JPS62254045A - Three-dimensional tomographer using electro-optical type x-ray detector - 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 Field of Industrial Application The present invention is directed to a three-dimensional tomography apparatus i! using an electro-optical X-ray detector. Regarding t.

[Prior Art] Computerized tomography is a method of determining planar segmentation of a target by measuring and analyzing the attenuation of a beam of radiation passing through a target camera along a number of sets of coplanar rays. 1. means the procedure used to form a two-dimensional map of physical quantities in two locations. As is usually the case, a complete device consists of four elements:
1-) a penetrating radiation source; (2) a detector capable of measuring the transmitted intensity of the radiation after passing through the target and calibrated to determine the unattenuated intensity of the radiation in the absence of the target; (X□3) Check the attenuation measurement value! a calculation device m for storing, processing, and converting into a digital map of the attenuation coefficient in the observation plane of Targetera i;
4) It must be equipped with a device for displaying the image formed thereby.

Tomography can be performed in a number of ways, but
The most widespread commercial use is in the field of medical radiography, in which a diagnostic map of a patient's skeletal and tissue structures is created.
32-41 of Physjes Today
Mr. L 5w1nde11 and 1; published on page. ! +
, Barrett's ``Computerized tomography;
Cross-sectional X-ray photography (ComputerizCd chomo
graphy:Taking 5ectjonal X
-Rays) J; l 982 American Science l” (American 5eieritist
) 70.576 by C. C. Jaffe, “Medical Imaging Technology (Medical Imaging Technology)”
g) J; and 1983 Computer
er) 1 published in 1000].

AJ, exandpr, “Array Processors in Medical Imaging”
Medical Imaging) J). Medical CT uses broadband bremsstrahlung radiation from an X-ray tube to create penetrating radiation, which typically
Measured by scintillation crystal and phototube. The measurements are stored in a programmed digital computer and analyzed using a method commonly referred to as convolutional (or filtered) back projection (hereinafter FBI').9 The density map derived by this analysis is displayed on the cathode ray tube as a two-dimensional cross-sectional image containing approximately 250×250 pixels, or vixels. Its resolution is approximately 111I11 and the accuracy of determination of the X-ray attenuation coefficient is 1%. In medical procedures:! Other special purpose tomographic probes using other forms of ionizing radiation such as gamma rays or electrons has also been developed.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a micro-tomography apparatus that uses an intense, well-collimated radiation beam to form two-dimensional images with improved spatial resolution. It is to be. 1 using this micro tomography device! ) The spatial resolution that can be obtained is as small as 0.5 microns, which is about 100 to 1000 times higher than the resolution obtained with conventional medical CT.

Yet another object of the invention is to increase the physical scale across the reconstructed image. The micro-tomography apparatus described herein can obtain images with twice as many pixels (physical scale) per plane light as conventional medical CT. The increase in physical scale across the image is particularly important in micro-tomography.

Another object of the invention is to reconstruct an object with a three-dimensional network of one point. Rather than obtaining takes in 1 to 20 adjacent planes as in medical tomography machines, the apparatus according to the invention obtains enough data to reconstruct an image in 100 or more adjacent IP planes. This allows three-dimensional information about the object to be obtained. In general terms, the device disclosed herein can be considered a three-dimensional X-ray microscope.

Improvements in resolution and physical scale in the reconstructed images are obtained by the design of the detectors used to measure the transmitted intensity and the computational techniques used to process the data. Yet another object of the invention is to obtain data using two contrast electro-optic detectors. This detector obtains full linear slices or full planar images of the image with large physical scales without statistically adversely affecting the count of each vixel across the image. By using electro-optical detectors to obtain large-scale, full-plane images, data collection time and the number of lines applied to the sample are significantly reduced. Furthermore, it is possible to construct an electro-optical detector with significantly higher spatial resolution than scintillation detectors used in medical CT. The number N of data points (resolved pixels) obtained in a line across the image is significantly thicker than in medical CT, and L, L many stacked! Since data can be obtained simultaneously in the V-plane, it is important to use a data inversion technique that requires only N2 operations rather than Nj operations to reconstruct the image. By using data inversion techniques that involve only N2 operations rather than N3 operations, the time required to process data can be reduced to more than 100 minutes in many cases.

Means to solve problems The present invention relates to an apparatus for forming tomographic images of objects.

The apparatus includes a plurality of sources of collimated radiation transmitted through the object and a contrast detector for detecting the attenuated transmitted radiation after passing through the object, the detector comprising an energy converter. . Also provided are means for obtaining projection data from the transmitted radiation and means for calculating from this projection data a reconstructed image of the attenuation coefficient of the object. In a preferred embodiment, the contrast detector described above is an electro-optical detector, which also comprises an image formatting device and a readout device. This detector has a detection quantum efficiency of 0
．． 05, the total signal dependent background is less than 10% of the signal from the unattenuated x-ray beam, the effective dynamic range is greater than 10, and the response non-uniformity between adjacent active vixels is less than 75%. Yes.

The geometric straightness deviation is less than 10 pixels in the recorded image.

Embodiments Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

TECHNICAL FIELD The present invention generally relates to computerized tomography devices. The device includes a detector that can be used in a mode that increases the spatial resolution in the reconstructed image to less than 10 microns. Certain types of detectors can be used to collect data in multiple stacked planes and reconstruct the internal structure of the sample onto a three-dimensional network of points. In this mode, the device disclosed herein functions as a three-dimensional x-ray microscope.

The detectors used in the present invention form a significant part of a wide range of electro-optical detectors that ionize radiation.

Electro-optical X-ray detector.

It can be broadly defined as a position-sensitive detector using components developed for optical image amplification and recording. The contrast properties of this type of detector make it possible to obtain significantly higher spatial resolution than was possible with scintillation detectors used in conventional tomography devices.

The present invention utilizes a collimated radiation beam. This radiation must be converted into an easily manageable quantity that can be used to form an image in a detector. Generally, radiation that meets this condition is ionizing radiation such as X-rays, gamma rays, neutrons and ultraviolet radiation.

Fluorescence conversion processes for visible light in which ionized states are not formed are also included in the definition of radiation that can be used in the present invention. Explanation: h. The present invention will be explained using X-rays as radiation.

In order to form an accurate tomographic image, it is necessary to obtain sufficiently noise-free data and a sufficient number of data passing through the target. (1,974, IEE [ETrans
, Nuel, Scj, Volume N5-21, Volume 21-
14°A, 5hepp and B published on page 43,
F.

Mr. Logan's ``Fourier reconstruction of the head''
UrjerReconstruction of a
l1ead 5ection) J and Roberg
e and [3, P, Mr. Flannery's 1'' present application (
1986, No. 196475).

