JP2003263502A - Self-organizing map for medical examination and forming method thereof, and method and program for displaying health condition - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To visually inform a health condition which is comprehensive judged from results of a medical examination including a plurality of examination items. <P>SOLUTION: By utilizing data relating to a plurality of items of a medical examination, a self-organizing map is created using an algorithm for a self- organizing map. The results of a recipient who had a medical examination are displayed on a self-organizing map for a medical examination using marks such as 'Yours'. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

TECHNICAL FIELD The present invention relates to application of a self-organizing map (SOM) to a medical examination.

[0002]

2. Description of the Related Art In a conventional health examination, first, a medical examination is performed to obtain data (examination data) of each medical examination item (there are a plurality of items such as GOT and cholesterol). Subsequently, each of the acquired examination data is classified into normal, caution, reexamination required, medical treatment required, etc. Then, the classification results were shown for each abnormal examination item, and the subsequent treatment was instructed.

[0003]

However, in the conventional medical examination, it was judged whether or not there is an abnormality for each examination item, but it can be obtained only after comprehensively considering a plurality of examination items. No judgment was made regarding health status or lifestyle.

The present invention makes it possible to judge the health condition obtained by comprehensively considering a plurality of examination items, and to make the judgment visually. An object is to provide a creating method, a health status display method using a self-organizing map, and a health status display program.

[0005]

The self-organizing map for health examination according to claim 1 uses data on a plurality of examination items to be examined in a health examination from a plurality of units having reference vectors. A search is performed for a winning unit having a reference vector most similar to the input vector, and the reference vector of each unit included in the neighborhood area of the winning unit, which is sequentially narrowed according to the number of learning, is sequentially reduced according to the number of learning. It is updated based on a coefficient and the input vector, and is created by repeating the search for the winner unit and the update of the reference vector until the learning number reaches a preset number. According to the first aspect, it is possible to judge the health condition by comprehensively considering a plurality of checkup items to be checked in the health checkup, and to make the judgment visually.

The self-organizing map for health examination according to claim 2 is characterized in that a line is drawn to divide the self-organizing map into a plurality of health states. According to the second aspect, the line for dividing the health condition is drawn on the self-organizing map, which makes it easy to judge the health condition.

The self-organizing map for health examination according to claim 3 is characterized in that the self-organizing map is painted in different colors according to each of a plurality of health conditions. According to the third aspect, the self-organizing map is colored in different colors according to the health condition, which makes it easy to judge the health condition.

In the self-organizing map for health examination according to claim 4, the data regarding the plurality of examination items are normalized data obtained by normalizing the actual examination data of each of the examination items. If the actual examination data of the examination item is within the preset normal range of the examination item, the actual examination data is normalized to 1, and the actual examination data of the examination item is the minimum value of the normal range. If it is smaller, the actual medical examination data is normalized to a value obtained by dividing the actual medical examination data by a preset first standardized value, and the actual medical examination data of the medical examination item is normal. When the value is larger than the maximum value, the actual examination data is divided by the preset second standardized value, and when the division result is less than or equal to the predetermined cutoff value for the examination item, the actual examination data is divided. Medical examination day Normalized to the division result, when the division result is greater than the head cut value is characterized the fact that the normalized data is created by normalizing the actual examination data to the head cut value. According to claim 4, since the cut-off value is set, the self-organizing map for health examination affects the abnormal value even when the actual examination data includes the abnormal value. It has not been received.

According to a fifth aspect of the self-organizing map for health examination, the cutoff value is set according to the examination item. According to the fifth aspect, the self-organizing map is created by using the cutoff value suitable for each examination item.

In the self-organizing map for health check according to claim 6, each of the units is associated with the self-organizing map for the unit created from the examination items of the unit. Characterize. According to claim 6, since the self-organizing map for each unit is associated with each unit, more detailed judgment (health condition, treatment method, etc.) becomes possible.

According to a seventh aspect of the present invention, there is provided a self-organizing map creating method for physical examination, wherein an input vector which uses data relating to a plurality of medical examination items to be examined in a medical examination is selected from a plurality of units having reference vectors. Search for a winner unit having a similar reference vector, and sequentially narrow the reference vector of each unit included in the neighborhood of the winner unit that is narrowed according to the number of times of learning. Characterized in that the self-organizing map for health examination is created by repeating the search for the winner unit and the update of the reference vector until the number of learning reaches a preset number. To do. According to claim 7, a health check that enables a health condition to be determined by comprehensively considering a plurality of check items to be checked in the health check, and also allows the determination to be visually performed. A self-organizing map for can be created.

