JPH09126987A - Particle image analyzer - Google Patents

Abstract

PROBLEM TO BE SOLVED: To attempt to improve the analyzing accuracy of a column and the like containing particles therein by adopting a constitution to simultaneously calculate the morphological feature amounts of the particle image area and particle internal structure area by a feature amount calculator by providing a ternary processor and a particle image labeling processor. SOLUTION: The stroboscopic light 11 of a microscope light source is lit, and projected to the particles in a flow cell 8 via a condenser lens 12. The transmitted image of the particle is introduced from an objective 13 to a TV camera 15, and transduced into an electric signal. The particle image is image- processed and feature-extracted by a feature extractor 16. Eventually, the particles in the sample are automatically sorted by a particle analyzer 17 by using the particle feature amounts. A central control unit 9 controls the entire apparatus, calculates the data and judges the analyzed result. The data necessary for the measurement is sent from an external data unit 3. A feature amount calculator 20 simultaneously calculates the morphological feature amounts of the area and the peripheral length at the respective labels, and stores them in a feature amount memory.

Description

Detailed Description of the Invention

[0001]

TECHNICAL FIELD The present invention relates to a particle image analysis device,
In particular, the present invention relates to a particle image analyzer that generates an image of particles contained in a fluid and classifies the particles based on the image data.

[0002]

2. Description of the Related Art In a conventional cell image processing apparatus such as that disclosed in Japanese Patent Laid-Open No. 1-242961, only white blood cells, whose internal structure is important, calculate only the morphological features of the whole particle. The analysis accuracy cannot be improved because the morphological feature amount representing the internal structure of the cylinders containing particles inside the white blood cell cylinders and the like contained in the urine sediment is not calculated.

[0003]

In the above-mentioned prior art, when the binarization process, the particle identification number calculation process, and the morphological feature amount calculation process are performed in series, it takes a long time to perform real-time processing. , If you try to calculate the morphological features of the structure inside the particle in real time,
A plurality of binarization processing units, a plurality of particle identification number calculation units, and a plurality of morphological feature amount calculation units are required, and the morphological feature amount of the structure inside the particles is adopted in terms of mounting area and cost. Therefore, the accuracy of analysis can be improved because the morphological feature quantity that represents the internal structure of the cylinders containing particles inside the white blood cells and the leukocyte cylinders contained in the urine sediment where the internal structure of the particles is important cannot be calculated. There was a problem that there was not.

An object of the present invention is to provide a means for calculating the morphological characteristic amount of the internal structure of particles, which is inexpensive and has a small mounting area, in real time processing. It is intended to improve the accuracy of analysis of cylinders containing particles inside white blood cell cylinders and the like contained in urine sediment.

[0005]

According to the present invention, in order to calculate a morphological feature amount of a grain internal structure region, an observation target region is divided into three light intensity distributions of a background region, a particle image region, and a grain internal structure region. A ternarization processing unit that divides the area is provided, and a particle image labeling processing unit that has a function of adding a structure identification flag for distinguishing whether or not the particle image is a particle internal structure area to the particle image identification number is provided. A feature amount calculation unit that simultaneously calculates the morphological feature amounts of the particle image region and the particle internal structure region is provided for each added particle image identification number.

According to the present invention, the observation area is divided into three light intensity distribution areas of the background area, the particle image area, and the particle internal structure area by the ternarization processing unit, and the binary image processing unit divides the binary image by the binary image processing unit. Since the structure identification flag for distinguishing whether or not it is the particle internal structure region is added to the particle image identification number based on the data calculated by the crystallization processing unit, the feature amount calculation unit distinguishes between the particle image region and the particle internal structure region. Morphological features can be calculated at the same time, real-time processing, inexpensive and small mounting area, the internal structure of the particles is important. The column containing the particles inside the leukocyte, the leukocyte column contained in the urine sediment, etc. It is possible to improve the analysis accuracy of classes.

[0007]

FIG. 1 is a block diagram showing the overall configuration of the present invention.
I will explain using.