1131 detection paths are determined by their DI angle φ and ? with respect to the coordinates fixed to the target, as shown in FIG. #Shot parameters can be viewed based on. In medical t17i layer imaging, a+lJ constant values are typically obtained with a fixed set of detectors placed along a ring around the patient, as shown in FIG. The x-ray source rotates around the ring and directs a collimated fan-shaped beam of radiation onto a series of detectors facing the source. The spread angle of the collimated Xn beam is wide enough so that the fan of the path from the source to the detector completely surrounds the target. In order to accurately reconstruct the entire target, the range of impact parameters must span the diameter of the target and the rotation angle must span at least half a complete revolution. The creating mode shown in FIG. 2 will be explained as a fan-shaped beam collimating operation.

Another mode of operation of the tomography scanner is shown in FIG. The collimated x-rays or other penetrating radiation strikes the target in a two-dimensional set of planar parallel paths recorded by a panoramic electro-optical detector. When using a two-dimensional contrast-enhanced optical detector, data in a number of stacked layers are measured simultaneously. 1316 can be performed from various angles by rotating both the source and the detector around the sample.
It can be said that this was obtained in a plane parallel collimated state. When a high spatial resolution of 0°5 to 225 microns is selected, the plane-parallel data acquisition mode is clearly preferred. When collecting data with high spatial resolution, the X-ray beam can be easily collimated for the plane-parallel mode shown in Figure 3, but the fan-shaped beam configuration does not allow for proper collimation. Have difficulty. Honmei Im'
We describe the use of electro-optical detectors to collect data in plane-parallel mode through O#. However, the invention is also applicable to fan beam mode data collection.

The time required for observation and the quality of the image formed by the apparatus of the present invention depend on the composition and size of the Targetera, the characteristics of the source, and the performance of the detector. In general, the observation conditions can be adjusted to optimize the source spectrum for a given sample, or by adjusting the sample size to optimize the source spectrum for a given source. The optimal viewing conditions are determined by signal-to-noise considerations for the tomographic image.

These signal-to-noise considerations impose constraints on detector performance in various ways, and the types of electro-optic detectors suitable for use in tomography scanners are limited to a wide range of electro-optical detector types. Limited to a small portion of the species. j9? to determine constraints on the types of electro-optical detectors available. Analyze the zt sound source in the layered image.

Noise in tomographic images is generated from two noise sources.

These are (1) noise in the data and (2) noise amplification caused by the inversion method.

The basic data is the measurement of the attenuation of Xv, typically passing through the target along many coplanar beams.
Consists of measurements of A-beam attenuation (1974, I[E
E[E Trans, Nuel, Sci, No. NS-
Published in Volume 21-Page 21-43, A, 5
Fourier reconstruction of the head by F. Hepp and F. Logan.
structure of a head 5
(see section )J). FIG. 1 defines the Cartesian coordinates in the frame of the target and the scanning device oriented to view the target from an angle φ. Here, (φ2L0) defines the position perpendicular to the beam path and the position along the beam path. For an accurate conversion, the attenuation measurements must have sufficient signal-to-noise ratio and the geometric extent of the scanned beam must be dense enough to give the desired resolution (φ , ti) It is necessary to fill the plane.

In the desired mode of operation to obtain high spatial resolution, the data are collected from M separate equally spaced observation angles in the range 0≦φ<π and N in the range −D/2<ti and D/2. is obtained with several separate equally spaced parallel impact parameters. where ■] is the projected diameter of the target. The recovered image is divided into pixels each corresponding to a size Δt. This Δt is Δt=D/N in the preferred embodiment. The resolution in reconstruction is the pixel size ΔL
, and the rotation between angles Δφ is
It must satisfy the condition that Δφ/2≦Δ, that is, M≧πN/2.

Data in this form can be used to reconstruct images of a plane of observation \() covering a region of grid F l of NxN pixels each corresponding to size t'. Therefore, ~ The target is mapped with ~N2 points using 2N' observations. For targets of unknown structure, we map this in the (φ, t) plane to reconstruct the image with a resolution corresponding to Δt, A range of degrees must be available. N' pixels of the section are normally displayed on a cathode ray tube, and these pixels are
81. rather than being physically separated. magnification +1 to
'1. - Be careful. However, for targets with known symmetries, fewer observations may be required, e.g.
It is only necessary that the projection from the observation of - is sufficient to reconstruct the image of the rotationally symmetric target.

In transmission tomography, the intensities of the incident beam (Io) and the detection beam (ID) are related to the attenuation along the path through the target.

If there is no scattering,
The relationship zl (1) holds,
where F(tl, - is the linear attenuation coefficient of the target and the integral over t2 crosses the beam path (see Figure 1). The quantity actually used in tomographic analysis is the optical Depth, i.e. defined as below the kite [projection J
P(φ.

ti).

P(φ,tt)4n[Io(φ,t,,)/IO(φ*
t1)] (2) This device sets both Io and ID to 4.
11. This measurement is performed by using a suitable calibration procedure.

The purpose of tomography is to convert −F(X+y) to its line integral P(φ,
L+)=/F d t to recover from the measured value of 2. Generally, in the inversion method, the damping coefficient F (X
*y) is 1 at a certain point as the linearly weighted sum of the measured projection data as follows:
The ear consists of

F(xvy)”22w(xvy;φm,tn) P(φ
m, tn) (3) n n Here, the weight W (X*y; φ, 1) depends on the position in within the target camera 1- and the scanning direction. No. (:3)
The equation shows that the scan data can be inverted at any point within the target to equal F(''+3').

The initial reconstruction method for medical tomography is based on the attenuation coefficient F
Reciprocal procedure to recover (X+y) (U.S. Pat. No. 3,7
78,614) is used.

This method starts with any initial trial solution,
Values for the projection data resulting from the trial images are derived by calculation. The difference between the determined data and the derived projected projection data is calculated by successively modifying the trial images until a satisfactory agreement between the calculated and observed images is obtained. It is used in

Then 1 convolutional back projection (after filtering) j projection (F
A far superior method called BP) was developed and used in tomography equipment (TEE in 1974).
E Trans, NucL, Sei, No. N5-2
Volume 1, pp. 21-43: t4astbrit,, A.

[The Fourier Reconstruction of the Head] by Shepp and B, F. Logan.
econstruejior+ of L111ead
5ection) J. and C.A.G. LeMay, U.S. Pat. No. 3,924,129).

Filtered back projection (FnP) has become a versatile implementation for commercial tomographic reconstruction.

This method can be used directly for both the fan-bee 11 data collection mode and the plane-parallel data collection mode.

Other reconstruction methods include the direct Fourier inversion method CDI”I
) is used. The essential advantage of this method over FI3P is that the number of mathematical operations required to invert the data to form an image of NXN pixel size is
In the P method, NXNXN Te & Ru (7) Te D for IC
PI method”Q Lt, N X N X log 2
(N). For example, 25 (
For an image containing EX 256 pixels, the DPI method inverts data 40 times faster than the FBP method. Also, when the image size is larger, the relative speed advantage becomes more pronounced.

