JPH0769306B2 - How to set normal value range in electrophoretic analysis - Google Patents

Description

TECHNICAL FIELD The present invention relates to a method of setting a normal value range in electrophoretic analysis.

[Conventional technology]

In clinical tests, the normal value range is one guide for evaluating the analytical values of various items of the analyzed sample.

The normal value range is a value that almost falls within this range for healthy people. Generally, by performing statistical processing on the analysis data of many healthy people, the target analysis data has a probability of about 95%. Is set to fall within that range. This normal value range is based on statistical probabilities to the last, and it cannot be said that the analysis value is abnormal because it is out of this normal value range or is normal because it is within the range. Is extremely important in evaluating.

A / G (albumin to globulin ratio), Alb in serum protein fraction by cellulose acetate membrane electrophoresis analysis
(Albumin), α ₁ , α ₂ , β, γ globulin (globe
For the fractional value of b), the normal value range is set and used for evaluation, like other biochemical analysis items. Conventionally,
The normal value range of each fraction value in this electrophoretic analysis was set by fraction% in most cases, but recently, fraction concentration value (g / dl) -total protein (TP) x fraction%. The number of cases set by-is gradually increasing.

FIG. 23 shows an example of use of the normal value range in electrophoretic analysis. Most automatic electrophoretic devices or densitometers display normal values by listing each fraction value in the protein fraction report. The display may be printed in advance on the report, or may be input to the automatic electrophoretic device or the densitometer with a tenki or the like, stored in the internal memory, and printed out on the report at the same time as the analysis result. In addition, the normal value range stored in the internal memory is compared with the analysis value.If the value exceeds the normal value range, H, ↑, etc., and if it is lower, L, ↓, etc. are printed in the judgment field. There is.

Thus, the normal value range in the conventional protein fraction is A /
G, each fraction%, and each concentration value (g / dl), which are only a part of the information contained in the electrophoretic densitogram obtained by protein fractionation. Regarding the information such as the peak position of each fraction, the position of each fraction point, and the subtle changes in the densitogram curve, due to the increase or decrease of the individual protein components (proteins of a finer classification than 5 fractions) contained in The normal value range has not been set, and its evaluation is made by the judgment required of the skill of a laboratory technician or clinician.

[Problems to be solved by the invention]

However, in the use of the normal value range by the conventional fractionation%, there were the following problems.

Tables 1 and 2 show the analysis results of patient A and patient B, respectively.

In the case of patient A shown in Table 1, when the fractional% was evaluated, the fractional value was almost in the center of the normal value range. ing. On the other hand, patient B shown in Table 2 has a lower Alb than patient A.
The fraction concentration values are reduced to almost half, but the other fraction concentration values are the same as those of the patient A.

Here, when the patient B is evaluated in the normal value range of the fraction%, it can be determined that all the fractions are out of the normal value range. However, the globs of α ₁ , α ₂ , β, and γ are not quantitatively different from those of the patient A, and no abnormality is recognized. Therefore, when evaluating in the normal value range of the fraction%, because the other fraction% is affected by the abnormal fraction, the overall balance is lost and the relative ratio is out of the normal value range. However, it is difficult to evaluate the quantitative fluctuation of each fraction. In addition, when the evaluation is performed in the normal value range of the fraction% as described above, an abnormal fraction may fall within the normal value range and the determination may be erroneous.

On the other hand, when the patient B is evaluated in the normal range of the fractional concentration value, it can be determined that the Alb has a low abnormal value and the other fractions do not change and are within the normal range.

However, even if the fraction density values are used, if the area ratio between the fraction points shown in the densitogram is the same, for example,
As shown in Fig. 24, the appearance of M (monoclonal) protein causes γ
Even if the descintogram curve shows an abnormal change in the fraction, it cannot be evaluated in the normal range of the fraction concentration value. Similarly, for example 25 increasing the amount of alpha _2, as shown in FIG β
Even in the example in which the amount of P is decreased, if the fractional point between α _{2 and} β changes toward α ₂ and the relative area ratio does not change, the fraction% even if the pattern is abnormal. , It cannot be evaluated from the separation concentration value.

Here, in order to evaluate the pattern abnormality of the descintogram shown in FIG. 24 and FIG. 25, the peak position of each fraction and
The evaluation is easy if the normal value range of the fraction point position or the normal value range of the concentration at each measurement point in the migration length direction of the densitogram is set as shown in FIG. Thus, alpha _2, be a value fractionation%, or the fractional concentration value falls to the normal value range of beta, alpha ₂ of the peak position, alpha _2-beta
It is possible to easily determine that there is an abnormality in the position of the fractionation point between them, and also that the concentration at each measurement point is outside the normal value range at the α ₂ peak, near the fractionation point, and at the β anode side .

However, in the conventional electrophoretic analysis, the migration length fluctuates due to subtle changes in the analysis conditions, so the reliability of each measurement point position in the migration length direction is not sufficient and 1 μm
Since it is not possible to precisely control a small amount of the sample coating amount as described below, the coating amount varies and the concentration information at each measurement point cannot accurately correspond to each protein concentration in the sample. In addition, since the densitogram is generally displayed by autospan processing so that the peak value of Alb becomes a predetermined value, the densitogram will only display the relative ratio of the concentration without the concentration information, and the reliability of Some concentration information cannot be obtained. Therefore, it has not been possible at present to set the peak position of each fraction, each fraction point position, and the normal value range of the density at each measurement point.

In addition, as a method of solving such a conventional problem,
The following methods are possible. That is, at present, in protein fractionation analysis, it is common to analyze multiple samples (10 to 30 samples) simultaneously on one support, so apply a normal sample as a standard to these samples and simultaneously analyze them. Analysis is performed, and a determination range similar to the normal value range is set based on the analysis result of the normal sample. However, in order to obtain a statistically reliable value as the judgment range in this method, sufficient knowledge based on experience and experience in selecting a reference sample is necessary, and the analysis result of the normal sample is always It must be constant. In addition, in order to use it for daily analysis, it is essential that the same analysis result be obtained even if the quantity of normal sample is secured and the lot is changed. Therefore, it is very difficult to simultaneously analyze normal samples and set the determination range.

