image producing step S21, is divided into a plurality of frame list data in order of acquisition, by using the data groups contained in each frame of the list data, to create a sinogram as processing target image P _m of the frame. In the guidance image producing step S22, by using a large number of data of the data group from the data group used in preparing the processing target image P _m for each frame in the processed image producing step S21, the processing target image P _m of the frame creating a sinogram as guidance image I _m corresponding to. The processing target image creation step S21 and the guidance image creation step S22 may be performed in any order, or may be performed in parallel.

In the filter processing step S23, noise removal processing is performed by the guided filter using the corresponding processing target image (input image) P _m and guidance image I _m of each frame, and a sinogram after the noise removal processing of the frame is created. Do. In the image reconstruction step S24, a tomographic image of the frame is reconstructed using the sinogram after the de-sizing process of each frame created in the filtering step S23.

List data, frame data group used in preparing a processing object image P _m, and the data group used in preparing a guidance image I _m, the relationship between the Figure 4 used in the first embodiment As described above. The processing target image P _m and the guidance image I _m used for the noise removal processing by the guided filter are different in that they are sinograms in the second embodiment, whereas they are reconstruction images in the first embodiment. In this embodiment, the guidance image generation unit 22 (the guidance image creating step S22), and the guidance image I _m corresponding to the processing target image P _m of the m-th frame, the maximum pixel value of the guidance image I _m is the processing target image It is preferable that normalization is performed so as to be equal to the maximum pixel value of P _m (the above equation (10)).

Next, simulation results will be described. In this simulation, as described with reference to FIG. 5 in the first embodiment, a sinogram (FIG. 5 (b)) of each frame is created based on the numerical phantom (FIG. 5 (a)), and further each frame Poisson noise was given to the sinogram of to create a noise-added sinogram (FIG. 5 (c)). The noise addition sinogram (FIG. 5 (c)) was processed image _{P m.} Moreover, after performing the noise removal process of Comparative Example 1, Comparative Example 2 or this embodiment using the noise addition sinogram (FIG. 5 (c)) as the processing object image P _m , the reconstructed image by the ML-EM method It was created. The number of iterations in the ML-EM method was 40 times. In the noise removal processing of the present embodiment, a normalized guidance image I ′ _m corresponding to the processing target image P _m of each frame is created by applying Poisson noise to a sinogram created using the entire list data.

  FIG. 27 is a graph showing a time change of MSE of reconstructed images in Comparative Example 1 and Comparative Example 2 obtained in the simulation and in the present embodiment. FIG. 28 is a graph showing a time change of PSNR of reconstructed images in Comparative Example 1, Comparative Example 2 and this embodiment obtained in the simulation. FIG. 29 is a graph showing temporal changes in SSIM of reconstructed images in Comparative Example 1, Comparative Example 2 and the present embodiment obtained in simulation. As can be seen from these figures, any index of MSE, PSNR and SSIM has better performance in the case of the noise removal process of this embodiment as compared to the noise removal process of Comparative Examples 1 and 2. It shows.

Also in the second embodiment, the same effect as that of the first embodiment can be obtained. That is, in the noise removal processing by the guided filter, a data group having a larger number of data than the data group used in creating the processing target image P _m of each frame is used to correspond to the processing target image P _m of the frame Create a guidance image I _m (or a normalized guidance image I ' _m ). As a result, even when the period of each frame is not constant or the number of data of each frame is not constant, it is possible to create a noise-removed tomographic image of each frame with higher performance. Also, when noise removal processing by Gaidido filter, in particular by using a normalized guidance image I _'m, it is possible to further improve the image quality of tomographic images. In addition, when using the guidance image I _m (or the normalized guidance image I ' _m ) in consideration of the movement of the PET ligand in the subject during the noise removal processing by the guided filter, the movement PET that changes the movement rapidly The image quality of the image can be improved.

Third Embodiment
FIG. 30 is a diagram showing a configuration of a radiation tomography system 1C of the third embodiment. The radiation tomography system 1C includes a radiation tomography apparatus 2 and an image processing apparatus 30. The image processing device 30 includes a sinogram creation unit 31, a guidance image creation unit 32, an initial image creation unit 33, a forward projection calculation unit 34, a back projection calculation unit 35, a processing target image creation unit 36, a filter processing unit 37, and a storage unit 38. including. A computer is used as the image processing apparatus 30.