The DPI method is based on a mathematical analysis that describes the two-dimensional Fourier transform of the target, and the one-dimensional Fourier transform of the projected image of the targeter 1- is identical. This result, known as the projection slice theory, applies to accurate continuous representations of targets and their projections. In practical applications, cut-off MIJ imaging only serves as a noisy discrete 11111 projection. The technology for inverting individual data using Fourier transform is W.
Rober6e and mouth, recently developed by Fiannery (patent application 1986/196475),
FIG. 4 illustrates the important steps involved in performing a Direct Fourier Transform (DPI). From the projection data, it is easy to determine the Fourier coefficients of the target along a series of discrete points arranged on a polar raster, i.e. at equal intervals along a number of sets of rays from the coordinate origin in frequency space. A coefficient is given at the point (see Figure 4). However,
In order to perform the reconstruction efficiently, it is necessary to know the Fourier coefficients along a number of sets of points distributed in frequency space in the Cartesian raster. Therefore, in order to efficiently use the Fourier method in tomography,
An interpolation procedure is required from the polar scale to the Cartesian raster. Inaccuracy in this interpolation procedure results in image defects and noisy amplification.

One key point in implementing DPI is to develop fast and highly accurate interpolation methods. Another important detail of this method is that the Fourier coefficients depend on the coordinate origin chosen for the spatial measurement. It is necessary to shift all transformations so that the origin corresponds to a common point. This is the origin.

The angle of rotation defining the viewing angle is given by the point at which the axis of rotation intersects the target plane. If the origin is not shifted, the phase of the Fourier coefficients will be scrambled in the inversion. Therefore, the basic steps of the 5 direct Fourier transform method are as follows.

(1) ID FFT: For a given angle of projection data, obtain a separate one-dimensional fast Fourier transform for the impact parameters. As a result, Fourier coefficients are given to the radiation in the signal space at equal intervals up to the maximum frequency consisting of the origin.

(2) Phase shift to target origin 2. Step (1) above
The phase of the coefficient obtained in step 2 is brought to match the position setting procedure that places the spatial coordinate origin on the axis around which the observation angle is rotated.

(3) Polar raster filling: repeat steps (1,) and (2) above for the projection data at each new viewing angle to establish the Fourier coefficients along a series of rays as shown in Figure 4. do.

(4) Interpolation to Cartesian grid: By interpolation, the values of Fourier coefficients at equally spaced points on a two-dimensional Cartesian grid are determined.

(5) Phase shift to the Cartesian origin I-: A phase shift is performed from the JJX point at the center of the target to the origin at the lower left corner of the rectangular area forming the image. This is a two-dimensional FF
required by the procedure for searching the origin at T.

(6) Inverse Fast Fourier Transform: An inverse FFT is used to transform from the frequency domain of the Fourier transform to the signal space that forms the image of the target.

In the basic form described above in steps 1-6: I)
The FI method can produce acceptable images only for targets with smoothly varying attenuation coefficients. However, due to many studies.

This method has been shown to produce unacceptable images for real targets, such as those found in medical applications, where there are sharp density changes between the skeleton and soft tissue. There is. A number of problems arise from the basic problem that the interpolation procedure is imprecise and furthermore, Fourier analysis introduces oscillatory defects when sharp discontinuities are encountered. These problems result in unacceptable distortions and defects in the reconstructed image.

Related problems also affect reconstructions obtained using the back projection jj method. In fact, a low-pass filter must be used to correct defects that may contaminate images formed by the back projection method without a filter.

Roberge and Flannery devised a means to improve the DPI method to the extent that it produces acceptable images that are qualitatively comparable to the results obtained with FBPA, yet retain the significant speed effects of DPI. (U.S. Patent Application No. 767, ≦)02). These stages are the fourth
In the figure, "Padding" and "Filter link force" are indicated.

Padding: Before step (1), the projection data is "padded" by adding additional data in both small and large impact parameters than actually observed. If the target does not extend beyond the m1FN region of the impact parameters, there is no need to estimate the values of the padded data; these are found to be exactly zero. Therefore, as required for optimal use of the 7-fold FFT without introducing approximations by padding and using additional known information,
Padding is used so that the number of data points in the projection is an integer power of two.

Padding also allows obtaining Fourier coefficient values at many points along the ray in a polar raster. Because additional resolution cannot be obtained.

The maximum distance from the origin of the frequency space to each point is not increased, but the number of points between the origin and the last frequency point is increased by the padding factor. For example, projection data with 256 impact parameters lit! If we FJ and then padded the data by adding zeros at 256 more points, we would get 257 values between the origin and the most illuminated point, whereas without padding we would get 257 values between the origin and the most illuminated point. There are 129 pieces. Furthermore, the midpoint value is approximately equal to the value obtained using higher order interpolation based on an analysis of the properties of the Fourier coefficients at frequencies intermediate between the individual values. To apply these interpolations, formulas at arbitrary intermediate points can be considered, but they are computationally expensive. Padding gives the same exact interpolation along the ray,
The FFT algorithm itself can be used to obtain points at many intermediate frequencies. The amount of padding can be adjusted to meet the needs of the particular target being analyzed. This step makes the interpolation procedure more accurate.

Filtering: A number of standard low-pass digital filters have been developed to eliminate or minimize defects corresponding to vibrations introduced by using Fourier methods. A standard low-pass filter, e.g. a Hanning filter with adjustable cut-off frequency
g) It has been found that high frequency vibrations in the image can be effectively removed by using a filter. This Hanning filter smoothly reduces the amplitude of the Fourier coefficients by a factor Y(s). This coefficient changes smoothly from 1 to O as the frequency increases from 0 to sc.

1/2(1)coS(7Cs/sc) If s<sc Y(s)E If <Os>se By selecting SC, the cutoff of high frequency changes can be selectively adjusted. (Note that the selection of low-pass filters in the FBP method serves a similar purpose.) Filtering can be applied to the Fourier coefficients at the poles or at the Cartesian grid or both. Filtering essentially corresponds to averaging the reconstruction over a length scale that is inversely proportional to the cutoff frequency. Viewed this way, it is readily apparent that filtering sacrifices resolution in order to improve the relative accuracy of the value of the xvQ attenuation coefficient in the reconstruction ↓. The cutoff frequency sc can be selected to selectively adjust the degree of smoothness.

Note that filter selection can further save computation hk. There is no need to evaluate the Fourier coefficients beyond the cutoff frequency or to perform an inverse Fourier transform on unnecessarily large silk coefficients. For example, 5 12 X , 51 corresponding to the maximum frequency 5 (512)
Assume that enough data has been obtained to construct an image lj on a grid of 2 pixels, but it is necessary to remove 1/2 of the frequencies in the filter stage. Therefore, the Fourier coefficients of the Cartesian grid only need to fill an array of 256 x 256 points, and the smaller silk 25
The inverse transform can be performed more quickly using f3×256 Fourier coefficients.

For data obtained using the above protocol, it only means that the reconstruction on a grid with point spacing (8X-Δy)≧1 is considered as one tsubo value.

Actually, in practical inversion methods, further smoothness is always introduced by a convolution or filter linking step that further reduces the effective resolution. The trade-off here is that low-pass filters increase signal-to-noise ratio at the expense of reduced resolution.

