阅读说明：本技术 散射估计方法和图像处理装置 (Scattering estimation method and image processing device ) 是由 山川善之 于 2017-07-25 设计创作，主要内容包括：该散射估计方法包括以下步骤：基于放射性图像(5)的散射线指标值(R)来决定用于对单次散射分布进行平滑化的卷积核的步骤(S4)；以及使将卷积核应用于单次散射分布来进行平滑化后的散射分布拟合到正电子放射断层摄影测量数据的步骤(S5)。(The scatter estimation method comprises the following steps: a step (S4) for determining a convolution kernel for smoothing the single scatter distribution on the basis of the scattered ray index value (R) of the radioactive image (5); and fitting the smoothed scatter distribution in which the convolution kernel is applied to the single scatter distribution to the positron emission tomography measurement data (S5).)

1. A method of scatter estimation comprising the steps of:

acquiring positron emission tomography measurement data and absorption coefficient data;

generating a radiological image from the positron emission tomography measurement data and the absorption coefficient data;

estimating a single-shot scatter distribution of radiation in the radiological image from the radiological image and the absorption coefficient data;

a convolution kernel determining step of determining a convolution kernel for smoothing the single scatter distribution based on a scattered ray index value of the radiological image; and

fitting the smoothed scatter distribution by applying the convolution kernel to the single scatter distribution to the positron emission tomography measurement data.

2. The scatter estimation method of claim 1,

the index value of the scattered radiation is a scattering fraction,

in the convolution kernel determining step, the convolution kernel is determined based on a comparison result of respective scatter fractions of the radiological image smoothed by using a plurality of parameters characterizing the convolution kernel.

3. The scatter estimation method of claim 2,

in the convolution kernel determining step, the convolution kernel is determined based on a magnitude of a change in the scattering fraction that accompanies a change in the plurality of parameters.

4. The scatter estimation method of claim 1,

in the step of estimating the single scatter distribution from the radiological image and the absorption coefficient data, the single scatter distribution is estimated by a single scatter simulation method.

5. The scatter estimation method of claim 1,

in the step of estimating the single scatter distribution from the radiological image and the absorption coefficient data, the single scatter distribution is estimated by a monte carlo simulation method.

6. The scatter estimation method of claim 2,

the convolution kernel specifies a weighted distribution determined by the parameter relating to chromatic dispersion, and the smoothing of the single scattering distribution is performed by a weighted average filter using the convolution kernel.

7. An image processing apparatus includes:

a control unit; and

an image data acquisition section that acquires positron emission tomography measurement data and absorption coefficient data,

wherein the control unit is configured to:

generating a radiological image from the positron emission tomography measurement data and the absorption coefficient data,

estimating a single-scatter distribution from the radiological image and the absorption coefficient data,

deciding a convolution kernel for smoothing the single scatter distribution based on index values of scattered rays of the radiological image,

fitting the smoothed scatter distribution by applying the convolution kernel to the single scatter distribution to the positron emission tomography measurement data.

Technical Field

The present invention relates to a scatter estimation method and an image processing apparatus, and more particularly, to a scatter estimation method and an image processing apparatus in measurement data of a positron emission tomography apparatus.

Background

Conventionally, a scatter estimation method and an image processing apparatus in measurement data of a positron emission tomography apparatus are known. Such a scatter estimation method is disclosed in, for example, patent document 1 and non-patent document 1.

Generally, in a positron emission tomography apparatus, radiation is scattered in a subject at the time of imaging, and an image obtained thereby sometimes includes noise. Therefore, as in patent document 1 and non-patent document 1, scattering estimation is performed to perform scattering correction on the measurement data. As the scatter estimation, estimation of a single scatter distribution assuming that the radiation is scattered only once in the subject is performed. The scattering of radiation also includes multiple scattering in which radiation is scattered multiple times within the subject.

Non-patent document 1 discloses, as a scatter estimation method, a multiple scatter distribution estimation method that models convolution of a single scatter distribution. That is, in the above-mentioned non-patent document 1, the single scattering distribution is convolved (convolution operation) to Simulate (Simulate) a multiple scattering distribution, thereby performing the scattering correction.

Patent document 1 discloses the following method: the single scattering distribution and the multiple scattering distribution are fitted by the least square method, and the multiple scattering distribution is estimated using a convolution kernel in which the amplitude and the width are varied. That is, in patent document 1, the scattering correction is performed by directly estimating the multiple scattering distribution.

Disclosure of Invention

Problems to be solved by the invention

However, the parameters of the convolution function disclosed in the above-mentioned non-patent document 1 are parameters determined empirically. Therefore, when the parameter is not appropriate, an error occurs between the estimation result and the actual measurement value. When an error is generated between the estimation result and the measured value, problems such as the following occur: the influence of scattered rays cannot be sufficiently removed to generate noise in the resulting image, or the scatter correction is excessively performed to generate a blank portion in the resulting image.

In the scatter estimation method disclosed in patent document 1, the parameters of the convolution kernel are determined by fitting the convolution kernel to the multiple scatter distribution. Since the multiple scattering distribution is estimated by the method of patent document 1 described above, there are the following problems: it takes a calculation cost, and the time for performing the scattering estimation becomes long compared with the case of simulating the multiple scattering distribution from the single scattering distribution.