The image processing device 30 reconstructs a tomographic image based on the list data. The image processing apparatus 30 also performs noise removal processing using a guided filter to create a tomographic image after the noise removal processing. The sinogram creation unit 31 divides the list data into a plurality of frames in the order of collection, and creates a sinogram of the frame using a data group included in each frame among the list data. Guidance image generation unit 32 generates a guidance image I _m for use in the noise removal processing by Gaidido filtering in the filter unit 37. The initial image creation unit 33 creates an initial image. The initial image is, for example, an image in which each pixel value is a positive constant value (for example, a value of 1), although it is arbitrary.

The forward projection calculation unit 34 performs forward projection calculation for each of the plurality of frames based on the input image to create a sinogram. The back projection calculation unit 35 performs back projection calculation based on the comparison between the sinogram created by the forward projection calculation unit 34 and the sinogram created by the sinogram creation unit 31 for each of a plurality of frames. The processing target image creation unit 36 performs image update calculation for each of a plurality of frames based on the back projection calculation result by the back projection calculation unit 35, and creates an image after the image update calculation as the processing target image P _m. . The filter processing unit 37 generates a processing target image P _m of each frame created by the processing target image creating unit 36 and a guidance image created corresponding to the processing target image P _m of the frame by the guidance image creating unit 32. using I _m, it performs noise removal processing by Gaidido filter to create an image after the noise removal processing of the frame.

The guidance image creation unit 32 corresponds to the processing target image P _m of the frame using a data group having a larger number of data than the data group used when creating the sinogram of each frame in the sinogram creation unit 31 among the list data. creating a reconstructed image as guidance image I _m for. Storage unit 38 stores list data, and stores the corresponding processed image P _m and guidance image I _m for each frame. Further, the storage unit 38 stores the tomographic image after the noise removal processing of each frame.

  In the forward projection calculation by the forward projection calculation unit 34, the back projection calculation by the back projection calculation unit 35, the image update calculation by the processing target image creation unit 36, and the noise removal processing by the filter processing unit 37, the number of repetitions reaches a predetermined number of repetitions Until, it is done repeatedly. The forward projection calculation unit 34 sets the initial image created by the initial image creation unit 33 as an input image at the first time of the iterative process. Thereafter, the forward projection calculation unit 34 uses the image after the noise removal processing by the filter processing unit 37 as an input image. The image after the image update calculation by the processing target image creation unit 36 when the number of repetitions reaches a predetermined number of repetitions, or the image after noise removal processing by the filter processing unit 37 is output as a tomographic image after noise removal processing Be done.

  FIG. 31 is a flow chart for explaining the image processing method of the third embodiment. The image processing method according to this embodiment includes a list data acquisition step S30 for acquiring list data collected by the radiation tomography apparatus 2, a sinogram creation step S31 performed by the sinogram creation unit 31, and a guidance performed by the guidance image creation unit 32. Image creation step S32, initial image creation step S33 performed by the initial image creation unit 33, forward projection calculation step S34 performed by the forward projection calculation unit 34, back projection calculation step S35 performed by the back projection calculation unit 35, and processing object The processing target image creation step S36 performed by the image creation unit 36, the filter processing step S37 performed by the filter processing unit 37, the repetition number update step S38 in which the value of the repetition number k is increased by 1, and the repetition number k is a predetermined repetition number And a determination step S39 of determining whether or not

In list data acquisition step S30, the list data acquired by the radiation tomography apparatus 2 is acquired. In the sinogram creation step S31, the list data is divided into a plurality of frames in the order of collection, and the data group included in each frame of the list data is used to create a sinogram of the frame. In the guidance image producing step S32, it creates a guidance image I _m for use in the noise removal processing by Gaidido filtering in the filter processing step S37. In the guidance image creation step S32, the processing object image P _m of the frame is handled using a data group having a larger number of data than the data group used in creating the sinogram of each frame in the sinogram creation step S31 among the list data. creating a reconstructed image as guidance image I _m for. In the initial image creation step S33, an initial image is created.

In the forward projection calculation step S34, forward projection calculation is performed on each of a plurality of frames based on the input image to create a sinogram. In the back projection calculation step S35, the back projection calculation is performed based on the comparison between the sinogram created in the forward projection calculation step S34 and the sinogram created in the sinogram creation step S31 for each of the plurality of frames. In the processing target image creation step S36, an image update calculation is performed on each of a plurality of frames based on the back projection calculation result in the back projection calculation step S35 to create an image after the image update calculation as the processing target image P _m. . In the filter processing step S37, the processing target image P _m of each frame created in the processing target image creating step S36 and the guidance image created corresponding to the processing target image P _m of the frame in the guidance image creating step S32 using I _m, it performs noise removal processing by Gaidido filter to create an image after the noise removal processing of the frame.

