JP2003339646A - Disease pickup system and knowledge base - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a disease pickup system by which the precision of 'disease pickup' is improved and a memory is utilized effectively. <P>SOLUTION: A single disease (= single record) is divided into fields consisting of a disease number field 101 (5 bytes), a disease name field 103 (30 bytes), a disease information field 105 (209 bytes), an index field 107 (3 bytes), and a surplus field 109 (9 bytes). <P>COPYRIGHT: (C)2004,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a disease pick-up system, and more particularly to a disease pick-up system for searching a disease name or the like from disease information.

[0002]

2. Description of the Related Art As a conventional technique for retrieving a disease name or the like from disease information, there are analysis techniques using multivariate analysis and Bayes theory. However, this was intended to obtain a correct diagnosis (confirmation diagnosis) by analyzing only a computer. In addition, they are in a limited field such as renal disease and hypertension, and do not cover the entire general internal medicine.

[0003]

DISCLOSURE OF THE INVENTION The present invention does not seek a final definitive diagnosis only by the analytical power (algorithm) of a computer, but uses abnormal information (subjective symptom / test finding) obtained from a patient, It is an object of the present invention to provide a disease detection system that can remind a doctor of a disease that does not reach (unimaginable) by using a computer. Another object of the present invention is to provide a disease detection system capable of covering not only a specific field of disease but also general medical science as a whole.

[0004]

In order to solve the above-mentioned problems, the disease pick-up system according to the present invention is a means for collating the disease information registered in the knowledge base with the disease information inputted from the outside. When both disease information matches, the symptom level of the disease information in the knowledge base,
Means for obtaining the weighted score weighted for each symptom level, and for all disease information that has been input, means for determining the symptom level and weighted score, and summing the weighted scores, and the total value of the weighted scores is the predetermined value. In the case of the above, a means for displaying the information registered in the knowledge base is provided.

The knowledge base according to the present invention is a knowledge base for obtaining a disease name having a predetermined possibility from the disease information of a patient, and a disease number field for registering a disease number and a disease name are registered. It has a disease name field, a disease information field in which disease information is registered, and an index field in which an index is registered. In the disease information field, one or more disease information arranged in order of symptom level is registered. In the index field, the arrangement position of the disease information of each symptom level and the value that determines the weighted score of each symptom level are registered.

[0006]

BEST MODE FOR CARRYING OUT THE INVENTION (1) Basic Principles of this System In a disease detection system which is an embodiment of the present invention, a score system is adopted as a knowledge base in order to improve the accuracy of “discovery of disease”. This system has the following two basic principles. Give more points to important subjective symptoms / test findings (hereinafter referred to as “disease information”) that have a high frequency of occurrence. The total score of all "disease information" should not be significantly dissociated. In order to realize the above, each knowledge base is provided with a unique "index" to ensure the equality of the individual knowledge bases.

(2) Structure of Knowledge Base First, a general knowledge base which is an embodiment of the present invention will be described. Knowledge base is 1 disease = 1 record (= 256
Byte) is the smallest unit. The reason for 1 record = 256 bytes is that this system has been around for about 20 years now, and it is a remnant of memory saving at the time of development (1982).

FIG. 1 is a data structure diagram of one knowledge-based disease according to an embodiment of the present invention. As shown in the figure,
In the embodiment of the present invention, one disease (= 1 record) is recorded as a disease number field 101 (5 bytes), a disease name field 103 (30 bytes), and a disease information field 105.
(209 bytes), index field 107 (3 bytes), and remainder field 109 (9 bytes).

In the present embodiment, the disease name “liver absces
The knowledge base of "s" will be used as an example. The knowledge base includes data “R203” in the number field 101, data “liver abscess” in the disease name field 103, and data “feverchilleucocyto” in the disease information field 105.
s / liverGTPjaundCRPALP / fatigpainESR ^ resist / diste
ndshocknaus @ FU / hepatomanorexLDHLAP / vomitrebound-
tcold / hypoproteinurshould / chestenesweathydrot / o
ligurenalungbleed / Xp (ascitchyperamy) / purpileusk
In CPKback ”, data“ 332 ”is registered in the index field 107. Disease information field 105
The symbol "/" in the above indicates the division of each section, but naturally it is not in the actual memory.

