JPS61154646A - Radiation tomographic measuring apparatus - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed Description of the Invention] [Technical Field of the Invention] The present invention relates to a radiation tomography device that uses radiation to obtain a tomographic image of an object to be inspected, and performs a tomographic inspection or measurement of the object from this image. It is related to.

[Technical background of the invention]

2. Description of the Related Art There is a radiation tomography measuring device called a combinator tomography scanner (hereinafter referred to as a CT scanner) as a device that can measure internal defects, composition, structure, etc. of an object non-destructively and with high precision.

This device includes, for example, a radiation source that emits fan beam X-rays that spread in a flat fan shape, and a radiation source that is placed facing this radiation source through an object to be inspected, and that a plurality of radiation sources are arranged in the direction in which the fan beam X-rays spread. The radiation source and the detector are rotated head-to-head in the same direction over a range of 180° to 360° in 1 degree increments, centering on the object to be inspected. After collecting X-ray absorption data from multiple directions on the tomographic plane of the specimen, a computer etc. performs image reconstruction processing to reconstruct the tomographic image. Since images can be reconstructed in as many as 2,000 gradations, it is possible to know the state of the tomographic plane in detail.

This type of CT scanner is so-called third generation, and it also includes a radiation source that emits pencil beam X-rays, a detector that faces this radiation source, and a shield that covers the radiation source and detector. Traverse along the cross section of the specimen
A so-called first generation, pencil beam xst-a narrow fan beam X-ray detector that has several detection elements The so-called second generation, which can be said to be an improved version of the first generation, uses the above-mentioned traverse scan and rotational scan to perform the above-mentioned traverse scan and rotational scan. There are various types of OCT scanners, such as the so-called @4 generation, which uses a radiation source that emits beam X-rays and rotates and scans only the radiation source.

Image reconstruction methods for CT scanners can be broadly classified into analytical methods and algebraic methods.

Of these, analysis method 1, especially FBP (Filtered Back Project), is currently used in terms of processing speed.
tlon) method is the mainstream. The FBP method performs a convolution (sum-of-products) operation on radiation absorption data using a filter function, and reconstructs an image by backprojecting the projection data obtained by this operation. Although the FBP method is a useful method, since it emphasizes the high frequency range of projection data in order to improve resolution, it also emphasizes quantum noise and the like.

If noise is limited to quantum noise, if the transmitted dose is small, the S/N ratio will be relatively poor υ, resulting in image deterioration. This development occurs when there is a material with high radiation absorption in the imaging area (inside the circle formed at the outermost edge of the X-ray beam by rotational scanning) and the transmitted radiation becomes significantly smaller. You can also see it getting worse. As described above, the conventional FBP method is not suitable for reconstructing images from such projection data when the radiation dose is small or when the radiation dose is significantly small in some parts. .

Therefore, using different filters depending on the radiation absorption rate,
Convolution is performed using a filter function that emphasizes high frequencies for highly reliable data, that is, data with low radiation absorption, and a filter function that does not emphasize high frequencies for data that has high radiation absorption. A method has been proposed in which the above-mentioned disadvantages can be suppressed even in the FBP method.

To explain the details, for example, using fan beam X-rays that have a spread that includes the cross section of the object to be inspected, the cross section of the object to be inspected is projected from multiple directions, and the transmitted X-rays are transmitted to a radiation detector with spatial resolution. and collects X-ray absorption data from multiple directions for the cross section of the object to be inspected via a data collection device. And this X-ray absorption data KljL
, various corrections such as REF (reference) correction, offset correction, and standard correction, and log conversion.

Filter processing is performed on this preprocessed data.

Here, the preprocessed σ X-ray absorption data is converted into projection data f
Let's call it (,). This projection data f (→ is as shown in FIG. 8, for example, and the data (
transparent data). Here, Pa is X at the time of output
It is the number of photons in the line beam. Now, the transparent data p(x
) The middle order will be high. This means that the projection data f(x
), the smaller f(→, the higher the reliability. Therefore, the threshold value TH4+TH2 is determined for the projection data f(x) according to the degree of reliability.

As shown in Figure 8, O≦f(d (THlf(x) has high reliability TH1<,
Reliability of f(x) < TH2f(x) TH2≦f(
x) The reliability of f(x) is small.

