JPH0446144B2 - - Google Patents

Description

[Detailed Description of the Invention] [Objective of the Invention] (Industrial Application Field) The present invention relates to an image reconstruction method used in an X-ray CT scanner device, etc., and in particular, to improve spatial resolution and increase the amount of calculation. This invention relates to an image reconstruction method that suppresses

(Prior Art) This type of X-ray CT scanner device will be explained using a typical third generation device as an example.
That is, as shown in FIG.
Inside, an X-ray source 11 that irradiates a fan beam XB and a detector 12 in which a number of unit detectors are arranged side by side in an arc shape are arranged facing each other with the subject Pat in between.
The pair of radiation source 11 and detector 12 maintains a facing relationship and is rotatable in the direction of the arrow about a center of rotation (not shown).

The data acquisition device 20 integrates the X-ray transmission data detected for each X-ray pass by each detector channel of the detector 12, and then converts each into a digital signal and makes it standby for subsequent processing as digital projection data. It's getting old.

The image reconstruction device 30 takes in projection data from multiple directions regarding the subject Pat from the data acquisition device 20, and calculates the degree of X-ray absorption at a position in the X-ray transmission direction of the subject Pat using, for example, a filtered back projection method. It is now possible to generate tomographic images that reflect the That is, the preprocessing unit 31 performs processing such as direct current component correction and X-ray intensity change correction on projection data, and in this case, fan beam data is rearranged into parallel beam data and signal processing is performed. In the apparatus, this rearrangement processing is assumed to be performed in the preprocessing section 21. The convolver 32 mainly performs blur recovery filter processing and centering processing, and the preprocessing unit 21 performs convolution (product-sum calculation) processing using a blur recovery filter for each projection direction based on the corrected projection data. and performs centering processing. The back projector 33 back-projects the convolution projection data for each projection direction by the convolver 32 onto the image memory 34 and superimposes the data, and generates a tomographic image on the image memory 34.

This image conversion device 40 includes an image memory 34
CT values in a desired range for the above image are converted into a grayscale image video signal and provided to a display device 40, where a tomographic image based on CT values in a desired range is displayed.

Also, a system controller 61 that performs overall control as a control device 60, a console 62 that gives operator instructions to the system controller 60, an X-ray controller 63 that performs X-ray control on the X-ray source 11 under the management of the system controller 61, A scanner controller 64 is provided which controls the rotation of the gantry 10 under the control of a system controller 61.

The above is an example of a configuration including one detector 12, but as shown in FIG. 19, two detectors 12A, 12B ( There is a configuration in which detector A and detector B) are arranged side by side. Here, the arrangement relationship between detector A and the detector is 1/4 of the width of one detector channel.
It is assumed that the Then, as the data collection device 20, a first
Equipped with a DAS (Data Acquisition System) 21A and a second DAS 21B corresponding to the detector B,
As shown in Fig. 20, by rotating the object Pat by 180°, it is actually a fan beam, but to make the explanation easier to understand, we can express it as a parallel beam. As shown in Fig. 21, the opposing data A ( Time t _A ,
It is possible to obtain data B (time t _B , number of samples N points).

On the other hand, conventional image reconstruction methods are as shown in Figure 22a.
There are two methods: one shown in Fig. 22b, and the other shown in Fig. 22b.

The method shown in FIG. 22a is generally employed in a configuration using a series of detectors. That is, the multidirectional transmission data collected in step P1 is subjected to preprocessing including fan/parallel conversion in step P2, and convolution is performed using a blur recovery filter function in step P3. At step P4, centering processing is performed to facilitate the subsequent back projection processing, at step P5, back projection processing is performed, and at step P6, an image is generated.