In the figure, 1 is a sample to be measured and 2 is a sample suction part. Reference numeral 4 represents a flow system control unit, and 5 represents a measurement sample discharge unit. 7 is a sheath liquid, 6 is a sheath liquid syringe, and 8 is a flow cell. 9 is a central control unit, 10
Is a strobe lamp drive unit, 11 is a strobe lamp, 12
Indicates a condenser lens. Reference numeral 13 is an objective lens, 15 is a TV camera, 16 is a feature extraction circuit, and 17 is a particle analysis unit. Reference numeral 14 is an output unit, and 3 is an external data unit.

The well-stirred measurement sample 1 is sucked by the sample suction unit 2, and the measurement sample discharge unit 5
Is extruded from the upper part of the flow cell 8. Simultaneously with discharging the measurement sample, the sheath liquid 7 is also pushed out to the flow cell 8 via the sheath liquid syringe 6.
As a result, the measurement sample becomes a flow wrapped in the sheath liquid, and flows down through the center of the flow cell 8 without disturbing the flow.

The particle still image is taken at a predetermined position of the flow cell 8. The strobe light 11, which is a microscope light source, is turned on, and the flow cell 8 is passed through the condenser lens 12.
The particles inside are irradiated. The transmission image of the particles passes through the objective lens 13 and is projected on the image pickup surface of the TV camera 15 to be converted into an electric signal.

The picked-up particle image has a feature extraction circuit 16
Then, the image processing and the feature extraction processing are performed, and finally, the particle analysis unit 17 automatically classifies the particles in the sample using these particle feature amounts. The central control unit 9 controls the entire apparatus, calculates data, and determines the analysis result. Data required for measurement is sent from the external data unit 3.

Capture of a particle still image will be described with reference to FIG.

When a measurement sample, which is a measurement particle, flows through the flow cell 8, a strobe lamp 11 as a microscope light source is used.
Is lit, and the particles in the flow cell 8 are irradiated through the condenser lens 12. The transmission image of the particles passes through the objective lens 13 and is projected on the image pickup surface of the TV camera 15 to be converted into an electric signal.

The above operation is repeated for the measurement time, and static particle images are collected until the processing of all the measurement samples is completed.

The automatic classification of particle images will be described with reference to FIG.

The picked-up particle image has a feature extraction circuit 16
The image processing and the feature extraction processing are performed in step (4), and finally, the particle analysis unit 17 performs automatic classification of the particle components using these particle feature amounts, and the final inspection result is collected by the central control unit 9.

In the image processing, image processing is performed in real time for each particle still image. Regarding the contents of the image processing, the binarization processing, the labeling processing, the particle characteristic calculation and the like are executed at high speed by the characteristic extraction circuit 16. Next, the plurality of obtained feature amounts of each particle are combined and the particle analysis unit 17
The pattern recognition process is executed with. When the pattern recognition processing is performed on all particles imaged during the measurement time and the particle classification processing is completed, the central control unit 9 based on these classification results.
The conversion result is converted into the particle concentration in the measurement sample, and the classification results are totaled.

The particle classification result is output to the output unit 14. When the result of the external measurement is reflected in the measurement condition of the apparatus, it is sent to the central control unit 9 via the external data unit 3.

A flow of ternarization processing of an image having an internal structure in a particle will be described with reference to FIGS. 2 and 3. FIG. 2 is a white blood cell image of which the internal structure is important, in which the nucleus is present in the cytoplasm. The threshold for distinguishing the background from the cytoplasm is Thl, the density threshold for distinguishing the background and the cytoplasm from the nucleus is THh, and the threshold of the cytoplasm and the nucleus obtained by binarizing the image data with these thresholds. It shows the binarized data of the region. FIG. 3 shows a label image of the white blood cell image of FIG. The label image is a basic process for performing particle analysis, and is an image in which the identification numbers of particles required when a plurality of particles are present are assigned to the constituent pixels, and the feature amount is calculated for each label. The nuclear region is also included in the binarized data of the cytoplasm, and the label image of the particle is created for this binarized data. The binarized data of the nucleus is used as a structure identification flag, and "1" is added to the label data "A". The background label data around the white blood cells is L = “0 0”, the cytoplasm has no binarization data of the nucleus, so L = “0 A”, and the region of the nucleus has the structure identification flag set L = “1 A”. "Is. In this example, the structure identification flag is added to the most significant bit, but it can be realized by adding it to any bit.

