JPH05233302A - Fuzzy inference knowledge verification method - Google Patents

Abstract

(57) [Abstract] [Purpose] It was described using the if part of the fuzzy rule described in the form if to then, and the membership function that defines the meaning of each proposition described in the then part. In a device having a function of performing fuzzy inference based on knowledge, a fuzzy inference result is visually displayed to make it easy to verify fuzzy knowledge. A fuzzy inference knowledge verification method is a fuzzy inference knowledge creating means 31, a means 32 for setting parameters for knowledge verification, a fuzzy inference function 33 and a fuzzy inference result value as a graph, or a two-dimensional image, or It is composed of a function 34 for displaying by replacing with three-dimensional graphics.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an automated system for checking work of a large amount of test data generated from a biochemical automatic analyzer, and more particularly to a verification system for creating check judgment knowledge.

[0002]

2. Description of the Related Art (1) Kasuga et al.
Name: Examination of disease type classification method using test values and its improvement by applying fuzzy theory: Proceedings of the 10th Joint Conference on Medical Informatics (Nov.1990), (2) Nishimura et al. 3 people: Fuzzy clinical examination database System: Proceedings of the 10th Joint Conference on Medical Informatics (Nov.1990), (3) Nishihata et al. 3
Name: Development and Evaluation of Contradictory Data Retrieval System: Journal of Japan Society for Clinical Laboratory Automation (Vol.11 No.5 198)
6), there is.

In the above prior art (1), when the membership function is set using the distribution map of measured values, the distribution map of measured values classified into the normal group and the abnormal group is used as the attribute of the measurement data. I am doing it.

Further, in the prior art (2), the form of the membership function is determined by using only the distribution of the normal group and statistically using the value of an integral multiple of the standard deviation.

Further, in the prior art (3), the measurement data is determined by utilizing the knowledge using the flowchart.

[0006]

DISCLOSURE OF THE INVENTION The above prior art (1),
In (2), there is no consideration for the verification means of the created discrimination knowledge. Further, in the prior art (3), basically, a relationship diagram of items a and b is created (eg, see FIG. 1), a regression equation is calculated from this, and a discrimination flow (eg. (See FIG. 2) has been created, but here too, no consideration has been given to the means for verifying the created knowledge.

The present invention is to provide a method of knowledge verification in which a relationship diagram of a plurality of items is displayed as a result of inference and which is easy to visually recognize. Specifically, it is to finally provide a method of displaying a relationship diagram of a plurality of items as shown in FIG. 1 as a result of the created knowledge.

[0008]

In order to solve the above problems, the following means are provided. (1) A fuzzy inference knowledge verification method is provided in which fuzzy inference result values using edited fuzzy rules and membership functions are replaced with graphs, 2D images, or 3D graphics for display. (2) The fuzzy inference result value has a function of using the representative value, the goodness of fit, or the calculation result using the representative value and the goodness of fit. (3) Various parameter setting means for performing fuzzy inference knowledge verification and arithmetic expression setting means are provided. (4) As the above-mentioned various parameters, set the numerical range of the condition (maximum value, minimum value), the division width, the value to be adopted as the fuzzy inference result value, and further, using the representative value and the goodness of fit as the arithmetic expression. A means for setting the formula was provided.

[0009]

The above-mentioned means operate according to the processing flow shown in FIG. (1) In step 31, fuzzy inference knowledge is created. (2) In step 32, parameters for knowledge verification are set. (3) In step 33, fuzzy inference is performed according to the parameters set in step 32. (4) In step 34, the fuzzy inference result value is replaced with a graph, a two-dimensional image, or a three-dimensional graphic for display.

The operation of each step will be described with an example. 1. Step 31 (Knowledge Creation) FIGS. 4A and 4B show examples of knowledge created in Step 31. This is described in the fuzzy rule FIG. 4 (a) expressed in the if-then type format for the numerical values of the item a and the item b, and the if part and the Zen part of the fuzzy rule. Membership function that defines the meaning of the proposition described in Fig. 4 (b). Specifically, the fuzzy rule-if portion is described by an ambiguous expression such as a high value or a low value of the item a, and the Zen portion is described by an ambiguous expression such as a high abnormality degree.