The present invention has been made to solve the above-described problems, and an object of the present invention is to provide a scatter estimation method and an image processing apparatus capable of reducing the time taken for scatter estimation and suppressing degradation of image quality due to scatter correction.

Means for solving the problems

As a result of intensive studies to achieve the above object, the present inventors have obtained the following findings: when the convolution kernel used for the scatter estimation is determined, if the convolution kernel is changed, the scattered ray index value changes, and the image quality of the obtained image changes. Based on this insight, the scatter estimation method of the first aspect of the invention comprises the steps of: acquiring positron emission tomography measurement data and absorption coefficient data; generating a radiological image from the positron emission tomography measurement data and the absorption coefficient data; estimating a single-shot scatter distribution of radiation in the radiological image from the radiological image and the absorption coefficient data; a convolution kernel determination step of determining a convolution kernel for smoothing the single scatter distribution based on a scattered ray index value of the radioactive image; and fitting the smoothed scatter distribution by applying a convolution kernel to the single scatter distribution to the positron emission tomography measurement data. In the present specification, the term "convolution kernel" refers to a filter function for performing convolution processing, and is synonymous with the term "convolution filter". In addition, the "radiological image" is an image showing a radiation source inside the subject. In addition, "absorption coefficient data" is image data indicating a structure within the subject. Thus, the single scatter distribution of the radiation can be estimated from the radiological image and the absorption coefficient data.

In the scatter estimation method of the first aspect of the present invention, as described above, the method includes the steps of: acquiring positron emission tomography measurement data and absorption coefficient data; generating a radiological image; estimating a single-shot scatter distribution of radiation in the radiological image; deciding a convolution kernel for smoothing the single scatter distribution based on the index value of the scattered rays of the radiological image; and fitting the smoothed scatter distribution by applying a convolution kernel to the single scatter distribution to the positron emission tomography measurement data. Thus, the multiple scattering distribution can be simulated using the single scattering distribution without directly estimating the multiple scattering distribution. Further, without using parameters determined empirically, the convolution kernel used for smoothing the single scatter distribution can be determined from the scattered ray index value of the radioactive image. Therefore, a multiple scatter distribution can be simulated from a single scatter distribution using an appropriate convolution kernel in a radiological image, and thus it is possible to suppress an error from occurring between an estimation result and an actual measurement value. As a result, the time taken for the scatter estimation can be shortened, and the degradation of the image quality due to the scatter correction can be suppressed.

In the above-described scatter estimation method according to the first aspect, it is preferable that the scattered ray index value is a scatter fraction, and the convolution kernel determining step determines the convolution kernel based on a result of comparison of respective scatter fractions of the radiological image smoothed by using a plurality of parameters characterizing the convolution kernel. Here, the inventors of the present application have obtained the following findings: the change in the image when the parameters of the convolution kernel are made to change is based on the change in the scatter fraction. In addition, the inventors of the present application have obtained the following findings: by comparing the scatter fractions, appropriate parameters of the convolution kernel can be easily determined. Therefore, according to the above-described configuration, by comparing the respective scatter fractions of the radiological image in which the parameters of the convolution kernel are varied and smoothed, the convolution kernel can be easily determined.

In this case, it is preferable that, in the convolution kernel determining step, the convolution kernel is determined based on the magnitude of a change in the scattering fraction caused by a change in the plurality of parameters. Here, the inventors of the present application have obtained the following findings: when the variation in the scattering fraction accompanying the variation in the parameters of the convolution kernel is large, noise in the resulting image is reduced. Therefore, with the above-described configuration, the convolution kernel can be determined more easily.

In the scatter estimation method of the first aspect described above, it is preferable that, in the step of estimating the single scatter distribution from the radiological image and the absorption coefficient data, the single scatter distribution is estimated by a single scatter simulation method. According to such a configuration, the single scattering distribution of radiation can be easily estimated from the radiological image and the absorption coefficient data. Further, the single scattering simulation method refers to a method of estimating a scattering distribution in a subject from a radiological image and absorption coefficient data.

In the scatter estimation method according to the first aspect described above, it is preferable that, in the step of estimating the single scatter distribution from the radiological image and the absorption coefficient data, the single scatter distribution is estimated by a monte carlo simulation method. Even with such a configuration, the single scattering distribution of radiation can be easily estimated, as in the case of estimating the single scattering distribution using the single scattering simulation method. The monte carlo simulation is a method of obtaining an approximate solution by repeating a simulation using a random number.

In the above-described scatter estimation method according to the first aspect, it is preferable that the convolution kernel defines a weighted distribution determined by a parameter relating to chromatic dispersion, and the single scatter distribution is smoothed by a weighted average filter using the convolution kernel. With this configuration, the multiple scattering distribution of radiation can be simulated from the single scattering distribution of radiation using a convolution kernel in which the weighting distribution is changed by a parameter related to dispersion. As a result, the processing time can be shortened as compared with the case of directly estimating the multiple scattering distribution. In addition, since the parameters of the convolution kernel are determined using actually measured data, it is possible to simulate the multiple scattering distribution using an appropriate convolution kernel, compared to a method in which the parameters are determined empirically. As a result, the accuracy of the scatter correction can be improved.

