JP3672399B2 - CT image reconstruction method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an image reconstruction method for a medical or industrial X-ray CT apparatus, and more particularly to a CT image reconstruction method by a multiprocessor that performs image reconstruction by a counter beam interpolation method using a plurality of unit arithmetic processors.
[0002]
[Prior art]
FIG. 1 is a block diagram showing an overall configuration of a conventional X-ray CT apparatus 1 that performs a general helical scan. In the figure, an X-ray CT apparatus 1 can carry a subject P into a gantry unit 9 having an X-ray source 3, a detector 5 and a data acquisition circuit (DAS) 7, and a central opening 9 a of the gantry unit 9. A bed 11, a host CPU 13 that controls the entire X-ray CT apparatus 1, a system memory 15 as a shared memory for temporarily storing collected or processed data, collected projection data and reconstructed images A large-capacity storage device 17, a system console 19, a system bus 21, a plurality of reconstruction units (hereinafter abbreviated as RU) 23, 25, and 27, and a communication control device 29 are provided. Configured.
[0003]
The RU includes, for example, a micro processing unit (hereinafter abbreviated as MPU) 41 for controlling the RU or processing data, a local memory 43 for storing data inside the RU, and a back in the reconfiguration operation. The BP 45 is a dedicated hardware for performing projection processing. And according to the desired reconfiguration arithmetic capability, it is set as the multiprocessor structure which can increase / decrease the number of RUs to mount, and each RU can operate | move in parallel.
[0004]
The example of FIG. 1 shows the case of an RU having a multiprocessor configuration of three unit RUs, in which RUs 23, 25, and 27 including three identical RU boards are mounted. In addition, since the back projection processing in charge of the BP in the RU is a cumulative addition operation on the image matrix, the BPs may be cascade-coupled by a dedicated bus called a pixel bus 31.
[0005]
In addition, there is a configuration example as shown in FIG. 4 in which a plurality of RUs are mounted on one substrate, and a reconfiguring device may be formed by using a plurality of substrates 33 in FIG.
[0006]
Next, reconfiguration processing by the multiprocessor will be described with reference to a flowchart and a data division explanatory diagram.
[0007]
FIG. 3 is a flowchart showing an outline of an example of processing contents when helical interpolation processing is performed in the X-ray CT apparatus of FIG. According to FIG. 3, the step of detecting the intensity of X-rays transmitted through the subject with a detector and performing AD conversion or the like to collect projection data (step S11), and the steps of various correction processing of the projection data (steps) S13), a step of counter beam interpolation processing (step S15), a step of convolution processing for performing superposition integration with a filter function (step S17), and a step of back projection processing for obtaining a tomographic image of the subject by back projection (step S17). Step S19), and display and recording of the obtained tomographic image (Step S21).
[0008]
FIG. 5 shows the raw data of the data collection stage collected by the DAS 7 and stored in the raw data area 15a of the system memory as a sinogram displayed in two dimensions in the channel direction and the view direction.
[0009]
By the way, the conventional method for dividing the two-dimensional data in order to execute the processing in step S13 and subsequent steps in FIG. 3 in parallel with a plurality of processing devices is the view division shown in FIG. The channel division shown in FIG. 5B was used.
[0010]
For example, in the convolution process in step S17 and the back projection process in step S19, the data is divided into each RU and processed by the view division in FIG. This is because the filter used in the convolution process must act on all channel data of the same view.
[0011]
On the other hand, in the correction process including the moving average process in the view direction, the channel division of FIG. 5B is often used. One example of the reason is that when the moving average processing in the view direction is performed by view division, when the view section subject to moving average crosses the data division boundary, between the RUs carrying the respective view division data This is because it is necessary to transfer data, and such data transfer complicates software and lowers processing capability.
[0012]
In addition, most correction processing can be performed by channel division in principle. If filtering is performed in the channel direction during the correction process, the ends of the divided areas may be overlapped according to the filter width as shown in FIG.
