JP2007068717A - Image reconstruction device of cone-beam ct - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To increase the speed of the reconstruction process of a cone-beam CT. <P>SOLUTION: This image reconstruction device tabulates all of convolution weighting factors and back projection weighing factors. This device also tabulates all of the coordinate positions of data to be back projected on a projection image and their interpolation coefficients. This device executes the reading of the tables in parallel with the computing of the convolution and the back projection. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a cone beam X-ray CT apparatus, and more particularly to an image reconstruction apparatus for cone beam CT that obtains a tomographic image by performing reconstruction processing efficiently.

  A conceptual diagram of the cone beam CT is shown in FIG. A surface sensor detects X-rays emitted from one X-ray source, absorbed in the body of the subject, attenuated and transmitted. Then, the projection image data for one round is acquired by rotating around the subject without changing the relative positional relationship between the X-ray generator and the surface sensor. At this time, instead of the X-ray generator and the surface sensor, the subject may be rotated once. A tomographic image is obtained by reconstructing the projection image data acquired in this way.

  As shown in FIG. 3, the reconstruction processing is performed by first performing convolution processing on the projection image data, and backprojecting the convolution-processed projection image data onto each pixel of the reconstruction image as shown in FIG. Done.

  In the CT reconstruction using cone beam, when the method called direct method is used instead of the method called fan-para conversion method to convert to parallel beam, convolution is applied by applying the convolution weight coefficient. It is necessary to backproject the projection data by applying a backprojection weighting factor.

If these weighting factors are taken as the convolution weighting factor W ₁ and the back projection weighting factor W ₂ , the geometric relationship is as shown in FIG.

と表すことができる。

It can be expressed as.

  The calculation of the weighting factor must be calculated according to each pixel of the projection image for the convolution weighting factor and according to each pixel and the projection angle of the reconstructed image for the backprojection weighting factor. Therefore, this weighting factor calculation, especially the back projection weighting factor, must be calculated every time the three parameters x, y, and φ change, and the amount of calculation becomes enormous, leading to an increase in reconstruction time. It has been.

In order to speed up such backprojection weighting coefficient calculation, for example, in Japanese Patent Laid-Open No. 9-187449, backprojection is first performed on a preset centering surface, and backprojection data of the centering surface is backprojected to each pixel. Thus, a technique for simplifying the weighting factor calculation is disclosed.
JP 09-187449 A

  Some speedup can be achieved by simplifying the weighting factor calculation by backprojecting to the centering plane. However, the weighting factor must still be calculated in the back projection onto the centering surface, and the speed is not sufficiently increased.

  In addition, not only the weighting factor but also a great amount of time is spent on the calculation for searching the projection image for the data to be backprojected in the backprojection processing and the interpolation coefficient of the searched data, which is a cause of the long reconstruction processing time. It has become.

  All the convolution weight coefficients and back projection weight coefficients are tabulated. Further, the coordinate positions on the projection image of the data to be backprojected and the interpolation coefficients thereof are all tabulated.

  As a result, the calculation of the weighting factor at the time of the reconstruction process, the search at the time of back projection and the calculation of the interpolation can be omitted, and the speed of the reconstruction process can be increased. The weight coefficient table, position, and interpolation coefficient table created here are very large tables of 2D, 3D, and 4D. Therefore, it is necessary to store this table in a large-capacity storage device such as a hard disk. However, since the reading time is delayed, reading, convolution and back projection operations are performed in parallel. As a result, the overhead of the table read time can be hidden, and the reconstruction process can be performed at high speed.

  According to the present invention, the reconstruction process of the cone beam CT apparatus can be speeded up. Therefore, the flow from imaging to diagnosis becomes smooth, and the throughput can be significantly improved.

  FIG. 1 is a schematic diagram showing an overall configuration of an example of an X-ray CT system according to the present invention.