A health condition display method using a self-organizing map according to claim 8 is a health condition displaying method using the self-organizing map, wherein a health condition is selected from a plurality of units having reference vectors. Searching for a unit having a reference vector that is most similar to the vector using the data regarding the examination item of the diagnosis target person, and displaying a predetermined mark at the position on the self-organizing map corresponding to the searched unit. Is characterized by. According to claim 8, since it is indicated at which position on the self-organizing map for the health examination the health condition of the health examination target person is shown, the health examination target person considers a plurality of examination items comprehensively. It is possible to visually judge the self-health status obtained by the above.

A health condition display program using a self-organizing map according to claim 9 is a health condition display program using a self-organizing map using the self-organizing map, and is referred to by a computer. From a plurality of units having vectors, a procedure for searching a unit having a reference vector that is the most similar to the vector using the data related to the medical examination item of the physical examination target, and the self-organizing map corresponding to the searched unit. And a procedure for displaying a predetermined mark at the upper position. According to claim 9, it becomes easy to expand sales of the self-organizing map applied to the health examination.

[0014]

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, preferred embodiments of the present invention will be described with reference to the drawings.

First, the configuration of a computer according to an embodiment of the present invention will be described with reference to FIG. Figure 1
FIG. 1 is a configuration diagram showing a configuration of a computer according to an embodiment of the present invention. The computer functions as a self-organizing map creation device and a health condition display device.

The computer 1 includes a CPU 2, a keyboard 3, a hard disk 4, a ROM 5, and a RAM 6.
And a liquid crystal display 7 and an interface 8. In addition to the keyboard 3, a pointing device such as a mouse may be attached as an input means.

The keyboard 3 is provided with various keys such as a symbol key for inputting various characters and symbols and a return key which is pressed when performing various processes. The liquid crystal display 7 is for displaying characters and the like input using the keyboard 3, and the liquid crystal display 7 includes a simple matrix type such as STN system and DSTN system and an active matrix such as TFT. There are things such as methods. Interface 8 is
It is for connecting the computer 1 and an external device.

The hard disk 4 is a storage device for storing applications and data.
Stores various programs (such as a health status display program) for creating a self-organizing map. The ROM 5 stores the code that the CPU 2 or the like first executes, the code for loading the OS from the disk, and the like. The RAM 6 has a work area in which the CPU 2 works.

[0019]

[Table 1]

Table 1 will be described below. In Table 1, the label is a coded representation of a person. As for the checkup item, it will be checked in the medical examination B
Examination items such as MI (Body Mass Index) and blood pressure (upper and lower blood pressure) are shown. Further, the normal value represents the range of normal values of each examination item. The upper standard value (the second standardized value) is a value used when normalizing the medical examination data when the medical examination data of the medical examination item is larger than the range of the normal value, and the lower standard value (the first standardized value). The value) is a value used when normalizing the examination data when the examination data of the examination item is smaller than the normal value range. The cut-off value is the normalized value regardless of the value obtained when the value obtained from the actual examination data exceeds the “certain value” using the standard value. It is a "certain value" when resetting to "value". For example, when the cutoff value is 1.5, the value obtained by dividing the actual examination data by the normalized value is 1.6.
In the case of, the normalized value is reset to 1.5. As can be seen from Table 1, this cutoff value is set according to each examination item. 2 for each check item on each label
Although three values are described, the upper row shows the normalized data, and the lower row shows the values actually examined in the health examination (actual examination data).

The CPU 2 normalizes the checkup data of the actual checkup items to create normalized data (normalized data creation process). Further, the CPU 2 creates a self-organizing map for health examination according to the algorithm of the self-organizing map (self-organizing map creating process). Furthermore, CPU2
Displays the health state on the self-organizing map for health examination based on the examination data of the examination item (health state display processing).

Further, the CPU 2 will be described in detail with reference to FIG. FIG. 2 is a functional block diagram of the CPU.

As shown in FIG. 2, the CPU 2 functions as a normalized data creating unit 21, a self-organizing map creating unit 22, and a health condition display unit 23.