(3) Description of Disease Information Field 3-1. Description of Data Structure and Symptom Level First, the disease information field 105, which is the core part of the present system, will be described. FIG. 2 is a data configuration diagram of the disease information field 105. As shown in the figure, the disease information field 105 has a memory allocation of 209 bytes. In the present embodiment, this disease information field 105 is divided into 11 sections of 19 bytes each. Then, the coded “disease information” is assigned to each section as “large symptom”, “medium symptom”, “small symptom” (hereinafter, “large”,
Fill in the order of "medium" and "small". Here, "large"-"medium"-"small" represents the symptom level of each disease information, and the disease information with a larger symptom level is an important subjective symptom / test finding having a higher frequency of occurrence. In the present embodiment, the symptom level of “disease information” to be registered is
"Large" (expression frequency 50% or more) based on expression frequency,
It was divided into three groups: "medium" (20% or more and less than 50%) and "small" (expression frequency less than 20%).

3-2. Disease information codes and their meanings The disease information was encoded in advance based on each English expression. An example of the encoded disease information is shown. The values in parentheses are the meanings of each code. fever, f
hil (cold, shake), leucocytos (leukocytosis), liv
er (liver dysfunction), rG (elevated gamma globulin), GTP
(Increased gamma GTP), jaund (jaundice), CRP (increased CRP),
ALP (serum ALP increase), fatig (fatigue), pain (abdominal pain), ESR ^ (accelerating erythrocyte sedimentation), resist (muscle defense), diste
n (abdominal distension), tend (abdominal tenderness), shock (nausea), @ (all inflammatory reactions present), FU
(Unexplained fever), hepatom (hepatomegaly), anorex (anorexia), LDH (elevated serum LDH), LAP (elevated serum LAP), vo
mit (vomiting), rebound-t (recoil pain), cold (cold-like symptoms), hypopro (hypoproteinemia), proteinur (proteinuria),
should (shoulder pain), chest (chest pain), tenes (stomach abdomen), sweat (sweating), hydrot (pleural effusion), oligur (oliguria), renal (renal dysfunction), lung (abnormal chest radiograph) ), Bleed (bleeding tendency), Xp (abnormal bone / joint radiograph), Xp (a (abdominal radiograph or ultrasound abnormality), ascit
(Ascites), itch (itch), hyperamy (hyperamylaseemia), purp (purpura), ileus (ileus), skin (rash), CPK (serum CPK increase), back (backache).

3-3. Disease Information Registration Method The disease information encoded as described above is shown in FIGS.
It is registered in each section in the form as shown in. “Large” Disease Information FIG. 3 shows data registered in the first to third divisions of the disease information field 105. In the present embodiment, “large” disease information is registered in the first to third sections. Referring to FIG. 3, data “feverchilleucocytos” is registered in the first section. The data is "fever-chil-leucocy
It can be decomposed as "tos" and each data is fev
It means er (fever), chil (cold, shake), leucocytos (leukocytosis).

Similarly, in the second section, "liverGTPjaundCRP"
The data "ALP" is registered. The data is "liv
er-rG-GTP-jaund-CRP-ALP ", and each data is liver (liver dysfunction), rG (gamma globulin increase), GTP (gamma GTP increase), jaund (jaundice), It means CRP (CRP increase) and ALP (serum ALP increase). Similarly, "fatigpainESR ^ resis" is displayed in the third section.
The data "t" is registered. The data is "fatig
-pain-ESR ^ -resist "can be decomposed, and each data is fatig (fatigue), pain (abdominal pain), ESR ^
Means such as (acceleration of red sedimentation) and resist (muscle defense). As described above, there are 13 pieces of "disease information" from fever (1st to 5th bytes of the first section) to resist (14th to 19th bytes of the third section) in "large".

"Medium" Disease Information FIG. 4 shows data registered in sections 4 to 6 of the disease information field 105. In the present embodiment, "medium" disease information is registered in the 4th to 6th sections. Referring to FIG. 4, data “distendshocknaus @ FU” is registered in the fourth section. The data is "disten-tend-shock-
naus-@-FU ”and each data is disten (abdominal distension), tend (abdominal tenderness), shock
(Shock), naus (nausea), @ (existence of all inflammatory reactions), FU (fever of unknown cause).

