JP4244065B2 - Urine component classification device - Google Patents

Description

  The present invention relates to an apparatus for classifying tangible components in urine by applying computer image processing and recognition techniques.

Conventionally, analysis of urine components, such as analysis of urinary sediment components, includes (1) centrifuging a urine sample, (2) removing the supernatant using an aspirator or pipette, or by decantation. (3) Apply a certain amount of the remaining residue to the slide glass and use the cover glass as a specimen. (4) Set on a microscope and form components (red blood cells, white blood cells, epithelial cells, cylinders, microorganisms, crystals. Salt), etc.). Usually, all these processes are performed manually, which places a heavy burden on the laboratory technician. In addition, there are large variations in the inspection results, and there are naturally individual differences in judgment, and there is a problem in accuracy (Japanese Patent Laid-Open No. 4-337460, Japanese Patent Laid-Open No. 5-296915, and Japanese Patent Laid-Open No. 5-32285).
JP-A-4-337460 JP-A-5-296915 JP-A-5-322885

  In order to solve these problems, in recent years, a flow cytometer is not used to prepare a specimen to be sampled. After mixing the staining solution with the test solution, the sample is flowed through the flow cell in a suspended state. There are automatic analysis methods such as optical and optical. However, in the method of automatic analysis by the above flow cytometer method, the formed components are always flowing, so it is very difficult to focus on each, and the focus is fixed at a fixed position. In addition, the flow cell has a certain thickness. Based on these facts, the above flow cytometer method can accurately measure only the formed component that has flowed to the focused position, and if the formed component is located slightly away from the focused position, The amount of deviation is blurred and an image measurement result different from that at the time of in-focus is obtained, and an erroneous determination result may be obtained.

  It is an object of the present invention to provide a highly accurate and accurate urinary substance classification device by having a function of automatically focusing on a tangible substance in a test liquid, particularly urine. .

As a result of intensive studies in view of the above circumstances, the present inventors have a step of imaging a test liquid and are equipped with a function of automatically focusing when imaging a formed component in the test liquid As a result, a clear and focused image was obtained, detailed classification was possible with high precision, and it was found that the drawbacks of the conventional component analyzer could be eliminated, and the present invention was completed. That is, the present invention has the following configuration.
(1) It has a means for imaging the test liquid on the light-transmitting plate or in the flow cell, and is equipped with an automatic focusing function for automatically focusing on the formed component in the test liquid in the means. A device for classifying urinary particles characterized by the above.
(2) It is possible to convert the light intensity of the formed component with respect to the formed component in the urine on the translucent plate or in the flow cell into an electric signal and automatically adjust the focus using the change of the electric signal. (1) The urine sediment classification apparatus having an automatic focusing function.
(3) Automatic focusing capable of detecting a phase difference with respect to a urinary component on a light transmitting plate or in a flow cell, and automatically adjusting the focus using the phase difference analysis (1) The urinary sediment classification device having a function.
(4) Built-in auxiliary light emitting device that emits auxiliary light for assisting detection of the focus state, calculated by converting the distance from the auxiliary light reflected from the urine components on the light transmission plate or in the flow cell (1) The urine sediment classification apparatus having an automatic focusing function capable of automatically adjusting a focus based on a distance.

  As described below, the use of the urine particle classification device of the present invention simplifies the work related to the classification of the particle component in the test liquid, reduces the burden on the laboratory technician, and increases the accuracy. Highly accurate analysis results can be provided.

  In the present invention, the means for automatically focusing the formed component classification apparatus includes a means for detecting the focused state of the formed component in the test solution and a means for driving the lens or camera to the in-focus position. It is preferable.

  As a specific method of detecting the focus state, the light intensity of the subject is converted into an electrical signal, the focus is detected by using a change in the electrical signal, the phase difference is detected and the phase difference analysis is used. A method for detecting the focus state, and in order to assist the detection of the focus state of the subject, a built-in auxiliary light emitting device is used to measure the reflected light from the subject, thereby converting the distance and based on the calculated distance There is a way to detect the focus. Among the above methods, the method of converting the light intensity of the subject into an electric signal is a so-called contrast method in which the light intensity of the image pickup device for photographing an image that is the subject is regarded as contrast and detection is performed by detecting a high frequency component of the contrast. It is done.

  In addition, as a method using phase difference analysis, a pair or a plurality of pairs of image detection sensors such as a line sensor are provided in the vicinity of an image pickup surface on which an image pickup device such as a silver salt film or a photoelectric conversion device is arranged. There is a phase difference detection method in which light beams at different parts of the light that is arranged and transmitted through the photographing lens are guided to different sensors, and the focus of the photographing lens is detected from the shift of the position of the subject image on the paired sensors. Furthermore, as a method of converting the distance from the camera to the subject from the reflected light from the subject and detecting the focal point based on the calculated distance, a built-in light emitting device that emits auxiliary light such as infrared rays is reflected and reflected from the subject. An infrared method that calculates an absolute distance to the subject based on triangulation that converts the incident angle of the infrared ray into a distance can be given. When adjusting the focus, there are a continuous moving method and a stepwise moving method as a driving method of the lens or the camera. In the continuous movement, the lens or the camera is moved by a very small distance every predetermined period. The focus state is evaluated in each cycle, and the moving position showing the best focus state is set as the in-focus state. In stepwise movement, the lens or camera is moved so that a plurality of images are obtained from a plurality of predetermined focal positions. The moving position indicating the best focus state from the obtained image is set as the in-focus state.

