JPH0355673A - Characteristic curve comparing method and organic area extracting method for mri image - Google Patents

Abstract

PURPOSE:To eliminate the influence of noises for comparison of characteristic curves and to extract an organic area of an MRI image by deciding the coincidence or discordance of two characteristic curves based on a membership function which applies the absolute value of the difference between both curves as a variable. CONSTITUTION:A memory has a function which applies the absolute value of the difference between a reference curve obtained from the characteristic curve of an echo signal presumed by a spin echo method owing to a multi-error and a reference characteristic curve of an extracted organ as a variable. At the same time, the absolute value delta is obtained for the difference between the reference curve and the actually measured reference characteristic curve of the extracted organ. Then the membership function value mu is obtained from the value delta. The measured echo signals area available to all degrees and therefore an evaluation function is obtained from the function value of all degrees. Then it is decided that the echo signal generating part is identical to the extracted organ as long as the evaluation function value is kept within a prescribed range. Thus an organ is extracted in an image, and an echo signal is extracted at each point of the relevant area. The mutual product of membership function values of all degrees or the sum total of function values of all degrees is used to the evaluation function.

Description

[Detailed description of the invention] [Industrial application field] The present invention relates to a characteristic curve comparison method and an extraction method.

[Conventional technology]

There are many cases in which it is determined whether two characteristic curves are the same or not. In some cases, various input signals become characteristic curves, and in other cases, characteristic curves are obtained by performing certain image processing and extraction. In many cases, this characteristic curve is compared with a reference characteristic curve to determine whether they are the same or similar. These characteristic curves are stored in memory or registers, and comparisons can be made by
Executed by mutual data comparison processing.

Now, a conventional method of comparing two characteristic curves will be explained with reference to FIG. The horizontal axis is the variable X, and the vertical axis is its function F(x). A standard characteristic curve A is given, and this and characteristic curve B are
Consider the case of determining whether or not are the same. For this determination, conventionally, two parallel curves C and D called @Wo are set on both sides of the standard characteristic curve A, and if all points of the characteristic curve B exist between C and D, then B is the same as A,
If even one point does not exist between C and D, it is determined that B is not identical to A.

[Problem to be solved by the invention]

However, if the characteristic protrudes outside the curves C and D, as in B1, for example, it is determined that B and A are not the same. There are many cases in which B is noise, and even if A and B are seen to be substantially the same within the human visual field, they will be determined to be not the same in terms of processing. Therefore, the method is susceptible to noise.

A spin echo method using multi-echo is a method for measuring echo signals in the field of MHI. In this method, after applying a 90° pulse, 180° pulses are applied one after another, and echo signals corresponding to each 180° pulse are received as many times as the number of 180° pulses.

That is, it is a method of applying one 90° pulse to obtain a plurality of echo signals one after another, and is used as an efficient method for determining the region of a subject.

In the multi-echo spin echo method, successive echo signals are obtained one after another, starting from the first echo signal. They are called primary echo, secondary echo, tertiary echo, etc. Moreover, the first, second, third, etc. are collectively called the order. (
The echo amplitude decreases as the order increases, as follows: (amplitude of first-order echo) (amplitude of second-order echo) (amplitude of third-order echo) Now, FIG. 3 shows an example of echo measurement using the spin echo method using multi-echo. The horizontal axis shows the echo order, and the vertical axis shows the echo signal (which may be considered as pixel density). Now let us assume that the characteristic curve E is a characteristic curve for a certain region, and that we want to find a region that is histologically close to this region (more precisely, an image of that region). Therefore, assuming that the measured characteristic of a certain part is F, we would like to determine whether this characteristic F is the same as characteristic E. In comparing the characteristics E and F, a comparison method as shown in FIG. 2 is used. If it is determined that they are the same, the site with characteristic F is recognized as being the same tissue as the site with characteristic E, and the site with characteristic E is extracted as an organ.

Here, what is m-device extraction? There is a memory for all the orders counted by the MHI, and the contents of this memory are displayed on the CRT, and the operator specifies the organ by looking at the displayed contents.
This means that all pixels belonging to this organ are extracted. Here, within the organ, since they are the same tissue,
The echo signals (referring to the data in memory) describe the same characteristics. Therefore, if echo signals with the same characteristics exist, that part can be identified as the organ to be extracted.
Note that in addition to this, a case where an organ image is obtained from measured data may also be included.

An object of the present invention is to provide a characteristic curve comparison method and an organ extraction method that are less susceptible to the influence of noise.