An image processing apparatus according to a second aspect of the present invention includes a control unit and an image data acquisition unit, the image data acquisition unit acquiring positron emission tomography measurement data and absorption coefficient data, the control unit being configured to: a radioactive image is generated from positron emission tomography measurement data and absorption coefficient data, a single scatter distribution is estimated from the radioactive image and the absorption coefficient data, a convolution kernel for smoothing the single scatter distribution is determined based on a scattered ray index value of the radioactive image, and a scatter distribution smoothed by applying the convolution kernel to the single scatter distribution is fitted to the positron emission tomography measurement data.

As described above, the image processing apparatus according to the second aspect of the present invention includes the control unit and the image data acquisition unit, and the control unit is configured to: a radioactive image is generated from positron emission tomography measurement data and absorption coefficient data, a single scatter distribution is estimated from the radioactive image and the absorption coefficient data, a convolution kernel for smoothing the single scatter distribution is determined based on a scattered ray index value of the radioactive image, and a scatter distribution smoothed by applying the convolution kernel to the single scatter distribution is fitted to the positron emission tomography measurement data. Thus, the multiple scattering distribution can be simulated using the single scattering distribution without directly estimating the multiple scattering distribution. Further, without using parameters determined empirically, the convolution kernel used for smoothing the single scatter distribution can be determined from the scattered ray index value of the radioactive image. Therefore, a multiple scatter distribution can be simulated from a single scatter distribution using an appropriate convolution kernel in a radiological image, and thus it is possible to suppress an error from occurring between an estimation result and an actual measurement value. As a result, the time taken for the scatter estimation can be shortened, and the degradation of the image quality due to the scatter correction can be suppressed.

ADVANTAGEOUS EFFECTS OF INVENTION

According to the present invention, as described above, it is possible to provide a scatter estimation method and an image processing apparatus capable of shortening the time taken for scatter estimation and suppressing degradation of image quality due to scatter correction.

Drawings

Fig. 1 is a block diagram showing an overall configuration of a positron emission tomography system including an image processing apparatus of an embodiment of the present invention.

Fig. 2 is a block diagram showing the overall configuration of an image processing apparatus according to an embodiment of the present invention.

Fig. 3 is a block diagram of a positron emission tomography apparatus according to an embodiment of the present invention.

Fig. 4 is a schematic diagram for explaining the distribution of radiation detected by the detector according to the embodiment of the present invention.

Fig. 5 is a flowchart for explaining a scatter estimation method according to an embodiment of the present invention.

Fig. 6 is a diagram (a) to (C) for explaining a method of determining a convolution kernel according to an embodiment of the present invention.

Fig. 7 is a diagram (D) and a diagram (E) for explaining a method of determining a convolution kernel according to an embodiment of the present invention.

Fig. 8 is a diagram (a) showing an image of an abdomen, a diagram (B) showing a scattering score of an abdomen image, and a diagram (C) showing a difference in scattering scores of abdomen images for explaining the scattering estimation method of the first example of one embodiment of the present invention.

Fig. 9 is a diagram (a) showing an image of a pelvic region, a diagram (B) showing a scattering score of the pelvic region image, and a diagram (C) showing a difference in scattering scores of the pelvic region image, for explaining the scattering estimation method according to the second example of the embodiment of the present invention.

Fig. 10 is a flowchart for explaining a scatter estimation method according to a first modification of the embodiment of the present invention.

Fig. 11 is a flowchart for explaining a scatter estimation method according to a second modification of the embodiment of the present invention.

Detailed Description

Hereinafter, embodiments embodying the present invention will be described based on the drawings.

(configuration of image processing apparatus)

First, the configuration of a positron emission tomography system 100 including an image processing apparatus 2 according to the present embodiment will be described with reference to fig. 1 to 3.

As shown in fig. 1, the positron emission tomography system 100 includes a positron emission tomography apparatus 1, an image processing apparatus 2, and a display unit 3.

The positron emission tomography apparatus 1 is provided with an imaging unit 10. The camera unit 10 includes a housing 11.

The positron emission tomography apparatus 1 includes a plurality of detectors 12. Specifically, the plurality of detectors 12 are disposed inside the housing 10. In addition, the plurality of detectors 12 are arranged so as to surround the object O. Each detector 12 detects radiation emitted from the object O. Each detector 12 is configured to: the detected radiation is converted into an electric signal, and the electric signal is transmitted to the control unit 20. In the present embodiment, a radioactive marker reagent is administered to the subject O in advance, and each detector 12 detects radiation emitted from the subject O. Examples of the radioactive labeling agent include 18F-FDG (fluorodeoxyglucose).

Fig. 2 is a block diagram showing the overall configuration of the image processing apparatus 2. As shown in fig. 2, the image processing apparatus 2 includes a control unit 20, an image data acquisition unit 21, and a storage unit 22, and the image data acquisition unit 21 acquires positron emission tomography measurement data and absorption coefficient data.

The control unit 20 is configured to generate a radiological image 5 (see fig. 8) from the positron emission tomography measurement data and the absorption coefficient data acquired by the image data acquisition unit 2. The control unit 20 is configured to estimate the single scattering distribution from the generated radiological image 5 and the absorption coefficient data. The control unit 20 is configured to simulate a multiple scattering distribution based on the estimated single scattering distribution. The control Unit 20 includes a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), and the like. Further, a configuration in which the control section 20 performs the scattering estimation will be described later.