  In the repeat count updating step S38, the value of the repeat count k is increased by one. At the determination step S39, it is determined whether the number of repetitions k has reached a predetermined number of repetitions. The forward projection calculation step S34, the back projection calculation step S35, the processing target image creation step S36, the filter processing step S37, and the repetition number update step S38 until it is determined in the judgment step S39 that the repetition number k has reached a predetermined repetition number. Do it repeatedly.

  In the forward projection calculation step S34, the initial image created in the initial image creation step S33 is used as an input image in the first process (k = 0) of the iterative process. In the forward projection calculation step S34, in the following (k ≧ 1), the image after the noise removal processing in the filter processing step S37 is used as an input image.

  If it is determined in the determination step S39 that the number of repetitions k has reached the predetermined number of repetitions, the image after the image update calculation in the processing target image creation step S36 or the image after the noise removal processing in the filter processing step S37 is It is output as a tomographic image after noise removal processing.

List data, frame data group used in preparing a processing object image P _m, and the data group used in preparing a guidance image I _m, the relationship between the Figure 4 used in the first embodiment As described above. The processing target image P _m used for the noise removal processing by the guided filter is a reconstructed image in the first embodiment and a sinogram in the second embodiment, whereas it is an image reconstruction in the third embodiment. It differs in that it is an image after the image update calculation in the middle of the repetitive processing. In this embodiment, the guidance image generation unit 32 (the guidance image creating step S32), the guidance image I _m corresponding to the processing target image P _m of the m-th frame, the maximum pixel value of the guidance image I _m is the processing target image It is preferable that normalization is performed so as to be equal to the maximum pixel value of P _m (the above equation (10)).

Next, simulation results will be described. In this simulation, as described with reference to FIG. 5 in the first embodiment, a sinogram (FIG. 5 (b)) of each frame is created based on the numerical phantom (FIG. 5 (a)), and further each frame Poisson noise was given to the sinogram of to create a noise-added sinogram (FIG. 5 (c)). The noise-added sinogram (FIG. 5C) is a sinogram created by the sinogram creation unit 31 (sinogram creation step S31). The normalized guidance image I ′ _m corresponding to the processing target image P _m of each frame gives Poisson noise to a sinogram created using the entire list data, and performs image reconstruction based on this noise-added sinogram Created. In the image reconstruction processing by the ML-EM method, the noise removal processing by the filter processing unit 37 (filter processing step S37) is performed after the image update calculation by the processing target image generation unit 36 (processing target image generation step S36). The number of iterations in the ML-EM method was 40 times.

  FIG. 32 is a view showing an image obtained by the noise removal processing of the third embodiment. This figure shows a tomographic image of the 30th frame (period of 5400 to 6000 seconds). Compared with the image (FIG. 10) obtained by the noise removal process of the first embodiment, the image (FIG. 32) obtained by the noise removal process of the third embodiment is further noise-removed. It has become a thing.

  FIG. 33 is a graph showing time change of MSE of reconstructed images in Comparative Example 1, Comparative Example 2, First Embodiment, and Third Embodiment obtained in the simulation. FIG. 34 is a graph showing a time change of PSNR of reconstructed images in Comparative Example 1, Comparative Example 2, First Embodiment, and Third Embodiment obtained in the simulation. FIG. 35 is a graph showing temporal changes in SSIM of reconstructed images in Comparative Example 1, Comparative Example 2, and the first and third embodiments obtained in the simulation. As can be seen from these figures, any index of MSE, PSNR and SSIM has better performance in the case of the noise removal process of the third embodiment compared to the noise removal process of the comparative examples 1 and 2. Is shown. Further, compared to the noise removal process of the first embodiment, the performance in the case of the noise removal process of the third embodiment is superior.

Also in the third embodiment, the same effect as that of the first embodiment can be obtained. That is, in the noise removal processing by the guided filter, a data group having a larger number of data than the data group used in creating the processing target image P _m of each frame is used to correspond to the processing target image P _m of the frame Create a guidance image I _m (or a normalized guidance image I ' _m ). As a result, even when the period of each frame is not constant or the number of data of each frame is not constant, it is possible to create a noise-removed tomographic image of each frame with higher performance. Also, when noise removal processing by Gaidido filter, in particular by using a normalized guidance image I _'m, it is possible to further improve the image quality of tomographic images. In addition, when using the guidance image I _m (or the normalized guidance image I ' _m ) in consideration of the movement of the PET ligand in the subject during the noise removal processing by the guided filter, the movement PET that changes the movement rapidly The image quality of the image can be improved.