The present invention has been made in view of such conventional problems, and it is possible to set a highly reliable normal value range capable of detecting an abnormality that cannot be judged from the fraction% value or the fraction concentration value. It is intended to provide a method for setting a normal value range in electrophoresis analysis.

[Means for solving problems]

In order to achieve the above-mentioned object, the present invention sets the normal value range in electrophoretic analysis by using electrophoretic image data obtained by performing electrophoretic analysis on each of a large number of samples, and analyzing each electrophoretic image data by using a densitogram. The migration length is normalized so that the migration lengths are constant with each other, and at the same time, the density values for at least one fraction are proportional to each other by using the migration length normalized migration image data. Values are normalized, desired desired data related to each of these normalized electrophoretic data are extracted and collected, and statistical processing is performed based on the collected data to perform the above-mentioned normalization. Waveforms of the upper limit value and / or the lower limit value of the normal value range for the concentration value or information calculated based on the concentration value, which connects consecutive points in the migration length direction of the obtained densitogram It is characterized in that setting.

[Action]

In this way, for each electrophoretic image data of a large number of samples for setting the normal value range, if the concentration value normalization process is performed using the data that has been subjected to the normalization process to make the migration length constant, Even if there is variation in the coating amount or migration length, migration image data equivalent to that obtained by analysis with a constant migration length and a constant coating amount can be obtained. Therefore, statistical processing is performed based on the normalized data, and the density value or the density value that connects consecutive points in the migration length direction of the densitogram obtained by the normalization processing is calculated based on the density value or the density value. By setting a waveform of the upper limit value and / or the lower limit value of the normal value range for information, the waveform becomes highly reliable. Therefore, together with this waveform, the densitogram of the actual subject and its electrophoretic image information are shown. By displaying the data curve, it is possible to detect an abnormality that cannot be determined by the fraction% value or the fraction density value.

〔Example〕

First, concrete outline steps in the method of setting the normal value range according to the present invention will be described below.

Collection of specimens from a large number of people (In principle, it is desirable to have healthy people. If a large number of specimens from healthy people are not collected, patients may be added, but among them, those from healthy people should be increased as much as possible. Electrophoretic analysis of samples Normalization of electrophoretic image data Extraction of target information from normalized electrophoretic image data Collecting extracted information from each sample Statistical distribution of collected information data Judgment of form If the distribution form is not a normal distribution, data conversion to approximate it to normal distribution Rejection of outliers from information data Setting of normal value range from data judged to be valid Data to approximate to normal distribution When conversion is performed, the inverse conversion of the normal value range to return to the original data value is performed. Thus, for example, the migration image densitogram of the peak position of each fraction and the position of each fraction point Setting the normal value range of the broadcast.

If the normal value range to be obtained is information that continuously changes in the order of adjacent data points,
After the above process is performed for each data point, the following steps are further performed.

The values in the normal value range of the obtained electrophoretic image information of each data point are arranged in order of the data points.

If the data of adjacent data points cannot be smoothly connected because the conversion forms performed in the data conversion are different, smoothing processing is performed as necessary.

The obtained normal value range data is connected in the order of data points.

By the above, a continuous normal value range is set.

Hereinafter, a case where the normal value range upper limit value, the same median value, and the same lower limit value of the information related to the densitogram are set from a large number of electrophoretic image data will be described.

There are several statistical methods for setting the normal value range, but the following method is typical as a method for healthy people.

1) Parametric method (data mean, standard deviation SD
2) Non-parametric method (using rank of data) 3) Probability paper method (using data plotted on normal probability paper) Moreover, when it is difficult to collect a large number of data of healthy people, patient data When setting the normal value range based on, there are 1) Hoffmann method, 2) CRRP method (Clinical Reference Range Program), etc.

An example of setting a normal value range by the parametric method will be shown below for healthy person data. In order to use the paratric method, it is necessary to collect at least 100 healthy subject data.

Collection of samples Collect as many samples of healthy people as possible as analytical data for calculating the normal range. A healthy person is selected by comprehensively judging the biochemical analysis results of as many items as possible, including the protein fractionation test of the collected samples, and the health condition.

Protein Fraction Analysis by Electrophoresis Electrophoretic analysis of selected healthy human samples is performed to obtain electrophoretic image data. If the protein fractionation test is carried out in, the analysis result of the sample of a person selected as a healthy person may be extracted from the analysis result.

Normalization of electrophoretic image data For each electrophoretic image data, find the Alb fraction peak and β fraction peak, and the Alb fraction peak is the 100th point from the anode,
The electrophoretic image data is normalized so that the β peak becomes the 200th data point and the area surrounded by the baseline and densitogram is proportional to TP.

Hereinafter, the electrophoretic analysis of the sample and the normalization processing of the electrophoretic image data will be described.

FIG. 1 is a schematic cross-sectional view showing the configuration of a main part of an example of a densitometer in an electrophoretic device for performing electrophoretic analysis of a sample.

Support 1 that has been dyed, decolorized and dried in a dyeing tank
Is conveyed by the feed roller 2 to the photometric unit 4 containing the decalin 3, where the photometric device 5 causes a constant speed,
For example, after the photometric scanning is performed at 8 mm / sec, the sheet is discharged by the sheet discharge roller 6.

The photometric device 5 is configured to receive the light from the light source 5a passing through the support 1 by the light receiving element 5b, and the photometric device 5 having the light source 5a and the light receiving element 5b is used as shown in FIG. The electrophoretic image 7 formed on the support 1 is photometrically scanned by moving the electrophoretic image 7 on the support 1 in the scanning direction b orthogonal to the transport direction a.