The normalized data creation unit 21 is a block for performing a normalized data creation process, and functions as a data creation unit 31 and a data correction unit 32. Data creation unit 31
Determines whether the actual examination data of the examination item is within the range of normal values, and if it is within the range of normal values, normalizes the actual examination data to 1. In addition, the data creation unit 3
1 is the case where the actual medical examination data is not within the range of the normal value and the actual medical examination data is smaller than the minimum value of the normal value, the actual medical examination data is set to a value below the standard value of the medical examination item (table. 1)) and normalize the actual examination data to the division result. Further, when the actual examination data is not within the range of the normal value and the actual examination data is larger than the maximum value of the normal value, the data creation unit 31 sets the actual examination data to the standard value of the examination item. Divide by the above value (see Table 1) and normalize the actual screening data to the division result.

The data correction unit 32 determines whether or not the value normalized by the data creation unit 31 is larger than a value predetermined for the examination item, and if it is larger, it is normalized by the data creation unit 31. Value is corrected to the cutoff value.

Here, the processing of the normalized data creation unit 21 functioning as the data creation unit 31 and the data correction unit 32 will be described by taking the specific values in Table 1 as an example.

The process of obtaining the normalized data "1" from the value "165" of the actual screening data of total cholesterol labeled 4fc will be described. The data creation unit 31 of the normalized data creation unit 21 determines that the value “165” of the screening data is within the range of the normal value of total cholesterol (greater than 150 and 199
The value “165” of the medical examination data is normalized to “1”.

The process of obtaining the normalized data "0.97" from the actual screening data value "146" of the total cholesterol labeled 3f3 will be described. The data creation unit 31 of the normalized data creation unit 21 uses the value “1
46 ”is not included in the normal value range of total cholesterol (range greater than 150 and less than 199) and is smaller than the minimum value in the range of normal value. Then, the data creation unit 31 determines the value“ 146 of the screening data ”. "0.97" is obtained by dividing "" by the value "150" below the standard value of total cholesterol, and "0.97" obtained by the division is taken as the normalized data.

The process of obtaining the normalized data "1.07" from the actual screening data value "212" of total cholesterol labeled 2f2 will be described. The data creation unit 31 of the normalized data creation unit 21 uses the value “2
12 ″ is not included in the normal value range of total cholesterol (range greater than 150 and less than 199) and is determined to be larger than the maximum value in the range of normal value. Then, the data creation unit 31 determines the value “212” of the examination data. "1.07" is obtained by dividing "" by the standard cholesterol value "200". Then, the data correction unit 32 of the normalized data creation unit 21 obtains "1. It is determined that 07 is less than or equal to the cutoff value "1.7", and the value of the normalized data is left as the division result "1.07".

The process of obtaining the normalized data "1.1" from the value "7" of the actual examination data for uric acid having the label 4m5 will be described. The data creation unit 31 of the normalized data creation unit 21 determines that the value “7” of the examination data is not included in the normal value range (less than 5.9) of uric acid and is larger than the normal value range. Then, the data creation unit 31 divides the value “7” of the medical examination data by the value “6” on the standard value of uric acid,
I get "1.17". Then, the normalized data creation unit 21
The data correction unit 32 of “1.17” obtained by the division
It is determined that the value of is larger than the cutoff value “1.1”, and the value of the normalized data is set to the cutoff value “1.1”.

The above-described processing is performed by the normalized data creating section 21 to obtain the normalized data for all the examination items for each label in Table 1.

The self-organizing map creating unit 22 is a block for performing a self-organizing map creating process, and includes an initial value setting unit 41, an input vector presenting unit 42, a winner unit searching unit 43, and a reference vector updating unit. 44 and the learning number determination unit 45.

The initial value setting unit 41 presets the size of the network (the number of arrays of units), the number of times the reference vector is updated for one input vector (total learning number) T, and the unit Near-phase shape (for example,
Either a rectangular grid type or a hexagonal grid type), an initial value Nc (0) of the neighborhood area, and an initial value α (0) of the learning rate coefficient are set. Further, the initial value setting unit 41 initializes a plurality of reference vectors and determines the presentation order of a plurality of input vectors.