Further, the membership function for each proposition is described. 2. Step 32 (Knowledge verification parameter set) The following four types of knowledge verification are set.

(1) Minimum value (2) Maximum value (3) Numerical value range (4) Value adopted as fuzzy inference result (representative value, goodness of fit, or calculation result) Displayed by minimum value, maximum value and numerical value range The size of the grayscale image to be processed is determined.

For example, when displaying a two-dimensional image, the size of the image is determined by (maximum value-minimum value) / numerical width. Here, as an example, both items a and b have a minimum value of 1, a maximum value of 8 and a numerical value width of 1.
The output image at this time has a size of 8 × 8. Furthermore, the value adopted as the fuzzy inference result is a representative value.

3. Step 33 (fuzzy inference) The fuzzy knowledge created in step 31 and step 3
Fuzzy inference for verification is performed according to the various parameters set in 2. The calculation of the goodness of fit for each item value will be described with the membership function of FIG. Here, the operation when the item a = 7 and the item b = 6 are set will be described.

First, when the numerical value 7 is set for the item a, the conformity to the proposition "a is high" is:
It becomes 1.0, and the goodness of fit for the proposition "a is medium" is 0.4. Further, when the numerical value 6 is set for the item b, the goodness of fit for the proposition "b is high" is 0.6, and the goodness of fit for the proposition "b is medium" is 0.4. As a result, finally, the goodness of fit for the proposition "high degree of anomaly" becomes 0.6,
The suitability for the proposition that "the degree of abnormality is medium" is 0.4. Then, the membership function for the abnormality degree is multiplied by the goodness of fit.

The barycentric value of the membership function thus created and its fitness are calculated to be 7 and 5, respectively. At this time, the representative value is adopted as the final fuzzy inference result depending on the parameter setting conditions. That is, the final result is 7.

As described above, verification fuzzy inference is sequentially performed from the minimum value to the maximum value of item a and item b.

4. Step 34 (display output) FIG. 6 shows an example in which the result of the inference performed in step 33 is displayed as a two-dimensional image. Each numerical value is converted into a density value and displayed. At this time, various image processing means such as linear interpolation can be applied to adjust the size of the image.

[0019]

Embodiments of the present invention will be described below in detail with reference to the drawings. The present embodiment detects a measurement abnormality in the measurement value output from the biochemical analysis device, and further verifies the discriminant logic created for instructing retesting.
In the present embodiment, an example of outputting a two-dimensional image using numerical values of two items will be described.

FIG. 7 is a block diagram of an apparatus to which the present invention is applied, which comprises a biochemical analyzer 71 having an automatic reinspection mechanism, a measurement discriminating device 72, a display device 73, a keyboard 74 and a mouse 75. ..

FIG. 3 shows a discriminant logic creation processing flow in this embodiment. (1) In step 31, fuzzy inference knowledge is created. (2) In step 32, parameters for knowledge verification are set. (3) In step 33, fuzzy inference for verification is performed according to the parameters set in step 32. (4) In step 34, a grayscale image is displayed by the grayscale image output function of the fuzzy inference result value.

Next, FIG. 8 shows an example of fuzzy knowledge. Here, as an enzyme contained in serum, the knowledge to judge the necessity of retesting based on the magnitude of the numerical values of GOT and GPT will be shown.

In fuzzy rules, GOT and GPT
And, the knowledge to instruct whether to re-inspect or not is described according to the state of each ratio. For example, in the fuzzy rule 81, "GOT is abnormal, GPT / GOT
If (ratio of 2 items) is abnormal, re-examine. The knowledge of "is described. The same applies to the following fuzzy rules.