FIG. 3 is a block diagram showing a configuration of a main part of an example of a data processing device that processes an output of the photometric device 5. This data processing device includes a logarithmic amplifier 12, an A / D converter 13, a CPU 14, and a memory 1.
5, keyboard 16, CRT 17, floppy disk 18 and printer 19 are provided. The photoelectric conversion output obtained from the light receiving element 5b by the photometric scanning is logarithmically amplified by the body number amplifier 12 and converted into absorbance, and then the sampling rate determined by the analysis conditions such as migration time in the A / D converter 13 ( Sampling is performed in synchronization with a clock corresponding to, for example, 12 msec, and converted into a digital signal, which is stored in the memory 15 under the control of the CPU 14, and the smoothing process which is a moving average process is performed on the sampling data stored in the memory 15. At the same time, a normalization process is performed after an auto-zero process for removing the base floating amount due to the light amount.

FIG. 4 shows a flowchart of an example of the normalization process. The number of data points related to the predetermined migration length in normalization was 350, and the peak position of Alb, which is stable as a prominent peak in the migration image at 100 data points, and the appearance position in the migration image at 200 data points. Match the stable β peak positions. That is, using the respective peak positions of Alb and β as reference points, these reference points are made to coincide with 100 data points and 200 data points of 350 data positions in the normalization processing, respectively.

The number of sampling data points in the scanning range of the photometric device 5 is more than 350 points.

In this normalization process, first, in step 1, the reference points, which are the peak positions of Alb and β associated with the migration length, are detected from the sampling data of the measurement sample stored in the memory 15. The detection of this reference point will be described below.

First reference point detection method From the sampling data of the measurement sample, as shown in FIG. 5, data that exceeds a predetermined threshold value such that both end points become the end points of the electrophoretic image is extracted, and both end points of the extracted data are extracted. To detect the presence or absence of a peak in a predetermined range l ₁ and l ₂ ,
The maximum concentration in the detected peak is the Al of the measurement sample.
The reference points are detected as the peak position of b and the peak position of β.

Second reference point detection method Sampling data of a measurement sample is sequentially integrated from both end points, and positions at which those values become a predetermined value or a predetermined ratio to a total integrated value are set as both end points of the electrophoretic image. Similarly to the first method, the peaks in a predetermined range are detected from both end points thereof, and the reference points are detected by using the peak positions as the respective peak positions of Alb and β. For example, if the cumulative value from the positive polarity side to the Alb direction of the migration polarity is set to 2% of the total cumulative value and the cumulative value from the negative electrode side to the γ direction is set to 1% of the total cumulative value, the migration length almost matches the visual length obtained. To do.

Third reference point detection method Normal serum was electrophoresed on the same support and measured together with the measurement sample. First, the peak value of Alb of normal serum was detected, and then Alb was detected.
From, the third peak position of the migration polarity in the negative electrode direction is detected as the β peak position. After that, the peak position of Alb is detected for the measurement sample, and then a peak is obtained near the position of the data points equal to the data points between the Alb peak position and the β peak position of the normal serum previously obtained from this Alb peak position. Then, the peak position is detected and the respective reference points are detected with the peak position of β. When detecting the peak position of Alb from the sampling data of the electrophoretic image, Alb is stable as a conspicuous peak, so a relatively high threshold value is set, and the maximum concentration value of data exceeding this is the peak of Alb. It may be detected as a position, or as described in Japanese Patent Application No. 60-22622 proposed by the present applicant, the maximum value D _M is detected from all sampled data, and then from the positive polarity side of the migration polarity. A peak of 1/16 or more of the maximum value may be detected as the Alb peak position P _M. Here, 1/16 removes the prealbumin peak, and empirically according to various pathological conditions so that P _M can be detected as the peak position of Alb even if D _M is not Alb but other components. It is a fixed value and not fixed.

By any one of the above detection methods, the peak position of Alb that is stable as a prominent peak position and the peak position of β that is stable in the appearance position are detected as reference points.

Next, normalize the X axis so that the peak positions of Alb and β detected in step II coincide with the positions of 100 data points and 200 data points on the X axis having 350 data points. To do. For example, the peak position of Alb in the measurement sample is 120, and the peak position of β is 230.
If the points are set, those data positions are made to coincide with the 100 data points and 200 data points on the X axis, and the other sampling data has the migration length of the reference point of the measurement sample of 110.
It corresponds to the number of data points (230-120), but corresponds to 100 data points (200-100) on the X-axis, so the data points on the X-axis are shifted according to the ratio.

Here, if there is no sampling data corresponding to the data point on the X axis, interpolation processing or extrapolation, that is, processing for aligning with both end data is performed.

This completes the normalization of the X-axis regarding the migration length.

Next, in step III, the values of the sampling data of 350 points on the X-axis normalized so that the integrated value is almost equal to the integrated value between the data positions of the corresponding measurement sample,
Number of data between reference points of measurement sample and number of data between these reference points on the X-axis (100 in this example)
The Y axis is normalized based on the ratio of For example, in the above example, the Alb peak position in the measurement sample is 120 data points,
Since the β peak position is 230 data points and the number of 110 data points between them is normalized to 100 data points, the value of the sampling data at each normalized data point is
It is multiplied by 110/100 and the Y axis is normalized.

The above processing is performed by reading the sampling data stored in the memory 15 under the control of the CPU 14, and the processed data is stored in the floppy disk 18.

Next, in step IV, the density is normalized so that the integrated value and the absolute amount of the density in each fraction correspond to each other. In the normalization of this concentration, the total protein concentration value or Alb concentration value of the measurement sample is measured in advance by a biochemical analysis device or the like, and the value is passed through the keyboard 16 or from the biochemical analysis device or the device. The data is input online or offline via the computer system for inspection connected to and stored in the floppy disk 18. Also, the reference integrated value for the unit concentration (1 g / dl) of the input absolute amount is also stored in the same manner. Alb sets the reference integrated value for 1 g / dl to 15,0, for example.
Set 00 (by A / D conversion value). Assuming that the concentration value of Alb is input as 4 g / dl, first obtain the integrated value of the Alb fraction of the normalized data, and then obtain the integrated value and the standard corresponding to the input Alb concentration value. Calculate the ratio with the integrated value. For example, the integrated value of normalized Alb fraction is 80,00
If it is 0 (in A / D conversion value), the input Alb concentration value is 4 g / dl and the standard integrated value is 4 (g / dl) x 15,000 = 6
Since it is 0,000, the ratio of 80,000 to 60,000 is Becomes Then, the data value at each point is multiplied by this ratio to complete the normalization of the density. In this density normalization process, it is also possible to simultaneously correct the difference in the dyeability of each fraction as described in Japanese Patent Application No. 60-36606 proposed by the present applicant. When the total protein concentration value is input, the same processing can be performed by obtaining the ratio of the reference integrated value corresponding to the input value to the integrated value of all the fractions. The normalization process is performed as described above, and the data is stored in the floppy disk 18.