On the other hand, the method shown in FIG. 22a is adopted in the configuration using the dual detectors shown in FIG. An example of this is Publication No. In other words, the multidirectional transmission data collected in step Q1, that is, the opposing data A
(time t _A , number of samples N points) and data B (time t _B ,
In step Q2, preprocessing including fan/parallel conversion is performed for the sample number N points. Up to this point, although the number of data has been doubled, the processing contents for data A and B are the same as in FIG. 22a. Then, in step Q3, the preprocessed data A and the data processed in step Q2 are combined and processed, and the number of data is doubled, and a blur recovery filter is applied in step Q4 as shown in FIG. 22a. Convolution is performed using a function, and in step Q5, centering processing is performed to facilitate the subsequent backprojection processing.
Back projection processing is performed in Q6, and an image is generated in step Q7.

Here, the method shown in FIG. 22a and the method shown in FIG. 22b will be explained regarding the high resolution and the amount of calculation that are the problems of the present invention.

Comparison of Nyquist frequencies that determine resolution In the reconstruction method shown in Figure 22a (hereinafter referred to as "normal reconstruction method"), which does not combine data, the second
Since data A and data B shown in FIG. 1 are processed as separate data, the highest frequency that can be reproduced is the Nyquist frequency Nyq determined by the sample interval P, as shown in FIG. 23a.

In the reconstruction method shown in FIG. 22b (hereinafter referred to as "CIA reconstruction method") that combines data, the 21st
Since data A and data b shown in the figure are combined and the sampling interval is set to 1/2P, the highest frequency that can be reproduced is twice the Nyquist frequency determined by the sampling interval P, as shown in Figure 23b. frequency i.e.
It is 2Nyq.

According to the above, the spatial resolution of the CIA reconstruction method is theoretically twice that of the normal reconstruction method.

Comparison of calculation amount for blur recovery filter processing A case where blur recovery filter processing is performed in the sample space will be explained. That is, in the normal reconstruction method that does not combine data, ^N2 multiplications and N(N-1) additions are required for one projection data (number of samples: N points).

On the other hand, in the CIA reconstruction method that combines data, the projection data (number of samples: 2N points) that combines data A and data B is multiplied ^4N times and added 2N (N-1) times. is necessary.

Here, if N is sufficiently large, the amount of calculation for blur recovery filter processing in the CIA reconstruction method will be four times the amount of calculation in the normal reconstruction method.

(Problems to be Solved by the Invention) As described above, in the conventional image reconstruction method, it is possible to improve the resolution by obtaining two opposing sets of data, but the amount of calculation increases four times, This was a problem because it went against the request for high-speed scanning.

Therefore, an object of the present invention is to provide an image reconstruction method that suppresses an increase in the amount of calculation.

Another object of the present invention is to provide an image reconstruction method that suppresses an increase in the amount of calculation and achieves high resolution.

[Structure of the Invention] (Means for Solving the Problems) The present invention is characterized by taking the following measures in order to solve the above problems and achieve the objects. That is, the present invention processes projection data based on transmitted X-rays from multiple directions on a subject through preprocessing, blur recovery filter processing, centering processing,
An image reconstruction method in which a tomographic image at an X-ray transmission position of the subject is generated by reconstruction processing by a filtered back projection method including back projection processing, wherein a first detector row and a second The detector row is shifted by 1/4 of the width of one detector channel, and the angle is 180°.
Using opposing first and second projection data obtained separately, the first projection data after the blur recovery filter processing is combined with the second projection data before or during the centering processing, and the second projection data is combined and reversed. It is characterized by obtaining data for projection processing.

The present invention is also characterized in that high-frequency emphasis processing is performed on the combined data before or during backprojection processing.

(Operation) According to such a configuration, first, before or during centering processing of the first projection data after the blur recovery filter processing, the first projection data and the opposing second projection data are combined. Therefore, if you sample with a sampling interval of 1/2P after this combination, the highest frequency that can be reproduced will be twice the Nyquist frequency determined by the sampling interval P, that is,
Since it is 2Nyq, high resolution is achieved, and blurring recovery filter processing, which increases the amount of calculation, is performed for each data before combining the data, which increases the amount of calculation. There is almost no increase in the amount of calculation compared to the case where no combination is used.