  The urinary component classification device of the present invention is applied to analysis of body fluids such as urine, blood, serum, plasma, spinal fluid, semen, prostate fluid, joint fluid, pleural effusion, ascites, secretion, etc. can do. In particular, it is effectively used for analyzing urine such as raw urine, concentrated urine, sediment after centrifugation, and the like.

  In the present invention, the component to be analyzed is not particularly limited as long as it is dispersed or suspended in the test solution. For example, when the test solution is urine, red blood cells (deformed red blood cells, various red blood cells), white blood cells (dark cells, light cells, bright cells), epithelial cells (squamous cells, transitional cells, tubular epithelial cells) , Circular epithelial cells, urethral columnar epithelial cells, prostate epithelial cells, seminal vesicle epithelial cells, endometrial epithelial cells, oval fat pad, cytoplasmic inclusion body cells, polygonal cells, etc.), cylinders (glass cylinders, Epithelial column, granule column, waxy column, fat column, erythrocyte column, leukocyte column, cell column, hyaline leukocyte column, hemoglobin column, hemosilylidin column, myoglobin column, amyloid column, protein column, vacuolar deformed column, platelet column, bacteria Cylinder, bilirubin cylinder, salt cylinder, etc.), microorganisms (fungi, bacteria, protozoa, sperm, etc.), crystals / salts (uric acid salt, phosphate, calcium oxalate, bilirubin, cystine, kore Terol, 2,8-dihydroxy-adenine crystals, etc.), other (intranuclear inclusions cells, fat granule cells, macrophages, and as solid components of atypical cells), etc. is subject to analysis.

  The present invention is a process for analyzing a sample image of a formed object such as a formed still image included in a test liquid on a light transmitting plate or in a flow cell, and performing a plurality of types of classification and counting in the test liquid And a function (autofocus function) for automatically focusing on the formed component in the test liquid when the specimen image is captured. The imaging region for capturing the sample image is not particularly limited, but it is preferable that the imaging region can be moved so that the imaging position can be changed. Although the movement may be performed manually, it is preferably mechanically performed using, for example, a servo motor, a stepping motor, a linear motor, or the like.

  Examples of the imaging means include a digital camera and a CCD color video camera. The magnifying means may optically magnify a still image or specimen image before imaging, or may magnify the captured image by digital processing or the like. Specifically, a zoom lens, an objective lens, or the like attached to the camera can be used.

  For the focus adjustment of an electronic camera such as a digital camera or a CCD color video camera, a so-called contrast method is preferable, in which the light intensity of an image pickup device for taking an image is taken as contrast and high-frequency components of contrast are detected. Compared to the method of measuring the distance to the subject and moving the focus lens in the photographing lens system according to the distance, this contrast method does not require a distance measuring optical system separate from the photographing optical system. It is excellent in that the structure can be simplified.

  The formed component in the test solution may range from several μm in size such as blood cells and bacteria to several hundred μm in size such as a cylinder. Therefore, it is preferable that the enlargement unit has two or more enlargement magnifications rather than only one. In this case, it is possible to analyze more accurately from a small formed component to a large formed component. Further, the magnification may be continuously changed. What is necessary is just to determine an enlargement magnification suitably according to a formation part.

  The identifying means classifies and identifies formed components in the captured image based on the form and the like. It is preferable to add a function for calculating the analysis result and a function for outputting the analysis result from the output device to the identification means. Moreover, it is preferable that the identification means is provided with a memory for temporarily storing the captured image. Examples of the identification means include a computer programmed to perform the above-described identification, an identification device configured with a theoretical circuit, and the like. Of these, the use of a computer as the identification means is preferable because all the operations such as the operation of each process, image processing, storage, calculation, and output can be performed on software.

  The identification means preferably has a learning recognition function. By having a learning recognition function, a measurement result with high accuracy and precision is provided. The identification means includes (1) color extraction that separates red, green, and blue into lightness and chromaticity, (2) binary image processing that includes hole filling, line segment writing, and image separation, and (3) image feature amount. Learn range specification (area, circularity coefficient, equivalent circle diameter, circumference length, absolute maximum length, ferret diameter X / Y ratio, maximum chord length X / Y ratio, minor axis length / major axis length ratio, etc.) And can be identified.

