JP2010061389A - Interactive interface for medical diagnostic support - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an interactive interface for displaying a relationship between the result of the analysis of check data of blood or the like and the degree of health for spontaneous health care so that it is easily understandable for a client and a doctor. <P>SOLUTION: An interactive interface is configured for medical diagnostic support for displaying a relationship between the result of the analysis of the check data of a client and the degree of health in the format of a decision tree, and provided with: a data input part 1 for accepting the input of data such as the check items, check values, the degree of health, and medical record information of a client; first and second and third characteristic storage parts 2, 3, and 4 for storing the data by featured value; a characteristic extraction part 5 for extracting a featured value for creating the decision tree; an inference part 6 for determining the decision tree by using the extracted featured value and the degree of health and the information of a disease possibility site; a display part 8 for displaying various screens; and a display control part for controlling the display part 8 to display the relationship between the analytic result of the check data of the client and the degree of health in the form of the decision tree determined by the inference part 6. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to an interface for medical diagnosis support, for example, to raise the client's awareness of health management using test items such as blood such as γGTP and total cholesterol and their test values, which are performed in a health checkup, for example. Concerning the interactive interface.

  Conventionally, blood tests are performed in health checkups, and about 10 items such as γGTP and total cholesterol are measured by an analyst using about 50 to 100 samples for each item, and the average value and normal range are measured and passed to the client. In general, whether or not a diagnosis result is in a normal range is printed on paper and mailed.

  However, when an abnormal value of a single test item (for example, TG (neutral fat) male under 50 years old, 40-170 mg / dl is normal), when a client is male and under 50 years old is 180 mg / dl It is impossible to judge one's health management only with an abnormally high value, and with another male, the abnormally low value is 38 mg / dl. ALT (GPT) and AST (GOT) are indicators related to the liver, but the relationship between these items is statistically meaningful for the client, what correlation is there for the doctor, and other diseases are possible The fact that knowledge about the relevance about sex etc. is not examined about a large amount of data. Therefore, there are currently the following problems.

  That is, the first problem is the actual situation that a normal value range of a single item is obtained in a population of about 50 to 100 people.

  The second problem is that when trying to analyze the relationship between examination items, the examination items differ depending on the client (for example, client A has 14 items of health examination items, but other 30 items of data). On the other hand, client B has data for these 30 items, but there is no data for 14 items for health examinations, etc., so the same number of test items cannot be collected for quantitative numerical analysis. It is.

  The third problem is that even if a quantitative numerical analysis between items is possible, an interface that is easy to understand for clients, doctors, laboratory technicians, researchers, etc. is not provided for medical examinations. It is said that we are dealing with clients based on empirical rules without looking at quantitative analysis.

  The fourth problem cannot be dealt with, for example, in the case of a disease in which a tumor progresses rapidly within 3 months, with only a regular health check once a year. Establishing a system that can test blood by simple means on a daily basis will allow doctors to diagnose health and dietary methods that prevent these diseases. Therefore, an interface that can describe the relationship between inspection items and display it in an easy-to-understand manner is required.

  The present invention solves the problems of the prior art as described above, and interactively displays the relationship between the analysis result of blood test data and the health level for voluntary health management in an easy-to-understand manner for clients and doctors. It is an object to provide an interface.

  In order to solve the above problems, first, the present invention provides an interactive interface for medical diagnosis support for displaying a relationship between an analysis result of a client test data and a health level in the form of a decision tree. Test items and test values, authentication ID, test date, health level, data input unit for receiving test data such as disease potential sites and client chart information, and client test items received by the data input unit A first feature storage unit that stores test data such as a test value, an authentication ID, a test date, a health degree, a disease-prone portion, etc. as a first feature amount, and a client test item received by the data input unit The test items related to each type of disease are classified into a plurality of classes, and the number of items having abnormally high values or abnormally low values included in each class is defined as the second feature amount. A second feature storage unit that stores information, a third feature storage unit that stores health degree information, a disease-possible site information, and client chart information received by the data input unit, and the first feature storage unit A feature extraction unit for extracting a feature amount for generating the decision tree from the feature amount, a feature amount extracted by the feature extraction unit, and a health degree and disease-possible part stored in the third feature storage unit An inference unit that obtains the decision tree using the information of; a display unit that displays various screens; and the first, second, and third feature storage units, and the decision tree obtained by the inference unit. A display control unit that controls the display means to display the relationship between the analysis result of the test data of the client and the health level in the form, and the display control unit has an abnormal high value for each test item and each class, Abnormally low The first display control mechanism that controls the display unit to distinguish the difference between the abnormally high value or the abnormally low value and the other normal value range, and the health level and the relationship for each client A second display control mechanism for controlling the display unit to display the number of items having a deep inspection item and an abnormally high value or an abnormally low value in the order of the relationship depth, and the first display control mechanism. There is provided an interactive interface for supporting medical diagnosis, characterized by having a third display control mechanism for associating display contents to be displayed with display contents controlled by the second display control mechanism by hypertext.