Here, the input vector is a vector whose elements are the normalized data of each examination item (BMI, blood pressure up, blood pressure down, total cholesterol, etc.) normalized by the normalization data creation unit 21, and Each label of 1 (2
fc, 2m4, ...). Further, the reference vector is a vector having the same dimension as the input vector and exists for each unit. Note that the input vector is x
(T) and the reference vector are described as m _i (t). t is
It is a discrete time coordinate (t = 0, 1, 2, ...) And corresponds to the number of learnings currently performed.

The input vector presenting section 42 selects and presents the input vectors x (t) one by one according to the presentation order determined by the initial value setting section 41.

The winner unit search section 43 calculates the Euclidean distance from the input vector x (t) for each of the plurality of reference vectors. Then, the winner unit search unit 43 finds the reference vector having the closest (most similar) Euclidean distance to the input vector x (t) based on the calculated Euclidean distance, and the found reference is found. The unit with the vector is determined as the winner unit.

The reference vector updating unit 44 updates the reference vector on the basis of the input vector x (t), and includes a neighborhood area calculating unit 44a and a learning rate coefficient calculating unit 44b.
And functions as the reference vector calculation unit 44c.

The neighborhood area calculation unit 44a includes an initial value setting unit 4
The following equation (1) is calculated using the initial value Nc (0) of the neighborhood area set by 1 and the learning number t of the learning currently performed, and the neighborhood area N in the learning currently performed.
Calculate c (t). This neighborhood area Nc (t) becomes narrower as the learning progresses. However, the neighborhood area Nc
The reference vector of the unit included in (t) is updated (Equation (3) below), and the reference vector of the unit not included in the neighboring region Nc (t) is not updated (Equation (4) below). Nc (t) = Nc (0) × (1-t / T) (1)

The learning rate coefficient calculation unit 44b is currently being operated with the initial value α (0) of the learning rate coefficient set by the initial value setting unit 41 and the total number of learning times T set by the initial value setting unit 41. The following equation (2) is calculated using the learning number t of learning, and the learning rate coefficient α (t) in the learning currently performed is calculated. The learning rate coefficient α (t) becomes smaller as the learning progresses. α (t) = α (0) × (1-t / T) (2)

The reference vector calculation unit 44c learns the learning rate coefficient calculation unit 44b for each of the reference vectors of the units included in the neighborhood region Nc (t) obtained by the neighborhood region calculation unit 44a. Rate coefficient α (t)
By calculating the following equation (3) by using the reference vector m _i
(T) is updated, and the updated reference vector m _i (t + 1)
To get Further, the reference vector calculation unit 44c calculates the following expression (4) for each of the reference vectors of the units that are not included in the neighborhood area Nc (t) to calculate the reference vector m.
_i (t + 1) is obtained. m _i (t + 1) = m _i (t) + α _i (t) × (x (t) −m _i (t)) (3) m _i (t + 1) = m _i (t) (4)

The learning number determination section 45 determines whether or not the learning number t of the learning currently being performed has reached the total learning number T set by the initial value setting section 41.

Further, the self-organizing map creating process of the self-organizing map creating unit 22 will be described with reference to FIGS. 3 and 4. FIG. 3 is a flowchart showing the procedure for creating a self-organizing map, and FIG. 4 is an explanatory diagram for explaining the algorithm of the self-organizing map.

In step S101, the initial value setting unit 41 of the self-organizing map creating unit 22 presets the size of the network, the total learning number T, the phase neighborhood shape of the unit, and the initial value Nc (0 of the neighborhood area. ), The initial value α (0) of the learning rate coefficient is set. Then, step S1
In 02, the initial value setting unit 41 of the self-organizing map creation unit 22 initializes the reference vectors so that the reference vectors are different from each other (FIG. 4).
Together with (a)), the presentation order of input vectors is determined.

In step S103, the input vector presenting section 42 of the self-organizing map creating section 22 presents one input vector x (t) in the order of presentation set by the initial value setting section 41 (FIG. 4 (b)). ).

In step S 104, the winner unit searching section 43 of the self-organizing map creating section 22 inputs the input vector x presented by the input vector presenting section 42.
Find the winning unit whose reference vector has the closest Euclidean distance to (t) (FIG. 4 (c)). However, in FIG. 4 (c), the unit having the vector denoted by the symbol w in the figure is the winner unit.

In step S105, the reference vector updating unit 44 of the self-organizing map generator 22 uses the learning rate coefficient α (t) for each of the reference vectors of the units included in the neighborhood area Nc (t) of the winner unit. The input vector x (t) is then updated (learned).