[0013]
Next, when the correction process by channel division is completed and the view division is switched to a suitable convolution process, the corrected data is once collected in the shared memory from the local memory of each RU (in FIG. 1, in the system memory Is shared memory).
[0014]
If all correction processing is performed by channel division, data transfer between the shared memory and each RU is
(1) Shared memory-local memory transfer for starting correction processing (2) Local memory-shared memory transfer for collecting correction processing results (3) Shared memory-local transfer for starting convolution processing Transfer between memory is enough. In FIG. 1, the system memory is a shared memory, but the required number of transfers does not change wherever the shared memory is used.
[0015]
In addition, although interpolation processing in helical scanning can be considered as a kind of correction processing, in FIG. 3, individual steps are used as the counter beam interpolation processing in step S <b> 15. As shown in FIG. 6, the helical scan is a scan in which the X-ray source draws a spiral trajectory relative to the subject. Projection data of a desired reconstructed cross section is generated in a pseudo manner by interpolation processing.
[0016]
Next, the counter beam will be described with reference to FIG. The opposed beams are a pair of X-ray beams that are opposite to each other in space except for the position in the body axis direction (slice direction) of the subject as shown in FIG. FIG. 7B shows the opposite beam viewed from the body axis direction of the subject.
[0017]
For example, the beam facing the beam G of the channel at the right end in the traveling direction of the fan beam with the spread angle φ exposed from the X-ray source 3 at the rotation angle position 0 ° is (180 ° + φ). ) In the leftmost channel in the traveling direction of the beam at the rotation angle position.
[0018]
Similarly, the beam facing the beam H of the leftmost channel in the traveling direction of the fan beam with the spread angle φ irradiated from the X-ray source 3 at the rotation angle position 0 ° is It becomes the beam K of the rightmost channel in the traveling direction of the beam at the rotation angle position of (° −φ).
[0019]
When the opposing beam is shown on the sinogram, the positional relationship shown in FIG. 8 is obtained, and the opposing beam of the fan beam indicated by the line segment GH has a distribution indicated by the line segment JK. The
[0020]
Further, as shown in FIG. 9, on the sinogram showing the projection data divided into three in A, B, and C in the channel direction, the opposing data of the data set indicated by the line segment Gg is at the position indicated by the line segment Jj. While the data indicated by Gg belongs to the area A, the opposite data set of this data belongs to a different area C.
[0021]
[Problems to be solved by the invention]
However, when performing reconstruction by a multiprocessor, when performing interpolation processing using an opposite beam in a series of interpolation processing using channel-divided data, the corresponding projection data set has a different RU (substrate). Since it is held in the local memory above, new data transfer is required between RUs, MPU software becomes complicated, overhead associated with data transfer occurs, and the time required for interpolation processing increases. There was a problem of inviting.
[0022]
In view of the above problems, an object of the present invention is to eliminate the overhead associated with the transfer of the corresponding projection data set in a reconstruction apparatus having a plurality of RUs, and to enable high-speed counter beam interpolation. It is.
[0023]
[Means for Solving the Problems]
In order to achieve the above object, the present invention has the following configuration.
That is, the invention described in claim 1 is a CT image reconstruction method in which a desired CT image is reconstructed by a plurality of arithmetic processing units based on projection data collected by a CT scanning device. The gist of the invention is that the detector is divided into a plurality of regions that are symmetrical with respect to the median channel number of the detector, and the regions that are symmetrical to each other are assigned to the same arithmetic processing unit.
[0024]
According to a second aspect of the present invention, in the CT image reconstruction method for reconstructing a desired CT image by a plurality of arithmetic processing units based on the projection data collected by the CT scanning device, the collected projection data is stored in the CT image reconstruction method. The gist of the invention is that the area is divided into two times the number of regions of the arithmetic processing unit symmetrical to the median of the channel number of the detector, and the symmetrical regions are assigned to the same arithmetic processing unit.
[0025]
The gist of the invention described in claim 3 is that in the CT image reconstruction method according to claim 2, the sizes of the divided symmetrical regions are equal.