  Second, the present invention is characterized in that, in the first invention, the second display control mechanism performs control so that other client data having similar items are displayed in association with each other. Provides an interactive interface for supporting medical diagnosis.

  Third, the present invention provides the first or second invention, wherein the inference unit performs matching using the decision tree to test data without a health level, and estimates the health level. An interactive interface for medical diagnosis support characterized by

  Fourth, the present invention provides the decision tree according to any one of the first to third inventions, wherein the inference unit obtains the health level and the disease possibility portion of the client stored in the third feature storage unit. The health level of the client and the potential disease site stored in the third feature storage unit, and the health of the collation result are provided. When the degree and the disease possibility part are different, the difference and the chart information about the client stored in the third feature storage unit are displayed, and the correct health degree and the disease possibility part are corrected. Provide an interactive interface for medical diagnosis support.

  According to the present invention, by adopting the configuration as described above, an interactive and easy-to-understand display of the relationship between the analysis result of blood test data and the health level for voluntary health management to clients and doctors, etc. An interface can be provided.

  In addition, according to the present invention, in a health checkup based on blood test items that have become established in Japan, even if the population has increased, it has been input correctly in advance so that the determination result of the health degree diagnosed by the doctor can be correctly input. Using test data such as blood that has been confirmed, use a decision tree that describes the relationship between the test item and the health level to determine whether the health level input value for the newly entered test data is correct. By collating the determination result, the input value, and the chart image data, a correct input value can be input (corrected), thereby making it possible to construct a large amount of database.

  Also, in the present situation, for example, if only the abnormal values of a single item are tabulated for inspection items exceeding 100 items, what kind of items are related to each other and the degree of health is determined by the client It is difficult to understand, but it is pointed out that doctors do not always clearly understand the relationship between non-specialized inspection items. There was no suggestion of a display interface to understand. According to the present invention, it is possible to provide a method that can identify at a glance which examination items contributed to the determination of the health level of each client under such a current situation.

  Hereinafter, an interactive interface (hereinafter also simply referred to as an interface) according to the present invention will be described in detail.

  This interface shows the relationship between the analysis result of blood test data and the health level, thereby enabling detailed analysis of each data. Here, the health level means an index used for voluntary health management so as not to get sick, and the stages are, for example, five levels of I, II, III, IV, and V for the client. It can be. In this case, stage I is an ideal health state, stage II is almost healthy, stage III is a state in which daily life needs to be improved, stage IV is a state in which screening and improvement of daily life are indispensable, and stage V is a fine state. A condition that requires medical examination and treatment. For doctors and researchers, it can be further divided into 8 stages of I, II, III, IVa, IVb, IVc, Va, and Vb.

  FIG. 1 is a functional block diagram showing a configuration example of an interactive interface according to the present invention, and FIG. 2 is a schematic block diagram of hardware when the interactive interface is realized by a personal computer.

  This interactive interface includes a data input unit 1, a first feature storage unit 2, a second feature storage unit 3, a third feature storage unit 4, a feature extraction unit 5, an inference unit 6, a display control unit 7, a display The unit 8 and the verification unit 9 are configured.

  The hardware configuration includes a CPU 11, a hard disk drive 12 loaded with an application program for executing the operation of the interactive interface, a RAM 13, an I / O 14 for exchanging data with the outside, a mouse 15, and a keyboard 16. And an optical drive 17 for driving an optical disk, a display 18 such as a CRT or a liquid crystal display, and the like.

  The data input unit 1 includes a client authentication ID, age, date of birth, examination date, examination item and examination value, number of examinations, health degree determined by a doctor, possible disease and part, medical record, etc. or electronic medical chart The input of image information or the like is accepted.

