JPH03108638A - Method of identifying particle aggregation pattern - Google Patents

Abstract

PURPOSE:To identify a highly reliable aggregation pattern by calculating characteristic values of every concentration profile in an image screen of a particle aggregation pattern. CONSTITUTION:A microplate 2 has a well 2a placed in a matrix on its upper face and mounted on a microplate base 3. A lighting device 5 is provided below the base 3 so that the radiated light is projected through the base 3 and the plate 2 to a TV camera 4 wherein a video signal V is input to an A/D converter 6 to be converted into a digital signal Z, which is recorded on an image memory 7 as image data. A CPU 8 calculates the center position of a particle aggregated substance based on the image data and selects one or more lines including the center position to calculate a concentration profile in each line. Then the CPU 8 calculates a plurality of characteristic values containing peak height in every concentration profile to identify an aggregation pattern and outputs it to a printer 9. A D/A converter 10 converts the image data into a video signal V' and displays on a monitor TV 11.

Description

Detailed Description of the Invention (Industrial Field of Application) This invention provides a particle system that captures an image of particle aggregates formed in a container and determines the aggregation pattern of the particle aggregates based on the image data. This invention relates to an agglomeration pattern determination method.

(Prior Art) Particle aggregation pattern determination methods have been widely used for determining blood type and detecting antigens and antibodies in blood. According to this method, blood to be tested is placed in a well (reaction container) of a microplate, the blood, etc. is agglutinated based on an immunological reaction, and the blood type is determined based on the agglutination pattern.

However, since these judgments were made by visual judgment by inspectors, there were problems such as individual differences among inspectors and lack of reproducibility even among the same inspector. Therefore, the above-mentioned visual judgment is performed by a device to give objectivity to the judgment result, to improve measurement accuracy, and to save labor in the judgment by automation. Those techniques are disclosed, for example, in Japanese Patent Application Laid-open No. 60-135748, Japanese Patent Application Laid-Open No. 81-21.
Publication No. 59411.

JP-A-62-105031, JP-A-63-582
It is disclosed in Publication No. 37 and the like. According to these techniques, wells of a microplate (including aggregates) are imaged with a television camera, the image data is subjected to predetermined image processing, and an aggregation pattern is determined from the processing results.

(Problems to be Solved by the Invention) However, the automatic determination technology proposed above has the following problems.

Since aggregates are not always formed at the center of the wells of a microplate, it is necessary to determine the center position of the aggregate for each well. In addition, as one of the above automatic determination techniques, for example, the above image data is binarized, and 2
There is a method that determines the agglomeration pattern by determining the area of the agglomerate from the valued data, but such a simple method tends to lead to erroneous determinations. For example, if the agglutination reaction is positive, the aggregate spreads thinly within the well of the microplate, the difference between the brightness of the aggregate and the brightness of its surroundings (i.e., the well) is small, and the binarization level slightly increases or decreases. Therefore, the area value of the aggregate obtained as described above changes greatly. In addition, binarization processing is closely related to the amount of illumination light,
When the amount of illumination light changes, the area value of the aggregate changes similarly to the above.

In addition to the above-mentioned determination method, there is also a method in which the outline of the aggregate is determined based on the degree of grayscale change in the image, and pattern determination is performed from that outline. However, this method is easily affected by disturbances, and the outline It is difficult to extract accurately.

As mentioned above, the automatic determination technology (method) related to the above proposal
has the above-mentioned problems, making it difficult to make accurate determinations, and making it impossible to determine blood types, etc., with reliability equal to or higher than that of experienced examiners.

(Object of the Invention) The present invention has been made to solve the above problems, and it is an object of the present invention to provide a highly reliable particle aggregation pattern determination lj method.

(Means for Achieving the Object) A first aspect of the present invention provides a particle aggregation method in which an image of a particle aggregate formed in a reaction container is captured and an aggregation pattern of the particle aggregate is determined based on the image data. A pattern determination method, in order to achieve the above object, includes a first step of determining the center position of the particle aggregate based on the image data, and determining the center position on the image plane of the particle aggregate. a second step of selecting at least one straight line and determining a concentration profile for each straight line; and a plurality of feature quantities of the concentration profile including at least a peak height for each concentration profile determined in the second step. and a fourth step of determining an aggregation pattern of the particle aggregate based on the feature amount.