(, - ■ - No. Elements - Revised Part - Q ~ Consideration - (No. 11 When explaining the noise-to-noise ratio, we define ΔX in terms of the interval characteristics of the effective resolution allowed by reconstruction. For the FBP method, ΔX is given by the band ri used in the five convolution steps (1977 J, Comput.

As5L, Published in Tomagr, Volume ■, Page 64-7/1, 80, A, Chesler, S, J
, Riedcrer and N. J.

Noise due to photon-coefficient statistics in 1-calculated X-ray tomography by Mr. Pelc
otonCounting 5 statistics i
n Computed X-Ray Tomogra
phy) J). In the DFI law method,
ΔX is close to the reciprocal of the cut-off frequency used for filtering.

The propagation of noise from the data to the reconstruction depends on the unique properties of the target itself, which gives rise to correlations in the reconstruction, but the general tendency is that the algorithm involved in the reconstruction Behavior can be analyzed. Define ω as the ratio between the signal-to-noise ratio in the data and the signal-to-noise ratio in the reconstruction.

[σF/Fl/[(7P/Pcomiω(D, ΔX,
Δt) ('l) Here, all the projection data are
Mark i1'! If it is normally distributed with deviation σ['', it is a typical size P that includes noise that can be explained,
, σF are assumed to be typical values of a linear damping coefficient and its standard deviation.

The analysis to determine ω is as follows for the F13P method:
D., A., ChesLer, S., ,1. Ri
edsrer and N. J. r'elc (1977 J. Comput. As5
"Noise due to photon coefficient statistics in computed X-ray tomography (Noise due to + oton Co
untin5 Sshiatistj, cs j, n
Computed X-f? ay Tomoira
phy) J) and for the DPI method, see J Roberge and B, Flannery.
(Patent application 1986/1964)
75). For both algorithms,
The amplification factor can be expressed as follows.

However, B is a first-order coefficient determined by details of the algorithm. Usually, inversion is applied near the resolution limit ΔX=ΔL accommodated by the scan data, so Equation 5
is the number of pixels per surface in the reconstruction K = I)
/∆X, and therefore ω is v' or ∆X;
C is determined. Imaging a target with high resolution while maintaining a constant signal-to-noise ratio requires more observations and greater precision.

For example, to double the resolution with a certain accuracy, 4
Not only must the ill'l determination be made along twice as many paths, but the image to noise ratio for each observation must be increased by V2°. Furthermore, Equation 1F) shows that the method of averaging the reconstruction (ΔX>Δt) reduces noise. For example, a reconstruction that averages 4 single points (Δx=281) reduces the noise by half.

At the beginning of this section, it was stated that noise in tomographic images arises from two sources. The above discussion quantified the degree of noise amplification introduced by the inversion method. Here, we consider the uncertainty of the data observation l-.

Observation of data 1. The uncertainty of
arises from the noise tg injected by the detector. Accuracy ρ
Two related measurements, the detection quantum efficiency and the detection quantum efficiency, can be used to quantify the amount of noise that the detector adds to the x-ray signal. These are defined below.

and So Si j where S means the integrated signal 7, σ is the integrated noise RM S, and the letters 0 and j appended are the outputs of the detector, respectively. and input X-ray No. 411.

Here, let S and σ be the number of X-ray photons in units L7, and the photon source follows Poison's radiation statistics, that is, the light f is emitted at a constant average rate at random times. Then, the two 811 constants are uniquely related by 1 to the following equation.

Here, the number of input X-rays b required to perform dI11 constant
-X is called the dose1. - In general, σ and no represent the dose S
l-do.

1 #1 variables (r, j, -1-1
...mouth). A typical r,i is the region over which the signal is integrated, the integration time, and the dose ゛4(, etc., ρC8x+r,>) is plotted as 1.
It is useful to derive the plot of I'1 (Sj, Rj), and vice versa. Therefore, the selection of a+II constant values if and Ij in 2-) is mainly determined by convenience.

Detection quantum efficiency has been used in the electro-optical industry as a characteristic of imaging devices for many years.

All of the efficiency of two-dimensional X-ray detector! Fixed value and 1. How to use it is S, M, Gruner and J, R,M.
ilch (i982 Tr.
ansaetiori Amer, Crystallo
graphie As5oe+, page 149), which accounts for the (=1) additive noise introduced by the detector for an ideal detector. The statistical output is defined to be equal to the statistical input belief, i.e. So Si −1−−=−−−(9) σ Oσ j In this case, H=1. The degree to which 1 is less than 1 indicates how far the detector falls short of the ideal.In the case of electro-optical X-ray detectors used in micro-tomography systems, 8 is 0. It is preferable that !l is larger than 0.5 (it is more preferable if !l is larger than 0.5.

If the detector is exposed for a time T, the total number of photons incident on the detector is Nl)Il)'r, and the expected standard deviation σN for N photons is If the number of should be Ω, then the number determined by 81 should be n
= ritoτ[)lnko. However, 2. τ is a normally distributed random variable. At this time.

The fixed projection also avoids Gaussian noise based on the equation:

p=p + τ [N1] No.11
(1,0
) jJ l, , p is the average expected value.For a typical target of diameter 1 and average linear attenuation coefficient 1·゛, this allows the assembled result to be the signal-to-noise ratio Determine the number of photons per detection required to form an image of the current pixel force by F/σF. Formula] and 2 are I) = FD and ND = No
It shows that exp(-F D). Therefore 1
Recall that, as is clear from equation 4, the amplification factor ω (see FIG. 5) essentially depends on the image size of the pixel.

Therefore, for a given pixel size and precision image σF/F, No depends only on the optical depth FD through the target.

φ-1 key - ■v key 1ht1 As shown in FIG. 5, No means optical depth F1. )=2
is minimum if . Although the graph of FIG. 5 is for the signal-to-noise ratio σF/F=O,O], similar properties are seen for all useful signal-to-noise ratios. This occurs because the actual signal depends on the number of photons absorbed by the target. Large optical depth FD >>
In the case of 2, No increases because some photons are transmitted.

For small optical depths FD<<2, No increases because some photons are absorbed.

Further substituting ω from Equation 5 to Equation 11, the surface protrusion or D/Δ
When the image is reconstructed into a grid of pixels of
The target of relative accuracy a F'/F = O, OL
In order to image the 100 x J-00 pixels, the number of pixels per pixel in the detector is
10' incident photons are imaged (when ΔX=Δt,).

In the explanation in section 2 of ``Continuously gg i; It has been shown that, along with the beam energy, it is highly dependent on the optical depth (FD), which varies widely. The optimal mlt'A condition occurs when FD=2. Optical depth range C, FD=0.2
Or xr o = 6te. A more preferred range of optical depth for sample observation is FD=0.8 to FD=3. This range of Il1 conditions is
It can be obtained by changing the x-ray beam energy or the sample size. The sample size is
If the number of pixels across the object is N and the spatial resolution of each pixel is Δt, then D=NΔt.