After the electrophoretic image data normalization processing is completed, the target information is extracted from the normalized electrophoretic image data.
In the subsequent steps, an example of setting the normal value range of the fraction point position between Alb and α ₁ will be described.

Extraction of electrophoretic image information Follow the concentration data sequentially from the Alb peak (100th point) of the normalized electrophoretic image data to the cathode side, and find the first minimum point.
The data position is defined as the fraction point position as the fraction point between Alb and α ₁ . Here, the density data at each data position of the normalized divided image data is represented by D: density, i: a number corresponding to a healthy person as a target (i = 1 to 200 for 200 people) = j: data position ( 1 to 350 from the anode side, j = 100 at the Alb peak,
If the β peak is represented by Di, j with j = 200), the minimum point is generally D _{i, j} <D _{i, j + 1} , as shown in FIG.
It is the position of j that satisfies Di, j <D _{i, j-1} . The position of this j, that is, the position of the fraction point between Alb and α ₁ is
When the peak is normalized to j = 100 and the β peak is set to j = 200, it is around j = 120.

Collection of extracted information For each sample, find the fraction point between Alb-α ₁
Collect them. Table 3 shows an example of the frequency distribution of the fraction point positions between Alb and α ₁ collected.

Judgment of statistical distribution To judge the distribution form of the target data, a. Display by histogram b. Judge by plotting data on normal probability paper c. Probit conversion data by probit analysis method Then, the shape of the distribution is assumed by the goodness-of-fit test method, and the X ² test is performed based on the deviation from the distribution of the data.

Here, a method example using the histogram of a is shown.

FIG. 7 shows the histogram of the frequency distribution shown in Table 3. It can be seen that this histogram is almost bilaterally symmetric and can be regarded as a normal distribution. That is, if the distribution is normal, the skewness is 0 and the kurtosis is 3, but the skewness and kurtosis of the distribution in FIG. 7 are And are very close to 0 and 3, respectively. Therefore, it can be determined that the distribution is normal.

Data conversion If the distribution cannot be regarded as a normal distribution, the data should be converted to approximate the normal distribution. In this example, the distribution form of the data may be regarded as a normal distribution, so
No data conversion is performed. An example of data conversion will be described later.

Rejection of outliers If there are exceptionally high or low values in the data, SD may be affected by that and may not show the correct value. In order to exclude these outliers, the data
Then, from SD, the upper limit of rejection; UL (Uppor Limit) = + 3SD The lower limit of rejection; LL (Lower Limit) = -3SD is calculated, and data larger than UL or smaller than LL is truncated. After that, for the remaining data, SD, UL, and LL are calculated again, and the data outside the range is rejected.
This is repeated until there is no data to be discarded.

In this example, (first time) = 122.89, SD = 1.59, UL = 127.67, LL = 11
From 8.11, the data at fraction point 128 is rejected.

(Second time) Calculating N = 412 excluding the fraction point position 128 = 122.89, SD = 1.59, UL = 127.61, LL = 118.1
From 5, there is no data to be rejected, which ends the rejection of outliers.

Setting of normal value range Abnormal value rejection is completed, SD is used. Upper limit of normal value = + 2SD = 126.03 Lower limit of normal value = -2SD = 119.73 is set. In this example, since the data conversion of is not performed, these values become the upper limit value and the lower limit value of the normal value range. Also, is set as the median of the normal value range.

From this, the normal value range of the fraction point position between Alb-α ₁ is 1
It was set as 19.73 to 126.03.

Above, although the setting example of the normal value range fractions point position between the Alb-alpha ₁ fraction, the same method with alpha _1-.alpha. between _2, alpha _2-beta
The position of the fraction point between β and γ can also be detected from the normalized electrophoretic image data of each sample, which is used for calculating the normal value range.

Also, instead of the minimum point, the maximum point is detected, or the maximum value between each fraction point is detected to obtain the peak position of each fraction,
It is also possible to calculate the normal value range of each fractional peak position by performing the same process for the relevant peak position of each sample. However, in the above example, the Alb peak is 100
Since the 200th point and the β peak are set as the reference points and normalized, the normal value range cannot be set for these peaks.

FIG. 8 shows an example of how to effectively use the normal range of each fractional peak position and each fractional point position calculated as described above.

In FIG. 8, the electrophoretic image data of the actual test sample is similarly normalized to display a densitogram, and the normal value range of each fractional peak position is overlapped with the densitogram and the normal value range is contacted with the densitogram. On the outer side, and the normal value range of each fractional point position on the inner side in contact with the densitogram. In this way, it is possible to easily determine the abnormality of each fractional peak position and fractional point position. When the detected peak position and the separated point position of the detected sample are out of the normal range, marks such as * are displayed above and below the normal value range. In FIG. 8, alpha _2, the position of the fractionation point between γ peak positions and alpha _2-beta indicates an abnormal deviates the normal value range.

Next, an example will be shown in which the normal value range at each data point of the density data of the densitogram is set as the information data of the migration image represented by the continuous data points associated with each other from the normalized migration image data.

The frequency distribution shown in Table 4 is j when the area between the densitogram and the baseline is set to 100,000 when TP is 7g / dl and the distance between each data point is 1 in the above normalization process. This is an example of the density data of the 290th point.