[0026]
DETAILED DESCRIPTION OF THE INVENTION
Next, embodiments of the present invention will be described in detail with reference to the drawings.
FIG. 2 is a sinogram showing the division of data when the multiprocessor CT image reconstruction method according to the present invention is applied to an X-ray CT apparatus composed of three RUs. The configuration of the X-ray CT apparatus to which the present invention is applied is the same as that of the conventional example shown in FIG. In the present invention, the projection data collected by the CT scanning device is divided into a plurality of regions that are symmetrical with respect to the median channel number of the detector.
[0027]
FIG. 2 (a) is a sinogram showing a first embodiment in which data is divided when three RUs, A, B, and C perform parallel processing. The data area is twice the number of RUs. It shows a method of dividing a channel into regions having an equal size.
[0028]
According to FIG. 2 (a), the data area is divided into six areas 51, 52, 53, 54, 55 and 56 each having an equal size, and each is symmetrical with respect to the median of the channel numbers. Each area is assigned to the same RU. That is, areas 51 and 56 are assigned to RU-A, areas 52 and 55 are assigned to RU-B, and areas 53 and 54 are assigned to RU-C.
[0029]
By assigning the data area in this way, when each RU performs counter beam interpolation, the corresponding data set is held in the local memory on the same RU. -Overhead for transferring the set to the same local memory is reduced, and high-speed reconfiguration processing is realized.
[0030]
FIG. 2B is a sinogram showing a second embodiment in which data is divided when three RUs, A, B, and C perform parallel processing. The data area is symmetrical with respect to the median channel number. A method for dividing a channel into a plurality of regions is shown.
[0031]
According to FIG. 2 (b), the data area is divided into five areas 61, 62, 63, 64, and 65. The areas 61, 62, 64, and 65 are equal in size, and the area 63 is It is twice as large as other areas. In addition, regions that are symmetrical with respect to the median channel number are assigned to the same RU. That is, areas 61 and 65 are assigned to RU-A, areas 62 and 64 are assigned to RU-B, and area 63 is assigned to RU-C.
[0032]
By assigning the data area in this way, when each RU performs counter beam interpolation, the corresponding data set is held in the local memory on the same RU. -Overhead for transferring the set to the same local memory is reduced, and high-speed reconfiguration processing is realized.
[0033]
FIG. 2C is a sinogram showing a third embodiment in which data is divided when three RUs, A, B, and C perform parallel processing. The data area is twice the number of RUs. It shows a method of dividing a channel into regions having unequal sizes.
[0034]
According to FIG. 2 (c), the data area is divided into six areas, 71, 72, 73, 74, 75, and 76, and the sizes of the areas symmetric with respect to the median value of the channel numbers are equal. Assigned to the same RU.
[0035]
That is, regions 71 and 76 are assigned the same size and the same RU-A, regions 72 and 75 are assigned the same size and the same RU-B, and regions 73 and 74 are the same size. And assigned to the same RU-C.
[0036]
By assigning the data area in this way, when each RU performs counter beam interpolation, the corresponding data set is held in the local memory on the same RU. -Overhead for transferring the set to the same local memory is reduced, and high-speed reconfiguration processing is realized.
[0037]
FIG. 2D is a sinogram showing a fourth embodiment in which data is divided when three RUs, A, B, and C perform parallel processing. The data area is symmetrical with respect to the median channel number. A method for dividing a channel into a plurality of regions is shown.
[0038]
According to FIG. 2 (d), the data area is divided into five areas, 81, 82, 83, 84, and 85, and the sizes of areas that are symmetric with respect to the median of the channel numbers are equal. ing. That is, the areas 81 and 85 have the same size, and the areas 82 and 84 have the same size.
[0039]
In addition, regions that are symmetrical with respect to the median channel number are assigned to the same RU. That is, areas 81 and 85 are assigned to RU-A, areas 82 and 84 are assigned to RU-B, and area 83 is assigned to RU-C.