In the tomography system disclosed herein, 7N ranges from 20 to 5000 pixels. A more valid range for N is 100 to 1000 pixels. The maximum resolution obtainable with the tomography system presented here is 0.5 microns, and the minimum resolution is approximately 5 m. In a more preferred mode of operation, the device comprises: 1. It provides a resolution (Δt) in the range of -100 microns. The minimum sample size accepted by this device is 1o pixels x
0.5 micron = 5 micron, more preferred minimum sample size is 100 pixels x 1 micron = 10
It is 0 micron. Optimal *a condition for samples equal to or larger than the minimum sample size (FD=
In order to operate the device near 2), the linear X-ray absorption coefficient (F) through the sample must be less than 10100 O', and more preferably less than 200 am-'. To obtain a linear X-ray absorption coefficient (F) below these limits, the X-ray energy must generally be greater than 1 keV, and more preferably greater than 5 kaV. These limits are described in the X-ray Handbook (Ilandbook) edited by E. F. Kaelble.
-rays) (MeGra, New York, 1967)
It is derived from the mass attenuation coefficient of X-rays, as found in a number of specific references, including (Published by Wtrill 1look Co.).

Sources of X-rays with energies of 1 to 5 kaV or higher include synchrotrons, rotating anodes, X-ray generators and
It is equipped with a tube. In order to use a double radiation source in the preferred embodiment of the micro-tomography system described herein, the radiation beam generated must be in a plane parallel to the predetermined spatial distribution. The nature of the construction algorithm requires parallelism of the radiation, as it must pass through only one row of pixels in the sample. Thus, the main ray (FIG. 3) passing through the sample must be followed by another ray passing through another point of the sample and S7 with the following precision:

ΔL α< −X 2 (+3) where α is the maximum angular deviation of the two principal rays passing through separate points in the sample, Δt, is the minimum resolved pixel of the image, and D is the This is the distance traveled through the sample. The divergence of two rays through the same point in the sample is limited by the penumbra drawn on the detector.

To maintain the desired resolution, the divergence of two rays passing through the same point in the sample must be of first order.

Δ L α′≦−−−−−(1/I) where α′ is the divergence angle of two rays passing through the same point in the sample, and S is the divergence angle of the two radiations passing through the same point in the sample, and S is the angle of divergence of the two radiations passing through the same point in the sample, and S is the distance.

Thirteen different collimation techniques can be used to obtain such parallelism and beam divergence for a rotating anode and x-ray tube. Collimating action can be achieved by using either a monochromator, a physical collimator, or a distance that limits the diagonal beam dispersion through the sample. The degree of collimation increases as the distance between the sample and the radiation source increases.3 Let S be the effective source size in the X-ray generator F.
genel"a jor, the distance the sample must be separated from the generator D ger+erator
(determined from Equation 14) is determined as follows.

Δ Also, the Collineau 1-operation is an X-ray mirror with grazing incidence,
This can be achieved by placing either a stacked synthetic multilayer monochromator, a flat crystal monochromator or a curved crystal monochromator in the beam. The arrangement of these X-ray optical elements is expressed by formula 1
4 and 15, while maximizing the flux through the sample as much as possible.

The choice of collimation method to use for the rotating anode and x-ray tube is determined not only by these requirements, but also by the spectral purity required for accurate image reconstruction. Spectral purity in a beam collimated by distance can be obtained by filtering the radiation. For example, a nickel filter can be used with a Cu x-ray tube to improve spectral purity. In the case of synchrotron radiation, the angular distribution of the radiation generated by the ring is sufficiently small that no additional collimating action is required in many cases. However, due to the high brightness of synchrotron radiation sources, it is usually preferable to use a monochromator to improve the spectral purity of the radiation.

11 reeds, 2 people, 1 rod - chivalry figure - Fu Q, Kionoka Jun.

There are a number of methods for directly recording two-dimensional X-ray images with a spatial resolution of microns to 10 microns. Two-dimensional X-ray images can be recorded directly on film with a resolution of 1 to 10 microns. Disadvantages 1; - The silver halide grain structure of the film introduces noise, making the maximum signal-to-noise ratio obtained unacceptable for tomography. Two-dimensional X-ray images can also be obtained with direct impact solid-state detectors such as charge-coupled devices and programmable read-only memories.

Again, disadvantageously, commercially available solid state devices tend to degrade after approximately 0.01 x-rays have been collected per pixel. To overcome these problems and record two-dimensional images with high resolution, an electro-optical X-ray detector is used that uses components developed for amplifying and recording optical images.

A typical detector, as shown in Figure 6, has four elements:
That is, it is composed of an energy converter, an optical gain element, a device for enlarging or reducing an image (ie, an image formatter 1 to a changing device), and a reading device. The function of the energy converter is to distribute the energy of X-ray photons, among other easily handled quantities. Typically, it is a fluorescent screen that generates visible light. In the case of be done. Typical gain elements are magnetic or electrostatic focusing image intensifiers or microchannel plates. The format of the output from the gain element (or phosphor) usually has to be changed before being recorded with an image readout device. The Forman 1- modification is necessary because the image from the gain element or phosphor is usually different in size from the image from the readout device. Therefore, it is usually necessary to combine the two images through a magnification or reduction mechanism. Enlarging or reducing the image can be performed electronically or optically. Changing the optical format is done using lenses or fiber optic coupling. In the case of electronic means, the format change can be performed by electrostatic focusing in the image intensifier. The readout device 1't is the most diverse element in electro-optical detectors. These range from vacuum tubes and solid state detector arrays to resistor devices (some of which are listed in Figure 1), converters, converters, A large number of different configurations of the 5-element and readout element are available, most of which can form a 1-1 visible image.

Only a few are suitable for use as the mouse detectors required for tomography. There is a limit to the number of electro-optical detectors that can be used for tomography due to the following detector functions. (]) quantity - (detection efficiency, (2) confidence-dependent background in the recorded image, (:3) effective dynamics of the detector, (4) spatial uniformity of response ( (quantum uniformity) (5) positional linearity (geometric linearity).These performance criteria require a certain range of The scope of applicability of these performance criteria is discussed below.

In the following description, the performance criteria apply to active pixels (ie, non-defective pixels). There will be a small number of defective pixels in any device. These pixels are those whose quantity f@ rate is reduced by a factor of about 10 or more from the average value. These are referred to as ``r-bad pixels'' and are specifically excluded from the following description. The number of bad pixels that can be present in the detector is limited such that no more than 20% of the pixel columns extending perpendicular to the axis of rotation of the system contain bad pixels in the sample image. In a preferred embodiment, no more than 1% of the pixel columns extending 1υ perpendicular to the axis of rotation of the system contain bad pixels in the sample image.

The required quantum detection efficiency of an electro-optical detector is determined by considering the signal-to-noise ratio of the signal that is constructed.