Furthermore, the invention of claim 2 has the following features in order to achieve the above object:
In the third step, as the feature quantity, in addition to the peak height, one or more of the area 1 variance, 1'2 uniform width) 1'-value width of the concentration profile is further determined for each concentration profile. After that, the plurality of feature amounts are statistically processed to determine the aggregation pattern of the particle aggregates.

Furthermore, the invention of claim 3 has the following features:
Concentration profiles on two predetermined straight lines that are orthogonal to each other on the image of the particle aggregate are determined, and the positions of the centers of gravity of each concentration profile are determined, and these positions are defined as the center position of the particle aggregate.

In addition, the invention of claim 4 has the following features in order to achieve the above object:
In the third step, the characteristic quantities include, in addition to the peak height, the area 1 variance and average width of the concentration profile for each concentration profile.

One or more of the '-1' In widths are also determined.

Furthermore, the invention of claim 5 has the following features in order to achieve the above object:
In the second step, four straight lines are selected, and the angles formed by the straight lines are a multiple of 45°.

Furthermore, in order to achieve the above object, the invention of claim 6, in addition to the invention of claim 5, statistically processes the feature amount in the fourth step to determine the aggregation pattern of the particle aggregates. are doing.

(Operation) According to the present invention, at least one straight line including the center position is selected on the image plane of the particle aggregate, and after the concentration profile on each straight line is determined, the peak height is calculated for each concentration profile. Since the characteristic amount of the concentration profile including , is determined, the concentration information of the particle aggregate is sufficiently reflected in the characteristic amount, and the influence of disturbance is suppressed. Furthermore, it becomes unnecessary to set a "binarization level".

(Embodiment) FIG. 2 is a configuration diagram of an agglomeration pattern determination device 1 to which an embodiment of the present invention can be applied. This device 1 has a microphone plate stand 3 on which a microplate 2 is placed and positioned. The microplate 2 has wells 2a arranged in a matrix on its upper surface,
It is mounted on the microplate table 3 by a transport mechanism (not shown). Further, an illumination device 5 is provided below the microplate stand 3, and the light emitted from the illumination device 5 passes through the microplate stand 3 and the microplate 2 and is provided above the microplate stand 3. The image is projected onto the television camera 4.

Then, the entire microplate 2 is photographed by the television camera 4 and input as a video signal V to the A/D converter 6, where it is converted into an 8-bit, 256-gradation digital signal Z according to the luminance. As shown in Figure 3,
Image data ID of 480 pixels vertically and 512 pixels horizontally (X, Y
) is recorded in the image memory 7. Further, the image memory 7 is connected to a CPU 8, and the image data 10 (X, Y) is read out to the CPU 8, various calculations (to be explained in detail later) are performed, and based on the calculation results, an agglomeration pattern is created. A judgment is made. The determination result is then output to a printer 9, a display (not shown), or the like. Further, the image data recorded in the image memory 7 is converted into a video signal V' by a D/A converter 10, and is provided to a monitor television 11 so that the image of the microplate 2 is displayed on the monitor television 11. It is composed of

Next, one embodiment of the present invention will be described with reference to FIG.

(1) Imaging of Sample In this embodiment, the microplate 2 is first mounted on the microplate table 3 by the transport mechanism. Then, an image of the microplate 2 is taken with the television camera 4 while the illumination device 5 is turned on, and the image data 10
(X, Y) is recorded in the image memory 7.

Based on the image data, the aggregation pattern of each well is sequentially processed in the following steps.

(2) Calculation of center position Inverted image data (7
Data corresponding to degree a) D(X, Y) is obtained and recorded in another area of the image memory 7.

D(X,Y)-10MAX-10(X,Y)-
(1) Here, the well region is a circular, blank, or square-shaped region centered on the predicted well center position and including the inside of the well. Also, "MAX is the image data ID
This is the maximum value of (X, Y), and corresponds to image data of the background of the aggregate (ie, around the aggregate).

Then, based on this inverted image data D(X, Y), the center position (1°j) of the aggregate formed in the well is determined (step Sl, first step). An example of such a method is as follows.