[Means to solve the problem]

The present invention uses a membership function whose variable is the absolute value of the difference between two characteristic curves to determine whether these two characteristic curves are the same.

Furthermore, the present invention used this membership function for organ extraction in MHI.

Furthermore, in the present invention, the membership function in MRI is defined as the difference between a reference curve obtained from a characteristic curve of echo signal data assumed under the multi-echo spin echo method and a reference characteristic curve of the extracted organ. This is a function that uses an absolute value as a variable. This function is prepared in memory, and on the other hand, the absolute value of the difference between the echo signal measured by the spin echo method using actual multi-echo and the reference characteristic curve of the extracted organ is calculated. , calculate the membership function value on the memory from this absolute value. Furthermore, since the measured echo signals exist for all orders, the membership function values also exist for all orders. Therefore, we created an evaluation function that determines whether the organ is extracted from the function values of all orders, and if this evaluation function value is within a predetermined range, the echo signal generation site is considered to be the same as the extracted organ, and the organ in the image is Extract. Since the organ extraction site is often spread out, the extraction process described above is performed on the echo signals at each point within the site.

Furthermore, the present invention uses the mutual product of membership function values for all orders or the total addition of membership function values for all orders as an evaluation function.

[Effect]

In the present invention, a membership function whose variable is the absolute value of the difference is used as an expression for determining whether two characteristic curves are the same or not. Furthermore, the present invention utilizes membership functions for organ extraction in MHI. Further, in the present invention, the absolute value of the difference between the echo signal and the reference characteristic curve is determined as the membership function, and the membership function value is previously determined on the memory from this absolute value. Then, calculate the function value for each order from the absolute value of the actual difference, calculate the evaluation function from all these orders, and perform organ extraction based on this evaluation function. Furthermore, the evaluation function of the present invention is given in the form of a total product or a sum. [Example] FIG. 4 shows membership functions in MRI use of the present invention. The horizontal axis δ indicates the absolute value of the difference, and the axis μ indicates the membership function value. This membership function value is a judgment curve for determining whether or not two characteristic curves, a reference characteristic curve and another characteristic curve, are the same, and a membership function value μ=1 means that the characteristic curves are completely the same. This means that δ is large, that is, 2
The larger the difference between the two characteristic curves, the smaller the membership function value μ, indicating that the two characteristics are different. This function is obtained from a reference curve and a standard characteristic curve within the possible range of the other characteristic curve. The membership functions in FIG. 4 are obtained for each degree.

Therefore, if the number of degrees is 4, there are four membership functions μ, μS, μ2, μ3, and μ4. Although there are examples of the functions μ1 to μ4 being the same, in MRI, μ1 to μ4 are often different. For example, μ4>μl because the magnitude of the echo signal obtained in the fourth order is smaller than that in the first order, and therefore the ambiguity increases as δ increases.

The reason why the membership function is set for each degree is that a membership function for determining identity can be created for each degree. That is, for example, if we look directly at the function μ of the first order, we can say that even in the first order, they are completely identical (δ' = 0), so that the two are quite different.
In some cases, it becomes large.

From this, the function μ weight is expressed by using the difference δ as a variable to express the relationship between these same and considerably different values. The same holds true for other orders.

Note that in order to obtain the membership function, it is necessary to know the reference characteristic curve representing the extracted organ.

Functions μ1 to μ4 are obtained that indicate how much deviation (difference) from this reference characteristic curve is considered to be the range of identity.

FIG. 5 shows an example for extracting whether an arbitrary part is part of an extracted organ. The characteristic G is a reference characteristic curve determined by the extracted organ, and the characteristic H is the scanned region. Characteristic H
Since this differs depending on the scanned region, the membership function shown in FIG. 4 is used to compare whether or not there is a range of identity.

For this purpose, the absolute value δ1 of the difference between the value R1 of the reference characteristic curve at the first order and the value XI of the characteristic curve at the scanning region is determined.

δr= l R+ x+ l --(1) On the other hand, according to Figure 4, the first-order membership function is μ1
From this function, find the μth function value where δ=δ1.

It is also possible to determine whether they are the same or not only from this first-order function value μI1. However, since the number of judgment data is too small, an evaluation function is introduced in this embodiment.

For this purpose, the absolute values δ2, δ3, and δ4 of the differences from the second order to the fourth order are also determined.