Therefore, as shown in Fig. 9, the projection data f(x) is divided into f3+f2pf1 from the part with the highest reliability based on the level, and then fcx) = fl(x)+fz(:r)
+ f3(d is satisfied.Here, fl(→
, fl→, f3(→ is also f(→ medium) based on the evaluation in FIG. Here, various filter functions e, h2, and h3 as shown in Fig. 10 are prepared during condelusion, and the filter function is selected according to the reliability of the data.Here, As shown in FIG. 10, the filter functions are, for example, at, b, filter function 1 which does not emphasize high frequencies, filter function h2 which slightly emphasizes high frequencies,
Is it a filter function that emphasizes high frequencies? In Lupa, the degree of high-frequency emphasis is changed by selecting one of these filter functions according to the reliability of the data to be convolved. In figure j, H
(c) is the spectrum of the filter function, and "T" is the Nyquist frequency.

Now, how the actual condelusion is performed will be explained with reference to FIG. 11. Since condelusion is a linear operation, there is no problem even if the projection data f(→ is divided into fl efz + fs according to the level. Therefore, the projection data f(→ is divided into f
By comparing the projection data θ using threshold values TH,, TH2, it is divided into fl # fz r fs in area units delimited by THl * TH2. Therefore, f becomes f=f1+f2+f3, and these f11f2
t fs are f, (X), fz (X) in Fig. 11, respectively.
, f, stops like (x). Therefore, each f1(x)
+ fz(→, f3(→), select a filter function according to its reliability, t h2. h, perform convolution, and add the convolved data f Hh = (t, ( :r)+fz(x)+f
3(x) )x h = f, (x) * h, + fz
(x)M h2+f3(x)*h, (X indicates convolution) is obtained. That is, depending on the reliability of the projection data, the filter function h1vh2rh
5 is selected, convolution is performed using this selected filter function, and by adding these, the degree of high frequency emphasis is changed for each data region depending on the level of reliability, and those that can emphasize high frequencies are Process it in a way that emphasizes it.

Then, the data fX obtained by this condelusion
A reconstructed image with improved resolution can be obtained without deteriorating image quality by applying h to the backgrojector 12 and performing a quadrature correction. Also fl, fz
, t, and TH2 can be simply divided into three equal parts of the dynamic range of the projection data. For example, projection data f (→ is from 32767 to -32
Assuming a dynamic range of 768, TH2=10923.

TH,=-10923. Then, from +32767 to +10923, smoothing processing is performed using filter function h1, and 109
High-frequency emphasis is performed by filter function h2 between 22 and -10923, and by h3 below -10924.

However, in reality, it is better to divide the area larger than the reference value "0" into three.

The reason for this is that since the material used as the reference material is selected with a good S/N, it can be determined that the S/N is good when the median value θ is below. In addition, for substances with high radiation absorption, such as industrial products whose main component is metal, the threshold value T
H4*TH2 is set to a high value depending on the composition, such as 10000.1000, for example. In this case, the filter function h2 is used when f(→ is in the range of 32767 to 10000, the filter function h2 is used when f(→ is in the range of 9999 to 1000, and the filter function is used when f(→ is in the range of 999 to -32768. Using such a method, it is possible to suppress noise components without reducing resolution, which was not possible with conventional FBP methods, and it can be achieved at a relatively low cost by simply adding simple logic and memory to conventional components. In addition, when reconstructing an image from projection data with a markedly insufficient dose, high-frequency enhancement is performed with emphasis on areas with a relatively high dose, and areas with a lot of noise are smoothed to generate an image, which is difficult to do with conventional methods. A clearer, higher quality image with fewer noise components can be obtained than the above.In the previous example, the division method of the projection data f(X) could be called simple additive division, but proportional additive division can also be considered. This is similar to Figure 9, f (x) = f 1 (x) + t 2 (X) +
Let f 3(x) be respectively. Although this proportional additive division has lower resolution than simple additive division, it is possible to relatively suppress noise components.