  104. X-rays generated from the 103. X-ray source controlled by the X-ray generator control unit are transmitted through the subject 102. The patient and detected by the 101. X-ray detector. The detected X-ray is input to the image input unit 105 as a projection image. The 103.X-ray source and the 101.X-ray detector 102. Collect projection images for each predetermined rotation angle while rotating about the patient as the center of rotation. Alternatively, it is also possible to rotate the 102. patient as the subject fixed to the rotary table while maintaining the positional relationship between the 103. X-ray source and the 101. X-ray detector.

  The input projection image at each rotation angle 107 is subjected to image processing such as pre-processing and reconstruction processing including correction of the X-ray detector and log conversion by the image processing unit, and a tomographic image group is created. The created tomographic image group is displayed on 109. a diagnostic monitor, 108. stored in an image storage unit, 111. output to 112. printer, 113. diagnostic workstation, 114. image database via network. Various operations such as a display window operation, a tomographic image switching display operation in the body axis direction, a cross-section conversion operation, and a three-dimensional surface display operation are performed by the operation unit 110.

  An embodiment that operates by being incorporated in the image processing unit of such a system will be described.

  Projection image data is input from the X-ray detector during or after imaging. A convolution process is performed on the projection image data.

  The convolution process is performed by taking the convolution of the horizontal line data of the projection image data and the one-dimensional data called a reconstruction function as shown in FIG. FIG. 6 shows the functions of Ramachandoran and Shepp & Logan, which are typical examples of the reconstruction function.

Further, this convolution process must be performed while applying the convolution weight coefficient W ₁ . This is expressed by the following equation.

ここで、ｇ_φ（Ｙ，Ｚ）は投影角度φの投影画像データ、ｈ（ｉ）は再構成関数、Ｇ_φ（Ｙ，Ｚ）は投影角度φのコンボリューション結果画像である。

Here, g _φ (Y, Z) is projection image data at the projection angle φ, h (i) is a reconstruction function, and G _φ (Y, Z) is a convolution result image at the projection angle φ.

FIG. 7 illustrates the concept of convolution processing performed while applying this weighting coefficient. Since the weight coefficients as shown in FIG. Does not depend on the projection angle phi, the weighting factor of one image worth it is necessary, the weight coefficient W ₁ is intended to determined if determined geometric relationship is as described above Yes, it does not change with every shooting. Therefore, a table of the weighting factor W ₁ is created when the geometric conditions for photographing are determined. If the convolution process is performed while reading this table calculated in advance during the convolution process, the calculation of the weighting factor can be omitted during the reconstruction, and the convolution process can be performed at high speed.

  Back projection processing is performed using the projection image data subjected to the convolution processing in this way.

  In the back projection process, as shown in FIG. 4, for each pixel of the reconstructed image, the coordinates of the point on the projected image that has passed through the pixel are obtained, and the pixel values of the four points in the vicinity of the coordinate position are obtained by interpolation and added. It is something that goes. The geometric relationship in the interpolation is shown in FIG. In FIG. 8, P1 to P4 are the center points of four pixels in the vicinity of the projection point. If the distance to the projection point is d1 to d4, the pixel values of the four neighboring pixels are Q1 to Q4, and the backprojection data is V

によって補間を行えばよい。

Interpolation may be performed by

  In this interpolation process, it is necessary to obtain d1 to d4 as the distances from the coordinate positions P1 to P4 of the four neighboring points and the projection points from the geometric positional relationship, but it corresponds to the pixels and projection images of all reconstructed images It takes an enormous calculation time to obtain the coordinate positions and projection points of the four neighboring points.

  However, since these data are determined only by the geometrical positional relationship, they do not change with every shooting. Therefore, when the geometrical conditions for photographing are determined, a table of d1 to d4 is prepared for the distances between the coordinate positions P1 to P4 of the four neighboring points and the projection points.

  However, regarding the coordinate positions of the four neighboring points, only the coordinates i and j of P1 need be tabulated. In addition, for the distances d1 to d4 with the projection point, this may be made into a table as it is,

のように補間係数に変換しておけば、補間の計算も

If it is converted into an interpolation coefficient like

と簡単になり、しかもテーブル化するデータもt1,t2の２つですむ。

The data to be tabulated is only two, t1 and t2.

