JPH0318459B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an automatic diagnostic device that performs logical diagnosis processing regarding metabolic abnormalities based on analysis data and specimen information from an analyzer and extracts names of metabolic abnormalities.

In recent years, research on inborn errors of metabolism has made remarkable progress. The congenital metabolic disorder is caused by genetic malfunction due to deficiency or inhibition of enzymes that act as catalysts at each reaction step on the metabolic pathway in the body. Early detection of this metabolic abnormality is of great significance, as early treatment is highly effective. It is estimated that there are more than 2,000 metabolic disorders, but it has been confirmed that several hundred of these can be diagnosed by analyzing amino acids, organic acids, sugars, steroids, etc. using various analytical methods. ing.

The present invention provides a new device that automatically diagnoses metabolic abnormalities based on analysis data and sample information from existing analyzers.

The present invention includes an analyzer, a first storage unit in which a healthy value table indicating a healthy value range (normal range) for each item is stored, a disease name related to each item, and a weight representing the degree of relationship of the disease name to the item. a second storage unit storing an item/disease name table indicating the disease name, and a disease name/item table indicating the items related to each disease name and the abnormality model weight and logical model weight indicating the degree of relationship of that item to the disease name. a third storage unit in which is stored, a fourth storage unit in which a disease name/symptom table indicating model symptoms related to each disease name and a weight indicating the degree of relationship of the model symptoms to the disease name is stored; Based on the normalized sample analysis data from the analyzer and the sample information, the normal value table is referenced to detect an abnormal item in the sample and calculate its degree of abnormality.・Refer to the disease name table to extract the suspected disease name of the specimen and calculate the degree of suspicion for the suspected disease name, and for each extracted suspected disease name, refer to the disease name/item table to determine the degree of suspicion for the abnormal item. is calculated based on the abnormality degree of the abnormality item that matches the sample and the model listed in the table, the weight of the abnormality degree model of the abnormality item, and the sum of the weights of the abnormality degree model of all abnormality items of the model. The anomaly degree model match degree, the anomaly degree of the anomaly item that matches the sample and the model listed in the table, the weight of the logical model of the anomaly item, and the sum of the weights of the logical model of all the anomaly items of the model. The logical model matching degree calculated based on This is an automatic diagnosis device that includes a calculation device that calculates a symptom model matching degree based on the severity, the weight of a model symptom of the symptom, and the sum of the weights of all model symptoms of the model.

FIG. 1 is a schematic diagram of an automatic diagnostic device shown as an embodiment of the present invention. Now, the automatic diagnosis method of the present invention will be explained in detail with reference to the figure. In this example, an amino acid analyzer was used as the analyzer.

In the figure, reference numeral 1 denotes an amino acid analyzer, and sample analysis data from the analyzer is stored in a disk memory 3 according to instructions from a central control unit (hereinafter referred to as CPU) 2. The CPU reads sample analysis data from the disk memory, and normalizes the data, that is, the data of each peak intensity and retention time of the chromatogram of the amino acid analyzer. In other words, in the amino acid analyzer, retention times may differ for the same eluted component depending on the elution conditions, so an original chromatogram with retention times determined according to each eluted component must be created in advance. , the retention time of each peak in the chromatogram of the specimen is normalized based on which of the peaks in the original chromatogram the retention time of each peak is within an allowable range. Then, in response to a command from the CPU, the standardized analysis data is input to the storage device 4 as a data file. In addition, specimen information such as patient attributes (age, gender, etc.), patient information (past symptoms, symptoms, medications, etc.), specimen number (specimen number), etc. is input into the data file by the manual.