In addition, the amplitude of the data obtained by combining the data after processing with the blur recovery filter becomes zero at the Nyquist frequency, and effective high resolution cannot be achieved. Resolution can be achieved.

(Example) An example of the image reconstruction method according to the present invention will be described below with reference to the drawings.

Figure 1a shows the algorithm of the image reconstruction method using steps R1 to R8.
Figure b shows the algorithm of the image reconstruction method from step S1 to step S8, and the differences between the two are high frequency enhancement processing R5 and S6 and centering processing.
Since there is essentially no difference between the two, only the processing order of R6 and S5 is changed, the method shown in FIG. 1a will be described here.

That is, as shown in FIG.
Step R2 performs fan/parallel conversion on the multidirectional transmission data collected in R1, that is, opposing data A (time t _A , number of samples N points) and data B (time t _B , number of samples N points). Contains and pre-processes. Then, at step R3, step R2
Convolution is performed using a recovery filter function separately for preprocessed data A and data B that have been respectively processed in . Up to this point, although the number of data has been doubled, the processing contents for data A and B are the same as the normal reconstruction method shown in FIG. 20a.

Here, the method of this embodiment, the normal reconstruction method, and the CIA reconstruction method will be compared with respect to blur recovery filter processing. That is, FIG. 2 shows the procedure of blur recovery filter processing in the method of this embodiment and the normal reconstruction method. As shown in FIG. 2a, the projection data at sample point N and Nyquist frequency Nyq is multiplied by convolution by the blur recovery filter shown in FIG. 2b (interpolation using the Sinc function). The processing result shown in FIG. 2c is obtained.

FIG. 3 shows the procedure of blur recovery filter processing in the CIA reconstruction method. As shown in Fig. 3a, the sample points are 2N points, and for the projection data of Nyquist frequency Nyq' = 2Nyq, Fig. 3b
The blur recovery filter shown in FIG. 3 is multiplied in convolution to obtain the processing result shown in FIG. 3c.

In the above, the method of this embodiment and the normal reconstruction method have 1/2 the number of sample points compared to the CIA reconstruction method, so the amount of calculation is 1/4. Therefore, calculations can be shortened.

In addition, in the CIA reconstruction method, data A is obtained at time t _A , then data B is obtained at time T _B after 180° rotation, and the data are combined after this time t _A and t _B. After blurring, blur recovery filter processing is performed. On the other hand, in the method of this embodiment and the normal reconstruction method, data A is obtained at time _tA , blur recovery filter processing is performed on data A at this time _tA , and then data A is processed at time _tB after rotation by 180°. Then, data B can be subjected to blur recovery filter processing.

In other words, in the method of this embodiment and the normal reconstruction method,
It is possible to perform blur recovery filter processing on data A without waiting for the 180° rotation to obtain data B, and in addition to reducing the amount of calculation mentioned above to 1/4, the overall reconstruction time is also shortened. It will be done.

Next, in step R4 of FIG. 1a, a combination process is performed on the data A and data B that have been subjected to the blur recovery filter processing to improve the spatial resolution.
That is, the data is combined using two detectors 12A and 12B shown in FIG. 17 as detectors,
As shown in Figure 4, a line passing through the focal point of the X-ray source (X-ray tube) and the center of rotation of the tube moves to the right or left of the center of the detector by h/4 (h: width of one channel of the detector = sample interval). The relationship is deviated by P).

Due to the above arrangement, as shown in Figure 5,
Data A is obtained from detector A, and then data B can be obtained from detector B by rotating 180°. Since the positional relationship between data A and data B is the same even if blur recovery filter processing is applied to each, if it is a parallel beam, data A and data B are combined as shown in Figure 6. . Also, if it is a fan beam, data A
and data B are combined as shown in FIG.

Figure 8 shows data A from data collection (step R1) to data combination processing (step R4).
and data B expressed in sample space and frequency space. That is, data A is collected at time _tA , data A after preprocessing is obtained as shown in Figure 8a, and this data A is multiplied by the blur recovery filter processing function shown in Figure 8b. be done.
As shown in FIG. 8c, data A after the blur recovery filter processing is generated.