As noted above, the number No of lights P required to reconstruct an image with a signal-to-noise ratio of is 1/! l” (12th
(formula). Therefore, as January decreases, the time required for a tomographic scan becomes significantly shorter. Month is 0, 0
5, the time required for tomography scanning is
Increased by a factor of 400 compared to an ideal detector. The reconstructed image is accepted as 4@l; noise ratio (
σF/F ~ 0.1 to 10%) <7? Since a very large number of photons are usually required to detect
is preferably greater than 0.05 to obtain practical exposure times. In a preferred embodiment, Jl is Q,
5, the exposure time is larger than that of an ideal detector (/l=,1
) is within a factor of 4 of the time obtained by The method for determining the detection quantum efficiency of an electro-optical X-ray detector is 8.4.
゜G r II n e r and 1. R, Much
By Mr. ]982 Trjins+1etions
of the Allerican Crysf,
all, oHra-pliie As5ociat, i
Published in o++ volume 18. .

Signal-dependent background is present in all electro-optic x-ray detectors due to scattering of radiation during the energy conversion process and detection by the electro-optic readout. The scattered radiation that produces the signal-dependent background has a component that is spatially correlated to the signal and a component that is spatially uncorrelated to the original signal. The optimal observation condition is that the optical density to Targetera 1 is approximately 2 (transmittance 13.
5%), the total signal-dependent background (correlation plus uncorrelation) at the pixels covered by the sample should be less than 10% of that obtained from the undattenuated X-ray beam. In a more preferred embodiment, the total dependent background (correlation + decorrelation) at the pixels covered by the sample is less than 2 parts of the signal derived from the undattenuated x-ray beam. The signal-dependent background of an electro-optical X-ray detector can be measured using various techniques. A simple technique is to cover t of the incident X-ray beam with a mask that completely absorbs 5 X-rays. In principle, there should be no signal in the area covered by the mask, and the signal level detected in the area behind the mask is clearly due to the correlated and uncorrelated background. Yes, if you move away from the edge of the mask.

Although the 411-dependent background changes slowly,
This is mainly due to uncorrelated signal levels. Near the mask M(=J, the signal-dependent background 1-
The spatial variation of is a direct measure of the correlation number scattering. Therefore, the signal-dependent background near the edge of the mask can be spatially (by filtering the 41 function)
5 Correlated and uncorrelated signal-dependent backgrounds can be quantified.

Test 11) The effective dynamic range of t is defined as the ratio of the maximum signal level recorded during the exposure to the sum of all (No. 79 dependent background and noise sources). (not a scientific process). The effective dynamic range is maximized by minimizing the noise from the C source rather than by statistical processing of the counts. Noise sources other than statistical A1 processing of counting include, for example, read noise and dark noise. Dark noise accumulates over time in the absence of an input signal and severely controls the maximum exposure time.

Readout (noise is added indefinitely to the signal during readout of the sensor used to record the optical image. The readout device with the lowest dark noise and readout noise is the charge-coupled device (CCD). The sensor readout noise and dark noise are at least smaller in magnitude than vacuum tube TV sensors such as vidicons. Such CCD sensors are the most preferred for the construction of contrast electro-optical X-ray detectors. [Under conditions 9, a signal of 13% of the unattenuated beam must be recorded, so the effective dynamic range of the detector is] 0
Must be bigger. In a more preferred embodiment, the effective dynamic range of the detector is greater than 50.

Most detectors do not have uniform sensitivity over their area. If response non-uniformity between adjacent pixels is not corrected out from the data, the reconstruction process will be flawed. Stable 1 between adjacent active pixels
0% non-uniformity can be easily calibrated out from the data. However, one calibration method becomes useless when the non-uniformity between adjacent active pixels is greater than 75%. The deficiencies of the calibration method for deeply modulated non-uniformities between adjacent pixels are due to detector stability considerations. Vibration, spatial drift, and time-dependent drift in sensitivity all reduce the stability of the detector, making deep modulated inhomogeneities between adjacent pixels not easily removed from the data. Therefore, the response non-uniformity between adjacent active pixels is preferably less than 75%. There are also limitations on locally averaged sensitivity changes over the detector surface. The locally i1L-averaged sensitivity is defined as the average value of the sensitivities of the active pixel and its immediate neighbors. Considering the signal-to-noise ratio, the locally averaged sensitivity is.

Preferably, it does not vary by more than a factor of 2 over the entire surface of the detector. As the locally averaged sensitivity changes over the surface of the detector, the signal-to-noise ratio at the detected Preferably, only a minimum number of X-ray photons are required to reconstruct the target to noise ratio.When the locally averaged sensitivity varies across the detector, the required number of X-ray photons is reduced to a minimum average corresponds most closely to what would be expected for sensitivity. The region of locally averaged minimum sensitivity generally corresponds to a defect in the detector. Therefore, minimizing the exposure time also To optimize the performance of the imaging system I, it is desirable that the variation in locally averaged sensitivity across the detector be less than 10, and in a more preferred embodiment less than 2.

When the geometric distortion from the true positional linearity of the pixels in the detector is too great, the reconstruction algorithm cannot adequately represent the original target. Defects result from uncorrected positional nonlinearities. This is because the impact parameter 1+ (FIG. 3) is systematically erroneously measured. Such erroneous measurements occur when the Cartesian grid imaged by the electro-optic detector deviates from linearity by more than one pixel position. If the deviation from true positional linearity is small, a distortion map can be applied to the data to correct the shock balance to obtain the data on a true Cartesian grid. This correction causes the jet to become distorted when the deviation from linearity is more than O pixel positions across the surface of the detector. Therefore, the deviation from true geometric linearity is Preferably smaller than 10 pixels in the recorded image. In a further preferred embodiment, the impact parameters are measured accurately at the surface of the detector and the maximum deviation from geometrical straightness is in the recorded image] -becomes less than a vixel.

12 The effect of the detector as noted: (]) quantity - r ~ detection efficiency, (2) signal-dependent background in the recorded image, (3) effective dynamic range of the detector, (4)
Spatial uniformity of response and (5) positional linearity severely limit the number of electro-optical detector types that can be used in tomography systems. Other important detector actions are:
(1-) Linearity of response vs. strength, (2) Stability of response vs. strength, (3) Spatial resolution and (4) Limits on count rate. The preferred range of operation of these detectors is discussed below.

The detector response versus x-ray intensity must be linearized to accurately measure the projection defined by Equation 2. The measurement of projections forms the basis of the tomographic inversion method. Projection ilt! can only be obtained accurately when the detector response is linearized.

The response of the detector versus the dose of incident x-rays can be linearized by a suitable calibration method when the deviation from linearity is less than 25% over the effective dynamic range of the detector. When the deviation from linearity exceeds 200% anywhere in the effective dynamic range of the detector, calibration techniques cannot adequately correct the data. The reason why the response of the detector cannot be linearized sufficiently when the deviation from linearity exceeds twice is because the response generally changes with time. In all electro-optical detectors, the response changes slightly over time.

This is due to the fact that the energy converter plate 1- is damaged by the radiation and the amplification of the gain element and readout device changes in a time-dependent manner.The time-dependent change in the detector response is The rate of change in the detected quantum efficiency between two consecutive calibration frames must be such that it is less than the desired signal-to-noise ratio. A calibration frame 1z is collected to measure the intensity Io of the incident X-ray beam, which is often obtained once per angular rotation of the target.
Sometimes it is obtained once per 180° rotation to Targera 1). To obtain a calibration frame, the sample is removed from the x-ray beam and the unattenuated x-ray intensity is measured.