  The back projection data coordinate position and interpolation coefficient table is a four-dimensional table using the x coordinate, y coordinate, z coordinate, and projection angle φ of the reconstructed image as parameters.

  When creating backprojection data, if the backprojection data is created while reading this pre-calculated table, the calculation of coordinate position and interpolation coefficient can be omitted, and backprojection data can be created at high speed. Can do.

The backprojection processing is performed by applying a weighting coefficient to the backprojection data created in this way and adding it to the corresponding pixel of the reconstructed image. FIG. 9 illustrates the concept of back projection processing performed while applying this weighting coefficient. As shown in this figure, the back projection weight coefficient W ₂ is calculated according to the x-coordinate, y-coordinate, and projection angle φ of the reconstructed image, and multiplied by the back-projection data. The back projection weighting factor W ₂ is determined by determining the geometrical relationship in the same manner as the convolution weighting factor W ₁ , the coordinate position of the backprojection data, and the interpolation factor. Absent. Thus, where the geometric conditions of the imaging is determined, it creates a the weighting factor W ₂ table. When the back projection process is performed while reading this table calculated in advance, the calculation of the weighting factor can be omitted, and the back projection process can be performed at high speed.

  As described above, the weighting coefficient, the coordinate position, and the interpolation coefficient of the convolution process and back projection process are all tabulated, so that it is not necessary to perform a special calculation at the time of reconstruction, and a high-speed reconstruction process can be realized. However, these tables are enormous because the convolution weight coefficient is two-dimensional, the back projection weight coefficient is three-dimensional, the coordinate position and the interpolation coefficient are four-dimensional.

  Therefore, it is impossible to store this table in the memory, so it is necessary to save it in a large-capacity storage device such as a hard disk. However, since reading time is slow, reading, convolution and back projection operations are performed in parallel. To increase the speed. This is illustrated in FIG. Here, parallel processing is realized by a double buffer using two buffer memories. FIG. 11 illustrates the flow of processing using such a double buffer. First, the first projection image data and each table data related to this projection image are read into the buffer memory 1 from the hard disk. Next, the buffer memory of the hard disk is switched to the buffer memory 2 and the second projection image data is read into the buffer memory 2. During this reading, the buffer memory 1 is connected to the CPU to perform convolution processing and back projection processing on the first projection image data. Next, the hard disk buffer memory is switched to buffer memory 1, the third projection image data is read into buffer memory 1, buffer memory 2 is connected to the CPU, and convolution processing is performed on the second projection image data. Perform projection processing. In this way, while sequentially switching the buffer memory, data reading, convolution processing, and back projection processing are performed in parallel.

  Here, the parallel processing of data reading and processing is performed for each projection image data, but parallel processing may be performed for each reconstructed image.

  High-speed reconstruction processing of cone beam CT can be realized by the above method.

The figure showing the system configuration example. The conceptual diagram explaining cone beam CT. The conceptual diagram explaining a convolution process. The conceptual diagram explaining back projection processing. The conceptual diagram explaining the geometric relationship in a reconstruction process. The conceptual diagram explaining a reconstruction function. The conceptual diagram explaining the convolution process which applies a weighting coefficient. The conceptual diagram explaining interpolation. The conceptual diagram explaining the back projection process which applies a weighting coefficient. The conceptual diagram explaining embodiment by parallel processing. The conceptual diagram explaining the flow of the process by parallel processing.

Claims (3)

  Using a convolution weight coefficient, a back projection weight coefficient, a coordinate position on the projection image of the back projection data, a means for holding the interpolation coefficient as a table, and a table that is used in the reconstruction process. An image reconstruction apparatus for cone beam CT, characterized by comprising means for performing reconstruction processing.   2. The cone beam CT image reconstruction apparatus according to claim 1, further comprising means for performing reading of the table and execution of reconstruction processing in parallel.   3. The cone beam CT image reconstruction apparatus according to claim 2, wherein the reading of the table and the execution of the reconstruction process are performed in parallel using a double buffer.

Images