Similarly, in the fifth section, "hepatomanorex LDH"
The data "LAP" is registered. The data is "hep
atom-anorex-LDH-LAP ”, and each data has the meaning of hepatom (hepatomegaly), anorex (anorexia), LDH (serum LDH increase), LAP (serum LAP increase). . Similarly, in the 6th section, "vomit rebound
The data "-tcold" is registered. The data can be decomposed into "vomit-rebound-t-cold", and each data means vomit (vomiting), rebound-t (recoil pain), and cold (cold-like symptoms). As described above, "medium" is disten (1st to 6th bytes of the 4th partition)
To cold (15th to 18th bytes of the 6th section), there are 13 pieces of "disease information".

"Small" Disease Information FIG. 5 shows data registered in the 7th to 11th sections of the disease information field 105. In this embodiment, 7-1
"Small" disease information is registered in the first section. Referring to FIG. 5, data “hypoproteinur should” is registered in the 7th section. The same data is "hypopro-proteinu
r-should ”, and each data is hypopro (hypoproteinemia), proteinur (proteinuria), shou
It means ld (shoulder pain).

Similarly, in the 8th section, "chestenesweathyd
The data "rot" is registered. The data is "che
It can be decomposed into "st-tenes-sweat-hydrot", and each data has meanings such as chest (chest pain), tenes (stomach belly), sweat (sweat), and hydrot (pleural effusion). Similarly, data “oligurenalungbleed” is registered in the 9th section. The data is "oligur-ren
al-lung-bleed ”and each data is oligur (oliguria), renal (renal dysfunction), lung
(Chest X-ray abnormal), bleed (bleeding tendency).

Similarly, in the 10th section, "Xp (ascitchype
The data "ramy" is registered. The data is "X-
pXp (a-ascit-itch-hyperamy "can be decomposed, and each data is Xp (bone / joint radiograph abnormal), Xp (a (abdominal radiograph or ultrasound abnormal), ascit
(Ascites), itch (itch), hyperamy (hyperamylaseemia) is meant. Similarly, in the 11th section, "pu
The data "rpileuskinCPKback" is registered.
The same data can be decomposed as "purp-ileus-skin-CPK-back", and each data is purp (purple), ileus
(Ileus), skin (rash), CPK (serum CPK increase), ba
Represents the meaning of ck (back pain). As mentioned above,
Back to "small" from hypopro (1st to 7th bytes of 7th partition)
There are 21 pieces of "disease information" up to (16th to 19th bytes of the 11th section).

(4) Description of Index Next, the "index" will be described. The first 2 bytes of the “index” are determined by the number of memory partitions allocated to “large” and “medium”. In the example above, 3 for "large"
3 partitions are allocated to the partition, "medium", and as a result up to this point, the "index" is <33X>.

Conversely, by referring to this index, the symptom level of each disease information can be derived. For example, the index of this embodiment is <332.
>, So the input disease information is the disease name field 1
When detected in the first to third divisions of the disease information field 105 of the record having the disease name A in 03, it indicates that the disease information corresponds to the disease level “large” of the disease name A.

Similarly, when the inputted disease information is detected in the 4th to 6th sections of the disease information field 105 of the record having the disease name A in the disease name field 103, the disease information is the disease of the disease name A. Indicates that the level corresponds to "medium". The last 1-byte number (X) of the "index" will be described later. When registering the "disease information" code, I tried to write it efficiently. This is for saving memory and effective use.

(5) Description of point system Next, the point system will be described. In the present invention, a score system is adopted for the final purpose of “discovery of disease”. This is an attempt of weighting "large", "medium", and "small" with points. This is because it is rational to give a high score to "disease information" that frequently appears and a low score to those that rarely appear. Moreover, the score varies depending on the disease even at the same symptom level. In the present invention, in order to realize this, 6 points were given to “Large” as a constant, 3 points were given to “Medium” as a constant, and 1 point was given to “Small” as a constant. Then, the individual score is calculated for each of the “large”, “medium”, and “small” coded “disease information”. Hereinafter, the details of the score calculation will be described as an example.