Next, the CPU 2 reads the standardized analysis data and specimen information from the data file 4, and reads the normal value range (normal value range)
Detects abnormal items and measures their degree of abnormality based on standardized analysis data by referring to a table of healthy values shown by age. That is, the CPU 2 creates a histogram of the peak intensity of each item of the standardized analysis data, and simultaneously inputs the normal value range of each item read from the first storage unit 5 into the histogram. In addition, the average value of the population of healthy data is also input into the histogram. Then, the histogram is displayed on the display device 6 (cathode ray tube and/or printer) as shown in FIG.
In Fig. 2, the numbers on the horizontal axis (No.) (1), (2),
(3), ... are peak numbers, among which (1) is threnion (Thr), (2) is alanine (Ala), (3) is glycine (Gly), (4) is valine (Val), ( 5) is the item methionine (Met),... The mark on the vertical axis indicates the peak intensity, the range between markers (-) indicates a healthy value range, and the mark x indicates the average value of the population of healthy data. As shown in the figure, items whose peak intensities are not within the normal range are abnormal items of the specimen. Next, as shown in Figure 3, the standard deviation of the peak intensity of each item from the average value of the population (marked on the vertical axis) and the standard deviation of the healthy value of each item are connected. (solid broken line) is displayed. Note that the solid horizontal axis represents the average value (X) of the population, and the broken horizontal lines represent 2σ and −2σ (σ ² is the population variance). Next, if the peak intensity of each item in the sample is larger than the population average value, calculate the ratio between that value and the healthy value, and if it is smaller, calculate the ratio between that value and the minimum healthy value. -5, -4, -3, -2, -1, 0, according to the specific range set for each item in advance based on in vivo metabolic efficiency
Leveled into 1, 2, 3, 4, and 5 levels. This leveled value is the degree of abnormality and is displayed in a histogram as shown in FIG. In this case, the abnormality degree of the item (normal item) within the healthy value range is set to 0.
Furthermore, CPU2 extracts only the abnormal items and
As shown in the figure, the display device 6 displays the results in descending order of + abnormality degree and in descending order of negative abnormality degree. The above processing is referred to as primary diagnosis. Next, the CPU 2 performs the following for the abnormal items detected in the first diagnosis:
The item/disease name table stored in the second storage unit 7 is referred to to extract the suspected disease name and calculate the initial suspicion level. As shown in Figure 6, the item/disease name table contains disease names (12, 7, 75, 4) related to each item (Thr, Val, Ala, etc.). ,...3, 6, 12, 3,
) and the weight (A ₁₁ , A 12 , A ₁₂ ,
A ₁₃ , ...A ₂₁ , A ₂₂ , ...) are stored.

The CPU 2 refers to the table to determine the abnormality items (Thr, Val, Ala, Arg,
Extract all disease names related to Met, Pn),
As shown in FIG. 7, a command is sent to the display device 6 to select, for example, three closely related items and display them on the side of each abnormal item. At this time, as shown in FIG. 7, Gal with an abnormality level of +4 and lactic acid with an abnormality level of -5 are also input as specimen information. At the same time, the real name of the coded disease name is displayed on a part of the display device. Next, the initial suspicion level is calculated for each suspected disease name extracted in this way.

If i is the abnormal item, j is the disease name, Z is the degree of abnormality, A is the weight, Amax is the maximum weight in the table, and Aij is the specific weight given to the disease name related to each item, then the degree of suspicion for disease name j is DDLj is DDLj＝Σ(Aij×｜Zi｜)/[Amax
×(Number of abnormal items in specimen)]……(1) For example, the degree of suspicion for disease name 12 (Kaede) is DDL12=A ₁₁ ×5+A ₂₂ ×4+A ₃₁ ×3+
.../(Amax×8).

In this way, the degree of suspicion for the disease names 12, 7, 75, . In this case, the disease name will be displayed using the real name. Perform the above operations in the second step.
This is called the next diagnosis.

Next, the CPU 2 sequentially refers to the disease name/item table stored in the third storage unit 8 for each initial suspected disease name (Kaede, Phenyl...) extracted in the second diagnosis. The model abnormality item of the extracted disease name (maple, phenyl, ...) is called up, and the abnormality item model list of the disease name is displayed on the display device 6, along with the list of abnormality items and suspected disease names of the specimen as shown in FIG. (The model list shown in Figure 9 is a maple list). Taking the disease name Kaede as an example, next we measure the degree of suspicion regarding the abnormal item, that is, the degree of suspicion agreement. When measuring the degree of suspicion agreement, the degree of agreement with the abnormality model and the degree of agreement with the logical model are measured. The degree of abnormality model agreement measures the degree of agreement between the sample and the actual statistical model, and the logical model agreement measures the degree of agreement between the sample and the theoretical model. In the disease name/item table, as shown in Figure 10, for each disease name (Kaede, Phenyl,...), related items (Thr, Val,..., Met,...) and the disease name of each item are listed. Anomaly model weights (B ₁₁ , B ₁₂ , ..., B _1o , ..., B ₂₁ , ...
) and logical model weights (C ₁₁ , C ₁₂ , ..., C _1o ,
..., C ₂ , ...) are respectively memorized.

The abnormality degree model match degree AG1j for disease name j is the abnormality item i (e.g.,
For the disease name Kaede, the abnormality degree of Thr, Val, Met) is the sum of the weights of the abnormality degree model of each abnormality item on the model list, and Mi is the abnormality degree model of abnormality item i that matches the specimen and model. AG1j=1−(Σ|Si−Mi|)/2 (2). For example, for the disease name Kaede, 1-[|5/(B ₁₁ +B ₁₂ +B _1o )-B ₁₁ /(B ₁₁ +B ₁₂ +B _1
o )｜+｜4/(B ₁₁ +B ₁₂ +B _1o ) −B ₁₂ (B ₁₁ +B ₁₂ +B _1o )｜+｜−4/(B ₁₁ +B ₁₂ +B _1
o ) − B _1o (B ₁₁ +B ₁₂ +B _1o ) |]/2.