Determination of systematic distribution type FIG. 9 shows the histogram of the example shown in Table 4. Unlike the one shown in the previous example, it shows an asymmetrical distribution as if the value is larger, and is different from the normal distribution. By the way, even if the skewness and kurtosis are calculated, the skewness = 1.24, kurtosis = 5.88, N = 413, = 19.2, SD = 4.2, which is very different from that of the normal distribution. In addition, if we calculate the skewness and x degree again by applying the outlier rejection based on ± 3SD to these data, we can see the skewness = 0.64 and kurtosis = 3.45. It is hard to say that they are close.

Data conversion Generally, as shown in the histogram in Fig. 9, when the data value is tailed toward the larger one, log conversion or 1 / nth power conversion It is known that can be approximated to a normal distribution, and conversely, if the data value is tailed toward the smaller side, it can be approximated to a normal distribution by n-th power transformation (x ² , x ³ ), 10 ^x power transformation. It has been known.

Table 5 shows the frequency distribution obtained by log-transforming the same data, and FIG. 10 shows the histogram.

As is clear from FIG. 10, the histogram obtained by the data conversion has a slight unevenness, but is almost symmetrical and close to the normal distribution. By the way, when skewness and kurtosis are calculated, skewness = 0.49, kurtosis = 3.64, N = 413, = 1.284, SD =
0.086, which is closer to the normal distribution than before log conversion.

Rejecting outliers Further, rejecting outliers with ± 3SD as the reference value.
As a result, skewness = 0.31, kurtosis = 3.14, N = 410, = 1.282, S
Even better results were obtained with D = 0.082, which means that the log transformation was able to approximate a normal distribution.

Setting of normal value = 1.282, SD = 0.082 obtained from the above, normal value range upper limit value = + 2SD = 1.446, normal value lower limit value = -2SD = 1.
Set 118. However, this is the normal range of the log transformed values, not the normal range of the original data values.

Inverse conversion of normal value range In order to return the normal value range calculated above to the normal value range of the original data value, inverse conversion is performed. In this example, since the data conversion was performed using log x, the inverse conversion is in the form of 10 ^x Normal value range upper limit value = 10 ^1.446 = 27.93 Normal value range lower limit value = 10 ^1.118 = 13.12 Set as.

The example of setting the normal value range has been described above by taking the density data of j = 290th point as an example. However, the normal value range can be set for the j = 1st to 289th points and the 29th to 350th points as well.

Also, not only the density data but also the information calculated based on the density data near each data point such as the unevenness judgment value of the densitogram can be used as a normal value by collecting the calculated values and performing the same processing. The range can be set.

Furthermore, when the calculated normal value range exists for each data point and these are continuously related to the adjacent data points, the normal value range upper limit value and lower limit value are graphed in order of each data point. By tying to the top, a curve showing the normal value range can be set. In this case, if the lines do not connect smoothly, the moving average method, weighted moving average method, or an appropriate curve function should be used if necessary because the conversion forms used for data conversion differ between adjacent data points. The data points are connected smoothly by performing smoothing such as the least square approximation of.

Normal value range curve obtained as described above, the densitogram in the actual test sample as shown in FIG. 11, and by overlapping display with the data curve showing the electrophoretic image information, its usefulness It can be further increased. By doing this, it is possible to clearly distinguish the abnormal portion of the density, change the line type, color, etc. of the densitogram of the portion exceeding the normal value range, hatch the range that exceeds,
The determination can be further facilitated by performing processing such as coloring. Incidentally, the densitogram of the actual test sample and the data curve showing the electrophoretic image information thereof are of course subjected to the normalization processing described in the setting of the normal value range and displayed.

The normal value range set according to the present invention is not only displayed redundantly with the densitogram and the like, but also used as a reference value for determination in the automatic analysis of pathological conditions, so that its effectiveness and reliability can be further enhanced.

Hereinafter, automatic analysis of pathological conditions using the normal value range set by the present invention will be described with reference to several examples.

In order to automatically analyze the pathological condition from the information obtained by the electrophoretic analysis, a flowchart of the pathological condition classification as shown in FIG. 12 has been proposed. In the following example, the pathological condition is analyzed by automatically detecting the characteristic points of the electrophoretic images of M protein (monoclonal protein), γ suppression and β-γ bridging from the densitogram.

First, the detection of M protein will be described.

M protein may be fractionated at any position on the densitogram, but most of them appear in the β fraction to the γ fraction and have a very narrow band width because they are monoclonal. Further, there are benign and malignant M proteins, and in many cases benign M protein appears in a form in which M protein is superimposed on the fractionation pattern, and malignant M protein appears in the fractionation pattern. Often accompanied by suppression of other proteins.

From these characteristics of M protein, M protein can be detected by detecting the presence or absence of a peak on the densitogram between β-γ fractions. That is, in the densitogram in which M protein does not appear, as shown in FIG. 13A, between β-γ fractions, there is a gentle valley between β-γ and a gentle peak of the γ fraction. In other words, when the change in the curvature is observed between the β-γ fractions, the pattern becomes convex upward, concave downward, convex upward, and concave downward. On the other hand, when a benign M protein appears, another peak exists between β-γ fractions, as shown in FIG. 13B, and the change in curvature becomes complicated. Also, if there is the appearance of malignant M protein,
For example, as shown in FIG. 13C, the pattern change between β-γ fractions is the same as in the normal case of FIG. 13A,
The appearance peak becomes sharper than that in the normal pattern, and a clear reduced portion in which the γ fraction is suppressed appears before or after the M protein peak.

Therefore, when simply detecting the M protein peak, the required migration length portion of the normalized densitogram, for example β
-When detecting the M protein peak between γ fractions,
3 from 200 data points (β peak position) of normalized data
It is only necessary to detect the presence or absence of a convex portion between 00 data points, but in order to determine whether it is a benign or malignant one, whether the γ fraction is suppressed near the peak position of M protein. Should be detected from the comparison with the normal value range in the corresponding part.