As shown in Figure 4, the expected center position of the aggregate (x
, yo), horizontal axis (X axis) and vertical axis (Y axis)
2 concentration profiles z' and z' on straight lines ρ and ρ parallel to , respectively, are determined.

y Note that the microplate 2 is positioned on the microplate table 3, and since the formation position of aggregates in the well can be predicted to some extent, the position
yo can be set in advance.

Furthermore, those concentration profiles z’, z’y The position of the center of gravity in the horizontal and vertical axis directions (x).

y) respectively. Specifically, position X according to the following equation.

Find each Yg.

Σ o(Xo +, yo) −(xo+k)k=−
nX! Yg ″ Σ D(xo+, yo) k=-n Σ D(xo, yo+k)・(yo 1k) k=-n ...(2A) ...(2B) Σ D(xo, yo+k) k=-n Here, to avoid the influence of the outer edge of the well, the value (2
The value (
n). Furthermore, the above operation may be repeated to obtain the center position (x, y) with higher accuracy. After this, following the calculation of the positions xg and yg, the X coordinate taX of the center of gravity position and the Y coordinate yg are rounded off to integers, and the values are converted to the central position of the aggregate (
t, j).

In this embodiment, the concentration profile 2' on the straight line N is
The position X g was calculated based on the position (x, yO)
The position Xg for each straight line is determined based on the concentration profile on a plurality of straight lines in the vicinity of The center position X may also be determined, which further increases the accuracy of the center position Xg. The same applies to the center position y.

(3) Derivation of concentration profile Next, on the aggregate image plane, select at least one straight line passing through the center position (i, j) found in step S1, and select the inverted line recorded in the image memory 7. Determine the density profile in each straight line based on the image data D (X, Y) (Fig. 5) (Step S2. Second process)
. In this embodiment, as shown in FIG.
+ j) and whose angles with the horizontal axis (X-axis) are 0@45°, 90°, and 135@, respectively.

J) Select N, J and set each linear amount 1. I.

45' 90 135 0 4
59, I am looking for the concentration profile at ρ90
135. Specifically, the inverted image data 9D in the image memory 7
(to i+, j) (However, k−・−, −2, −1, 0,
1, 2, -'F) by accessing the straight line 1
The concentration profile at ° can be determined. 7th
Figure A shows the results. Furthermore, when the inverted straight line data D (14k, j-k) in the image memory 7 is accessed, the density profile on the straight line r145 is determined. FIG. 7B is a diagram showing the results. Furthermore, each straight line 1. 90 for the concentration profile at 1
135 Each is found in the same manner as above. In addition, although the horizontal pitch of the data in FIG. 7A and FIG. 7B is the same, the horizontal pitch of the data in FIG. 7B is actually J2 times that in FIG. 7A, and these density profiles When calculating the characteristic mW, which will be described later, from the above, it is necessary to perform a predetermined correction. Note that this will be explained in detail later.

(4) Calculation of feature amount Next, for each density profile obtained in step S2, the feature amount of the density profile is calculated (step S3. third step). In this example, as the feature quantity, the peak height, 6 areas, 1 variance, or the average width are determined for each density profile as follows.

(4-1) Peak Height H The peak height H6 is determined based on the concentration profile shown in FIG. 7A. In addition, H45 was determined based on the concentration profile shown in Figure 7B, and the peak heights H and H were also determined as follows:
Find it in the same way.

90 135 (4-2) Area S The area s, sss is given by the following formula. 0 45' 90' 135 I ask.

In order to avoid this, the value (2 m) is set to be sufficiently smaller than the diameter of the well 2a. Note that formula (313').

The coefficient (-J'T) in (3D) is a correction coefficient for correcting the drawing pitch.

(4-3) Variance V The variance v , v , v , v is calculated as 0 45 90 135 using the following formula.

■o ′...(4A) S -, W-・Σ D (to 1+, j-k)
...(3B)45k,.

v −2・ 5 ・・(4B) S −J: “・ Σ D (in 14, j+k)
...(3D) 135 km-1
% However, as above, the influence of the outer edge of well 2a V90 -...(4C) V-2・35...(4D) Note that the coefficient (- 2)
is a correction coefficient for correcting the drawing pitch.