Also, as shown in FIG. 12 (&), the projection data is f(d
= t 1(x) + t, (x) + t jx
) and using the threshold value TH4*TH2,
), (C), and (d). In this method, in the previous method, each of the items in the area partitioned by the threshold level was calculated using the corresponding filter function, and the level area valve for addition was a proportional distribution method, but here Check which threshold value THt t TH2 has been reached for each position of the radiation detection element, and depending on the level, the projection data f which indicates a value of TH2 or more, and (x), only the filter function of hl can be used to determine whether THt or more is TH2 or less. Projection data f〆 indicating the value of →
In this case, the projection data f, which shows a value less than or equal to T, is a level-corresponding division that is convolved only with the filter function of h2, and this is faithful to the reliability of the projection data. It is possible to use a suitable division method.

By using such a method, it becomes possible to perform image reconstruction of a cross section of the object to be inspected, which includes X-ray absorption ranging from large to small X-ray absorption, using the FBP method.

However, in this way, the above method is a method in which the entire possible range of data is divided into levels in advance according to the reliability of the data, and a filter function corresponding to each level is used. Threshold values TH4 and TH2 for this purpose were set in advance. Therefore,
It is fine if the data values are distributed over the entire area, such as the data indicated by symbol I in Figure 13, but if the data values are distributed in a region below the vicinity of the threshold value T, as indicated by symbol ■, it is fine. In the case of absorption rate data, the threshold value TI (areas of 2 or higher is completely useless.And even for such useless areas, in the case of the above method, the presence or absence of data of this level or higher is determined using the threshold value TH2. This is a wasteful process as it involves investigating.

Furthermore, it is not possible to apply a fine-grained filter function to the entire range of data values that can be handled without considering the distribution of data values.

Therefore, there is a limit to the improvement of image quality, and even though it is possible to obtain better resolution depending on the distribution of data, this is restricted.

[Purpose of the invention]

The present invention was made in view of the above circumstances, and its purpose is to prevent F
Using the BP method, it is possible to suppress quantum noise and obtain high-quality reconstructed images without significantly deteriorating resolution, and the threshold value can be varied according to the distribution state of the values of X-ray absorption data. It is an object of the present invention to provide a radiation tomography measuring device that can use a filter function having frequency characteristics according to a threshold value, thereby making it possible to further improve the resolution according to the reliability of data.

[Summary of the invention]

That is, in order to achieve the above object, the present invention projects radiation from each direction on a set tomographic plane of an object to be inspected,
Radiation projection data for each projection direction is obtained by detecting the transmitted radiation with spatial resolution, and image reconstruction is performed using these radiation projection data to reconstruct a image corresponding to the radiation absorption rate of each position on the tomographic plane. In the image acquisition device, as a means of reconstructing the image, the maximum value of the projection data is checked, and the reliability of the data due to noise is calculated by determining the difference between the maximum value and the value of a standard material. means for determining a threshold for dividing the projection data into predetermined sections; means for dividing the projection data according to reliability using the threshold; determining filter functions with different frequency characteristics according to the reliability and projecting for each of the divided projection directions; and means for condelusing and synthesizing the data with a filter function corresponding to the reliability of the data to obtain data for back projection;
The maximum value of the projection data is examined, the difference between the maximum value and the data value of a material with standard reliability is determined, and the interval is divided into predetermined sections, and the projection data is divided according to the reliability based on the relationship with the noise component. At the same time, filter functions with different degrees of high-frequency emphasis are determined depending on the reliability of the projection data of this divided area, and condelusion is performed using this, and by compositing, the more reliable the projection data, the higher the By obtaining region-enhanced backprojection data and backprojecting it to obtain a reconstructed image, noise components can be suppressed and resolution can be improved in the FBP method.

[Embodiments of the invention]