The image data acquisition unit 21 is configured to acquire positron emission tomography measurement data acquired by the positron emission tomography apparatus 1 and absorption coefficient data converted from morphological data of the object O acquired by a tomography apparatus (CT apparatus) or the like. The absorption coefficient data may be data converted from data acquired by a magnetic resonance imaging apparatus (MRI apparatus).

The storage unit 22 is configured to store a program or the like used when the control unit 20 estimates the scattering of radiation. The storage section 22 includes an HDD (hard disk drive) or the like.

The display unit 3 is configured to display the radiological image 5 generated by the control unit 20. The display unit 3 includes a liquid crystal monitor and the like.

Fig. 3 is a cross-sectional view taken along the 200-and-200-line of the housing portion 11 of the positron emission tomography apparatus 1 in fig. 1, the detectors 12 are arranged in a circumferential shape inside the housing portion 11 as shown in fig. 3, a pair of radiation emitted in directions opposite to each other from the radiation source of the object O is detected by the pair of detectors 12 arranged oppositely with the radiation source interposed therebetween, the center of a virtual circle on which the detectors 12 are arranged is shown as the center 4 in fig. 3 for convenience of explanation, the radius of the virtual circle is shown as d in fig. 3, and the angle between the center 4 and the detectors 12 is shown as α in fig. 3.

Fig. 4 is a graph PD, a graph SD, and a graph T showing the count number of radiation (intensity of radiation) detected by the detector 12. The horizontal axis of the graph of fig. 4 represents the distance from the radiation source, and the vertical axis represents the number of radiation counts. The graph PD is a graph showing a relationship between the count number of radiation detected by the detector 12 and the distance from the radiation source. The graph SD is a graph showing the relationship between the count number of radiation from scattered rays and the distance from the radiation source among the radiation detected by the detector 12. The graph T is a graph in which the graph SD is removed from the graph PD. That is, the graph T is a graph showing the relationship between the count number of radiation from which scattered rays are removed and the distance from the radiation source, among the radiation detected by the detector 12.

When the radiation passes through the inside of the object O, the radiation is scattered by the tissue or the like inside the object O, and therefore, the distribution of the radiation actually detected by the detector 12 is as shown in the graph PD obtained by combining the graph T and the graph SD. Therefore, in order to obtain radiation emitted from the object O, it is necessary to perform scatter correction for removing scattered radiation as shown in the graph SD from the signal detected by the detector 12.

The scattered radiation includes single scattered radiation obtained by primary scattering in tissue or the like inside the object O, and multiple scattered radiation obtained by multiple scattering inside the object O. In the present embodiment, the control unit 20 estimates the single scattering distribution from the radiological image 5 and the absorption coefficient data by a single scattering simulation method. Then, the control unit 20 is configured to perform the scatter correction by estimating a scatter distribution (graph SD) in consideration of the multiple scatter distribution from the estimated single scatter distribution. The single scattering simulation method is a method of estimating the scattering distribution (graph SD) inside the object O from the radiological image 5 and the absorption coefficient data.

(method of performing Scattering estimation)

Next, a scattering estimation method according to the present embodiment will be described with reference to fig. 5 to 9.

First, a flow of the scattering estimation method in the present embodiment will be described with reference to fig. 5.

In step S1, the image data acquisition section 21 acquires positron emission tomography measurement data and absorption coefficient data. The acquired data form is a list data form, a sinogram form, or the like, but may be any form of data. The data used may be either two-dimensional data or three-dimensional data.

Next, in step S2, the control unit 20 generates a radiological image 5 from the positron emission tomography measurement data and the absorption coefficient data. Any method may be used to generate the radiological image 5, and for example, the control unit 20 is configured to generate the radiological image 5 using a reconstruction algorithm (FBP), an iterative reconstruction algorithm (OSEM, MLEM), or the like. The control unit 20 is configured to apply at least sensitivity correction, coincidence counting correction, and absorption correction when generating the radiological image 5. The occasional coincidence count correction is performed using a known technique such as a delayed coincidence counting method, an estimation method based on a single count, or the like. The control unit 20 is configured to perform noise reduction processing when the radiological image 5 is generated. As the noise reduction processing, for example, a gaussian Filter, Non-Local Means Filter (Non-Local mean Filter) may be applied. In addition, the noise reduction processing may be performed at any timing among preprocessing applied to the projection data, processing during reconstruction, and post-processing after reconstruction. Next, the process proceeds to step S3.

In step S3, the control unit 20 estimates a single scatter distribution from the generated radiological image 5 and the acquired absorption coefficient data. Next, in step S4, a convolution kernel for smoothing the single scatter distribution is determined based on the scattered ray index value SF of the radiological image 5. Further, the scattered ray index value SF of the radiological image 5 refers to a ratio of radiation from scattered rays, specifically, a scattering fraction RF, among all radiation detected by the detector 12.