Next, the logical model agreement degree AG2j for disease name j is Ei, which is the weight of the abnormality model in Si, and Ti, which is the weight of the abnormality model in Mi. Then, AG2j=1-(Σ|Ei-Ti|)/2...(3). For example, for the disease name Kaede, 1-[|5/(C ₁₁ +C ₁₂ +C _1o )-C ₁₁ /(C ₁₁ +C ₁₂ +C _1
o ) | + | 4 / (C ₁₁ + C ₁₂ + C _1o ) −C ₁₂ / (C ₁₁ + C ₁₂ + C _1o ) | + | −4 / (C ₁₁ + C ₁₂ +
C _1o ) − C _1o / (C ₁₁ + C ₁₂ + C _1o ) |]/2.

The above-mentioned two degrees of coincidence are sequentially determined for each disease name extracted in the secondary diagnosis. These matching degrees take values from 0 to 1, and 0 means complete mismatch, and 1 means complete match. In the case of the degree of abnormality model matching, it can be diagnosed that the closer it is to 1, the closer it is to the actual model, and the closer it is to 0, the farther it is. Furthermore, in the case of logical model matching, it can be diagnosed that the closer it is to 1, the closer it is to the theoretical model, and the closer it is to 0, the farther it is. Based on instructions from the CPU 2, the disease names and the degrees of coincidence between these two disease names are displayed on the display device 6 in descending order of degree of coincidence (see FIG. 11). The above operations are referred to as tertiary diagnosis.

Next, the CPU 2 generates a model related to the extracted disease name from the disease name/symptom table stored in the fourth storage unit 9 regarding the suspected disease name (Kaede, Phenyl, ...) extracted in the second diagnosis. The symptoms (△, 〇, □, ◎) are called up and displayed on the display device 6 along with the symptoms of the specimen (△, ×, ☆, ◎) and their severity (+2, +3, +2, +4) from the specimen information. FIG. 12 shows this display, and the real names of the symbolized symptoms are also displayed. As shown in Figure 13, the disease name/symptom table includes each disease name (Kaede, Phenyl...), model symptoms related to each disease name (△, 〇, □,..., ◎), and the disease name of the model symptom. The weights (D ₁₁ , D 12 , D ₁₂ ,
D ₁₃ , ..., D _1o , ...D, D ₂₁ , ...) are stored. The disease name/symptom table 9 is then referred to for each suspected disease name of the specimen, and the degree of symptom matching is determined. Symptom matching degree AG3j for disease name j is defined as Hi for symptom i that matches between the specimen and the model in the disease name.
(Example: △ and ◎ for disease name Kaede) / sum of weights of all model symptoms related to disease name j,
If Qi is the weight of the model symptom of symptom i that matches the sample and model for disease name j/the sum of the weights of all model symptoms related to disease name j, then AG3j=1−(Σ(|Hi− Qi｜)/2 ...(4).

For example, when searching for the disease name Kaede, 1-[|2/D ₁₁ +D ₁₂ +D ₁₃ +D _1o )-D ₁₁ /(D ₁₁ +D ₁₂
+D ₁₃ +D _1o ) +4/(D ₁₁ +D ₁₂ +D ₁₃ +D _1o )-D _1o /(D ₁₁ +D ₁₂ +D
₁₃ +D _1o )]/2. This degree of symptom matching is determined for each suspected disease name extracted in the secondary diagnosis. The degree of coincidence takes a value from 0 to 1, and the closer it is to 1, the closer it is to the theoretical model, and the closer it is to 0, the farther it can be diagnosed. The degree of coincidence obtained in this way is shown in Figure 14,
The disease names are displayed on the display device 6 in descending order according to a command from the CPU.

Next, CPU2 inputs the disease name determined in the second diagnosis.
A linear function f (α, β, γ, δ) whose variables are the initial suspicion level DDLj, the abnormality model agreement degree AG1j obtained in the tertiary diagnosis, the logical model agreement degree AG2j, and the symptom model agreement degree AG3j is integrated. Calculate as evaluation value TVj. The time constants α, β, γ, and δ are parameters determined empirically. Expressing this in the formula, TVj = α (DDLj) + β (AG1j) + γ (AG2j)
+δ(AG3j)...(5) is used for comprehensive diagnosis of disease name extraction. 1st
FIG. 5 shows the comprehensive evaluation value of each disease name calculated in this manner, displayed on the display device 6 together with the disease name in descending order of the value. Also, Figure 16 shows, for each disease name,
The overall evaluation value T, initial suspicion level A, abnormality model matching level B, logic model matching level C, and symptom matching level D are displayed as bar graphs.