An embodiment of the present invention will be described below with reference to the drawings. Although the third generation CT scanner will be explained here as an example, it can also be applied to other generations of CT scanners. FIG. 1 is a block diagram showing the basic configuration of this apparatus, and numeral 1 in the figure is the scanner body. Reference numeral 2 denotes an X-ray source (radiation source) that emits fan beam X-rays having a predetermined spread width. 3 is a radiation detector installed opposite to this X-ray source 2. This radiation detector 3 has a large number of minute radiation detection elements arranged in parallel in the direction of spread of fan beam X-rays, and has a high spatial resolution. X-ray intensity from an X-ray source can be detected using Here, the X-ray path connecting the X-ray source 2 and each radiation detection element is called an X-ray path, and each radiation detection element outputs a signal corresponding to the intensity of the radiation on the X-ray gas. The scanner body 1 also has a rotating mount that holds the X-ray source 2 and the radiation detector 3 facing each other around the imaging area and rotates and scans them in one direction at predetermined angle increments. are doing. Reference numeral 4 denotes an object to be inspected placed in the photographing area of the rotary frame, and reference numeral 5 denotes an X-ray controller for controlling the tube current, tube voltage, X-ray exposure, etc. of the X-ray source 2. Reference numeral 6 denotes a scanner controller using a CPU (central processing unit), which controls the rotation of the rotating pedestal of the scanner body 1. A system controller 7 controls the entire system. Reference numeral 8 denotes a console, which is used by the operator to give various commands and the like to the system, and is capable of inputting various operation commands and data to the system.

Reference numeral 9 denotes a data collection device which receives the output signals of each radiation detection element of the radiation detector 3 and converts the signals.
Output as X-ray absorption data. Reference numeral 10 denotes a preprocessing device, which receives absorption data for each gronoection collected by the data collection device 9 and performs preprocessing such as logarithmic transformation, r-in correction, and offset correction on the absorption data. .

Reference numeral 11 denotes a converter, which converts the preprocessed data output from the preprocessing device 10. It is something that can be done.

Reference numeral 12 denotes a back projector that reconstructs the tomographic image by back projecting the converted data in the glojection direction. The above 10.11.12 constitutes a reconstruction device. Reference numeral 13 denotes a memory for storing data of this back-projected reconstructed image, and reference numeral 14 stores CT values in a desired range (data corresponding to the level of radiation absorption) from among the data stored in the memory 13. For example, it is an image converter that outputs a black and white grayscale image. 15 is a CR that receives the output of this image converter 14 and displays it as an image.
It is a T indicator.

In such a configuration, when the operator first operates the console 8 to start up the system and start scanning (starting imaging), the system controller 7 controls the scanner core )o-96, Scanner body 1
The rotary mount is controlled to rotate in predetermined angle increments (for example, 0.6°), and the X-ray controller 5 is controlled to adjust the tube current and tube voltage to predetermined values every time the rotation is performed in the predetermined angle increments. It is applied to the X-ray source 2 for a predetermined time width. From this, X-ray source 2 sequentially sends 7 am beams X@F
B is exposed. Rotation center position of rotating mount (imaging area)
The object to be inspected 4 is disposed on the rotating frame, and since the X-ray source 2 and the detector 3 are mounted on the rotary stand facing each other across the center of rotation, the X-ray source 2 Fan beam X1iFB
This Fan Beam X@1B
The radiation transmission value of each X-ray beam at /C is detected by each radiation detection element of the radiation detector 3 and converted into an electrical signal.

This converted signal is collected by a data acquisition device 9, and subjected to logarithmic conversion, r-in correction, offset correction, etc. in a preprocessing device 10 for each 1f logiexion (one photographing direction). This processed data (X-ray absorption data in each X-ray pass for each Zorological Extraction) is stored in the memory 13, and is also given to the converper 11 for convergence.
CT values of individual pixel positions are obtained by back projecting with the next back projector 12, and the tomographic image is reconstructed based on these CT values. The reconstructed image is stored in the memory 13, and by giving commands from the console 8, the image converter 14 displays CT values in a desired range on the CRT display 15 with adjustment according to the CT values as necessary. Display. As a result, the reconstructed image is displayed as a black and white grayscale image.

Next, the processing of the reconstruction device, which is the central part of the present invention, will be explained using FIGS. 2(a) and 2(b).

The reconstruction process can be classified into three types: preprocessing, filter processing, and back projection, and the preprocessing section 10. Conga Lupa 11. It is executed by the back projector 12. That is, the preprocessing unit 10 performs various corrections such as REF (reference) correction, offset correction, and standard correction, and log conversion on the X-ray absorption data, and also performs log conversion on the X-ray absorption data.
performs an operation that returns the filtered data in the same direction as the data was collected.

The above two processes are general tables, so details will be omitted.

Next, details of the filter processing in the congelper 11, which is the main part of this embodiment, will be explained.