Next, in step S5, the smoothed scatter distribution (graph SD) in which the convolution kernel is applied to the single scatter distribution is scaled to fit the scatter distribution to the lower bottom portion of the positron emission tomography measurement data. In addition, the scaling method comprises the following steps: threshold processing is performed on an absorption coefficient sinogram generated from absorption coefficient data (absorption coefficient image), and regions inside and outside the object are determined. Then, a region of interest (ROI) is set in the region outside the object, and the number of counts in the ROI is determined so as to match the measurement data. The scaling is performed for each slice, but may not be performed for each slice.

Next, in step S6, the fitted scatter distribution (graph SD) is removed from the positron emission tomography data, and the radiological image 5 is reconstructed.

(method of determining convolution kernel)

Next, a method of determining the convolution kernel by the control unit 20 will be described with reference to fig. 6 and 7.

As shown in fig. 6 (a), in the present embodiment, the control unit 20 generates the radiological image 5 from positron emission tomography data and absorption coefficient data. The control unit 20 is configured to estimate the single scattering of the radiological image 5 from the generated radiological image 5 by a single scattering simulation method.

Specifically, the control unit 20 is configured to perform the scattering correction using the scattering distribution (graph SD) obtained by modeling by the following equation (1).

[ numerical formula 1]

S(r，θ，z)＝∫S_s(r-u，θ，z)k(r)du…(1)

Here, S is a scattering distribution (graph SD) in consideration of multiple scattering. In addition, S_sIs a single scattering distribution. In addition, r is a number indicating the order of the detectors 12 in the radial direction. In addition, θ is a number indicating the order in the angular direction in which the detectors 12 are arranged. In addition, z is a number indicating the order in the slice direction. That is, z is a number indicating the order of stepping in the body axis direction of the object O. In addition, k is a convolution kernel.

In the present embodiment, the convolution kernel is used to define a weight distribution determined by a dispersion-related parameter P (see fig. 7 (E)). The single scattering distribution is smoothed by a weighted average filter using a convolution kernel. That is, in the present embodiment, the control unit 20 is configured to: the single scattering distribution is smoothed by a weighted average filter, and a scattering distribution (graph SD) in which the multiple scattering distribution is considered is obtained as a convolution kernel. The weighted average filter is, for example, a gaussian function (gaussian filter), and is shown by the following equation (2).

[ numerical formula 2]

Here, σ is a standard deviation of the gaussian function.

In the present embodiment, the control unit 20 determines the convolution kernel based on the comparison result of the respective scatter fractions RF of the radiological image 5 smoothed by using the plurality of parameters P representing the convolution kernel. Specifically, the control unit 20 determines the convolution kernel based on the magnitude of the change in the scattering fraction RF caused by the change in the plurality of parameters P. In the present embodiment, the parameter P of the table convolution kernel is σ in the above expression (2). The parameter P is a standard deviation (σ) of the graph SD shown in fig. 7 (E).

The control unit 20 is configured to reconstruct the plurality of radiological images 5 by applying a convolution kernel in which the parameter P is changed to the single scatter distribution. Specifically, as shown in fig. 6 (B), the control unit 20 uses the parameter P to which it is applied₁～P₅To reconstruct a plurality of radiological images 5. In the present embodiment, the parameter P₁～P₅The setting is increased at intervals of a predetermined value. In addition, it is known that the multiple scattering distribution is distributed in a wider range than the single scattering distribution, and therefore the parameter P₁It is preferable to set the value to be larger than the parameter P of the convolution kernel in the case where smoothing is not performed.

The control unit 20 is configured to obtain the scattering fraction RF from the adjacent radiological image 5. The scattering fraction RF is a value indicating the ratio of radiation from scattered radiation among all the radiation detected by the detector 12, and can be obtained by the following equation (3).

[ numerical formula 3]

Here, i is a number indicating the order of the smoothing parameters P. In addition, S_iThe estimated scatter coincidences are counted. In addition, P_cRD is random count data for measured instantaneous coincidence count data.

FIG. 6 (C) is a graph plotting the parameter P as P₁～P₅A graph 6 of the scatter fraction of the reconstructed radiological image 5 is performed. In the graph 6, the horizontal axis represents the value of the parameter (bin value) and the vertical axis represents the scattering fraction. As shown in fig. 6 (C), the control unit 20 is configured to generate and plot each scattering fraction RF₁～RF₅OfFig. 6. Scattering fractional RF in (C) of FIG. 6₁～RF₅Respectively by a parameter P₁～P₅The values of the scatter fraction RF of the reconstructed radiological image 5 are performed. As shown in fig. 6 (C), it is understood that the scattering fraction RF decreases as the parameter P increases. Furthermore, the bin length is about 8 mm.

The control unit 20 is configured to create a graph 7 shown in fig. 7 (D) by taking the difference Q between the respective scattering fractions RF. The difference Q between the scattering fractions RF is a value obtained by taking the difference between the scattering fraction RF having a small number (i in the above expression (1)) indicating the order of the parameters P and the scattering fraction RF having a large number among the adjacent scattering fractions RF, and can be obtained by the following expression (4).

[ numerical formula 4]

dSF_i＝SF_i-1-SF_i…(4)

Here, dSF_iIs the difference Q of the scattering fractions RF.