FIG. 4 shows an example of the configuration of an image processing circuit including a ternarization processing unit 18, a labeling processing unit 19, a feature amount calculation unit 20, a feature amount memory 20, and a classification processing unit 17. The ternarization process consists of a binarization process A and a binarization process B. The binarization process A is set to Thl to perform binarization of the cytoplasm, and the binarization process B is set to THh to set the nucleus The region is binarized and the background, cytoplasm, and nucleus are ternarized. The output of the binarization process A has the particle identification number calculated by the particle identification number calculation unit, and the output of the binarization process B is added to the particle identification number through a shift register that delays the time required for the particle identification number process. Then, the label image is generated. Feature amount calculation unit 2
In 0, the morphological feature amount such as the area and the perimeter is calculated at the same time for each label and stored in the feature amount memory.

If the particle identification numbers are the same, the classification processing unit 17 can determine that the particles have the same cytoplasm and nuclei, and can perform analysis including the feature amount of the structure inside the particles, and classify the particles. The accuracy can be improved.

In the present embodiment, an example of a flow type particle analysis apparatus is shown, but it is applicable to an apparatus that analyzes an object using image processing, such as an apparatus that analyzes an image of a sample in which a sample is applied on a slide glass. Is also effective.

Further, although the processing using the ternarization processing is described in the present embodiment, by adding the processing of quaternarization or more, the feature quantity of the internal structure can be further increased, and the classification accuracy can be further improved. .

[0024]

According to the present invention, a binarization circuit for distinguishing the internal structure region of a particle is added, and a processing unit for adding this data to a particle identification number is provided. The morphological features of the particle image area and the internal structure area in the particle can be calculated at the same time by the configuration, and the internal structure of the particle is important inside the white blood cells, the white blood cell casts contained in the urinary sediment in the urinary sediment component, etc. It is possible to improve the analysis accuracy of cylinders including particles.

[Brief description of the drawings]

FIG. 1 is a block diagram of a flow-type particle image analyzer.

FIG. 2 is an explanatory diagram of an example of a white blood cell image.

FIG. 3 is an explanatory diagram of an example of a label image of white blood cells.

FIG. 4 is a block diagram of image processing.

[Explanation of symbols]

16 ... Feature extraction circuit, 17 ... Particle analysis unit, 18 ... Ternary processing unit, 19 ... Labeling processing unit, 20 ... Feature amount calculation unit, 21 ... Feature amount memory.

Claims (3)

[Claims] 1. A sample containing particles is caused to flow together with a sheath liquid in a flow cell, and an observation target area in the flow cell is irradiated with light to generate a particle image of the particles flowing in the observation target area, and the particle image is analyzed. In the particle image analysis device for classifying the particles, the observation area is divided into three light intensity distribution areas, that is, a background area, a particle image area, and a particle internal structure area. It has a particle image labeling processing unit that has the function of adding a structure identification flag that distinguishes whether or not it is a particle internal structural region, and the morphology of the cell for each particle image identification number to which the structure identification flag is added. A particle image analysis device characterized in that a morphological characteristic amount of a particle image region and a particle internal structure region is calculated by a characteristic amount calculation unit for calculating a characteristic amount. 2. A sample containing particles is applied on a slide glass, and an observation target area on the slide glass is irradiated with light to generate a particle image of the early-stage particles in the observation target area, and the particle image is analyzed. In the particle image analysis apparatus for classifying the particles, the observation area is divided into three light intensity distribution areas, that is, a background area, a particle image area, and a particle internal structure area. It has a particle image labeling processing unit that has the function of adding a structure identification flag that distinguishes whether it is a particle internal structure region or not, and the morphology of the cell for each particle image identification number to which the structure identification flag is added. A particle image analysis device characterized in that a morphological characteristic amount of a particle image region and a particle internal structure region is calculated by a characteristic amount calculation unit for calculating a characteristic amount. 3. The particle object identification according to claim 2, further comprising a multi-value quantization processing section for dividing the observation object area into a plurality of light intensity distribution areas, and distinguishing whether the particle image identification number is a particle internal structure area or not. A particle image identification number calculation unit having a function of adding a flag, and a feature amount calculation unit that calculates a morphological feature amount of a cell for each particle image identification number to which a structure identification flag is added. A particle image analyzer for calculating morphological features of a region and a particle internal structure region.