The filtering process of the controller 11 is performed through steps A, B, and C in FIG. 2(a), and the filtering process can be roughly divided into the following two methods.

(1) A method of individually dividing and filtering the projection data of each glojection. In other words, for each glojection's projection data f(,) (preprocessed X-ray absorption data), A method of examining the distribution, dividing it into a predetermined number of regions according to the distribution, and applying a filter function corresponding to each divided region to its projection data for filtering processing.

(2) A method in which the projection data of each projection is analyzed to obtain a standard filter function, and the data of all projections (or a specific projection is divided and filtered using this function).

That is, the frequency distribution of these data values is determined based on the individual distributions of projection data 1 (,) of all projections or projections in a specific direction, and the data values of the distribution are divided into a predetermined number of regions according to this frequency distribution. , find the filter function corresponding to each segmented area, and calculate υ by each filter function.
A method of filtering all or specific projection data f(x) for each segmented area.

Furthermore, the following two methods can be selectively used for region segmentation of projection data or frequency distribution in these filter processing methods (see B-1 and B-2 in FIG. 2(b)).

(g) A method of finding the extreme values of projection data, etc., and simply dividing the area up to the reference level into a predetermined number of parts.

(B) A method of examining changes in the distribution of projection data, etc., and dividing the data according to the change.Furthermore, the following two methods can be selectively used to adapt the filter function corresponding to each segmented area (see Figure 2 (b). )
(See C).

(Q. A method of calculating a filter function based on a threshold value of a region segment or a value of interest (for example, an intermediate value) within a region.

(b) A method of selecting a filter function given empirically in advance based on a threshold value for region division or a value of interest within the region.

These area segmentation methods and filter function adaptation methods can be selected and used depending on the form of the image.

As an example, the filter processing method (
An apparatus that uses method (1), (4) for region segmentation, and (0) for filter function adaptation will be explained with reference to FIGS. 3 to 7.

In this embodiment, the distribution of projection data f (→), which is preprocessed X-ray absorption data, is examined to find out its maximum and minimum values, and the threshold 'r is set to determine its width and divide it into thirds.
a1#TH2 is determined, a filter function corresponding to the threshold value is calculated, and filter processing is performed.

However, in the case of a general OCT scanner, the periphery of the object to be inspected is air, which is the material with the lowest X-ray absorption.
IN) is always detected around the object to be inspected, and it is safe to assume that its data value will always show the same value if the CT scanner is always calibrated to predetermined conditions.

Therefore, if the projection data with the maximum X-ray absorption (this is designated as Wα) is known, the threshold values TH, TH2 can be determined from the relationship between MIN and MAX as shown in FIG.

The figure shows an example of the distribution range of projection data values.Here, assuming that the numerical range that the CT scanner system used can handle is 32,767 to -32,768, the projection data is distributed in the range of -32,768 to about 20,000. An example is shown below. Data with a MIN of -32768 is air with the lowest X-ray absorption, and data corresponding to ■ are assumed to be contained in the test object subjected to measurement, such as lead, iron, copper, tungsten, etc. .

A data value of 0 means Xa that can be measured with a CT scanner.
This corresponds to a measured value of the middle intensity within the intensity range, and corresponds to projection data of a substance such as water.

In a CT scanner, a phantom in which water is placed in a hollow cylindrical container made of acrylic or the like is used as a standard test object for calibration, and calibration is performed while observing the state of a reconstructed image of this phantom.

The phantom is used to obtain reference data, and after calibration, the level of the signal component is sufficiently higher than the noise component for materials with smaller X-ray absorption, mainly water, so the frequency characteristics are determined as a filter function. Noise is not noticeable even when using one that extends sufficiently to the high range.

For example, if the data value of water, which is at a level where this noise is not noticeable, is set as STD, and the threshold values TH1 and TH2 are set so as to divide the data distribution range from this STD to the box into three parts, the calculation for finding the threshold value is easy. Furthermore, it is possible to adapt to a data value distribution range that would not appear unless careful consideration is given to the frequency characteristics of the filter function used to obtain a reconstructed image of the best quality.

Therefore, by dividing the area from ■ to STD into three, the filter function hi
(→ Assign (1=o, 1.2.3).