(4-4) Average Width W The average widths w , w , w , w are calculated using the following formula: 0 45 90 135 .

Wo -3/Ho...(5^)w
-s 7H...(5B) 45 45
45 W -3/H... (5C) 90 90 90 W -3/H... (5D) 1.35 135 135 In addition to these, the special quantities of each density profile include half width, (3rd moment, 4th moment), etc. However, as a result of actual verification, it was found that the above four features were sufficient in terms of processing speed and accuracy.

Then, the range (
- maximum value - minimum value) and determine whether or not their range exceeds a preset value. If it is determined that the value exceeds the set value, it is determined that the data is abnormal, and the class determination described below is not performed for that sample. Cases that are determined to be “data abnormalities” include:
For example, there are cases where the microplate 2 vibrates during the agglutination reaction, resulting in an asymmetrical shape of the aggregate.

On the other hand, when the range between each feature is within the set value, the average values H, S, and V between each feature are determined according to the following equation.

Find W.

H=(H+H+H+H)/4...(6A)0 45
90 135 S-(So1S45+S9o+S13,)/4 (6B
)V-(V+H+H+H)/4=(6C)0
45 90 135 W-(W +H+H+H)/4...(81))0
45 90 f35 (5) Class determination Next, the values H and S obtained as described above.

Based on V and W, it is determined which class of aggregation pattern the aggregate matches with among the five classes indicating positive or negative agglutination reaction, i.e., -±"+"++, "empty". (Step S4. 4th
step), the agglutination reaction within the well 2a of the microplate 2 is classified into five classes: "-"±", "+", "++", and "empty".

As a method for the above determination, for example, S RI (=Sta
There are methods such as a method using a discriminant function, a method based on Mahalanobis distance calculation, and the like. Note that all of these methods are conventionally well-known methods. In this example, SRI
An algorithm is used to determine the class. The procedure will be briefly explained below.

In order to perform class determination using the SRI algorithm,
First, prior to the actual inspection, prepare a large number of samples whose class has been determined by visual inspection in advance, and use the aggregation pattern determination device 1 shown in FIG.
~S3), the feature quantities H and S of those samples are calculated.

Is it necessary to find p V, , W (P-1, 2, 3,...)?

Then, each pp from the feature quantities H, S, V, and W
Find the average value and standard deviation of each feature in the p class.

That is, the average value H, S, Vct ct CI' WCl and standard deviation σ111' S l' of the feature quantity (peak height 2 area 1 variance and average width) based on each feature quantity of the sample corresponding to class -
V I ' σw1 is determined respectively. Similarly, for the other classes “±”, “+”, ’++”, and “sky”, the feature values (peak height 1 area 1
Find the mean value and standard deviation (variance and mean width).

Next, when determining the class of the unknown sample, the values H, S, and V obtained in step S3 are used.

Substituting W into the following equation, the distance from each class i'fD Cl
.

Find D, D, D, D.

C2C3C4C5 (blank below) ... (7A) ... (7B) ... (7C) ... (7D) ... (7E) In addition, formula (7B), (7C), (7D),
(7E), Ho2~HC5° ”C2~SC5
"C2~vC5'WC2~WC5" is the average value of each feature (peak height 1 area 9 variance and average width) of each class "±", "+", ++, and d, and σ
H2~σ115'S.

σ85' V2 V5' σw2ゞσw5 is the standard c table deviation of each feature (peak height 9 area 2 variance and average width) of each class “±”, ’+”, “++”, and “sky” .

Then, the class corresponding to the smallest value among these distances DC1''C2, DC3''C4°Dc5 is determined to be the class of the object to be inspected.

As mentioned above, in this example, class determination was performed using the SRI algorithm, but it goes without saying that the method is not limited to this, and the method using the above-mentioned discriminant function or the method based on Mahalanobis distance calculation may also be used. good. In particular, since the SRI algorithm is established on the premise that there is no correlation between the feature quantities, the determination accuracy decreases if there is a correlation between the feature quantities. On the other hand, a method using a discriminant function or a method based on Mahalanobis distance calculation can accurately perform class determination regardless of the presence or absence of a correlation between each feature.