In the membership function, a function for each proposition described in the fuzzy rule is described. For example, in the membership function 82, if the item GOT is 40 or less, it means that it is definitely normal, and if it is 100 or more, it means that it is definitely abnormal. Indicates that the goodness-of-fit / abnormality fitness changes in the relationship shown in the figure. The membership function 83 is similar. Further, in the membership function 84, if the correlation value between the items GOT and GPT is 0.05 or less, or 11 or more, it means that it is definitely abnormal, and for the numerical value between 0.1 and 9, , Means normal. Further, the membership function 85 means that if the numerical value is 3 or less, the re-inspection is not surely performed, and if the numerical value is 7 or more, the re-inspection is surely performed.

The procedure for verifying this knowledge according to the present invention will be described below. First, the parameters for knowledge verification are set. For GOT, the minimum value is 0 or more and the maximum value is
It is set to 200. For GPT, the minimum value is 0 or more and the maximum value is 200. Further, the numerical value width is 1, and the value adopted as the fuzzy inference result is the representative value.

According to the parameters set above, the GOT
, Numerical values from 0 to 200 and GPT from 0 to 200 are input, fuzzy inference is performed, and a representative value corresponding to each numerical value is calculated. An example in which the result of fuzzy inference is output as a two-dimensional image is shown in FIG.

In the grayscale image shown in FIG. 9, basically, a low density portion (black portion) is seen along the x-axis (GOT axis) and the y-axis (GPT axis). This shading distribution gradually increases toward the origin, and further, centering on the origin,
It shows that the density gradually increases toward the x = y line segment. In this figure, the one with the lower density (black part)
Means that the sample is surely re-examined, and that the high-concentration part (white part) in the central part does not re-examine.

The fuzzy knowledge is verified by the grayscale image displayed on the screen in this way. If it is judged that the displayed fuzzy image is insufficient for the fuzzy knowledge, the fuzzy rule and the membership function are re-edited.

The above is an embodiment for verifying fuzzy knowledge using the present invention.

In the present embodiment, the result of the created fuzzy inference knowledge is visually judged to facilitate the verification of the fuzzy knowledge.

[0031]

As described above, according to the present invention, it is possible to display the fuzzy inference result in a visually easy-to-understand manner, thereby facilitating verification of fuzzy knowledge.

[Brief description of drawings]

FIG. 1 is an explanatory diagram showing a correlation between two items.

FIG. 2 is an explanatory diagram showing a flow of determination knowledge.

FIG. 3 is an explanatory diagram showing a processing procedure of the present invention.

FIG. 4 is an explanatory diagram showing an example of fuzzy knowledge.

FIG. 5 is an explanatory diagram showing a fuzzy inference operation.

FIG. 6 is an explanatory diagram in which a fuzzy inference result is output as a two-dimensional image.

FIG. 7 is an explanatory diagram showing a device configuration to which the present invention is applied.

FIG. 8 is an explanatory diagram showing fuzzy inference knowledge.

FIG. 9 is an explanatory diagram illustrating an output grayscale image.

[Explanation of symbols]

 31 ... Creation of fuzzy knowledge, 33 ... Activation of fuzzy inference.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Tomonori Mimura 882 Ige, Katsuta City, Ibaraki Prefecture Hitachi Ltd. Measuring Instruments Division

Claims (4)

[Claims] 1. A fuzzy rule described in the form of if-zen part, and a membership function that defines the meaning of each proposition described in the if part and then part of the fuzzy rule. In a device having means for performing fuzzy inference based on knowledge, the fuzzy rule, and means for editing the membership function, a graph of fuzzy inference result values using the edited fuzzy rule and the membership function, Alternatively, a fuzzy inference knowledge verification method characterized by displaying by replacing with a two-dimensional image or three-dimensional graphics. 2. A fuzzy inference knowledge verification method in which the fuzzy inference result value according to claim 1 indicates a representative value, a goodness of fit, or an operation result using a representative value and a goodness of fit. 3. A fuzzy inference knowledge verification method having various parameter setting functions and an arithmetic expression setting function when performing the fuzzy inference knowledge verification according to claim 1. 4. The various parameters described in claim 3 are: (1) numerical range of conditions (maximum value, minimum value), (2)
A division width, (3) a value to be adopted as a fuzzy inference result value, and the arithmetic expression is a fuzzy inference knowledge verification method indicating an arithmetic expression using a representative value and a goodness of fit.