More specifically, the neighborhood area computing unit 44a of the reference vector updating unit 44 computes the neighborhood area Nc (t) by computing the above equation (1). Each unit having the reference vector included in this neighborhood region Nc (t) is a unit updated in the learning currently performed. For example, in FIG. 4D, the reference vector in the area surrounded by the frame f in the figure is the reference vector to be updated. The learning rate coefficient calculation unit 44b of the reference vector updating unit 44 calculates the learning rate coefficient α (t) by calculating the above equation (2). Then, the reference vector calculation unit 4 of the reference vector update unit 44
4c uses α (t) calculated by the learning rate coefficient calculation unit 44b and the input vector x (t) for each reference vector included in the vicinity region Nc (t) calculated by the vicinity region calculation unit 44a. Then, the equation (3) is calculated and updated (FIG. 4 (d)). The reference vector calculation unit 44c also performs a process of calculating the reference vector m _i (t + 1) by calculating the above equation (4) for each of the reference vectors of the units that are not included in the neighborhood area Nc (t). .

In step S106, the learning number determination unit 45 of the self-organizing map creating unit 22 determines whether the learning is performed T times with respect to the input vector presented in step S103, the total learning number T times set in step S101. To determine. If it is determined that the learning has not been performed the total number of times of learning T times (S106: NO), since the number of times of learning that has been initialized has not been performed, the process returns to step S104 and the processes of steps S104 to S106 are performed. Is done. On the other hand, when it is determined that the learning has been performed T times in total (S106: YES), step S10
The processing shifts to 7.

In step S107, the self-organizing map generator 22 determines whether or not learning for updating the reference vector has been performed for all input vectors. If it is determined that learning has not been performed for all input vectors, that is, there is an input vector that has not been learned (S107: NO), learning is performed for input vectors in the next presentation order. Returning to the processing, the processing from step S103 to step S107 is performed. On the other hand, when it is determined that learning has been performed for all input vectors (S107: YES), the process of creating the self-organizing map ends.

By performing the processing according to the algorithm of the self-organizing map described above on the input vector having the normalized data shown in Table 1 as an element, the self-organizing for health check as shown in FIG. A map is created. The self-organizing map for health checkup shown in Fig. 5 shows renal disorder, diabetes,
It is divided into a plurality of areas such as liver damage, and lines are drawn at the boundaries of these areas. This was divided by considering the screening data in Table 1. Figure 5
An area that is not included in any of the areas corresponds to an area where health is not impaired. Further, a region where a plurality of regions overlap (for example, a region where an alcohol drinker and a liver disorder overlap) is a region where both health conditions are impaired.

Since the self-organizing map shown in FIG. 5 is created according to the algorithm of the self-organizing map from the input vector including all the examination items, the self-organizing map shown in FIG. It is possible to obtain a health condition (a health condition in which abnormal examination items cannot be obtained individually) only when comprehensive consideration is given to the examination items. Further, since a plurality of health conditions are divided by drawing a line on the self-organizing map, it is possible to easily judge the health condition visually.

Returning to the description of the CPU 2, the health condition display unit 23 of the CPU 2 is a block for performing the health condition display processing, and includes the mark position determination unit 51 and the mark display control unit 5.
Functions as 2.

The mark position determining unit 51 uses the normalization data (the above-mentioned normalization data creating unit 21 to create the normalization data created from the actual screening data input by using the input screen for the health screening data shown in FIG. 6). A reference vector having the closest Euclidean distance to the vector having the created normalized data) as an element (the reference vector updated by the reference vector update unit 22 described above) is searched for. Then, the mark position determining unit 51 determines the position on the self-organizing map corresponding to the unit having the reference vector having the shortest Euclidean distance as the position of the mark indicating the health condition of the health examination target person. The input screen shown in FIG.
By moving the scroll bar, the input areas for all the examination items are displayed.

The mark display control section 52 controls the display of a mark (for example, "Yours") at the position on the self-organizing map determined by the mark position determining section 51.