(6) Illustrative Description of Score Calculation In this embodiment, the score of disease information = constant / number of sections. Therefore, in the case of the "liver abscess" of the present embodiment, since "large" occupies 3 partitions, the score of "disease information" of each "large" is 6/3 = 2 points. Similarly, since "medium" occupies 3 sections, the score of "disease information" of each "medium" is 3/3 = 1 point. In other words, the “disease information” that has a high frequency of appearance is assigned a high score, but if there are a relatively large number of these and the occupied section is large, the score of each “disease information” will be low. That is.

From the above, in the “large” and “medium” cases, the “statement of equality of individual knowledge bases” mentioned at the beginning is stated.
I tried Here, the total score of each "disease information" of "large" and "medium" is calculated. First, as calculated previously, the score of "disease information" included in "large" is 2 points for each. Since there are 13 pieces of "disease information" in "Large", the total of "Large" is 26 points (= 2 x 13). Since there are 13 "disease information" in "medium", the total of "medium" is 13 points (= 1
X13). After all, it is confirmed that the total score of each “disease information” of “large” and “medium” is 39 (= 26 + 13).

(7) Description of the last 1 byte of the "index" Next, the last 1-byte number (X) of the "index" will be described. First, the number of “small” occupied sections will be described. We have already explained that the "disease information" field is already divided into 11 sections of 19 bytes each. Therefore, "small"
The use area of is naturally determined by subtraction. However, in all the records (disease), all the “small” use sections are not completely filled with the “disease information”.
Therefore, we decided to give the following meaning to the last 1-byte number (X) of the "index".

As mentioned above, the present invention first described two principles. Recall that one of them was that the total score of "disease information" did not significantly dissociate in order to ensure the equality of individual knowledge bases. The last 1-byte number (X) of the “index” is
It is a number that satisfies this principle. In the case of the example of “liver abscess” of the present embodiment, it is specified that X = 2 in the form of “index” = <332>, and each “disease information” of “small” is 1/2 Point (= 1 /
X).

As shown in FIG. 5, in the embodiment of the present invention, the “small” section includes 21 pieces of “disease information”, and the total of “small” is 10.5 points (1/2 × 21). )It turns out that.
By the way, in this example, the total score of "disease information" is 49.5 points (39
+10.5). In this way, by manipulating the last 1-byte number (X) of the “index”, the total score of each record (disease) should be between 45 and 55 points (preferably as close to 50 points as possible). I made it fit.

(8) Explanation of other adjustments 8-1.1 Adjustment by average score of "disease information" However, due to the large number of total numbers of "disease information", the total score is 45 to 55. The points may deviate significantly up and down. At this time, the average score of 1 "disease information", that is,
<(Total score) / (total “disease information”)> is 1.0 to 1.5
I devised it to fit between the dots. Incidentally, in the example of “liver abscess” of the present embodiment, it is 1.05 (= 49.5 / 47). Although these numbers were assumed in advance, they were eventually obtained empirically when creating the knowledge base. From our trial experience so far, we are convinced that this deliberate manipulation helps ensure that the individual knowledge bases are more equal.

8-2. Adjustment by index In creating a knowledge base for many diseases, there is no description of the occurrence frequency, and for that reason, we collect it in various cases by ourselves, but it is difficult to divide the expression frequency into three. I noticed that it exists a lot (about 1/3). An example of such a disease is given below. At this time, "index" is <0ab> (a and b are positive integers) for convenience, "medium" (expression frequency 20% or more), "small" ( The expression frequency was less than 20%).

At this time, in order to ensure the equality of individual knowledge bases, the total score obtained is multiplied by 2.5 (= (6 + 3 + 1) / (3+
1)) was corrected and the total number of points after the correction was about 45-
The average score of "disease information", that is, <(corrected total score) / (total "disease information"number)>, so that the score falls within 55 points
Was devised so that the score would be between 1.0 and 1.5 points. 2.5
The formula for multiplying (= (6 + 3 + 1) / (3 + 1)) is that the maximum number of external inputs for "large" is 6, the maximum number of external inputs for "medium" is 3, and the number of external inputs for "small" is It is based on <total maximum number of external inputs / (maximum number of external inputs of “medium” + maximum number of external inputs of “small”) ”, assuming that the maximum number of external inputs is 1. An example of this case is shown below.