During the processing of the data A, data B is collected at time t _B after the collection of the data A by 180° rotation, and is multiplied by the blur recovery filter processing function in the same way as the data A, as shown in FIG. As shown in FIG. 3, data B after the blur recovery filter processing is generated. Then, these data A after blur recovery filter processing and data B after blur recovery filter processing are combined using the data combination positional relationship shown in FIGS. 6 and 7, resulting in the data shown in FIG. is sampled at a sampling interval of P/2 to obtain the data shown in FIG. 8f.

As can be seen from the above explanation, up to step R4, according to the method of this embodiment, data A and data B are processed separately using a blur recovery filter, and then the data are combined to achieve high resolution. By performing sampling, it has become possible to achieve higher resolution without increasing the amount of calculation. In this case, as shown in Figure 8f,
Since the amplitude becomes zero at the Nyquist frequency Nyq', theoretically it is not possible to increase the resolution by more than twice as much as with the CIA reconstruction method (of course, the CIA reconstruction method can actually more than double the spatial resolution) It is not possible.). In order to eliminate the above-mentioned problems, that is, to achieve a spatial resolution that is practically equivalent to that of the CIA reconstruction method, high-frequency emphasis processing is performed as step R5.

In the high-frequency emphasis processing in step R5, the high-frequency emphasis filter shown in FIG. 9b is applied to the data sampled at a sampling interval of P/2 for the combined processed data shown in FIG. 9a (same as FIG. 8f). By multiplying, it becomes possible to obtain data in which high frequencies are emphasized, as shown in FIG. 9c.

Comparing the data up to step R5 up to this point, that is, immediately before the centering process, the normal reconstruction method is as shown in Figure 10a, and the CIA reconstruction method is as shown in Figure 10b. The reconstruction method is shown in FIG. 10c.

Next, centering processing will be explained as step R6. In the centering process, as shown in Figure 11, the data after the blur recovery filter process is
It is projected onto the axis at equal intervals. FIG. 12 shows an actual calculation example of centering processing, and there are linear interpolation methods, quadratic interpolation methods, cubic interpolation methods, Sinc function interpolation methods, etc., and in this example, the linear interpolation method is used. That is, ray(k) indicates the data after the blur recovery filter processing, C(j) indicates the data obtained by the centering processing,
It is determined using the following linear interpolation formula (1).

C(j)=(α・ray(k)+(1-α)・ray(k-1))/3 ² …(1) However, S is the distance between C(j) and F, α is β /
Δψ.

In the present embodiment reconstruction method and CIA reconstruction method, the 12th
The Δψ shown in the figure is only 1/2 that of the normal reconstruction method, and the content of the centering process is the same for all three methods. In this case, the spatial resolution can be improved by shortening the interval between C(j).

Next, the back projection process will be explained as step R7. This is the same as what is done in the normal reconstruction method and CIA reconstruction method. This is superimposition processing by projection, and as a result, an image is generated as step R8.

As described above, according to this embodiment, by utilizing the fact that the positional relationship between data A and data B remains the same before and after processing with the blur recovery filter, blur recovery is performed separately for data A and data B. Since the opposing data are combined after filtering, if the sample interval is set to 1/2P after the combination, the highest frequency that can be reproduced will be twice the Nyquist frequency determined by the sample interval P, or
2Nyq, thus achieving higher resolution.

In addition, in the blurring recovery filter processing that causes an increase in the amount of calculation, the blurring recovery filter processing that causes an increase in the amount of calculation is performed for each data before combining the data. ) The amount of computation hardly increases compared to

Furthermore, the data after combining the data after processing with the blur recovery filter has its Nyquist frequency
Since the amplitude becomes zero at Nyq', theoretically it is not possible to achieve a resolution more than twice as high as with the CIA reconstruction method. Almost the same resolution can be obtained.