Below, referring to the flowchart shown in FIG. 14, β−
An example of the M protein detection process between the γ fractions will be described.

First, in step I, the degree (convex value) of the convex portion of the pattern between the predetermined data points corresponding to the β-γ fractions of the normalized densitogram of the test sample is calculated. The method of calculating this convex value will be described below.

First convex value calculation method The number of data points k on each side of the data point j
The detection width is set to 2k, and as shown in FIG.
How much the densitogram protrudes with respect to the straight line connecting the data value D _{jk at the} point and the data value D j + k at the point _{j + k} is evaluated by the area S of the portion surrounded by the straight line and the densitogram. In this case, for example, the area S when the trapezoid is approximated
Is It is represented by. The area S can be obtained by various quadrature methods such as Simpson other than the trapezoidal approximation.

Here, the value of k is preferably 3 to 6, and if it is smaller than this, a minute change can be read, but fine noise is also picked up, and if it is large, the value of S itself becomes large. , The effect of smoothing makes it stronger against noise, but fine changes are lost. The value of k is the same in the second to fourth calculation methods described below.

Looking at the change in the value of S thus obtained, not only M protein having an independent peak but also a trace amount of M protein having no clear peak as shown in FIG. 16 can be detected. .

Second convex value calculation method A function S / 2k is set with respect to S obtained by the first calculation method. Since the value of S / 2k represents the average height of the convex portion in the detection width 2k, the degree of dependence on the detection width 2k is lower than that of S (S is 2k in the apex of a triangle, for example). Corresponding to the degree of convexity on the densitogram strongly appears.

Third convex value calculation method A function S / (2k) ² is set for S obtained by the first calculation method. The value of S / (2k) ² is the ratio of the detection width 2k and the average height S / 2k, and represents the degree of convexity per unit detection width. If the shapes of the convex portions are similar, the detection width 2k is set. Regardless of the same value. Therefore, the detection width 2k in this case determines the degree of smoothing.

Fourth convex value calculation method The convex value is calculated by the second derivative. In this case, since the functional expression of the densitogram is unknown, a function approximation is performed from the sequential data value of the detection width 2k by the method of least squares, etc., and the second derivative value of the obtained approximate functional expression is used as a convex value. To do. Below, y = ax ² + bx when the detection width is 5, 7 and 9 data
Shows the approximate second derivative F ″ _(j) when the least-squares approximation is performed on the parabola generally represented by + c.

In case of 5 data (2k = 4) F ″ _(j) = (2D _j−2 −D _j−1 −2D _j −D _{j + 1} + 2D _{j + 2} ) / 7 In case of 7 data (2k = 6) F ″ _(J) = (5D _j-3 −3D _j-1 −4D _j −3D _{j + 1} + 5D _{j + 3} ) / 42 For 9 data (2k = 8) F ″ _(j) = (28D _j-4 ＋ 7D _j-3 -8D _j-2 -17D _j-1 -20D _j -17D _{j + 1} -8D _{j + 2} ＋ 7D _{j + 3} ＋ 28D _{j + 4} ) / 462 These values are essentially the detection width 2k. It is independent, so 2k will determine the degree of smoothing as in the third convex value calculation method.

After the convex value is obtained by any one of the above calculation methods, it is then determined in step II whether the convex value is larger than a predetermined threshold SL _1, and when the convex value is SL ₁ or less, the M protein peak is determined. If it is determined that there is no SL1 and it exceeds SL ₁ , then in step III, in order to detect whether or not the convex portion is an M protein peak, the half-value width of the convex portion (the number of data is within a predetermined range SL ₂ ~ S
Determines whether the L _3. Here, the full width at half maximum is SL ₂ to SL _3.
When it is out of the range, it is determined that there is no M protein peak, and when it is in the range, it is considered that there is an M protein peak and whether the convex value is larger than a predetermined threshold value SL ₄ (> SL ₁ ) in step IV next. To judge. In this judgment process, the convex value is
If SL ₄ or less, M protein peak was judged to be suspected, and SL ₄
When it exceeds, it is assumed that there is a clear M protein peak, and then the peak data position of the convex portion is detected in step V in order to detect the presence or absence of γ suppression.

In the above processing, the densitogram of the test sample was normalized according to the flowchart of FIG. 4 so that the integrated value was 100,000 with respect to the total protein concentration value of 7 g / dl, and the convex value was calculated by the second calculation method. As a result of various experimental studies, k = 3 to 6
When the S / 2k value was 30 or more, M protein having almost independent peaks could be detected, and a trace amount of M protein peak having no clear peak could be detected at 10 to 30. In addition, by setting the range of the full width at half maximum of the convex portion to 10 to 20 data numbers, the main peak of the γ fraction (half width at 20 or more) and the β _1c peak appearing near the valley between the β and γ fractions, It was possible to reliably separate the peak of fibrinogen (in the case of plasma analysis) (half-width of 10 or less) and the M protein peak.

In FIG. 14, when the position of the peak data of the convex portion as the M protein peak is detected, the data of the test sample and the corresponding data set by the present invention are displayed before and after the peak data point detected in step VI. By comparing with the normal value range at the point, it is determined whether there is a point below the normal value range, and if there is a point below the normal value range, it is determined that there is γ suppression and that malignant M protein (myeloma) is present. If not, it is determined that there is a benign M protein.

Next, an example of β-γ bridging detection processing will be described.

β-γ bridging means that the valley between β-γ is the γ fraction (I _g
G, I _g M, I _g A) is a phenomenon that becomes buried due to a polyclonal increase of G, I _g M, I _g A) and the separation between β-γ becomes unclear. When this phenomenon is extreme, as shown in FIG. 17, β-γ is smoothly connected, and there is no fractional point between β-γ. In conventional autospan densitograms, β-γ bridging was found in some of the densitograms showing β-γ bridging, and β-γ bridging decreased even if the γ fraction was normal. It showed the same pattern as above, and it could not be judged unless the fractional concentration (g / dl) was also checked for the distinction.