[0040]
By assigning the data area in this way, when each RU performs counter beam interpolation, the corresponding data set is held in the local memory on the same RU. -Overhead for transferring the set to the same local memory is reduced, and high-speed reconfiguration processing is realized.
[0041]
Although the preferred embodiment has been described above, this does not limit the present invention. For example, the number of reconfigurable units RU may be any number, and when four RUs are used to divide the area into twice that area, it may be divided into eight equal areas.
[0042]
Further, as shown in FIG. 4, even when a plurality of RUs are mounted on one board, the division and allocation method does not change.
[0043]
【The invention's effect】
As described above, according to the first aspect of the present invention, the projection data collected by the CT scanning device is divided into a plurality of regions that are symmetric with respect to the median of the channel numbers, and the symmetric regions are subjected to the same calculation. By allocating to the local memory belonging to the processor, the data pairs that make up the opposing data set exist in the same local memory, eliminating the data transfer overhead during opposing beam interpolation, and performing interpolation processing at high speed There is an effect that can be.
[0044]
According to a second aspect of the present invention, in the CT image reconstruction method by a multiprocessor that reconstructs a desired CT image by a plurality of arithmetic processors based on projection data collected by a CT scanning device, By dividing the projected data into areas that are symmetric with respect to the median channel number and twice the number of the arithmetic processors, and allocating the symmetrical areas to local memories belonging to the same arithmetic processor, the opposite data set Since the data pairs constituting the same exist in the same local memory, there is no data transfer overhead at the time of opposed beam interpolation, and the interpolation process can be performed at high speed.
[0045]
According to the invention of claim 3, by making the size of the divided areas equal, each reconfigurable arithmetic processor and local memory having the same configuration and capacity can be used. There is an effect that flexibility and repeated use of the same pattern substrate are possible.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of an X-ray CT apparatus to which the present invention is applied.
FIG. 2 is a sinogram showing area division of projection data according to an embodiment of a CT image reconstruction method by a multiprocessor according to the present invention.
FIG. 3 is a flowchart showing an outline of data processing in a helical scan CT apparatus that performs counter beam interpolation;
FIG. 4 is a block diagram showing a case where a plurality of reconstruction processing units are mounted on one board.
FIG. 5 is a sinogram showing a data division method in reconstruction processing, and shows (a) view division, (b) channel division, and (c) channel division overlapping between adjacent processing units, respectively.
FIG. 6 is an explanatory diagram of a helical scan.
FIG. 7 is an explanatory diagram of a counter beam.
FIG. 8 is a diagram for explaining the position of an opposing beam on a sinogram.
FIG. 9 is a diagram for explaining the location of a corresponding data set in processing using an opposing beam.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 ... X-ray CT apparatus, 3 ... X-ray tube, 5 ... Detector, 7 ... Data acquisition system, 9 ... Gantry, 11 ... Bed, 13 ... Host CPU, 15 ... System memory, 17 ... Mass storage device, DESCRIPTION OF SYMBOLS 19 ... System console, 21 ... System bus, 23, 25, 27 ... Reconfiguration unit (RU), 29 ... Communication control device, 31 ... Pixel bus, 41 ... Microprocessor (MPU), 43 ... Local memory, 45 ... Reverse Projection processing circuit (BP).

Claims (3)

In a CT image reconstruction method for reconstructing a desired CT image by a plurality of arithmetic processing devices based on projection data collected by a CT scanning device,
A CT image reconstruction method characterized in that the collected projection data is divided into a plurality of regions that are symmetric with respect to the median channel number of the detector, and regions that are symmetric to each other are assigned to the same arithmetic processing unit. In a CT image reconstruction method for reconstructing a desired CT image by a plurality of arithmetic processing devices based on projection data collected by a CT scanning device,
The collected projection data is divided into regions that are symmetric with respect to the median channel number of the detector and are twice as many as the arithmetic processing unit, and symmetrical regions are assigned to the same arithmetic processing unit. A CT image reconstruction method. The CT image reconstruction method according to claim 2, wherein sizes of the divided symmetrical areas are equal.

Images