As shown in graph 7 of (D) of fig. 7, the difference Q of the scattering fractions RF is a graph having an upwardly convex shape. The control unit 20 is configured to set a parameter P when the difference Q of the scattering fractions RF becomes maximum (maximum) as a parameter P (σ) applied to the convolution kernel. Further, if the number of parameters P is too large, the difference in the difference Q of the respective scattering fractions RF becomes small, and there is a possibility that it is difficult to select an appropriate parameter P due to noise or the like. In addition, if the number of parameters P is too small, a range in which the difference Q of the scattering fraction RF is maximum may not be actually included, and an appropriate parameter P may not be selected. Therefore, the number of parameters P is slightly changed depending on the size of the object O, and is preferably set to approximately five or so.

As shown in fig. 7 (E), the control unit 20 is configured to: the smoothed scatter distribution (graph SD) is smoothed by applying a convolution kernel to the single scatter distribution and is fitted to the lower bottom portion of the positron emission tomography measurement data. Specifically, the control unit 20 performs threshold processing on the measurement data to determine the regions inside and outside the object. Thereafter, the control unit 20 is configured to: a region of interest (ROI) is set in the region outside the object, the scattering distribution (graph SD) is scaled in such a manner that the number of counts in the ROI coincides with the measurement data, and fitted to the lower bottom portion of the positron emission tomography measurement data.

Further, the control unit 20 is configured to: the fitted scatter distribution (graph SD) is removed from the positron emission tomography data, and the radiological image 5 is reconstructed.

Next, an example of finding a method for determining the optimum parameter P for the convolution kernel in the present embodiment will be described.

[ first embodiment ]

An experiment for determining the parameter P characterizing the convolution kernel of the first embodiment is described with reference to fig. 8.

In the first embodiment, the parameter P is changed for the radiological image 5 taken of the abdomen to perform the scatter correction, the plurality of radiological images 5 are reconstructed, and the plurality of radiological images 5 are compared to determine the optimum convolution kernel.

Fig. 8 (a) is a diagram showing a radiological image 5 obtained by changing the parameters of the convolution kernel at the time of imaging the abdomen to perform scatter correction and performing reconstruction. The image shown in the line (D) in fig. 8 (a) is an axial (body axis cross section) image of the abdomen, and the image shown in the line (E) is a coronal (coronal cross section) image of the abdomen. Fig. 8 (B) is a graph 6 showing a change in the scattering fraction in the radiological image 5 when the parameter of the convolution kernel is changed. In fig. 8 (B), the horizontal axis represents parameters of the convolution kernel, and the vertical axis represents the scattering fraction. Fig. 8 (C) is a graph 7 showing the difference in the scattering fraction when the parameter P of the convolution kernel is changed. In fig. 8 (C), the horizontal axis represents parameters of the convolution kernel, and the vertical axis represents the difference in the scattering fraction.

In the first embodiment, as shown in fig. 8 (a), the radiological image 5 is reconstructed with the parameters P set to 5.0bin, 7.5bin, 10bin, 15bin, and 20bin without smoothing. Note that the non-smoothing means that the scattering correction is performed with the parameter P set to 1.5 bin.

In the example shown in fig. 8 (a), the radiation image 5 reconstructed by setting the parameters P to 5.0bin and 7.5bin without smoothing has excessive scatter correction, and the contour of the image is blurred or a blank portion can be seen in the image. In addition, the radiation image 5 reconstructed by setting the parameters P to 15bin and 20bin has insufficient scatter correction, and noise is generated in the image.

When the scattering fraction RF is obtained from each radiological image 5, a relationship as shown in fig. 8 (B) can be obtained. Further, a graph 7 as shown in fig. 8 (C) is prepared by taking the difference Q between the respective scattering fractions RF.

When each radiological image 5 is compared based on the graph 7 shown in fig. 8 (C) and the radiological image 5 shown in fig. 8 (a), it is understood that the image quality of the radiological image 5 is improved as the difference Q of the scattering fractions RF is increased, and the image quality of the image when the difference Q of the scattering fractions RF is maximized is optimal.

Further, in the first embodiment, as shown in fig. 8, when the parameter P is set to 10bin, the outline of the image is clearly displayed and no blank portion is visible, and in addition, a result of reducing noise can be obtained. That is, in the first embodiment, the result is obtained that the image quality is the best in the case where the parameter P is set to 10 bin.

[ second embodiment ]

Next, an experiment for deciding the parameter P characterizing the convolution kernel of the second embodiment will be described with reference to fig. 9.

In the second embodiment, the parameter P that best characterizes the convolution kernel is obtained from the radiological image 50 of the pelvic region by the same method as in the first embodiment.

Fig. 9 (a) is a view showing a radiological image 50 obtained by performing scatter correction by changing the parameter P of the convolution kernel at the time of imaging the pelvic region and performing reconstruction. The image shown in the line (D) in fig. 9 (a) is an axial (body axis cross-sectional) image of the pelvis portion, and the image shown in the line (E) is a coronal (coronal cross-sectional) image of the pelvis portion. Fig. 9 (B) is a graph 60 showing a change in the scattering fraction in the radiological image 50 when the parameter of the convolution kernel is changed. In fig. 9 (B), the horizontal axis represents parameters of the convolution kernel, and the vertical axis represents the scattering fraction. Fig. 9 (C) is a graph 70 showing the difference in the scattering fraction when the parameter of the convolution kernel is changed. In fig. 9 (C), the horizontal axis represents parameters of the convolution kernel, and the vertical axis represents the difference in the scattering fraction.