In this example, as shown in the flowchart in FIG. 18, first, a portion exceeding the normal value range is detected between the β-γ peaks of the normalized densitogram of the test sample. In this detection process, first, in step I, data of portions corresponding to predetermined β peaks and γ peaks are extracted from the normalized densitogram of the test sample, and normal portions at the portions corresponding to the extracted data are extracted. Compare with the value range and detect the part exceeding the normal value range.

Here, when the portion exceeding the normal value range is not detected, it is determined that there is no β-γ bridging, and when the portion exceeding the normal value range is detected, next, in step II, the exceeding portion is β-γ bridging. In order to determine whether or not, it is determined whether or not the width (number of data) of the exceeding portion is equal to or larger than the reference width. That is, since β-γ bridging shows a polyclonal increase, the range over the normal value range is wide. On the other hand, β _1C appearing near the valley between β-γ fractions
The widths of fibrinogen and fibrinogen are narrower than those of β-γ bridging even if there are changes beyond the normal range. Therefore, the width exceeding the normal value range is equal to or larger than the reference width, for example, β-γ.
By detecting whether or not there is 60% or more of the number of data between peaks, β-γ bridging can be reliably separated from β _1C and fibrinogen. Here, when the width exceeding the normal value range is less than or equal to the reference width, it is determined that there is no β-γ bridging, and when it is greater than or equal to the reference width, then the increase is the presence of the M protein peak in step III. In order to distinguish whether it is due to β-γ bridging or due to β-γ bridging, M protein detection step I, II or steps I to IV shown in FIG. The presence or absence of a peak is detected, and when an M protein peak is detected, the increase is due to M protein and β-
It is determined that there is no γ-bridging, and when the M protein peak is not detected, it is determined that β-γ-bridging is present, as the increase is a polyclonal increase.

In addition, as in the flowchart shown in FIG.
First, when the peak of M protein is detected and β-γ bridging is detected in the absence of M protein peak,
Step III can be omitted in FIG.

By the above processing, β-as a polyclonal increase
γ bridging can be detected with high accuracy.

Next, reading detection processing will be described by taking Alb fractionation as an example.

The Alb fraction is generally composed of a single protein, the pattern is almost symmetrical with respect to the peak position, the mobility is stable, and the concentration is high, so it is the most prominent fraction in the electrophoresis image. However, during administration of high jaundice serum, high lipid serum, antibiotics, etc., bilirubin, free fatty acids, and drugs bind to Alb to cause a change in mobility, and as shown in FIG. In many cases, the reading of the Alb fraction is asymmetrical.

In this example, reading the Alb fraction to the positive electrode side
The detection is performed according to the flowchart shown in FIG. First, in Step I, it is detected from the normalized densitogram of the test sample whether or not there is a portion exceeding the normal value range on the positive electrode side from the Alb peak position (100th data point), and if there is no reading, no reading is performed. If so, in the next step II, a discrimination value representing the degree of symmetry of the Alb fraction is calculated in order to distinguish whether the increase is an overall increase of Alb or an increase due to reading. The method of calculating the discriminant value will be described below.

First Discriminant Value Calculation Method As shown in FIG. 21A, the integrated value ΣI from the normalized densitogram of the test sample to the Alb peak from the positive electrode side, A
The integrated value ΣII from the lb peak to the fraction points of Alb to α ₁ is calculated, and the ratio ΣI / ΣII is set as the discriminant value.

Second Discriminant Value Calculation Method As shown in FIG. 21B, the center of gravity G of the Alb fraction is obtained, and the shift between the center of gravity point _j g and the Alb peak position _j p ( _j p− _j g)
Is the discriminant value.

Third discriminant value calculation method A predetermined ratio to the peak concentration of Alb or a desired value is set as the threshold SL as shown in FIG. 21C, and the range in which individual Al fraction data exceeds the threshold SL Detect
The ratio L ₁ / L ₂ of the widths L ₁ and L ₂ between the end point positions on the positive electrode side and the negative electrode side and the Alb peak position _j p in the detected range is taken as the discriminant value.

If the discriminant value is obtained by any one of the above calculation methods,
Next, in step III, the discriminant value of the test sample is
It is determined whether or not the normal value range obtained by the same calculation method is exceeded, and when it does not exceed, it is determined that there is no reading, and when it exceeds, it is determined that there is reading on the Alb positive electrode side.

The determination results regarding M protein, β-γ bridging, and Alb reading analyzed as described above are calculated based on the normalized data of the test sample for each fraction%, A / G
The ratio, each fractional density value, and its normalized densitogram are displayed on the CRT 17 in FIG. 3 and are recorded in the printer 19 in predetermined columns of the report 20. When displaying the densitogram on the CRT 17 and the report 20, the densitogram of the test sample and the pattern associated with the densitogram in the normal value range may be displayed in an overlapping manner.