The spatial resolution of the detector is determined by the formatting technique used to connect the main mechanism, energy converter plate, and readout device. The ability to adjust the detector resolution by simple format 1 changes is
This is an important advantage of electro-optical X-ray detectors over the common scintillation detectors used in tomography.

Formats] - can be changed using electronic or optical systems. The simplest of all these formatting techniques is a lens system that reads out the fluorescent screen and couples the light directly into the device without an intervening image intensifier. This system is well suited for enlarging images formed on high-resolution fluorescent screens;
Spatial resolution comparable to the wavelength of light (approximately Q, 5 microns) can be obtained. This limit in spatial resolution can only be achieved with very high resolution fluorescent screens, which consist of a honeycomb array of fluorescent plugs with dimensions comparable to or smaller than the desired spatial resolution. By forming the phosphor screen as a honeycomb array of 8 individual phosphor cells (cellular phosphors), which are ideally formed as can. In this case, the resolution limit is determined by the aperture of the stacked lens system of 0.6 to 0.8, which gives a final spatial resolution of about 0.5 microns. Lower resolution can be obtained by lowering the magnification of the lens system, and the coarsest resolution can be obtained by using a reduction lens system rather than a magnification lens system. Easy 711!
7) There is a limit to the maximum storm intensity that can be obtained with a Nons coupled electro-optical detector.

The effective reduction range constraint is due to the inherent limitations in light collection efficiency from the lens. According to the luminance theory of light, the maximum light collection efficiency Le of an ideal lens for observing the Lambertian luminance distribution from a phosphor is expressed as Le = M''(M<
<1). The actual L/lens light collection efficiency I,e is:
considerably smaller than this limit. The light transmission efficiency of most lenses is Le = (NA)2 or Le = (M /
2 f )''. For large demagnifications, this limit indicates that the lens transmits only a small fraction of the number of photons generated on the fluorescent plate. If only a small fraction of the number of photons reached the readout 7 device, the detection m-on efficiency decreases significantly. This decrease is large enough to take it outside the preferred range of D>0.05. Theoretically, , the constraints seen on shrinkage in lens coupling format changes can be overcome with optical fiber coupling.The light collection efficiency of the lens is much lower than the theoretical limit imposed by the luminance theory of light, but the A reduced bundle of fibers approaches this limit strictly due to its high numerical aperture.The light transfer efficiency of a reduced optical fiber bundle is determined by its input (output) numerical aperture being
If ut(N Aoutput), L e = (
N A 1nput) 2 or 1, e= (M
The improved light collection efficiency of the fiber optic bundle results in a higher detection quantum efficiency than that obtained with a similar format modification of the lens-coupled system. However, the effectiveness of the fiber-optic coupling The flexibility is limited by the distortion of the fiber backing that occurs during the manufacture of the reduced bundle. Typically, the reduced bundle is drawn from an optical fiber blank that is formed by sintering a bundle of individual fibers. In this process, Both local and extended defect structures occur.

Local defect structures include distortion of the fiber backing within the fiber splint 5 bundle and misalignment defects in the east backing. Extended defect structures that occur during the extraction process cause bottle cushion and barrel distortion in the image sent through the reduction bundle. In many cases, these geometric distortions lead to positional nonlinearities greater than 10 pixels and are therefore unacceptable for tomography systems. To overcome the format reduction constraints of the lens system and to be immune to distortions introduced by optical fibers.

It is advantageous to incorporate an intensifier stage between the energy converter plate (phosphor) and the readout device. Available intensifiers include microchannel plates and magnetically and electrostatically focused intensifiers. All these intensifiers can be used in such a mode that the X-ray image is sent without changing the Format 1 to the lens that reduces the image. In this mode, the intensifier advantage () overcomes the limitation on detection efficiency imposed by the effective light collection efficiency of the lens system used to reduce the image. Also, when using electrostatic focusing, image reduction can be conveniently performed within the intensifier. This technology has advanced significantly with the development of the SIT tube, and 2:]- reductions are now widely used. The custom-made electrostatic focusing intensifier is 2:1. However, the maximum reduction ratio is limited by the aberration 1: of the electrostatic focusing lens.

Electro-optical detectors provide a flexible and modular solution to soft x-ray detection. The second explanation below is based on a certain company that specifies the configuration of a detector that can be used in a tomography apparatus.

When designing an electro-optical X-ray detector for tomography, it is preferable to use a charge-coupled device (CCl') as a readout device. Recently, solid-state charge-coupled devices have been developed, which provide superior contrast imaging.
Optical sensor technology is becoming popular. The CCD sensor is
It is attractive in the field of X-ray detection because it offers significant improvements over other TV sensors such as vidicon, isocon, and oncon. The readout noise and dark noise are at least smaller in magnitude than in the case of vacuum 'ii' TV sensors, and the detection quantum efficiency of the sensor is improved;
be done. Commercially available C (shoulder) sensors operate at temperatures below -75"C and have a readout noise of 50 M1 when
children/pixel and the dark noise is 5" per minute-pixel." In addition, the CCD sensor is
Largest dynamic range (saturated signal/rms readout, noise) of all electro-optical sensors
shows. The saturation (i) of many CCD sensors reaches 10' electrons/vixel1, which gives a dynamic range for signal detection of approximately 1.0'. ), , cannot saturate to the 3Δ objective, problems arise when the saturation level is approached. For signal levels below the saturation level, CCD sensors exhibit very high linearity of response per degree of optical input. Furthermore, since the photosensitive element is fixed to the CCD chip, geometrical distortions associated with electronic beams read out from vacuum TV tubes can be avoided.

↓ An X-ray microtomography system configured with an electro-optical X-ray detector is shown schematically in FIG. X
The radiation is generated from a Cu focused x-ray tube manufactured by Philips Electronic Instruments. The tube was aligned to be a point source of X-rays to the experimental setup and operated with a power of i, 5 KW. X
The distance between the source and the sample was chosen to be 20 mm.

The tube was positioned so that the line connecting the center of the point xH source and the center of the portion of the sample to be imaged was perpendicular to the axis of rotation of the sample. The separation (t, ake-o-ff) angle of the sample relative to the anode shadow projected by the tube is about 5°, which gives an effective size of the x-ray source of about 750 microfron x
The diameter was set at 50 microns. The sample was held in a rotatable goniometer within the x-ray beam. The rotation stage is controlled by a step motor.

This step motor was capable of moving steps in increments of O,,01'. For a total rotation of 36 Q' of the rotating stage, the swing of the rotating shaft was less than 〇-5 radians.