In other words, the projection data f(→ is f(→<STD).(→8TD(f(→
<THl, (→TH1(r(i≦TH2, h
, (X)TH2(r(d) is assigned as h, (X). One filter function f used here.

(X) to f3(X), fo(→ is the filter with the best high-frequency characteristics, fl(x)→f2(→→f3(
This means that the filter has a low frequency high-frequency characteristic in the order of →.

Here, we will discuss an exceptional case in which all the projection data fits into MAX or STD.

In this case, since all the projection data used for reconstruction is below the STD, a reconstructed image with less noticeable noise can be obtained even if a filter function whose frequency characteristics are sufficiently extended to high frequencies is used, so dividing filtering using a threshold is necessary. There is no.

Everything is fine. (You can perform condelusion with →.

Next, a division method between STD and MAX in projection data having a normal data value distribution excluding the above-mentioned exceptions, that is, a projection data area division method (6) will be explained.

As explained earlier, the quantum noise targeted here depends on the square root of the number of transmitted photons, and the projection data f(x
) is the logarithm of the transparent data stable → f(:c)=kZn #-
Therefore, strictly speaking (transparent data is PIITD > PTill > PT
There is a relationship of II2 > P Yu. Further, k indicates a constant.

) and taking the square root, PIITD has a better S/N ratio than PMAx (8N ratio (PIITD) > S
N ratio (PMAx)), so the TH is divided into three equal parts.
When l and TH2 are obtained, the following formula is obtained.

Therefore, from the relationship f(→=kinu), P (x) is obtained. Therefore, if the formulas 4 and 5 are set, S
/N is divided into equal parts. Of course, the projection data f
(Divide into three at the point → and use the following 6th formula.

It is also conceivable to simplify it as in the seventh equation.

4”-8TD...(6)TH,
= 3+STD MAX -5TD TH2= MAX -3...(7) As is clear from the formula, these are calculated by dividing the width of the distribution range from STD to MAX into three, subtracting this from MAX, and calculating This is a method of performing addition to obtain TH2, TH.

The above is the method for setting the threshold value.

Next, a method for setting the filter function, that is, a method (C) for adapting the filter function corresponding to each segmented area (threshold value) will be explained.

This setting method is strongly empirically oriented, and depends on the appropriate image depending on the type of material that makes up the object to be inspected, that is, whether the observer wants to see the shape and dimensions of the object or the composition distribution. It is difficult to theorize quantitatively because the appropriate images vary.

For this reason, the filter function described above. (X) to h 3(
The design will be such that the high-frequency characteristics change smoothly over X), and this will be determined qualitatively.

Here, the relationship between the threshold value I(T and the filter function h(:c)) will be explained.

When the filter function h (X) is described in the frequency domain, it becomes as shown in FIG. 4 (!L) l (b).

Figure 4 (,) is the function proposed by Ramachandran, with 1
It has a form in which ω1 is truncated at the Nyquist frequency ω.

Figure 4(b) is the one proposed by Kuepp and Logan, which has a longer high frequency range than the function proposed by Ramachandran, so it is possible to obtain a clearer image, but on the other hand, , it has the disadvantage of being easily multiplied by noise.

In this way, filter functions can be distinguished based on their frequency characteristics. For example, in FIG. 4, among the respective values of H (c) when ω, Ramachandran's filter can be said to be a filter that emphasizes the high frequencies more.

Now, the relationship between frequency characteristics and threshold values will be explained here. Qualitatively, this can be explained by the fact that filter functions ■3 to H8 strongly emphasize high frequencies as shown in FIG. What is important here is that these filter functions H (c) are equal to the threshold TH
is a function of .

That is, H(G)) = Ho(+-)) X G(ω, TH)
・(8) However, ■. (c) represents the standard characteristic, G(ω.

TI() represents a function of the angular frequency ω and the threshold value 'rH.

In the embodiment, the filter function H6 (c) with standard characteristics is determined from the She, Zo, and Rhogan equations (where ω is the Nyquist frequency), and the 10th equation is used for G(ω, TH).

(When ω≧ωA) Incidentally, the function of Equation 10 is shown in FIG.