8-3. Example of knowledge base adjusted by index One of the things adjusted by index is the knowledge base of the disease name "systemic lupus erythematosus". In the disease name field 103 of the knowledge base, "sy
The data “052” is registered in the index field 107, and “stemic lupus erythematosus”. In this case, the total number of points after correction is 65.75 (= (3/5 × 23 + 1/2 × 25)
× 2.5) 1 "disease information" The average score is 1.37 (= 65.75 / 4
8). Data "R29" is registered in the disease number field 101.

The following data is registered in the disease information field 105. The disease information field 105
The symbol "/" in the above indicates the division of each section, but naturally it is not in the actual memory. As can be seen by referring to the index field 107, there is no “large” disease information, and the “medium” disease information is registered from the first section to the fifth section of the disease information field 105. The disease information registered in "Medium" is shown below. hearthralganemurmrG / weighcoldyspnskin (n / feveryth
emyalgESRay ECGLDHbleedizzchest / pancytophotosedemRA

Further, "small" disease information is registered from the sixth section onward in the disease information field 105. "small"
The disease information registered in is shown below. leucoproteinurliver / renalungpainconvuls / livedfat
igCRPnausYY myasthrombocytopMPH / ulcer (oralopecough @ / lymphops
ychFUlymphn As described above, the data in each section can be decomposed into disease information. In this specific example, 23 disease information items are registered in "middle" and 25 disease information items are registered in "small". There is. Furthermore, the "medium" constant = 3 and the "small" constant = 1.

(9) Capture Description of Knowledge Base The above is an overview of the knowledge base according to the embodiment of the present invention. The number of disease records at the beginning of development (April 1984) was 428, and now there are 989 records. There are also knowledge bases that are not disease units, such as "inability to analyze" or "drug ingestion (accident or willfulness / crime)."
There is also a summary of diseases that show almost the same "disease information" such as "duodenal or gastric perforation (also consider colon perforation / idiopathic sigmoid perforation)".

Further, this knowledge base is regularly based on cases actually experienced, case reports in various journals, statistical figures in books, etc. within a range satisfying the first two principles described above. Has been updated. In addition, the number of disease records will be gradually added in the future. The purpose of creating this knowledge base is to search for possible diseases for the subjective and subjective findings presented by the target patient and supplementary laboratory findings such as blood chemistry tests, and present the results to relevant persons. Therefore, it does not take into consideration factors such as the frequency of onset of disease, the manner of onset, difference due to progress, gender, gender difference, prevalence age, race, regional disparity, etc. I will add some things. These should be considered in the actual differential diagnosis in the field.

(10) Description of Disease Search Algorithm 10-1. Disease Search Algorithm Processing Procedure Next, a disease search algorithm using a knowledge base will be described. Step 1) For a target disease, "disease information" of a patient having the disease is input to a predetermined computer from the outside. A disease search program using the search algorithm is installed in this computer. Step 2) Input one or more "disease information"
And the 989 knowledge bases created above are brute force sample matched. Step 3) Points weighted for each symptom level (in the case of "liver abscess" in the present embodiment, "large" = 2 points,
Based on "medium" = 1 point and "small" = 0.5 point, the total points of each of the knowledge bases of 989 are calculated. Step 4) The knowledge base of 989 is sorted using the total points as a key. Step 5) If there is a knowledge base having more than the cut-off points prepared in advance, information (disease number, disease name, etc.) registered in the knowledge base is listed.

10-2. Process Procedure of Disease Search Algorithm (Specific Example) Next, the disease search algorithm will be described using specific disease information. Step 1) Disease information (= fever (fever), ALP (serum AL)
P increase), anorex (anorexia), hypopro (hypoproteinemia), itch (itch), CPK (serum CPK increase) are input to a predetermined computer from the outside. Step 2) Specimen collation of the six input "disease information" and the knowledge base. Here, a case where only the knowledge base of “liver abscess” shown in FIG. 3 to FIG. 5 is sample-matched will be described.
The sample is also collated with the knowledge base of 9.