In the second embodiment, as shown in fig. 9 (a), the radiological image 50 is reconstructed by setting the parameters P to non-smoothing (1.5bin), 5.0bin, 7.5bin, 10bin, 15bin, and 20 bin.

In the second embodiment, as shown in fig. 9, the scatter correction is excessive with respect to the radiological image 50 reconstructed by setting the parameters P to non-smoothing (1.5bin), 5.0bin, 7.5bin, and 10bin, and a blank portion can be seen in the image. In addition, the radiation image 50 reconstructed by setting the parameter P to 20bin has insufficient scatter correction, and noise is generated in the image. In the example shown in fig. 9, when the parameter P is set to 15bin, no blank part is observed, and a result of reducing noise can be obtained. That is, in the second embodiment, the result is obtained that the image quality is the best in the case where the parameter P is set to 15 bin.

In addition to the above examples, a total of about 20 experiments were performed. In any of the experiments, the result that the image quality of the radiological image 5 obtained under the parameter P at which the difference Q of the scattering fractions RF becomes maximum was the best was obtained.

(effects of the embodiment)

In the embodiment of the present invention, the following effects can be obtained.

In the present embodiment, as described above, the scatter estimation method includes the steps of: acquiring positron emission tomography measurement data and absorption coefficient data; generating a radiological image 5 (radiological image 50) from the positron emission tomography measurement data and the absorption coefficient data; estimating a single scatter distribution of radiation in the radiological image 5 (radiological image 50) from the radiological image 5 (radiological image 50) and the absorption coefficient data; a convolution kernel determination step of determining a convolution kernel for smoothing the single scatter distribution based on the scattered ray index value SF of the radiological image 5 (radiological image 50); and fitting the smoothed scatter distribution (graph SD) with a convolution kernel applied to the single scatter distribution to the positron emission tomography measurement data. Thus, the multiple scattering distribution can be simulated using the single scattering distribution without directly estimating the multiple scattering distribution. In addition, without using the parameter P determined empirically, the convolution kernel used for smoothing the single scatter distribution can be determined from the scattered ray index value SF of the radiological image 5 (radiological image 50). Thus, a multiple scatter distribution can be simulated from a single scatter distribution using an appropriate convolution kernel in the radiological image 5 (radiological image 50). As a result, the time taken for the scatter estimation can be shortened, and the degradation of the image quality due to the scatter correction can be suppressed.

In the present embodiment, as described above, the scattered ray index value SF is a scattering fraction RF, and the convolution kernel determining step determines a convolution kernel based on a comparison result of the respective scattering fractions RF of the radioactive image 5 (radioactive image 50) smoothed by using the plurality of parameters P representing the convolution kernels. Thus, the convolution kernel can be easily determined by comparing the respective scatter fractions RF of the radiological image 5 (radiological image 50) smoothed by varying the parameter P of the convolution kernel.

In the present embodiment, as described above, the convolution kernel is determined based on the magnitude of the change in the scattering fraction RF caused by the change in the plurality of parameters P. This makes it possible to determine the convolution kernel more easily.

In addition, in the present embodiment, as described above, in the step of estimating the single scattering distribution from the radiological image 5 (radiological image 50) and the absorption coefficient data, the single scattering distribution is estimated by the single scattering simulation method. Thereby, the single scatter distribution of the radiation can be easily estimated from the radiological image 5 (radiological image 50) and the absorption coefficient data.

In the present embodiment, as described above, the convolution kernel defines a weighted distribution determined by the dispersion-related parameter P, and the single scattering distribution is smoothed by the weighted average filter using the convolution kernel. This makes it possible to simulate a multiple scattering distribution of radiation from a single scattering distribution of radiation using a convolution kernel in which the weight distribution is changed by the dispersion-related parameter P. As a result, the processing time can be shortened as compared with the case of directly estimating the multiple scattering distribution. Further, since the parameter P of the convolution kernel is determined using actually measured data, it is possible to simulate a multiple scattering distribution using an appropriate convolution kernel, compared to a method in which the parameter P is determined empirically. As a result, the accuracy of the scatter correction can be improved.

In the present embodiment, as described above, the image processing apparatus 2 includes the control unit 20 and the image data acquisition unit 21, the image data acquisition unit 21 acquires positron emission tomography measurement data and absorption coefficient data, and the control unit 20 is configured to: a radioactive image 5 (radioactive image 50) is generated from positron emission tomography measurement data and absorption coefficient data, a single scatter distribution is estimated from the radioactive image 5 (radioactive image 50) and the absorption coefficient data, a convolution kernel for smoothing the single scatter distribution is determined based on a scattered ray index value SF of the radioactive image 5 (radioactive image 50), and a scattering distribution (graph SD) smoothed by applying the convolution kernel to the single scatter distribution is fitted to the positron emission tomography measurement data. Thus, the multiple scattering distribution can be simulated using the single scattering distribution without directly estimating the multiple scattering distribution. In addition, the convolution kernel used for smoothing the single scatter distribution can be determined from the scattered ray index value SF of the radiological image 5 (radiological image 50) without using the parameter P determined empirically. Thus, a multiple scatter distribution can be simulated from a single scatter distribution using an appropriate convolution kernel in the radiological image 5 (radiological image 50). As a result, the time taken for the scatter estimation can be shortened, and the degradation of the image quality due to the scatter correction can be suppressed.