However, in Equation 10, ωA and ωB are ω = ωN・(
1-1 stop) ω door = ωN (K is a constant). That is, the higher the threshold TH is, the smaller ω is than ωN, the higher the gain in the high range is, and the higher the high range is not emphasized. In this way, the threshold values TH1, TH2
After determining the projection data, divide the projection data into four parts, find the optimal filter function according to the S/N for each division, use this to perform filter processing on the projection data, and obtain zorozoexi1 data. These processes are performed on the projection data in each direction, and these data are sent to the back projector 12, where this zoloprojection data is back-projected for each glojection in a direction corresponding to the direction at the time of projection. Reconstruct the image.

This con? FIG. 7 shows a flowchart of the processing procedure performed by the projector 11 and the back projector 12.

In other words, regarding the measurement cross section of the object to be inspected: Q6 ~ 180°
Transmission data from each direction is obtained and preprocessed. If data is collected using a fan beam, then parallel beam conversion is performed to obtain projection data in the form of parallel beam data for each projection direction (st@p 1 ).

Next, the work area of ConDelpa 11 is entered. That is, 5t
In ep 2, the maximum value of all the projection data for each projection direction converted into the above-mentioned nonoraral data, that is,
Find the maximum X-ray absorption MAX. Then, 5tep
3, and it is determined whether this MAX is greater than the standard reliability level STD.

If not here, the projection data has a good S/N, so there is no need for divisional filtering. Therefore, in 5top 4, convolution is performed using only the filter function Ho (Ha) with good high frequency characteristics, and then the inverse Perform image reconstruction by projecting.

st@p If it is determined in 3 that MAX is larger than STD, the process goes to step 5, where the width between STD and MAX is checked and thresholds TH1 and T are set for dividing this into three.
Find H2.

Next, enter 5top 6, threshold TH,, TH2 and ST
D.

Filter function h1(x) suitable for WAX *h〆→
, h, (→.

Make h 0 (X) -, h 5 (→.

Next, in step 7, the projection data is divided into the regions described above, and convolution is performed using the filter function corresponding to each region to create projection data for each projection direction.

Next, pass the work to the pack projector 12 and perform 5 steps.
8, each of the above Glojek/Wn data is back-projected and reconstructed.

In this way, this embodiment examines the maximum value of the collected projection data for the imperial palace surface in each direction, and calculates the standard reliability value that can be guaranteed from the S/N relationship and the above value. The width of the numerical values between the maximum values was investigated, and the threshold value for dividing the width into three was determined, and the high-frequency characteristics were adjusted as a filter function for the data of each divided area according to the above-mentioned reliability of each area. Find an appropriate filter function for each region, perform filtering using a filter function with good high-frequency characteristics for regions whose reliability is sufficiently guaranteed, and perform image reconstruction using the processed data. This is what I decided to do.

Therefore, depending on the level of the maximum value of the collected projection data,
In addition, for data with higher X-ray absorption than standard reliability data, the data value is divided into regions with a division width corresponding to the width of the distribution of the value, and reliability related to the S/N of the data in each region is calculated. Since filtering can be performed by obtaining a filter function tailored to the degree of reliability, fine-grained filtering can be performed according to the degree of reliability. Emphasis is now possible, and resolution can be maximized to match the reliability of the data.

Next, as a filter processing method of ConDelpa 11, the above (2)
) method is used, the above method (B) is used for region segmentation, and the above method (B) is used for the filter function adaptation method.
This will be explained using figures.

In this embodiment, the frequency distribution measuring unit A is based on the signal value of the projection data f (→ that is, a signal that has been discretized and quantized corresponding to each detection element of the radiation detector (Fig. 2 (C)). The frequency distribution of the signal values is determined and this measurement is performed on the projection data in all directions or in a specific direction.The frequency distribution dividing section B
The amount of change is determined, and the projection data value for which this amount of change is equal to or greater than a predetermined value is determined as a threshold value. The filter function generation unit C for each frequency category stores various filter functions obtained empirically or by calculation according to the above-mentioned formula 8 in the internal memory in advance, and uniquely generates the filter functions based on the threshold value. Select the number of filters that correspond to . Then, based on this filter function, the projection data f(→) is sequentially filtered and supplied to the projector.

Note that the present invention divides the projection data f (x) into predetermined sections based on its distribution, and performs filter processing by obtaining a filter function corresponding to the section. C') (D) and configurations that obtain similar results can be freely adopted to achieve similar effects.

