JPS63174642A - Mri apparatus - 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] [Industrial application field]

The present invention relates to an MHI apparatus that performs imaging using a nuclear magnetic resonance phenomenon (NMR phenomenon), and more particularly to an MRI apparatus that performs Fourier transform to reconstruct an image.

[Conventional technology]

When reconstructing an image (two-dimensional or three-dimensional image) by performing two-dimensional (or three-dimensional) Fourier transform on collected data, the Fourier transform treats the sampled data string as a periodic function due to its nature. However, data actually collected by an MRI apparatus is not originally a periodic function, and sampling is often performed using a relatively small number of points. In such a case, discontinuous points will appear at both ends of the truncation width, and unnecessary frequency components will be added to the Fourier transform result, that is, the frequency domain. This is called Lea, kage.
In order to suppress this, filtering is generally performed on the sample point sequence. A Hannig filter is known as a typical example of this filter.

[Problems to be solved by the invention]

However, since MR figurines are reconstructed by two-dimensional or three-dimensional Fourier transformation, it is necessary to perform filtering two-dimensionally or three-dimensionally. dimension) and memorize it. It takes a lot of time to create two-dimensional filter information, and a large capacity memory is also required to store it. Therefore, conventionally, a method is often adopted in which filter processing is omitted, or even when filter processing is performed, it is limited to a one-dimensional filter. Omitting it in this way does not allow the aforementioned leakage to be sufficiently suppressed, and this leakage also causes artifacts to be produced in the reconstructed image. An object of the present invention is to provide an improved MRI apparatus that can perform filtering on collected data without requiring two-dimensional (or three-dimensional) filter information and can perform ideal filtering at high speed. do.

[Means to solve the problem]

The MHI device according to the present invention includes means for determining the distance between each data point and a predetermined reference point in the collected data matrix, and a filter value given as a function of the distance from the reference point is stored in advance for each distance. and means for reading a filter value corresponding to the distance from the storage means and applying it to the data of each data point.

[For production]