In summary, a unit having a reference vector whose Euclidean distance is closest to a vector whose elements are normalized data created from actual examination data input using the input screen shown in FIG. 6 is marked. It is searched by the position determination unit 51. Then, the mark is displayed at the position on the self-organizing map corresponding to the searched unit under the control of the mark display control unit 52. The position of the mark on the self-organizing map makes it possible to visually understand the health condition obtained only after comprehensively considering a plurality of examination items. That is, instead of notifying an abnormal value of the examination item for each examination item as in the conventional case, the comprehensive health state of a plurality of examination items is judged and the health state is visually indicated. For example, when the mark is included in the area of renal damage, instead of indicating to the physical examination subject which of the examination items has an abnormal value,
It will indicate that the patient has a direct renal disorder. Even if the mark is included in an area where health is not impaired, it can be determined that there is a possibility of illness in that area if it is in the vicinity of any area, It is possible to take measures to maintain good health after understanding what kind of illness is likely to occur.

According to the embodiment described above, the self-organizing map is created according to the algorithm of the self-organizing map by using the normalized data obtained by normalizing the actual examination data in the health check, and this is created. Since the position of the physical examination target on the map is displayed on the self-organizing map, the physical examination target visually indicates the health condition of the self determined by comprehensively considering multiple examination items. You can figure it out.

Further, since the cut-off value of the medical examination item is set, even if the actual medical examination data includes an abnormal value, the self-organizing map for health examination has the abnormal value. Is not affected by. In addition, since we set the cutoff value for each examination item,
The self-organizing map can be created by a cutoff value suitable for each examination item.

Although the preferred embodiments of the present invention have been described above, the present invention is not limited to the above-described embodiments, and various design changes are possible within the scope of the claims. Is. For example, in the above-described embodiment, the health condition on the self-organizing map is clarified by drawing a line that divides the plurality of health conditions on the self-organizing map. The health status on the self-organizing map may be clarified by separately painting the self-organizing map. For example, areas that are not in health are white, diabetes is red, renal damage is blue, and so on. It should be noted that the self-organizing map may be processed so that it is possible to easily and visually determine in which health condition the position on the self-organizing map is included.

Further, in the above embodiment, the neighborhood region Nc
This is an example in which the formula (1) is used in the calculation of (t), but this is an example, and the calculation method of the neighborhood region Nc (t) has heretofore been performed by creating a self-organizing map. Other calculation methods such as the calculated calculation method may be used. Further, the case where the formula (2) is used in the calculation of the learning rate coefficient α (t) is an example, and the learning rate coefficient α
The calculation method of (t) may be another calculation method such as the calculation method that has been used for creating the self-organizing map. Furthermore, the case where the formula (3) is used when updating the reference vector is an example, and the update method of the reference vector is the update that has been performed so far in the creation of the self-organizing map. Other update methods such as a method may be used.

Further, FIG. 5 in the above-mentioned embodiment.
For each unit (corresponding to one reference vector) of the self-organizing map (first layer) shown in Fig. 7, a self-organizing map (second layer) corresponding to it is created as shown in FIG. To do. This self-organizing map has a health state classified in more detail than the self-organizing map shown in FIG.

In creating this self-organizing map,
First, the medical examination items determined according to the health condition on the self-organizing map (first layer) of the unit (for example, the medical examination items when included in the renal disorder area are nitrogen urate (BU
N), creatinine, uric acid) to create normalized data from actual screening data. Then, based on the created data, the reference vector is updated according to the algorithm of the self-organizing map, and the self-organizing map of the second layer is created. In this way, a two-level self-organizing map is created.

In this two-level self-organizing map, first, when the input screen as shown in FIG. 6 is displayed on the display and the data of the examination items are inputted, the self-organizing map shown in FIG. 5 is displayed. Your health status is displayed as a mark such as "Yours". Then, when the displayed mark is clicked, an input screen of the examination item for the second layer self-organizing map corresponding to the clicked position is displayed on the display. Then, when the data of the examination item is input, the self-organizing map created corresponding to the position (unit) of the mark is displayed on the display, and the self-organizing state is displayed on the self-organizing map. The position of is displayed with a mark such as "Yours". By creating a two-level map in this way,
It becomes possible to obtain more detailed information from the result of the health examination. If the first input screen allows you to enter the required examination items for all layers, you can click one of the marks such as “Yours” without displaying the input screen in the middle. The self-organizing maps in the lower hierarchy will be displayed in sequence. In addition, if you further increase the number of layers and click the position of the self on the self-organizing map of the last layer, for example, a list of therapeutic drugs, immediate hospitalization, "Let's see for a while"doctor's prescription May be configured to be displayed.