  The first feature storage unit 2 stores test data such as a client test item, test value, authentication ID, test date, health level, disease possibility site, etc., in the data received by the data input unit 1 as a first feature amount. Remember as. The second feature storage unit 3 classifies the test items related to each type of disease among the test items of the client into a plurality of classes according to the level, and takes an abnormally high value or an abnormally low value included in each class The number is stored as the second feature amount. The third feature storage unit 4 stores health level, information on a disease-possible site and client chart information (such as a chart or electronic chart image information). These first to third feature storage units 2, 3, and 4 are assumed to be a database (DB).

  The feature extraction unit 5 extracts from the first feature storage unit 2 a feature quantity for creating a decision tree representing the relationship between the analysis result of the client test data and the health level.

  The inference unit 6 performs well-known machine learning or pattern recognition based on the feature amount extracted by the feature extraction unit 5 and the information on the health level and the disease-probable site stored in the third feature storage unit 4. The decision tree is obtained using an identification function. In the present embodiment, C4.5 is used as an example. Prior to analysis at C4.5, the data to be used is shaped by a filter so that it can be processed at C4.5. The analysis result in C4.5 appears in the form of a decision tree in text format.

  The display control unit 7 performs display control so that the result obtained by the inference unit 6 is displayed on the display unit 8. The text-format decision tree represented by the inference unit 6 is rewritten into a hypertext format so that it can be displayed on the Web browser. At that time, various link information and linked HTML files are dynamically generated, and the original data and the like can be accessed and referenced from above the decision tree. As a result, easy-to-understand data analysis can be performed using an interactive interface on a Web browser. Therefore, the display control unit 7 includes a first display control mechanism 7A, a second display control mechanism 7B, and a third display control mechanism 7C.

  The first display control mechanism 7A displays a distinction between the abnormal high value or abnormal low value and the other normal value range for the number of items taking abnormal high value or abnormal low value for each inspection item and class for each client. 8 is displayed.

The second display control mechanism 7B causes the display unit 8 to display the degree of health and the number of items having an abnormally high value or an abnormally low value in order of the relationship depth for each client.
A second display control mechanism for performing control as described above,
The third display control mechanism 7C associates the display contents controlled by the first display control mechanism 7A with the display contents controlled by the second display control mechanism 7B in hypertext so that the display unit 8 can link the display contents. To.

  Next, the operation of the interactive interface of this embodiment will be described with a specific example. FIG. 3 is a schematic flowchart of the operation.

  The data received by the data input unit 1 is data as shown in FIGS. These data are taken from an external inspection database, and new data is added to construct a large amount of database.

  The data includes client identification ID, test date, sex, date of birth, age, test items with more than 100 items and test values, number of tests, health level, disease, potential disease site, etc. The thing which consists of is input and it is made into a database. The result of the doctor's judgment is input for the health level, the disease, the likely disease site, and the like. These data are first stored in the first feature storage unit 2 as the first feature amount.

  In addition, the control unit (not shown) classifies test items related to specific diseases, such as hyperlipidemia and tumor, into a plurality of classes in the test items of the client, and abnormally high values or abnormalities included in each class. The number of items having a low value is obtained and stored in the second feature storage unit 3 as the second feature amount.

  Further, the third feature storage unit 4 stores the health level, the disease, the disease-possible site, and the image information and text information of the medical record or electronic medical record in which they are written.

  The feature extraction unit 5 extracts feature amounts for creating a decision tree from the first feature storage unit 2.

  The inference unit 6 analyzes the feature amount extracted by the feature extraction unit 5 and the information on the health degree and disease-possible part stored in the third feature storage unit 4 in C4.5. Since the result appears in the form of a decision tree in text format, it is rewritten into hypertext format so that it can be displayed on a Web browser. An example of a decision tree rewritten in the hypertext format is shown in FIG. An example of the decision tree displayed on the Web browser is shown in FIG. As can be seen from the figure, here, test items closely related to the degree of health diagnosed for each client and the number of items having test values, abnormally high values, or abnormally low values are displayed in the order of depth of relation.

  Here, various link information and linked HTML files are added so that they can be dynamically generated, and the original data and the like can be accessed and referenced from above the decision tree. An example in which the link destination is described in the hypertext format is shown in FIG.

  The first to third display control mechanisms 7A, 7B, and 7C perform control for displaying the decision tree created by the inference unit 6 and data related thereto on the display unit 8.