Step 3-1) Points weighted for each symptom level ("large" = 2 points, "medium" = 1 point, "small" = 0.5 points)
Calculate total points based on. In this case, fever (fever) and ALP (serum ALP increase) are 2 points each, anorex
(Anorexia) is 1 point, hypopro (hypoproteinemia), itch (itch), CPK (serum CPK increase) is 0.5 points each. Step 3-2) As will be described later, the formula for calculating the number of foot points is a function of the number of external inputs (g). In this system, the maximum number of external inputs is determined to be 10, and in this example, the total point is = (2 × 2 + 1 × 1 + 0.5 × 3) × 10/6 = 10.83. Step 4) The knowledge base of 989 is sorted using the total points as a key. Step 5) If the total score (10.83 points) of the knowledge base of "liver abscess" is equal to or greater than the cut-off score prepared in advance, the information (disease number "R203", disease name ""liverabscess").

(11) Definition of the number of cut-off points in C language FIG. 6 is a diagram in which the number of cut-off points is written in a program expression in C language. A description of the constants and variables used in the figure is given below. “Cc1” ・ ・ ・ It takes 0 if no keyword is specified, and takes -1 if there is no keyword specified. "L1" ... Operates when "index" is <0ab>. “G”: Number of “disease information” input externally.

11-1. Explanation of variable “st” “st”: This is a variable for adjusting the number of cut points in four stages. At the first stage, which has the lowest number of cuts, st = 0. In this case, the number of cut points = 0. It works only when the number of external inputs is one. When the number of foot cut points is the second step, st = 1, the number of foot cut points = s
Become an evere base. When the number of cut points is the third step, st = 2, the number of cut points = m
It becomes an ild base. When the number of foot cut points is the 4th stage, st = 3, the number of foot cut points = s
Become an everer base.

The individual constants and variables other than the above are based on those calculated from pseudo data created in advance using random numbers, but those currently used are empirically obtained from many case experiments. It is the given number. Although the embodiment of the present invention has been described above, the above description is for explaining the present invention, and the present invention is not necessarily limited to this.

[0042]

As described above, according to the diagnosis support system of the present invention, it is possible to improve the accuracy of “discovery of disease” and to effectively use the memory.

[Brief description of drawings]

FIG. 1 is a data configuration diagram of one record according to an embodiment of the present invention.

FIG. 2 is a data configuration diagram of a disease information field according to the embodiment of the present invention.

FIG. 3 is a diagram showing an example of data registration of “large” disease information.

FIG. 4 is a diagram showing an example of data registration of disease information of “medium”.

FIG. 5 is a diagram showing an example of data registration of “small” disease information.

FIG. 6 is a diagram showing the number of cut points in C language.

[Explanation of symbols]

103 Disease name 105 Disease information 107 Index

Continued front page    (72) Inventor Genichi Kato             2-44-8 Motomiya, Otsu City, Shiga Prefecture (72) Inventor Osamu Ishii             6-12-10 Nagayama, Tama-shi, Tokyo

Claims (2)

[Claims] 1. Means for collating disease information registered in the knowledge base with disease information input from the outside, and when both disease information match, the symptom level of the disease information in the knowledge base And means for obtaining a weighted score weighted for each symptom level thereof, for all disease information input, means for determining the symptom level and the weighted score, and summing the weighted scores, and the weighted score And a means for displaying information registered in the knowledge base when the total value of the above is greater than or equal to a predetermined value, the disease pick-up system. 2. A knowledge base for obtaining a disease name having a predetermined possibility from disease information of a patient, a disease number field for registering a disease number, a disease name field for registering a disease name, and disease information. Has a disease information field for registering an index, and an index field for registering an index. In the disease information field, one or two or more pieces of disease information arranged in order of symptom level are registered, and in the index field, A knowledge base characterized in that the location of the disease information at each symptom level and the value that determines the weighted score for each symptom level are registered.