(Modification)
The present invention is not limited to the above embodiment, and various modifications are possible. Any two or more of the noise removal processes of the first to third embodiments may be combined. For example, noise removal processing may be performed in the sinogram as in the second embodiment, and then noise removal processing may be performed in the reconstructed image as in the first embodiment.

  Although the radiation tomography apparatus 2 is a PET apparatus in the above embodiment, it may be a SPECT apparatus.

  1A to 1C: radiation tomography system, 2: radiation tomography apparatus, 10: image processing device, 11: processing target image creation unit, 12: guidance image creation unit, 13: filter processing unit, 15: storage unit, 20: ... Image processing apparatus 21 processing target image creation unit 22 guidance image creation unit 23 filter processing unit 24 image reconstruction unit 25 storage unit 30 image processing apparatus 31 sinogram creation unit 32 ... Guidance image creation unit, 33 ... Initial image creation unit, 34 ... Forward projection calculation unit, 35 ... Back projection calculation unit, 36 ... Processing target image creation unit, 37 ... Filter processing unit, 38 ... Storage unit.

Claims (13)

An apparatus for creating a tomographic image after noise removal processing based on list data collected by a radiation tomography apparatus, comprising:
A processing target image creation unit that divides the list data into a plurality of frames in the order of collection, and creates a reconstructed image as a processing target image of the frame using a data group included in each frame among the list data;
As a guidance image corresponding to the processing target image of the frame using a data group having a larger number of data than the data group used when creating the processing target image of each frame in the processing target image creation unit among the list data A guidance image creation unit that creates a reconstructed image;
The noise is removed by the guided filter using the processing target image of each frame created by the processing target image creating unit and the guidance image created corresponding to the processing target image of the frame by the guidance image creating unit. A filter processing unit that performs processing to create a tomographic image after noise removal processing of the frame;
An image processing apparatus comprising: An apparatus for creating a tomographic image after noise removal processing based on list data collected by a radiation tomography apparatus, comprising:
A processing target image creation unit that divides the list data into a plurality of frames in the order of collection, and creates a sinogram as a processing target image of the frame using a data group included in each frame among the list data;
As a guidance image corresponding to the processing target image of the frame using a data group having a larger number of data than the data group used when creating the processing target image of each frame in the processing target image creation unit among the list data A guidance image creation unit that creates a sinogram,
The noise is removed by the guided filter using the processing target image of each frame created by the processing target image creating unit and the guidance image created corresponding to the processing target image of the frame by the guidance image creating unit. A filter processing unit that performs processing to create a sinogram after noise removal processing of the frame;
An image reconstructor configured to reconstruct a tomographic image of the frame using a sinogram of each frame generated by the filter processor;
An image processing apparatus comprising: An apparatus for creating a tomographic image after noise removal processing based on list data collected by a radiation tomography apparatus, comprising:
A sinogram creation unit that divides the list data into a plurality of frames in the order of collection, and creates a sinogram of the frames using a data group included in each frame among the list data;
An initial image creation unit that creates an initial image;
A forward projection calculation unit that performs forward projection calculation based on an input image for each of the plurality of frames to create a sinogram;
A back projection calculation unit that performs back projection calculation based on a comparison between a sinogram created by the forward projection calculation unit and a sinogram created by the sinogram creation unit for each of the plurality of frames;
A processing target image creation unit that performs image update calculation for each of the plurality of frames based on the back projection calculation result by the back projection calculation unit, and creates an image after the image update calculation as a processing target image;
Among the list data, using a data group having a larger number of data than the data group used when creating the sinogram of each frame in the sinogram creation unit, the reconstructed image is used as a guidance image corresponding to the processing target image of the frame. Guidance image creation unit to create,
The noise is removed by the guided filter using the processing target image of each frame created by the processing target image creating unit and the guidance image created corresponding to the processing target image of the frame by the guidance image creating unit. A filter processing unit that performs processing to create an image after noise removal processing of the frame;
Equipped with
The forward projection calculation by the forward projection calculation unit, the back projection calculation by the back projection calculation unit, the image update calculation by the processing target image generation unit, and the noise removal processing by the filter processing unit are repeatedly performed.
The forward projection calculation unit sets the initial image as the input image at the first time of the repetitive processing, and thereafter sets the image after noise removal processing by the filter processing unit as the input image.
Image processing device. The guidance image creation unit creates a guidance image corresponding to the processing target image of each frame by performing normalization such that the maximum pixel value of the guidance image is equal to the maximum pixel value of the processing target image.
The image processing apparatus according to any one of claims 1 to 3. When the guidance image creating unit creates a guidance image corresponding to the processing target image of each frame, a data group including a data group used when creating the processing target image of the frame in the processing target image creating unit Using