Here, in the multi-layered self-organizing map, an example is shown in which the first layer is a comprehensive health evaluation (such as the self-organizing map of FIG. 5), the second layer is each disease, and the third layer is a treatment. Write down. For example, if the comprehensive evaluation of the first layer is in the area of liver injury,
The classification of the second layer is type B, hepatitis C virus antigen / antibody test, liver ultrasound image evaluation items, biliary enzyme levels, drinking history, etc., and the classification of the third layer is follow-up, lifestyle improvement, Administration of liver protection agents, interferon therapy,
For example, internal medicine / surgical treatment for cancer. Further, when the comprehensive evaluation of the first layer is in the hypertensive region (not shown in FIG. 5), the classification of the second layer is an evaluation item of an echocardiographic examination, an electrocardiographic finding,
Chest X-ray examination evaluation items, renal lenogram evaluation items, etc., and the third layer classification is follow-up, lifestyle improvement, antihypertensive agents (administered by drug), etc. Also, the first
If the comprehensive evaluation of the layers is in the hyperlipidemic region (not shown in FIG. 5), the classification of the second layer is the lipoprotein fraction measurement value, thyroid hormone test value, liver enzyme value, etc. The classification of layers is follow-up, lifestyle improvement, and treatment of diseases that cause internal medication (branching into each drug) (hypothyroidism, obstructive jaundice, etc.).

The self-organizing map of the first layer is a self-organizing map for comprehensive health evaluation (the self-organizing map of FIG. 5, etc.), and the self-organizing map of the second layer is a self-organizing map for lifestyle. It may be a map (map of daily activities, eating habits, smoking habits, drinking habits, etc.).

Furthermore, values other than the values shown in Table 1 may be used as the value of the head cut of the examination item. Further, although the normalized data is used when creating the self-organizing map, the self-organizing map may be created using the actual examination data itself. Further, the input screen shown in FIG. 6 may be displayed on a display provided in a terminal connected to the computer 1 via the public line network, and the medical examination data input from the terminal may be transmitted to the computer 1.

[0066]

As described above, according to the first aspect of the present invention, it is possible to make a judgment of the health condition by comprehensively considering a plurality of examination items to be examined in the medical examination, and to make the judgment visually. It will be possible to do.

According to the second aspect of the present invention, the self-organizing map is drawn with a line that divides the health condition, which makes it easy to judge the health condition.

According to the third aspect of the invention, the self-organizing map is colored in different colors according to the health condition, which makes it easy to judge the health condition.

According to the fourth aspect, since the cutoff value is set, even if the actual examination data includes an abnormal value, the self-organizing map for health check has the abnormal value. It is not affected by the value.

According to the fifth aspect, the self-organizing map is created by using the cutoff value suitable for each examination item.

According to claim 6, since each unit is associated with the self-organizing map for the unit,
More detailed judgment (health status, treatment method, etc.) becomes possible.

According to claim 7, it is possible to make a judgment of the health condition by comprehensively considering a plurality of examination items to be examined in the medical examination, and to make the judgment visually. It is possible to create a self-organizing map for physical examination.

According to the eighth aspect, it is indicated at which position on the self-organizing map for the health checkup the health condition of the health checkup subject, so that the health checkup subject comprehensively selects a plurality of examination items. It is possible to visually judge one's own health condition obtained in consideration.

According to the ninth aspect, it is easy to expand the sales of the self-organizing map applied to the medical examination.

[Brief description of drawings]

FIG. 1 is a configuration diagram showing a configuration of a computer according to an embodiment of the present invention.

FIG. 2 is a functional block diagram of a CPU.

FIG. 3 is a flowchart showing a procedure for creating a self-organizing map.

FIG. 4 is an explanatory diagram for explaining an algorithm of a self-organizing map.

FIG. 5 is a diagram showing a self-organizing map.

FIG. 6 is a diagram showing an example of a screen for inputting examination data.

FIG. 7 is an explanatory diagram for explaining a two-layer self-organizing map.