Furthermore, all the judgment criteria of this system are based on the numerical values in each table, and are separated from the system's program flow, and also has a back-up file mechanism for the table, so the user can revise the contents of the table. By doing so, any diagnostic logic can be set. Further, in the above embodiments, analysis data of amino acids was taken as an example, but analysis data such as organic acid data, sugar data, diffusion data, guanidine data, etc. can also be used.

According to the present invention, it is possible to accurately diagnose a disease name, and since the degree of agreement with a model pattern is quantified, the degree of suspicion of the disease name of a specimen is clarified, which is extremely effective for diagnosis.

[Brief explanation of the drawing]

FIG. 1 is a schematic diagram of an automatic diagnostic apparatus shown as an embodiment of the present invention, and FIGS. 2 to 16 are images displayed on a display device to supplement the explanation of the operation of the present invention. 1: Amino acid analyzer, 2: Central control unit (CPU), 3: Disk memory, 4: Data file, 5: First storage unit, 6: Display device,
7: second storage unit, 8: third storage unit, 9: fourth storage unit.

Claims (1)

[Claims] 1 Analyzer, healthy value range for each item (normal range)
a first storage unit in which a healthy value table is stored, an item indicating a disease name related to each item and a weight representing the degree of relationship of the disease name to the item;
a second storage unit in which a disease name table is stored;
A third storage unit stores a disease name/item table showing items related to each disease name, abnormality model weights and logical model weights indicating the degree of relationship of the item to the disease name, and a model related to each disease name. a fourth storage unit in which a disease name/symptom table indicating symptoms and weights indicating the degree of relationship of the model symptoms to the disease name is stored; Detect an abnormal item in the specimen and calculate its degree of abnormality by referring to the healthy value table, and extract the name of the suspected disease of the specimen by referring to the item/disease name table for the detected abnormal item and extract the name of the suspected disease. The degree of suspicion is calculated, and for each extracted suspected disease name, the disease name and
By referring to the item table, the degree of suspicion regarding the abnormal item can be determined by calculating the abnormality degree of the abnormality item that matches the sample and the model listed in the table, the weight of the abnormality degree model of the abnormality item, and the weight of the abnormality degree model of the abnormality item, and the weight of the abnormality degree model of the abnormality item Anomaly degree model matching degree calculated based on the sum of the weights of the abnormality degree model of the abnormality item, the abnormality degree of the abnormality item that matches the sample and the model listed in the table, and the weight of the logical model of the abnormality item, It is calculated based on the logical model matching degree calculated based on the sum of the weights of the logical model of all the abnormal items of the model, and further, for each extracted suspect disease name, by referring to the disease name/symptom table, An automated system comprising an arithmetic unit that calculates the degree of symptom model matching based on the degree of matching symptoms in the models listed in the table, the weight of the model symptom of the symptom, and the sum of the weights of all model symptoms of the model. Diagnostic equipment. 2 Analyzer, healthy value range for each item (normal range)
a first storage unit in which a healthy value table is stored, an item indicating a disease name related to each item and a weight representing the degree of relationship of the disease name to the item;
a second storage unit in which a disease name table is stored;
A third storage unit stores a disease name/item table showing items related to each disease name, abnormality model weights and logical model weights indicating the degree of relationship of the item to the disease name, and a model related to each disease name. a fourth storage unit in which a disease name/symptom table indicating symptoms and weights indicating the degree of relationship of the model symptoms to the disease name is stored; Detect an abnormal item in the specimen and calculate its degree of abnormality by referring to the healthy value table, and extract the name of the suspected disease of the specimen by referring to the item/disease name table for the detected abnormal item and extract the name of the suspected disease. The degree of suspicion is calculated, and for each extracted suspected disease name, the disease name and
By referring to the item table, the degree of suspicion regarding the abnormal item can be determined by calculating the abnormality degree of the abnormality item that matches the sample and the model listed in the table, the weight of the abnormality degree model of the abnormality item, and the weight of the abnormality degree model of the abnormality item, and the weight of the abnormality degree model of the abnormality item Anomaly degree model matching degree calculated based on the sum of the weights of the abnormality degree model of the abnormality item, the abnormality degree of the abnormality item that matches the sample and the model listed in the table, and the weight of the logical model of the abnormality item, It is calculated based on the logical model matching degree calculated based on the sum of the weights of the logical model of all the abnormal items of the model, and further, for each extracted suspect disease name, by referring to the disease name/symptom table, The degree of symptom model matching is calculated based on the degree of the symptoms that match in the models listed in the table, the weight of the model symptom of the symptom, and the sum of the weights of all model symptoms of the model, and An automatic diagnosis device comprising an arithmetic device that calculates a linear function with variables including the initial suspicion level of a disease name, an abnormality level model matching degree, a logical model matching degree, and a symptom model matching degree as a comprehensive value.