In FIG. 7, the description per leaf of the decision tree is changed from the result of C4.5. That means
... Total protein (TP)> 7.1: 4b (9/1)
Then, “If the total protein (TP)> 7.1, it can be read that the health level is 4b (IVb), and (9/1) is 9 correctly classified, It can be read that it was mistakenly classified as one.

  The nodes in the leaves are linked by looking at various relationships during the above processing.

  For example, when a link where “CA72-4> 4: 5a (2/0)” in FIG. 7 is clicked, a table showing data shown in FIG. 9 appears. This table shows the relationship indicated by the decision tree that appears on the decision tree. In the upper part, the health level and links to each client are displayed.

  In the example of FIG. 9, the link to another client is written as <1004.2006-09-19, 5a>, etc., which is “client ID 1004, examination date September 19, 2006, health "5a (Va)".

  Furthermore, in FIG. 9, when the link of another client is clicked, the inspection result of that client is displayed as shown in FIG.

  In FIG. 10, the abnormally high value is actually shown in red and the abnormally low value is shown in blue. This allows you to instantly know where to focus your attention.

  As described above, an interactive interface is realized.

  Further, according to the present invention, various application developments can be performed. For example, if there is an error in the data, it can be corrected.

In this case, the matching unit 9 is used. The collation unit 9 classifies the health degree and disease possibility part of the client stored in the third feature storage unit 4 into which health degree and disease possibility part using the decision tree obtained by the inference part 6. Check if it should be. As a result of the collation, when the health level and the disease-possible site to be classified are different, the collation unit 9 sends the client ID in the configuration ID column shown in the upper part of FIG. 11 to the third display control mechanism 7c of the display control means 7. In addition, the fact that the data has come to a class (decision tree branch) different from the health level is displayed in red, for example. Then, by clicking on the client ID displayed in red, the chart image data and the inspection data stored in the third feature storage unit 4 are displayed. FIG. 12 shows an example in which inspection data is displayed.
By looking in detail at the chart image data and the examination data, it is possible to know whether the classification by the decision tree is wrong or whether the health level is wrongly entered. For example, when data as shown in FIG. 12 appears, abnormally high values and abnormally low values are displayed in color. If it seems that there is no such bad data, it is considered that the classification is wrong. If abnormal data is included in this, it is considered that the health level was entered incorrectly, and the health level is reconsidered. In the case of correction, an authoring tool (editor environment) to be corrected as CAD is created, and correction is performed thereby.

  In addition, the interactive interface of the present embodiment can estimate the health level by matching the test data of the client having no health level on the decision tree. This diagnosis support is particularly effective when estimating the health level when the same client performs a plurality of tests.

  In this case, the inference unit 6 stores the first feature quantity and the second feature storage unit 2 that are captured via the feature extraction unit 5 and stored in the first feature storage unit 2 without health. The already obtained decision tree is used as a rule set for determining the health degree using the second feature amount. Then, the decision tree is traced according to the inspection data, and the place where the decision is reached is defined as the health level.

  Similarly to the above, various links are generated during the analysis, and an interface that can easily refer to the original data set or the like is dynamically generated. The generated code can also be linked on the Web browser in the HTML format.

  FIG. 13 shows an example of the result of estimating the health level. A list of estimated health levels is shown on the left side, and the decision tree used for estimation is shown on the right side. In the decision tree, 5a (0/0/1) is written, which means that one piece of data is estimated for a leaf having a health level of 5a (Va). You can also jump to the leaf of the decision tree from the list on the left. For example, as shown in FIG. 14, the health degree estimated from the data collected from ID 1138 (January 22, 2007) is the upper right portion “... Neutral fat (TG)> 105: 5a (0/0/4) ) "In the leaves.

  Then, by clicking on the leaf link there, a list of the inspection data of the clients on the link can be obtained (FIG. 15).

  Similarly, this page shows the range of data used to estimate health. When referring to individual data, the desired data can be obtained by clicking the link at the location of the client ID (FIG. 16).

  Thereby, it is possible to check the examination data on the assumption of the estimated health level and check whether the estimation is correct.

  Further, the reliability of the health estimation can be rechecked as follows, for example.