[Explanation of symbols]

1 computer 2 CPU 21 Normalized data creation section 22 Self-Organizing Map Creation Department 23 Health status display 31 Data creation department 32 Data correction section 41 Initial value setting section 42 Input vector presentation section 43 Winner Unit Search Section 44 Reference Vector Update Unit 44a Neighborhood area calculation unit 44b Learning rate coefficient calculator 44c Reference vector operation unit 45 Learning frequency determination unit 51 Mark position determination unit 52 mark display control unit

   ─────────────────────────────────────────────────── ─── Continued front page    (71) Applicant 502083152             Usami Makoto             4-7-3 Azalea Hill, Tarumi-ku, Kobe City, Hyogo Prefecture (72) Inventor Heizo Tokutaka             4-637 North Kitayamacho, Tottori City, Tottori Prefecture (72) Inventor Kikuro Fujimura             1-357 Koyamacho Nishi, Tottori City, Tottori Prefecture (72) Inventor Yoshiro Maniwa             571-1 Miyata, Wadayama Town, Asago District, Hyogo Prefecture (72) Inventor Makoto Usami             4-7-3 Azalea Hill, Tarumi-ku, Kobe City, Hyogo Prefecture

Claims (9)

[Claims] 1. A winning unit having a reference vector that is most similar to an input vector using data relating to a plurality of examination items to be examined in a medical examination is searched from a plurality of units having a reference vector, and the winning unit is searched according to the number of times of learning. The reference vector of each unit included in the vicinity region of the winner unit that is sequentially narrowed is updated based on the learning coefficient and the input vector that are sequentially reduced according to the number of times of learning, and the number of times of learning is set in advance. A self-organizing map for physical examination, which is created by repeating the search for the winning unit and the updating of the reference vector until the number of times is reached. 2. The self-organizing map for medical examination according to claim 1, wherein the self-organizing map is drawn with a line that divides the self-organizing map into a plurality of health states. 3. The self-organizing map for health examination according to claim 1, wherein the self-organizing map is colored in different colors according to each of a plurality of health states. 4. The data regarding the plurality of examination items includes
Normalized data obtained by normalizing the actual medical examination data of each of the medical examination items, and when the actual medical examination data of the medical examination item is within a preset normal range of the medical examination item, the actual medical examination data The data is normalized to 1, and when the actual examination data of the examination item is smaller than the minimum value of the normal range, the actual examination data is preset to the first value.
Normalized the actual medical examination data to a value obtained by dividing by the standardized value of, when the actual medical examination data of the medical examination item is larger than the maximum value of the normal value, the actual medical examination data is preset. When the division result is less than or equal to the cutoff value predetermined for the examination item, the actual examination data is normalized to the division result, and the division result is the cutoff value. The health check according to any one of claims 1 to 3, wherein the normalized data is created by normalizing the actual medical examination data to the cutoff value when larger. Self-organizing map for. 5. The self-organizing map for health examination according to claim 4, wherein the cutoff value is set according to the examination item. 6. The self-organizing map for the unit, which is created from the examination items of the unit, is associated with each of the units of the self-organizing map. 5. The self-organizing map for health check according to any one of 5 above. 7. A winning unit having a reference vector that is most similar to an input vector utilizing data relating to a plurality of examination items to be examined in a medical examination is searched from a plurality of units having a reference vector, and the winning unit is searched according to the number of times of learning. The reference vector of each unit included in the vicinity region of the winner unit that is sequentially narrowed is updated based on the learning coefficient and the input vector that are sequentially reduced according to the number of times of learning, and the number of times of learning is set in advance. A method for creating a self-organizing map for medical examination, characterized in that the search for the winning unit and the updating of the reference vector are repeated until the number of times is reached to create a self-organizing map for medical examination. 8. The self-organizing map for health examination according to any one of claims 1 to 6, or the self-organizing map for health examination according to claim 7, A method of displaying a health condition using a self-organizing map, wherein a unit having a reference vector that is most similar to a vector that uses data related to a medical examination subject's examination item is searched from a plurality of units having a reference vector. Then, a predetermined state is displayed at a position on the self-organizing map corresponding to the searched unit, and a health condition display method using the self-organizing map. 9. The self-organizing map for health examination according to any one of claims 1 to 6, or the self-organizing map for health examination according to claim 7, A health condition display program that uses a self-organizing map, and has a reference vector that is the most similar to the vector that uses the data related to the examination items of the health checkup target from among multiple units that have reference vectors in the computer. Health condition display program using a self-organizing map for executing a procedure for searching a unit and a procedure for displaying a predetermined mark at a position on the self-organizing map corresponding to the searched unit .