The image processing apparatus according to any one of claims 1 to 4. When the guidance image creating unit creates a guidance image corresponding to the processing target image of each frame, data before and after the data group used when creating the processing target image of the frame in the processing target image creating unit Use groups
The image processing apparatus according to any one of claims 1 to 5. A radiation tomography apparatus for collecting list data for reconstructing a tomographic image of a subject;
The image processing apparatus according to any one of claims 1 to 6, wherein a tomographic image after noise removal processing is created based on the list data collected by the radiation tomography apparatus.
Radiation tomography system comprising: A method of creating a tomographic image after noise removal processing based on list data collected by a radiation tomography apparatus, comprising:
A processing target image creation step of dividing the list data into a plurality of frames in collection order, and creating a reconstructed image as a processing target image of the frame using a data group included in each frame among the list data;
As a guidance image corresponding to the processing target image of the frame using a data group having a larger number of data than the data group used when creating the processing target image of each frame in the processing target image creation step among the list data A guidance image creation step of creating a reconstructed image;
The noise is removed by the guided filter using the processing target image of each frame created in the processing target image creating step and the guidance image created corresponding to the processing target image of the frame in the guidance image creating step. A filter processing step of processing to create a tomographic image of the frame after noise removal processing;
An image processing method comprising: A method of creating a tomographic image after noise removal processing based on list data collected by a radiation tomography apparatus, comprising:
A processing target image creation step of dividing the list data into a plurality of frames in the order of collection, and creating a sinogram as a processing target image of the frame using a data group included in each frame among the list data;
As a guidance image corresponding to the processing target image of the frame using a data group having a larger number of data than the data group used when creating the processing target image of each frame in the processing target image creation step among the list data A guidance image creation step of creating a sinogram;
The noise is removed by the guided filter using the processing target image of each frame created in the processing target image creating step and the guidance image created corresponding to the processing target image of the frame in the guidance image creating step. A filtering step of processing to create a sinogram after noise removal processing of the frame;
An image reconstruction step of reconstructing a tomographic image of the frame using a sinogram of each frame created in the filtering step;
An image processing method comprising: A method of creating a tomographic image after noise removal processing based on list data collected by a radiation tomography apparatus, comprising:
A sinogram creation step of dividing the list data into a plurality of frames in the order of collection, and creating a sinogram of the frame using a data group included in each frame among the list data;
An initial image creation step of creating an initial image;
A forward projection calculation step of performing forward projection calculation based on an input image for each of the plurality of frames to create a sinogram;
Backprojection calculation step of performing backprojection calculation based on a comparison between the sinogram created in the forward projection calculation step and the sinogram created in the sinogram creation step for each of the plurality of frames;
A processing target image creation step of performing image update calculation for each of the plurality of frames based on the back projection calculation result in the back projection calculation step, and creating an image after the image update calculation as a processing target image;
Among the list data, using a data group having a larger number of data than the data group used when creating the sinogram of each frame in the sinogram creation step, the reconstructed image is used as a guidance image corresponding to the processing target image of the frame. A guidance image creation step to be created;
The noise is removed by the guided filter using the processing target image of each frame created in the processing target image creating step and the guidance image created corresponding to the processing target image of the frame in the guidance image creating step. A filtering step of processing to create an image after noise removal processing of the frame;
Equipped with
Repeatedly performing the forward projection calculation step, the back projection calculation step, the processing target image creation step, and the filtering step;
In the forward projection calculation step, the initial image is set as the input image at the first time of the iterative processing, and thereafter, an image after noise removal processing in the filter processing step is set as the input image.
Image processing method. The guidance image creation step creates a guidance image corresponding to the processing target image of each frame by performing normalization such that the maximum pixel value of the guidance image is equal to the maximum pixel value of the processing target image.
The image processing method according to any one of claims 8 to 10. When the guidance image creation step creates a guidance image corresponding to the processing target image of each frame, a data group including a data group used when creating the processing target image of the frame in the processing target image creation step Using
The image processing method according to any one of claims 8 to 11. When the guidance image creation step creates a guidance image corresponding to the processing target image of each frame, data before and after the data group used when creating the processing target image of the frame in the processing target image creation step Use groups
The image processing method according to any one of claims 8 to 12.