In the case of such minute shaking, the five rotational axes remain fixed in space and do not move with the rotation of the sample. In the goniometer, the sample was aligned so that it remained within the lateral field of view as it was rotated. The sample attached to the goniometer is
A 750 micron diameter hollow glass tube was packaged with approximately 200 micron diameter silica spheres and a 10 micron tungsten wire extending along the axis of the tube. A phosphor conversion plate 1- was placed at the rear of the sample 211F. The fluorescent plate was a 5 micron thick layer of evaporated CsI doped with thallium. The light emitted from the phosphor conversion plate was imaged by a photographic lens onto a charge-coupled device. In order to maximize the detection quantum efficiency, the f number -]
, , , 4 photographic lenses were selected. The magnification of the lens system was adjusted so that the 8 micron pixels of the 5 fluorescent plates 1-1-1 were magnified to 30 microns, which is the pixel axis of CCl). C C1,) used in this experiment is 33 F3
R C with X 540 active pig cell elements
It was ASID-501. This means that in such a geometry of 7 instruments aligned so that the columns of the instrument are parallel to the axis of rotation of the sample, the angular deviation α of the two seed rays passing through separate points on the sample is 2°. 5XiQ"-" radians, which is 2ΔI:; / D = 4, 26 X10-2, as required by Equation 13.
less than radians. The angular deviation α' of two rays passing through the same point in the sample is 3.75 x-ray o-' radians, which is as required by Equation 14:
Δt/S=4. less than X1.0-' radians. Therefore, data is obtained in multiple stack planes in the 1r11'L row mode. The number of evenly spaced ma angles required for data collected in this mode, M, defines the number of equally spaced discrete parallel impact parameters across the target as N
given that.

yc must be greater than N/2. The target is 750 microns in diameter, the span of each pixel is 8 microns or 4'l+, and the number of impact parameters across the target is approximately 85. To satisfy the protocol for data collected with the NIZ plane-parallel mote, 240 equally spaced viewing angles (Δφ) were selected to scan the target.

The detection quantum efficiency of the detector was determined to be 0.75 and the total signal dependent background was found to be less than 2% of the signal from the undattenuated X-ray beam. This detection M
The effective dynamic range of the +i configuration was a factor of 80, and the maximum non-uniformity of response between adjacent active pixels was less than 5%. The lens coupling fluorescent plate 1~ to the CCD did not introduce any detectable image distortion, and the maximum deviation of geometric linearity was less than 1 pixel in the recorded images. By exposing for 3 minutes at each observation angle, data with a signal-to-noise ratio of 0.3% was detected J1+
Obtained with +. Data from each observation angle is digitized with a 6-bit analog-to-digital converter.
processed by a computer and recorded on magnetic tape.

The image was then reconstructed using a direct Fourier inversion method (DI''I) with a number of operations of N2 rather than N3.

FIG. 8 is a cross-sectional view at the point shown in FIG. 1, showing projection data obtained by one WJID of a hollow glass tube. It is clear that the 200 micron glass sphere is clearly visible in the cross-sectional image as well as the 10 micron tungsten wire. The spatial resolution obtained with this reconstruction is more than 25 times better than that obtained with conventional medical CAT scanners.

[Brief explanation of drawings]

FIG. 1 is a diagram defining the path passing through the observation plane of targeter 1~, in which the path I7 between source S and detector
The angle φ with respect to the fixed Cartesian axis (XIIXZ)
This is the time to show what is determined by. Figure 2 shows the observation path typical of a typical medical CT scanner using the "fan beam" observation mode, where individual observations are made at points (φ) as shown in the top panel. FIG. 3 is a diagram showing how a route is formed; FIG. 3 shows the observation path in a scanner using the j Np row beam J ml observation mode, in which a parallel collimated radiation beam is directed to a target laser in a number of lamination planes and with an impact parameter of S. Simultaneously hit, the target is rotated to perform L observations at various observation angles φ, resulting in separate observation paths in one plane to the point (1°φ) as shown in the top panel. This is a diagram. FIG. 4 is a diagram illustrating the steps of the direct Fourier inversion method, showing the relationship between targeter 1- and its projection in signal space, and showing the Fourier transformation of the target in polar and Cartesian coordinates. FIG. 5 shows the number of incident photons required per projection measurement into the prong 1 as a function of the optical depth F l) passing through the target laser 1, and in the case of position counting statistics, To give sufficient accuracy to reconstruct the tomographic image, with relative accuracy σi:/F = 0.01, N
If o incident rays r are required and ω2 is taken as the ratio between the relative precision of the image and projection data and directly Fourier inverted a 100×100 pixel image, then ω2 is ]0
It is a figure showing that it is larger. FIG. 6 is a diagram showing the general components of an electro-optic detector for X-rays, listing specific components for the energy converter, optical gain element, format converter and electro-optic readout device. It is. FIG. 7 is a schematic diagram of an X-ray microtomography device configured with an electro-optical detector. Figure 8 shows one *(Ip) of a 750 micron diameter glass capillary in its second part;
Loaded with 00 micron silica spheres and a single-chamber tungsten wire running along the length of the tube, the spatial resolution of the image (Ip) is approximately 10 microns, and three wires drawn across the upper image. A cross-sectional reconstruction of the tube at the position indicated by the line is shown at the bottom of the drawing, the cross-section of the silica bead can be easily seen in the image, and a small dark spot on the left side of the tube cross-section is 10 microns tungsten. It is a diagram showing a part corresponding to a wire. φ...Observation angle 05 kan, 7...Impact parameter drawing 111'r (inside '1.・no good history) FIG 1 Position angle Φ FIG, 2 1 Go to the end Required human flux vs. optics Depth electro-optical X-ray detector - Energy converter - Phosphor (optical and photoelectronic) - Gain element - Microchannel plate - Electrostatic convergence intensifier - Magnetic convergence intensifier - Format change - Optical lens - Optical fiber bundle - Image reduction intensifier Electro-optical readout device - Image orthocon - Image isocon - Vidicon - Vidicon-5IT
-Silicon diode vidicon-C1l)
- Resistive anode CCD FIG, 6: Mouth FIo, 8

Claims (10)

[Claims] (1) A device that forms a tomographic image of an object by irradiating a plurality of collimated radiation beams onto the object, including: a) a contrast detector that measures the attenuated transmitted radiation after passing through the object; b) means for obtaining projection data from said transmitted radiation; and c) means for calculating a reconstructed image of the attenuation coefficient of said object from said projection data. A device characterized by comprising: (2) The apparatus of claim 1, wherein the contrast detector is an electro-optical detector, further comprising an image format changing device and a readout device. (3) The device of claim 2, wherein the active pixels of the detector have a detection quantum efficiency greater than 0.05. (4) an active pixel of the detector has a total signal dependent background of less than 10% of the signal from the unattenuated x-ray beam and an effective dynamic range of 1
3. The device according to claim 2, wherein the device is greater than 0. 5. The apparatus of claim 2, wherein the active pixels of the detector have a response non-uniformity of less than 75% between adjacent active pixels. 6. The apparatus of claim 2, wherein the active pixels of the detector have a geometric linearity deviation of less than 10 pixels of the recorded image. (7) The device according to claim 1, further comprising a radiation source. (8) Claim 7, wherein the radiation source is an X-ray.
The equipment described in section. 9. The apparatus of claim 7, wherein the radiation source is firmly connected to the electro-optic detector. (10) The apparatus according to claim 9, wherein either the object or the radiation source rotates about an axis.