It should be noted that the present invention is not limited to the above-described embodiments, and many changes and modifications can be made. For example, although the fractional analysis of serum proteins has been described in the above example, the present invention can be effectively applied to the analysis of isozyme activity values by electrophoresis. Further, in FIG. 20, by detecting a portion exceeding the normal value range on the negative electrode side from the Alb beak position and performing the same processing,
Alternatively, by detecting a portion below the normal value range on the positive electrode side from the Alb peak position and detecting that the discriminant value representing the symmetry falls below the normal value range.
It is also possible to automatically determine the appearance of the reading of the Alb fraction on the negative electrode side. Further, the presence / absence of this reading can be determined not only in the Alb fraction but also in other fractions. The calculation method of the discriminant value of symmetry in this case is the same as the above first to third discriminant value calculation methods.
As shown in Fig. 22A, the difference between the peak position jp of the fraction and the center position jc of the fraction, which is the center position of both fraction points, is used as the discriminant value, or as shown in Fig. 22B, the center of gravity of the fraction is determined. The discrepancy between the position jg and the center position jc of the fraction is used as a discriminant value, and as shown in FIG. 22C, the data D _{j at the} data points separated by k from the center position jc of the fraction or the peak position jp on both sides. _{The difference between (ck)} and D _{j (c + k)} , or the difference between D _{j (pk)} and D _{j (p + k)} , the integrated value of the square of the difference, or the correlation thereof can be used as the discriminant value. In addition, M between the β-γ fractions described above
The method for determining the protein and the inhibition of the γ fraction associated with its appearance and the detection determination of bridging is the same for the determination of the increase or decrease of each individual protein in other Alb, α ₁ , α ₂ fractions, the appearance of minor peaks, etc. Can be applied. Further, in the normal value range, the upper limit value and the lower limit value can be set asymmetrically. Further, in the above-mentioned example, the smoothing process of the sampling data from the densitometer is performed before the normalization process, but it may be performed after the normalization process. Further, in the above example, the density normalization process is performed last, but this process may be performed before the X-axis normalization process. Furthermore, the reference point in the migration image is not limited to the Alb peak position and β peak position,
It is also possible to use the peak positions of other fractions, the end points of electrophoretic images, and the like. Further, a one-dimensional array sensor or a two-dimensional array sensor can be used as the light receiving element that constitutes the measuring device. Further, the normalized data of the test sample is compared with the normal value range to automatically determine the increase / decrease of the components in the test sample, or M
Detection of proteins etc., automatic analysis of the presence or absence of a predetermined component in the test sample, or detection of reading etc. from the comparison of data in a predetermined range and normal value range of the normalized data and detection of the test sample It is also possible to automatically analyze the presence or absence of a predetermined component therein.

The effects of the above-described embodiment are as follows.

(1) By performing the normalization, it is possible to improve the reliability of the original data for calculating the normal value range.

(2) By setting the normal value range from a large number of data, statistically reliable data can be obtained.

(3) A normal value range can be set for each data point of the densitogram.

(4) As compared with the case where a normal sample serving as a standard is analyzed at the same time, it is not necessary to secure and store the normal human serum serving as a standard.

(5) By making the set normal value range overlapped and displayed by a densitogram or a curve showing electrophoretic image information obtained from the densitogram, the determination becomes easy.

(6) In performing the automatic analysis of the pathological condition from the densitogram, by using the normal value range set according to the present invention as a criterion for judgment, more accurate pathological condition analysis becomes possible.

〔The invention's effect〕

As described above, according to the present invention, with respect to each electrophoretic image data of a large number of samples for setting the normal value range, the normalization process for making the electrophoretic length constant and the normalization process proportional to the concentration are performed, and For statistical information based on the normalized data, connect the continuous points in the migration length direction of the densitogram obtained by the normalization processing to the concentration value or the information calculated based on the concentration value Since the upper limit value and / or the lower limit value of the normal value range is set, it is possible to obtain a highly reliable waveform representing the normal value range. Therefore, together with this waveform, a densitogram of an actual subject, By similarly normalizing and displaying the data curve indicating the electrophoretic image information, it is possible to detect an abnormality that cannot be determined by the fraction% value or the fraction concentration value.

[Brief description of drawings]

FIG. 1 is a schematic cross-sectional view showing a configuration of a main part of an example of a densitometer in an electrophoretic apparatus for performing electrophoretic analysis, FIG. 2 is a view showing a scanning direction of an electrophoretic image, and FIG. 3 is a data processing apparatus. FIG. 4 is a block diagram showing the configuration of the main part of FIG. 4, FIG. 4 is a flowchart showing an example of normalization processing, FIG. 5 is a diagram for explaining an example of reference point detection, and FIG. FIG. 7 is a diagram for explaining, FIG. 7 is a histogram of extracted minimum points, FIG. 8 is a diagram showing an example of use of a normal value range set by the present invention, and FIG. 9 is a histogram of density data at predetermined data points. , Fig. 10 is a histogram of the data obtained by log-converting the concentration data in Fig. 9, Fig. 11 is a diagram showing an example of overlapping display of the densitogram of the test sample and the normal value range, and Fig. 12 is a conventional example. Proposed pathology classification FIGS. 13A to 13C are diagrams showing densitogram patterns between β-γ fractions, FIG. 14 is a flowchart showing an example of the M protein detection process, and FIG. 15 is a graph showing convex values shown in FIG. FIG. 16 is a diagram for explaining an example of a calculation method, FIG. 16 is a diagram showing an example of M protein on a densitogram that can be detected by the calculation method, FIG. 17 is a diagram showing an example of β-γ bridging, FIG. 18 is a flowchart showing an example of β-γ bridging detection processing, FIG. 19 is a diagram showing an example of Alb reading, FIG. 20 is a flowchart showing an example of Alb reading detection processing, and FIGS. C is a diagram for explaining three calculation examples of the discriminant value of symmetry shown in FIG. 20, and FIGS. 22A to 22C are diagrams for explaining other three calculation examples of the discriminant value of symmetry, 23 to 26 are views for explaining a conventional technique. 1 ... Supporting body, 2 ... Feed roller, 3 ... Decalin, 4 ... Photometric unit, 5 ... Photometric device, 5a ... Light source, 5b ... Light receiving element, 6 ... Paper ejection roller, 7 ... Electrophoretic image, 12 …… logarithmic amplifier 13 …… A / D converter, 14 …… CPU 15 …… memory, 16 …… keyboard 17 …… CRT, 18 …… floppy disk 19 …… printer, 20 …… report

Claims (1)

[Claims] 1. Normalization of migration length is performed so that each migration image data has a constant migration length on a densitogram by using migration image data obtained by performing electrophoresis analysis on each of a large number of samples. At the same time, the electrophoretic image data obtained by normalizing the electrophoretic length is used to perform a concentration value normalization process so that the concentration values for at least one fraction are proportional to each other. Related desired data are extracted and collected, statistical processing is performed on the basis of the collected data, and the densitogram obtained by the normalization processing is connected to consecutive points in the migration length direction, the concentration A method of setting a normal value range in an electrophoretic analysis, which comprises setting a waveform of an upper limit value and / or a lower limit value of a normal value range for information calculated based on a value or a concentration value.