JPH02272680A - Image reconstruction system - Google Patents

Abstract

PURPOSE:To simplify the circuit constitution of an arithmetic part by mutually cascade-connecting plural BP blocks and executing the pipeline parallel processing and accumulative addition of the same lines of different projections to execute the multiplex processing of BP interpolating operation. CONSTITUTION:Respective back projection(BP) blocks 6 are mutually cascade- connected and a BP operation part 4 executes the pipeline parallel processing and accumulative addition of the same lines of different projections to execute multiplex processing. Since the BP multiplexing direction coincides with the projection direction, a local memory to be connected to the post stage of the BP operation part 4 can be omitted, and since the same BPs are set up in different lines, storage capacity for one projection is sufficient for a back projection input memory(BIM) 7. Consequently, the operation part 4 can be simply constituted.

Description

Detailed Description of the Invention [Objective of the Invention] (Industrial Application Field) The present invention performs back projection interpolation calculation using a plurality of back projection blocks when reconstructing an image based on collected data obtained by CT scan. The present invention relates to image reconstruction methods that perform multiple processing, and particularly to improvements in multiprocessing methods.

(Prior Art) A conventional image reconstruction device to which this type of image reconstruction method is applied processes different lines of the same projection in parallel in order to perform multiple processing of back projection interpolation calculations.

For example, as shown in FIG. 3, each back projection block (BP block) is provided with a back projection input φ memory (BIM) and a local memory (LM).

Then, in each BP block, each time the BIM receives data from the centering calculation unit (CENT), calculations are performed on the same line in a different projection, and summation is performed on the LM. It was something to do. In this case, the fourth address on the IM corresponding to the coordinates on the image calculated in the BP block is
As shown in the figure, the projection (PROJI~PROJ
N) are cumulatively added and multiplexed. In addition, the third
In the figure, A indicates an address and D indicates data.

(Problem to be Solved by the Invention) However, when performing multiple processing as in the conventional image reconstruction method of this type, it is necessary to provide a dedicated LM for each BP block and temporarily save data using this LM. For this reason, the capacity of the LM limits the number of multiplexes and the processing unit. Furthermore, each BP block requires an adder, and the BIM that stores pre-BP data must have a relatively large capacity, which are disadvantageous in terms of circuit configuration. Moreover, it is necessary to temporarily store the calculation results in the BP block in the LM, and then cumulatively add the calculated values for the same point in each projection, making it impossible to further reduce the processing time during multiprocessing. Ta.

The present invention has been made with attention to such circumstances, and its purpose is to simplify the circuit configuration of a calculation unit that performs multiple processing of BP interpolation calculations, and to significantly shorten the processing time of the multiple processing. The object of the present invention is to provide an image reconstruction method that can perform the following tasks.

[Structure of the Invention] (Means for Solving the Problem) In order to achieve the above object, the present invention performs BP interpolation calculation using a plurality of BP blocks when reconstructing an image based on collected data obtained by CT scan. In the image reconstruction method that performs multiple processing, the multiple processing is performed by cascading the BP blocks, performing pipeline parallel processing on the same line of different projections, and performing cumulative addition at the same time. That is.

(Function) With the image reconstruction method according to the present invention, since the BP multiplexing direction becomes the projection direction, it is possible to omit the LM at the subsequent stage of BP calculation, which was required in the past.
Since the BPs are on different lines, the BIM only requires the storage capacity for one project, so the arithmetic unit can have a simpler configuration than the conventional one. Furthermore, the cycle of cumulative addition as a BP block can also be incorporated into the pipeline processing, thereby making it possible to omit the time for cumulative addition.

(Embodiment) FIG. 1 is a block diagram schematically showing an image reconstruction unit in an XIJ*CT scanner according to an embodiment to which the image reconstruction method of the present invention is applied.

The image reconstruction unit in this embodiment includes a centering operation block 1, a centering output memory (COM) 2, and a centering operation block 1, a centering output memory (COM) 2, and a centering operation block 1, in order to reconstruct an image by a direct method, especially a centering back projection method.
, a BP calculation control section 3, and a BP calculation section 4.1M5.

The centering calculation unit 1 performs calculation processing for centering data collected from an X-ray detector (not shown), and sequentially stores the processing results in C0M2.

The BPeL calculation control unit 3 sends an instruction signal (I signal) to the BPIfL calculation unit 4 in response to a control signal from a host CPU (not shown), thereby controlling the BP calculation unit 4.

The BP calculation unit 4 includes a plurality of BP blocks 6, and each BP block 6 includes a BIM 7, a BP calculation circuit 8, and a shift register 9.

Here, the 81M7 receives the centering data once stored in the C0M2 together with the address of the centering calculation block 1, and can exchange data with the BP calculation circuit 8 according to the address from the BP calculation circuit 8.

The BP calculation circuit 8 executes BP calculation based on the centering data from the 81M7, and is connected in cascade between the BP blocks 6.
It forms a pipeline parallel processing circuit which is connected to sequentially shift and receive the I signal from the arithmetic control unit 3 using a shift register. Then, the address (A) from the final stage BP calculation circuit 8 is added to 1M5 via the shift register 10 and selector 11, and the address (A) is added to 1M5 via the shift register 10 and selector 11.
) of the same line of different projections with
The result of the operation is added to 1M5 via an adder 5a.

Therefore, in the BP calculation unit 4, pipeline parallel processing is performed on the same line of different projections, and this is cumulatively added at the same time, so that the 1MS address corresponding to the image coordinates calculated on the BP block 6 is transferred to As shown in each projection (PROJI~P
Multiplexing is done on the same line through ROJN).

As mentioned above, by cascading the BP blocks 6 and setting the BP multiplexing direction as the projection direction, the outputs of the previous stage can be added in real time at the next stage, which eliminates the need for the post-stage BP calculation, which was previously required. L of
M can be omitted, and BIM7 can have a storage capacity for one project. Furthermore, since the cumulative addition cycle is also incorporated into the pipeline processing, the time required for this process appears to be omitted.

[Effects of the Invention] As explained above, the image reconstruction method of the present invention connects a plurality of BP blocks in cascade, performs pipeline parallel processing on the same line of different projections, and simultaneously performs cumulative addition. , BP interpolation calculations are multi-processed, it is possible to simplify the circuit configuration of the calculation unit that multi-processes the BP interpolation calculations. Further, there is an advantage that the processing time for the multiple processing can be apparently significantly shortened compared to the conventional method.

[Brief explanation of drawings]

FIG. 1 is a block diagram schematically showing an image reconstruction unit in an embodiment to which the present invention is applied, FIG. 2 is a schematic diagram showing a multiplexing state in an embodiment of the present invention, and FIG. FIG. 4 is a block diagram illustrating an outline of a conventional image reconstruction unit, and FIG. 4 is a schematic diagram illustrating a conventional multiplexing state. 1... Centering calculation block 2... Centering φ output memory (CO
M) 3... Back projection (B P) calculation control section 4... Back projection (B P) calculation section 5.
...Image memory (IM) 6...Back projection block (BP block) 7...Back projection input number memory (BIM) 8...Back projection (BP) calculation circuit 9
．． 10...Shift register 11...Selector

Claims (1)

[Claims] (1) When reconstructing an image based on collected data obtained by CT scan, in an image reconstruction method that multi-processes back projection interpolation calculations using multiple back projection blocks, the back projection blocks are connected in cascade to each other. and pipeline parallel processing of the same line of different projections,
An image reconstruction method characterized in that the multiple processing is executed by performing cumulative addition at the same time.