In addition, in the above embodiment, in step S2 (second upper step), four straight lines "'45.'90°ρ" are selected, and each straight line has ρ45"90'I35
Although the concentration profile at 0p is determined, the number of selected straight lines is not limited to 4. After selecting only one straight line (for example, straight line p) and determining its concentration profile, the above Pattern determination may be performed as follows. However, in order to improve the precision of the feature amounts, it is desirable to select at least two watersides.

As described above, the aggregation pattern can be determined without binarizing the image data or extracting the outline of the aggregate formed in the well 2a from the image data.

Therefore, in this embodiment, "
There is no need to set a "threshold value" and the operation is simple. Also, for the same reason as above, there is little influence from changes in the light amount of the illumination device 5. In addition, in the above embodiment, contour extraction processing is not performed. Therefore, there is almost no influence of disturbance.As a result, according to this embodiment, highly reliable aggregation pattern determination can be performed.

Furthermore, since class determination is performed using a statistical method from a plurality of feature quantities, the reliability of class determination becomes even higher.

Furthermore, the entire microplate 2 is imaged by the television camera 4, and the center position of the aggregate formed in each well 2a is determined as described above based on the image data (step S1). The aggregate is in each well 2a.
Even if it deviates from the center position of the aggregate, there is no need to adjust the position of the microplate 2 again to find the center position, and the center position of the aggregate can be found as is.

(Effects of the Invention) As described above, according to the invention of claim 1, after selecting at least one or more straight lines including the center position on the image plane of the particle aggregate and determining the concentration profile on each straight line, For each of these concentration profiles, the feature values of the concentration profile including the peak height are determined, and the aggregation pattern of particle aggregates is determined based on the feature values, which improves the reliability of aggregation pattern determination. can.

Further, according to the second aspect of the invention, since the aggregation pattern of the particle aggregates is determined by statistically processing a plurality of feature amounts, the reliability of the aggregation pattern determination can be further improved.

[Brief explanation of drawings]

FIG. 1 is a flowchart showing an embodiment of a particle aggregation pattern determination method according to the present invention, and FIG. 2 is a block diagram of an aggregation pattern determination apparatus to which an embodiment of the present invention can be applied. 3 and 5 are diagrams each showing image data recorded in the image memory, FIG. 4 is an explanatory diagram for explaining a method of calculating the center position of an aggregate, and FIG. , FIG. 7A and FIG. 7B are diagrams each showing a density profile. 2... Microplate, 81-S4... Step 2a... Well,

Claims (6)

[Claims] (1) A particle aggregation pattern determination method that captures an image of a particle aggregate formed in a container and determines an aggregation pattern of the particle aggregate based on the image data, the method comprising: a first step of determining the center position of the particle aggregate; a second step of selecting at least one or more straight lines including the center position on the image plane of the particle aggregate and determining a concentration profile on each straight line; a third step of determining feature quantities of a plurality of concentration profiles including at least a peak height for each concentration profile obtained in the second step; and determining an aggregation pattern of the particle aggregates based on the feature quantities. A method for determining a particle aggregation pattern, comprising: a fourth step of determining. (2) In the third step, in addition to the peak height, any one of the area, variance, average width, and half width of the concentration profile is selected for each concentration profile as the feature quantity.
Claim 1: further determining the agglomeration pattern of the particle aggregates by further determining a plurality of characteristic quantities and statistically processing the plurality of characteristic quantities.
The method for determining a particle aggregation pattern described above. (3) The concentration profiles in two predetermined straight lines that are perpendicular to each other on the image of the particle aggregate are determined, and the positions of the centers of gravity of each concentration profile are respectively determined, and these positions are determined as the center position of the particle aggregate. 1. The method for determining a particle aggregation pattern according to 1. (4) In the third step, in addition to the peak height, any one of the area, variance, average width, and half width of the concentration profile is selected for each concentration profile as the feature quantity.
2. The method for determining a particle aggregation pattern according to claim 1, wherein the particle aggregation pattern determination method further determines at least one. (5) The method for determining a particle aggregation pattern according to claim 1, wherein four straight lines are selected in the second step, and the angles formed by the straight lines are a multiple of 45°C. (6) The method for determining a particle aggregation pattern according to claim 5, wherein in the fourth step, the feature amount is statistically processed to determine the aggregation pattern of the particle aggregate.