From these absolute values δ2, δ3, δ4, membership function values μ21, μ3, μ subtraction are determined from FIG.

The evaluation function is as follows.

(a) In the case of the product, As the evaluation function J, the integration results of μth to μ4I are used.

J=μ■0μ21 'μ31 'μ4I゜゜”゜(3)
A reference value is set for this evaluation function J, and (J>reference value)
If so, the point of the scanned region is determined to be the same as the reference characteristic curve, and is extracted as one point of the extracted organ.

(mouth), in the case of sum The total addition result of μ■ to μ subtraction is used as the evaluation function.

J = μ+t + pz+ + μ31 + μa+
-<4) A reference value is set for this evaluation function J, and if (J > reference value), the point of the scanned region is determined to be the same as the reference characteristic curve, and it is determined that it is one point of the extracted organ. Extract as.

The above evaluation functions of the product and sum of (a) and (guchi) may be used depending on the extraction site. In the case of the product, relatively μ
Since the mutual magnitude relationship appears in the value J, it is convenient for comparing characteristic curves that can be determined to be the same when there is relatively no mutual magnitude relationship. In the case of a sum, since the mutual magnitude relationship between μ is not so greatly reflected in the result, it is convenient for comparing characteristic curves that can be determined to be the same even if the characteristic curves have a large mutual magnitude relationship between μ.

FIG. 6 shows an embodiment of the processing apparatus of the present invention. This processing device has a bus 10 and a CPU connected to this bus IO.
, main memory 2, trackball 3, high-speed calculation circuit (dedicated hardware) 4, calculation memory 5, display memory 6, C
It consists of 7 RTs and 8 recording media (files). The CPUI performs bus control, general image processing, and man-machine processing. Main memory 2 stores programs and various data. The trackball 3 is used for instructions on the display screen on the CRT 7.

The high-speed arithmetic circuit 4 performs the processing of this embodiment (determining the difference, determining the membership function value from its absolute value,
(Creating evaluation functions, making judgments, and extracting organs) is performed exclusively for this purpose. The calculation memory 5 is used to store various intermediate data for this processing.

The display memory 6 stores an MHI display image and causes the CRT 7 to display it.

The recording medium 8 stores various data and image data. The MRI image exists in this medium 8, and during processing, the CPU and high-speed dedicated circuit 4 read and use it. The result is stored again in the medium 8.

FIG. 7 shows how MRI images are stored in the calculation memory 5. The first to fourth echo images are stored in the four memories 50 to 53. These echo images are those read from the recording medium 8.

On the other hand, one of the four images is being displayed on the CRT 7. The operator performs an organ extraction operation while looking at the display screen.

This organ extraction operation is as follows. At the time of this operation, the membership function for the extracted organ has already been created and the memory 8
Assume that it is stored in the form of a table in a part of .

Looking at the screen, the operator indicates the organ he wants to extract. This is the PI point. The CPUI reads this P, point,
Based on the premise that the PI is extracted from medium 8,
The membership functions μ1 to μ4 belonging to the points are read out and stored in the memory 5.

Next, while looking at the screen, the operator successively specifies (scans) the locations that are believed to be the extracted organs. For example, q1 point e q
Specify 2 points, etc. one after another. For each designation, the dedicated circuit 5 accesses the memories 50 to 53 to obtain the absolute values δ, δ2, δ3, and δ4 of the differences for each order at the designated point.

Next, the dedicated circuit 5 calculates the membership function values μ1−, μ2 . ,μ3. , μ4+, create an evaluation function J, and compare it with the reference value. If it is above the standard value,
This designated point q is recognized as an extracted organ, and extraction is performed.

FIG. 1 shows a flowchart of the extraction process of this embodiment. First, the position iflP+ is specified (step 20). Next, find the absolute value δi of the difference for each order (step 2l
). δ+=IRt! ,1...(5
) However, i is the order, and in this embodiment, i==1 to 4.

Next, from the table, use δ as a variable to find the membership function values μ■, μ2I・μ, μ4I, and find the product.
It is determined whether the value of this product is greater than the reference value 1 (step 22). If it is larger than the reference value, "1" is written in the point P of the extracted organ memory (part of the memory 5), and if it is smaller than the reference value, "O" is written in the point Pi (step 23 .24)
．． If the above processing is performed for all indicated points, only the area along the size of the extracted organ will be stored as “1” in the extracted organ memory.
'' is obtained, and the extracted organ is thus identified. Organ extraction can also be referred to as specifying this organ, but the value in this memory is "1", and the organ area is Therefore, this memory may be used to extract the image corresponding to 11 1 1+ from the memory shown in Fig. 6.The image extracted in this way is an extracted organ in the true sense. becomes.