When considering performing two-dimensional (or three-dimensional) filtering to suppress leakage, it is possible to apply the same filter value (weight) to data at points equidistant from the reference point on the data matrix. I can see that it is working. Therefore, focusing on this, filter values are stored for each distance between each data on the collected data matrix and the reference point, and the distance between each data point on the data matrix and the reference point is , read out the filter value stored above at this distance, and apply it to the data at that point.
This means that the necessary filtering has been performed. In this way, you only need to calculate the distance on the data matrix, read it from the storage means, and apply the read filter value to the data, making it easy to create two-dimensional or three-dimensional filter information. You can avoid spending a lot of time on it or requiring a large amount of memory to store it. Therefore, ideal filtering can be performed at high speed. [Embodiment] In FIG. 1, register 1.2 contains a column number i and a row number j indicating a data position on a data matrix, which are sent from a host computer or microprocessor.
is set. Then, with these i and j, memory 3
．． 4 was accessed and stored in memory 3.4, the corresponding Lx2 on the Ly2 table
, Ly2 are read out. Each of these positions is added by an adder 5, and the resultant L2 (=Lx2+Ly
2) is stored in register 6. Next, the corresponding value on the square root table in the memory 7 is read out by this L2 and sent to the register 8. Accordingly, the filter value (weight) W is read out from the filter table in the memory 9, and this W is stored in the register 10. Here, it is assumed that the data collected by the MRI apparatus are arranged in a two-dimensional matrix as shown in FIG. In this matrix, the X direction is the frequency direction, and
The direction is the phase direction (different from the direction in real space). To simplify the explanation, the matrix is assumed to be 8×8. Ideal filtering applies the same filter value (weight) to data located equidistant from reference point 0. That is, for data located at a distance IL from the reference point O, in the case of a Hanning filter, the filter value W of W = 1/2 + (1/2) CO8 ((Z L ) as shown in FIG. Therefore, it is first necessary to find the distance between the reference point O and the position of each data.Therefore, from the column number i and row number j representing each data position, calculate the distance from the reference point 0. Find the distance Lx in the X direction and the distance Lx in the y direction, square them, and calculate L='rLower Calyx]τ.Memory 3.4 contains data corresponding to i and j in advance. Since each part of Li2 and Li2 calculated by Since the distance between the data position (i, j) and the reference point O is determined in this way, this value of L is used as address information for accessing the filter table in the memory 9. An operation equivalent to giving the value of L to the function W (L) shown in FIG. The value W becomes a filter value for weighting the data position i, j). Therefore, filtering can be performed by applying this value to the data. Therefore, leakage can be suppressed by performing this operation on all the data arranged in the matrix. In addition, Lz2, Li stored in memory 3.4.9
2. Each table of W can be rewritten arbitrarily, so if this part is rewritten by the microprocessor or host computer that controls it,
In addition to being able to change the position of the reference point 0, the shape of the filter to be applied can also be freely selected from the above. Furthermore, it can be extended to cases where the collected data is arranged in a three-dimensional matrix. That is, as shown in FIG. 4, a register 11, a memory 12, and an adder 13 are added to the configuration of FIG. Here, register 11
It is assumed that the data plane number k is stored in . In other words, k is the number of the data plane when a number of two-dimensional matrices (data planes) in which data is arranged two-dimensionally are arranged three-dimensionally as shown in FIG. In this way, even in the case of a three-dimensional data matrix, ideal filtering is achieved by applying the same filter value to data located equidistantly from the reference point 0. That is, the same filter value is applied to data located on the surface of a sphere having the same radius centered on the reference point 0. For data at a position P on the data plane, it is sufficient to find the distance NL from the reference point 0 to the point P, find the filter value W corresponding to it, and apply W to the data at this position P. This embodiment differs from the above embodiment only in the configuration for determining L, but after L is determined, it is the same. To be exact, L
Once you find z, the rest is the same. L2=Lx2+Ly2+Lz2, and this calculation can be performed. Since Lz is the distance in two directions from reference point 0 to point P, the data plane number k
The address of the memory 12 is specified by . This memory 12 stores in advance a table of values corresponding to the square of the distance in two directions from the reference point 0 to each data surface. Therefore, the value of Lz2 is read from this memory 12 and sent to the adder 13, and the value of Lx2+Ly2 from the adder 5 is
In addition, the above calculation can be performed by holding the result in register 6. In this way, the square value, square root value, and filter value are stored in each memory and only read out, and the calculation is just a simple addition, so when the filter value is calculated and stored as two-dimensional or three-dimensional information. The amount of calculation and amount of memory required can be significantly reduced compared to . For example, when the data matrix is 256×256, the memory capacity can be reduced to about 1750 to 1/100. 3
In the case of a dimensional data matrix, this improvement rate is even greater, even below the above-mentioned value of 1710. Further, it is also easy to change the characteristics of the filter to be applied.

【Effect of the invention】

According to the MRI apparatus of the present invention, two-dimensional or three-dimensional ideal filter processing is performed by focusing on the fact that the same filter value should be applied to data equidistant from a reference point on a data matrix. can be done quickly. Therefore, it is effective in reducing artifacts in reconstructed images.

[Brief explanation of the drawing]

Figure 1 is a block diagram of one embodiment of this invention, Figure 2 is a graph representing a filter function with distance as a variable, Figures 3 and 3 are diagrams representing a two-dimensional data matrix, and Figure 4 is another embodiment. The block diagram of FIG. 5 is a diagram representing a three-dimensional data matrix. 1.2.6.8.10.11...Register, 3.4.
7.9.12...Memory, 5.13...Adder.

Claims (1)

[Claims] (1) A means for determining the distance between each data point and a predetermined reference point in the collected data matrix, and a storage means in which a filter value given as a function of the distance from the reference point is stored in advance for each distance. and means for reading a filter value corresponding to the distance from the storage means and applying it to data of each data point.