  For a learning sample included in the same branch of the decision tree, a template is created by a statistical pattern recognition method, and a distance (or similarity) between the template M and the sample Xi included in the branch di (Xi, M) ( i = 1, 2,..., K), and the distance (or similarity) between the sample Xj (I not = j) and the template M other than the branch di (Xj, M) (j == 1, 2) ,..., N−K) (N is the total number of learning samples) and find I = Max {i} where di <dj, and only i = 1, 2,. To create M '.

  Then, regarding unknown X, if d (X, M ′) <d (X_I, M ′), the health degree of X∈M is determined.

It is a functional block diagram which shows the example of 1 structure of the interactive interface by this invention. It is a typical block diagram of the hardware at the time of implement | achieving an interactive interface same as the above with a personal computer. It is a schematic flowchart of operation | movement. It is a figure which shows the example of the taken-in data. It is a figure which shows the example of the taken-in data. It is a figure which shows an example of the decision tree rewritten by the hypertext format. It is a figure which shows an example of the decision tree displayed on the web browser. It is a figure which shows the example which described the link destination in the hypertext format. It is a figure which shows the table which shows data. It is a figure which shows the display of the test result of the other client linked. It is a figure which displays that the data different from a health degree came to a certain client ID. It is the figure where the data which makes it judge that classification was wrong was displayed. It is a figure which shows the result of having estimated the health degree. It is a figure which shows the leaf in which the health degree estimated from the data of a certain client ID collection exists. It is a figure which shows the list of the test | inspection data of the client on the said leaf. It is a figure which shows a mode that an individual data is referred.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Data input part 2 1st characteristic memory | storage part 3 2nd characteristic memory | storage part 4 3rd characteristic memory | storage part 5 Feature extraction part 6 Inference part 7 Display control part 7A 1st display control mechanism 7B 2nd display control mechanism 7C Third display control mechanism 8 Display unit 9 Verification unit

Claims (4)

An interactive interface for medical diagnosis support that displays the relationship between analysis results of client test data and health level in the form of a decision tree,
A data input unit that accepts input of test data and test data of the client, and test data and test values of the client, test ID, test date, health level, disease possibility site, and the like;
A first feature storage unit that stores, as a first feature quantity, test data and test values of the client accepted by the data input unit, test values, authentication IDs, test dates, health levels, disease potential sites, and the like;
The test items related to each type of disease among the test items of the client received by the data input unit are classified into a plurality of classes, and the number of items taking the abnormally high value or abnormally low value included in each class is set to the second A second feature storage unit for storing the feature amount;
A third feature storage unit for storing information on health level, disease-possible site and client chart information received by the data input unit;
A feature extraction unit for extracting a feature quantity for creating the decision tree from the first feature storage unit;
An inference unit that obtains the decision tree using the feature amount extracted by the feature extraction unit and the health degree and disease-possible part information stored in the third feature storage unit;
A display unit for displaying various screens;
Control to access the first, second, and third feature storage units and to display on the display means the relationship between the analysis result of the client test data and the health level in the form of a decision tree obtained by the inference unit A display control unit for performing
The display control unit
For each client, control is performed so that the number of items that have abnormally high values or abnormally low values for each inspection item and class is displayed on the display unit by distinguishing the difference between the abnormally high value or abnormally low value and other normal value ranges. A first display control mechanism;
A second display control mechanism for controlling to display the number of items having a health level and an inspection item closely related to the health level and an abnormally high value or an abnormally low value in the order of the relationship depth, for each client;
A medical diagnosis support feature comprising a third display control mechanism for associating display contents controlled by the first display control mechanism with display contents controlled by the second display control mechanism in hypertext. Interactive interface for.   2. The interactive display for supporting medical diagnosis according to claim 1, wherein the second display control mechanism controls to display other client data associated with similar items in association with each other. interface.   The medical reasoning support according to claim 1, wherein the inference unit performs matching using the decision tree on test data without a health level, and estimates the health level. Interactive interface.   A collation unit for collating which health degree and disease potential part of the client stored in the third feature storage part are classified into which health degree and disease possibility part are determined using the decision tree obtained by the inference part. And the third feature storage unit stores the difference between the health level and the disease potential part of the client stored in the third feature storage unit and the health level and disease possibility part of the matching result in the third feature storage unit. The medical chart support information for medical diagnosis according to any one of claims 1 to 3, wherein medical record information about the stored client is displayed and corrected to a correct health level and a disease-probable site. interface.