Extracted organs can be used as three-dimensional display data.

Note that there are many cases in which characteristic curves appear in cases other than MHI, and the membership function of this embodiment can be applied as is as a method for determining whether two characteristic curves are the same. The concept of the evaluation function can also be applied as is. [Effects of the Invention] According to the present invention, it is possible to determine whether or not two characteristic curves are the same by using a membership function whose variable is the absolute value of the difference between two characteristic curves. can do.

Furthermore, according to the present invention, organs can be extracted with high accuracy without being affected by noise, especially from image data measured under the pin-echo method using multi-echo for MRI images. .

[Brief explanation of drawings]

Figure 1 is an example diagram of the processing flow of the present invention, Figure 2 is a diagram showing a conventional comparison method, Figure 3 is a diagram of the relationship between MRI images and orders in the spin echo method, and Figure 4 is a diagram of the present invention. FIG. 5 is an explanatory diagram of the absolute value variable of the difference for determining the membership function value. FIG. 6 is a diagram showing an embodiment of the processing device of the present invention. FIG. 7 is a diagram showing the order image memory. ■...CPU, 4...High-speed calculation circuit (dedicated hardware), 5...Memory for calculation.

Claims (1)

[Claims] 1. A characteristic curve comparison method for determining whether a reference characteristic curve and another characteristic curve have the same shape, (b) various arrangements of the reference characteristic curve and another characteristic curve; From a range of reference curves that can be used, find a membership function whose variable is the absolute value of the difference in height between the two curves at multiple points between the characteristic curve and the reference curve, and create a table. Find the absolute value of the difference between another characteristic curve and the reference characteristic curve, find the membership function value tabulated above from this absolute value, and (c) determine whether the two curves are the same from the function value. A characteristic curve comparison method characterized by: 2. In a characteristic curve comparison method for determining whether a standard characteristic curve and another characteristic curve have the same shape, (a) comparing the standard characteristic curve and another characteristic curve with reference curves in various possible ranges; Find a membership function whose variable is the absolute value of the difference in height between two curves at multiple points, and create a table.
(b) Find the differences at multiple points between another actual characteristic curve and the reference characteristic curve, and determine the membership function values at the multiple points tabulated above from the absolute values of the differences at the multiple points;
(c) An evaluation function value is determined according to the evaluation function from the membership function values at the plurality of points, and it is determined whether the two curves are the same based on the magnitude relationship between the evaluation function value and a reference value. Characteristic curve comparison method. 3. The characteristic curve comparison method according to claim 2, wherein the evaluation function is a product of membership function values at a plurality of points. 4. The characteristic curve comparison method according to claim 2, wherein the evaluation function is a sum of membership function values at a plurality of points. 5. The characteristic curve comparison method according to claim 1, 2, 3, or 4, wherein the characteristic curve is a relationship curve between echo order and pixel density on an MRI image obtained by a multi-echo spin echo method. 6. In an MRI image organ region extraction method that extracts a specific organ from an MRI image obtained by spin echo method using multi-echo, (a) MR using the echo order as a parameter;
(b) A reference characteristic curve that serves as a reference for the relationship between echo orders and MRI pixel densities for organs to be extracted, and organs to be extracted at various possible ranges of echo orders. Based on the degree of variation in MRI pixel density for , a membership function with the absolute value of the difference in height of two curves at multiple points as a variable is calculated and created in a table. Obtain the absolute value of the difference at a plurality of echo order points between the characteristic curve of the MRI image at any organ position above and the reference characteristic curve, and calculate the plurality of orders tabulated above from the absolute values at the plurality of order points. Find the membership function value at a point, (d) find the evaluation function value from the membership function value at the plurality of degree points according to the evaluation function, and determine whether the two curves are the same from the magnitude relationship between this evaluation function value and the reference value. A method for extracting an organ region from an MRI image, comprising: determining whether or not the two are the same, and (e) extracting the arbitrary organ position determined to be the same as an extracted organ position. 7. The method for extracting an organ region from an MRI image according to claim 6, wherein the evaluation function is a product of evaluation function values of a plurality of echo order points. 8. The method for extracting an organ region from an MRI image according to claim 6, wherein the evaluation function is a sum of evaluation function values of a plurality of echo order points.