(modification example)

The embodiments and examples disclosed herein are to be considered in all respects as illustrative and not restrictive. The scope of the present invention is indicated by the scope of the claims, is not indicated by the description of the above embodiments and examples, and includes all modifications (variations) that fall within the meaning and range equivalent to the scope of the claims.

For example, in the above-described embodiment, in the step of estimating the single scattering distribution from the radiological image 5 (radiological image 50) and the absorption coefficient data, the single scattering distribution is estimated by the single scattering simulation method, but the present invention is not limited thereto. For example, in the step of estimating the single scattering distribution from the radiological image 5 (radiological image 50) and the absorption coefficient data, the single scattering distribution may be estimated by a monte carlo simulation method. Even with such a configuration, the single scattering distribution of radiation can be easily estimated, as in the case of estimating the single scattering distribution using the single scattering simulation method. The monte carlo simulation is a method of obtaining an approximate solution by repeating a simulation using a random number. In addition, any method may be used as long as the single scattering distribution can be estimated.

In the above-described embodiment, the example in which the parameter P of the convolution kernel is determined based on the difference Q of the scattering fractions RF is shown, but the present invention is not limited to this. For example, the parameter P of the convolution kernel may be determined based on a differential value of the scattering fraction RF.

In the above-described embodiment, the example in which the parameter P that becomes the maximum value (maximum value) is used to determine the convolution kernel is shown in the graph 7 (graph 70) of the difference Q of the scattering fraction RF, but the present invention is not limited to this. The parameter P around the maximum value (maximum value) may be used to determine the convolution kernel. The convolution kernel may be determined using a parameter P in the vicinity of the minimum value (minimum value) or the minimum value (minimum value) of a graph obtained by plotting the inverse of the difference Q of the scattering fractions RF.

In the present embodiment, an example is shown in which the parameter P is determined by performing the processing of step S2 to step S5, but the present invention is not limited to this. For example, as shown in fig. 10, the processing of step S2 to step S5 may be repeated. That is, in step S7, it is determined whether or not the convolution kernel has been optimized a predetermined number of times. If the convolution kernel is not optimized a predetermined number of times, the process proceeds to step S2. When the convolution kernel is optimized a predetermined number of times, the process proceeds to step S6. The prescribed number of times is a number of times that no change can be observed in the optimization of the convolution kernel, and is, for example, three or four times. With this configuration, the accuracy of the scatter correction can be improved.

In the present embodiment, an example is shown in which the parameter P is determined by performing the processing of step S2 to step S5, but the present invention is not limited to this. For example, as shown in fig. 11, the parameter P may be determined in the first iteration, and the step of determining the parameter P may be omitted in the subsequent iterations. With this configuration, the processing time can be shortened and the accuracy of the scatter correction can be improved.

In the above-described embodiment, the example in which the positron emission tomography measurement data and the absorption coefficient data are acquired to generate the (radiological image 50) has been described, but the present invention is not limited to this. For example, the following may be configured: a pre-generated radiological image 5 (radiological image 50) and absorption coefficient data used in the generation of the radiological image 5 (radiological image 50) are acquired, and these data are used to perform scatter estimation.

In the above-described embodiment, a case is assumed in which data is acquired after positron emission tomography is performed and scatter estimation is performed when an image is reconstructed, but the present invention is not limited to this. For example, the scatter estimation can be performed in real time while imaging is performed by the positron emission tomography apparatus 1.

In the above-described embodiment, an example is described in which data obtained by converting the form data of the object O is acquired as the absorption coefficient data, but the present invention is not limited to this. For example, the image data acquiring unit 21 may be configured to: form data of the object O is acquired and converted into absorption coefficient data by the control unit 20.

In the above-described embodiment, an example is shown in which the graph 6 (graph 60) in which the scattering fraction RF is plotted and the graph 7 (graph 70) in which the difference Q of the scattering fraction RF is plotted, is created, but the present invention is not limited to this. The following may be configured: the parameter P for determining that the difference Q of the scattering fractions RF is a maximum value (maximum value) is not created in the graph 6 (graph 60) and the graph 7 (graph 70).

In the first and second embodiments, the parameter P is varied by setting the range of the parameter P to 5.0bin to 20bin, but the present invention is not limited to this. The range of the parameter P may be set within a range matching the size of the object O.

In the present embodiment, the standard deviation (σ) of the gaussian function is used as the parameter P of the convolution kernel, but the present invention is not limited to this. For example, the full width at half maximum (FWHM) of the gaussian function may also be used. Further, a half-value-width-half-maximum (HWHM) of the gaussian function may be used as the parameter P. In addition, 1/10-value width of gaussian Function (FWTM) may be used as the parameter P.

Description of the reference numerals

1: a positron emission tomography apparatus; 2: an image processing device; 3: a display unit; 5. 50: a radiological image; 6. 60: a plot of the scattering fraction; 7. 70: a plot of the difference in scattering fractions; 20: a control unit; 21: an image data acquisition unit; 22: a storage unit; p: a parameter; q: a difference in scattering fractions; SF: a scattered ray index value; RF: the fraction of scattering.