The processing steps shown in Fig. 1a and b are shown in Fig. 13a to e for the purpose of improving processing accuracy, shortening processing time, and saving memory capacity required for calculation.
Two or three processing steps can be integrated, as shown in FIG. As shown in FIG. 13a, the high frequency enhancement processing step R5 and the centering processing step R6 in FIG. 1a are integrated to create centering processing steps R5 and 6 including high frequency enhancement.
It can be done. As shown in Figure 13b,
The data combination processing step R4, high frequency enhancement processing step R5, and centering processing step R6 in FIG.
6. As shown in FIG. 13c, the centering process step R6 and backprojection process step R7 in FIG.
It can be done.

Furthermore, as shown in FIG. 13d, the data combination processing step S4 and the centering processing step S5 in FIG. 1b can be integrated to form centering processing steps S4 and 5 including data combination. As shown in FIG. 13e, the high frequency enhancement processing step S6 and back projection processing step S7 in FIG. 1b can be integrated to form back projection processing steps S6 and 7 including high frequency enhancement. .

Using the integration method shown in Figures 13a-e,
The algorithm shown in Figures 1a and B is as shown in Figure 14a.
- c and Figures 15 a - c.

Here, a method of integrating the data combination processing step S4 and centering processing step S5 shown in FIG. 13d to form centering processing steps S4 and S5 including data combination will be described in detail. Previously, FIG. 7 shows the opposing positional relationship between data A and data B in a fan beam, and the combination of these opposing data was explained. This method can only be applied to line CT scanner equipment. Here, an X-ray CT scanner with an appropriate fan angle, number of channels, and number of projections is one that can combine data without the need for interpolation and satisfies the following relationship. It is.

2×(Fan angle/Number of channels)=2π/Number of projections An X-ray CT scanner device with a general device configuration requires interpolation as described below. That is, in FIG. 16, consider finding a ray R indicated by a broken line in order to combine data. However, since the relationship in the above equation is generally not satisfied, there is no R ray to be found.
Therefore, select r1 and r2 that are close to the desired R, and use the following
R can be determined by the interpolation formula (2).

R=(1-a)・r1+a・r2 a=γ/ΔQ...(2) To summarize the above, in general X-ray CT scanner equipment, if processing is performed separately, one interpolation is performed in combination processing. Requires. Furthermore, one interpolation process is required in the centering process. In other words, interpolation is required twice, and the error increases with each interpolation.

Here, we will discuss the relationship between interpolation and processing accuracy. That is, when a computer (image reconstruction device) attempts to express a certain real number a, since the length (bit length) of data handled by the computer is finite, it includes an error ε.

A=a+ε However, A is a number that can be expressed by a computer. Therefore, in the above-mentioned interpolation process, an error occurs due to the above.

R=(1-a)・f1+a・f2=(1-A+ε)・f1+(A-ε)f2=(1-A)・f1+A・f2+ε・(f1-f2) Here, (1-A)・f1+A·f2 is a result calculated by a computer, and ε·(f1−f2) is an error, and this error increases each time interpolation is repeated.

Therefore, it can be seen that the less interpolation processing is performed, the smaller the error will be. Therefore, by combining the data combination processing and centering processing and performing interpolation processing once, the error can be reduced.
This also applies to FIGS. 13a, b, c, and e.

FIG. 17 shows an example in which data combination processing and centering processing are combined and interpolation processing is performed once, using equations (1) and (2) above. That is, ray(k-1), ray(k), ray(k+1)
This is the data after the blur recovery filter processing, where ray(k) is the opposing data (R in FIG. 16) to be created by the processing. And r1, r2
is the opposing ray closest to the opposing data ray (k) to be created, and is selected from adjacent projection data. a and α are interpolation ratios.

Then, when trying to find c(j) using equations (1) and (2) above, ray(k)=R=(1-a)・r1+a・r2(1)′ C(j)=( α・ray(k)+(1-α)・ray(k-1))/S ² …(2)′ When (1)′ and (2)′ are combined, the following equation (3) is obtained. .