Further, in each of the above embodiments, three divisions were used as the number of divisions, but the division may be further subdivided in order to achieve a finer effect. However, in this case, it is necessary to set a corresponding filter function in advance according to the divided area. Furthermore, as the number of divisions increases, the processing time required for condelusion calculation also increases.

Furthermore, although filter processing is performed in the real domain in this embodiment, this method can be similarly applied to the frequency domain. This is because the Fourier transform is a linear transformation, so if we let (f(→F(ν)) (where, Re represents the pulley 8 transformation), then = Re(f)
, (→)+9(f〆→)+9(t,(x))=F,(
This is because the following holds true: ν)+Fノν)+F3(ν). In that case, the frequency representation of the filtered projection data is (h(t)) as the sum H(ν) as F(ν) × H(ν) = F, (ν) × also (ν) + F no ν) XHノν)+F, (
ν)×Castle ν).

Furthermore, the threshold values used in the above embodiments are merely examples, and can be set as appropriate according to the data of the actual reference material and the composition of the object to be inspected, without being limited thereto.

As for the setting method, it is sufficient to set the optimum one while checking the state of the reconstructed image. The present invention can be applied not only to a CT scanner using X-rays but also to a 90T scanner using other radiation sources such as radioisotopes.

〔Effect of the invention〕

As detailed above, according to the present invention, even in cases where the amount of radiation reaching the detector is extremely low, such as when inspecting an object that contains a substance that absorbs a large amount of radiation, image quality can be reduced by the FBP method. It is possible to provide a radiation tomography measuring apparatus having the characteristics of being able to perform image reconstruction without any interference, and also being able to obtain images with as high a resolution as possible depending on the composition of the object to be inspected.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of the device of the present invention, FIG. 2 is a diagram for explaining the present invention in detail, and FIG. 3 is a diagram for explaining the basic concept of the threshold setting method of the present invention. , FIG. 4 is a diagram showing the filter function in terms of frequency characteristics, FIG. 5 is a diagram showing an example of the filter function used in the present invention, and FIG. 6 is a diagram showing G(ω
, TH), Fig. 7 is a 70-chart showing the method of the present invention, Fig. fs8 is a diagram for explaining the relationship between projection data and threshold, and Fig. 9 is a diagram showing the relationship between projection data and threshold. Figure 1 for explaining the relationship with the data area
Figure 0 is a diagram showing an example of a filter function used in a conventionally proposed method, Figure 11 is a diagram for explaining the state of convolution used in a conventionally proposed method, and Figure 12 is a diagram showing an example of a filter function used in a conventionally proposed method. FIG. 13 is a diagram for explaining an example of the relationship between projection data and a threshold value. DESCRIPTION OF SYMBOLS 1... Scanner main body, 2... X-ray source, 3... Detector, 7... System controller, 9... Data acquisition device, 10... Preprocessing device, 11... Conderno 12...Pa, Kuzoro projector, 13...Memory, 15...CRT display. Applicant's representative Patent attorney Takehiko Suzue Figure 1 Figure 2 B-I B-2 211 (C) (d) TH+ TH27H3f(x) Figure 3 (a) Figure 5 (b) G ( cu, TH) 0 ωA (AJB Fig. 7 Fig. 8WA 9th evil 111il

Claims (1)

[Claims] Radiation is projected from each direction on a set tomographic plane of the object to be inspected, and the transmitted radiation is detected with spatial resolution to obtain radiation projection data for each projection direction, and image reconstruction is performed using these radiation projection data. and obtaining a reconstructed image corresponding to the radiation absorption rate of each position on the tomographic plane, the image reconstruction means means for determining a threshold value for dividing the projection data into predetermined sections based on the distribution of the projection data. , a means for determining a filter function with a corresponding frequency characteristic for each of the predetermined sections, and a means for convolving and synthesizing the projection data for each projection direction with the corresponding filter function for each of the predetermined sections to obtain data for back projection. What is claimed is: 1. A radiation tomography measuring apparatus comprising means for obtaining back projection data in which higher frequencies are emphasized for projection data with higher reliability, and back projecting the data to obtain a reconstructed image.