C(j) = (α・(1−α)・r1+α・a・r2 +(1−α)・ray(k−1)/S ² …(3) Using the above equation (3), create a data set. It is possible to perform alignment processing and centering processing in an integrated manner.
It becomes possible to suppress an increase in errors.

The present invention can be implemented with various modifications other than those described and illustrated above without departing from the gist of the present invention.

[Effects of the Invention] As described above, in the present invention, the first detector row and the second detector row are set to 1/1/2 of the width of one detector channel.
Opposed first, obtained 180° apart, shifted by 4
By using the second projection data and combining the second projection data with the first projection data after the blur recovery filter processing before or during the centering processing to obtain data for back projection processing. , the first projection data after the blur recovery filter processing is combined with the opposing second projection data before or during the centering processing, so the sample interval is set to 1 after this combination. When sampling at /2P, the highest frequency that can be reproduced is twice the Nyquist frequency determined by the sampling interval P, that is, 2Nyq. Therefore, high resolution can be achieved. In addition, in the blurring recovery filter processing, which causes an increase in the amount of calculation, the blurring recovery filter processing, which causes an increase in the amount of calculation, is performed for each data before combining the data, so the amount of calculation is greater than when the data is not combined. The effect is that there is almost no increase in

In addition, the data after combining the data after processing with the blur recovery filter has zero amplitude at the Nyquist frequency, and effective high resolution cannot be achieved. By performing high frequency enhancement processing, this problem can be resolved and higher resolution can be achieved.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing an algorithm of an embodiment of the image reconstruction method according to the present invention, and FIG. 2 is a diagram showing blur recovery filter processing in the normal reconstruction method and the reconstruction method of this embodiment in sample space and frequency space. Figure, 3rd
The figure shows the blur recovery filter processing in the CIA reconstruction method in sample space and frequency space.
Figure 5 shows the relationship between the wire tube and detector, Figure 5 shows the positional relationship of data, Figure 6 shows the combination of data in parallel beams, and Figure 7 shows the combination of data in fan beam. 8 is a diagram showing that the spatial resolution is improved by data combination processing in the reconstruction method of this embodiment, and FIG. 9 is a diagram showing high frequency enhancement processing in the reconstruction method of this embodiment. Figure 10 is a diagram showing a comparison of data immediately before centering in the normal reconstruction method, CIA reconstruction method, and reconstruction method of this embodiment. Figures 11 and 12 are diagrams showing centering processing. Figure 13 is a diagram showing the entire process. Figures 14 and 15 are diagrams showing algorithms that adopt processing integration, Figure 16 is a diagram explaining data combinations that require interpolation, and Figure 17 is a diagram showing data combinations. 18 is a diagram showing the configuration of a general X-ray CT scanner device, and FIG. 19 is a schematic configuration of an X-ray CT scanner device with dual detectors applied in the present invention. Figures 20 and 21 are diagrams explaining opposing data, Figure 22 is a diagram showing the normal reconstruction method and the CIA reconstruction method, and Figure 23 is a diagram showing the normal reconstruction method and the CIA reconstruction method. FIG. 2 is a diagram explaining the Nyquist frequency at . 11...X-ray source (X-ray tube), 12A...detector A, 12B...detector B, 20...data acquisition device, 30...image reconstruction device, 40...image conversion device, 50... …Display measures.

Claims (1)

[Claims] 1. Projection data based on transmitted X-rays from multiple directions on a subject are reconstructed by a filtered back projection method including preprocessing, blur recovery filter processing, centering processing, and back projection processing. , in the image reconstruction method that generates a tomographic image at the X-ray transmission position of the subject, the first detector row and the second detector row are set to 1/4 of the width of one detector channel. Using opposing first and second projection data obtained 180 degrees apart, the first projection data after the blur recovery filter processing is centered. An image reconstruction method characterized by combining data to obtain data for back projection processing. 2. The image reconstruction method according to claim 1, wherein high frequency enhancement processing is performed on the combined data before or during the back projection processing.