  Further, the apparatus of the present invention is preferably provided with means (image storage device) for storing the result of processing the captured image and identifying the various components. Examples of the means for storing include auxiliary storage devices such as a magneto-optical disk, a fixed disk, a digital video disk, and a CD-ROM having a large storage capacity. When a display is used as the output device, depending on the menu selection, in addition to measurement results and image data, time, current device status (measurement status of each sample, measurement end scheduled time of each sample or selected sample or A function that displays the required remaining time, the amount of waste liquid in the waste liquid tank, the remaining amount of pure water tank, the remaining amount of each reagent, the remaining amount of detergent, etc., and the enlarged display so that only selected information can be read from a remote location You may add the function to do.

  Hereinafter, embodiments of the apparatus of the present invention will be described more specifically based on the drawings. In addition, this invention is not limited to this Example.

Example 1
FIG. 1 is a diagram showing an example of the urine particle analyzer according to the present invention, and shows the urine particle analyzer. In FIG. 1, the test solution is passed through the flow cell 1 to detect the formed component. The imaging area of the flow cell 1 has a wall surface formed of a translucent plate and is disposed on the front surface of the CCD camera 5.

  An objective lens 4 is attached as an enlargement means to a CCD camera 5 as an imaging means. The objective lens 4 is movable in the light source optical axis direction, and its movement is adjusted by the drive control circuit 7 based on the judgment of the arithmetic circuit 6. The identification means includes an image processing / storage circuit 8 and a control calculation circuit 9. The output device 10 is a display and a printer showing the results of various identifications. Next, processing performed in the apparatus of the present invention will be described in time series.

  First, a required amount of a urine sample as a test solution is dispensed from a sample container into a reaction container using a probe. In order to dispense more uniformly, the urine sample is agitated before dispensing. If a container with a barcode attached in advance is used as a sample, the barcode is read in advance with a barcode reader. If the presence / absence, type, and specimen number of a sample are read by using a barcode, the subsequent processing data and the sample can be easily identified. Next, a required amount of the staining solution is injected into the reaction vessel, and the reaction solution of the test solution and the staining solution is collected from the reaction vessel and injected into the flow cell 1. The dyeing reaction time and reaction temperature are arbitrarily set.

  Next, the test liquid continuously flows in the flow cell and reaches the imaging region in the flow cell. In order to image the test solution, the light emitted from the light source 2 travels on the optical axis, passes through the condenser lens 3, and is collected on the test solution in the flow cell 1. A formed sample image in the test solution is formed at the imaging position by the objective lens 4, and the sample image at this imaging position is projected as an image on the imaging surface of the CCD camera 5 and subjected to photoelectric conversion. The drive control circuit 7 instructs the drive unit to move the objective lens 4 left and right by a very small distance every predetermined period until the in-focus state is detected by the arithmetic circuit 6. Then, the variance value of the brightness level obtained by the arithmetic circuit 6 in each cycle is determined as an evaluation value, and when the evaluation value becomes the minimum value, the movement position of the objective lens 4 when the evaluation value becomes the minimum value is detected. It is determined that the in-focus state is reached, and the automatic focus adjustment is completed by instructing the drive control circuit 7 to set the objective lens 4 to the in-focus position. An image obtained by the CCD camera 5 is digitized by an A / D converter in the image processing / storage circuit 8. The image processing / storage circuit 8 stores the digitized image data.

  Next, the image processing / storage circuit 8 inputs the stored image data to the control calculation circuit 9. The control calculation circuit 9 calculates the image feature amount (for example, area, circularity coefficient, equivalent circle diameter, circumference length, absolute maximum length, ferret X / Y ratio, maximum chord length X / Y ratio, short axis length / long axis). Length ratio etc.) are extracted as primary parameters. The image processing / storage circuit 8 inputs these primary parameters and secondary parameters generated by the combination operation thereof to the control calculation circuit 9.

  The control calculation circuit 9 classifies the formed component using a neural network. The neural network performs learning using a large amount of data based on expert judgment in advance, and optimizes the coupling coefficient between the neurons. Therefore, the control calculation circuit 9 causes the control calculation circuit 9 to automatically classify the formed component using the statistical learning recognition method instead of the neural network operation using the input primary and secondary parameters. Also good.

  Further, the control calculation circuit 9 stores the classification result and the image data in the output device 10. In this embodiment, a magneto-optical disk having a large storage capacity is used as the output device.

Example 2
[Urine Specimen Collection Step] After stirring 200 urine stock solutions without centrifuging, 0.75 ml of each was dispensed into a predetermined reaction tube.
[Staining step] 0.25 ml of S (Sternheimer) staining solution is added to the reaction tube into which the urine specimen has been dispensed, and stirred.
[Imaging step] The urine specimen mixed with the staining solution flows in the flow cell and reaches the imaging region. A CCD camera set in front of the imaging region captures 100 fields of view (strongly expanded field of view: HPF) of magnified images corresponding to a 10 × internal lens and a 40 × objective lens.
[Image processing / storage step] The captured image is stored in a magneto-optical disk or a fixed disk. The captured image is classified and counted into various components using the learning recognition function, and the result is transmitted to the control calculation step. Further, after storing the image, the test solution moves and proceeds again to the focus process, the imaging process, and the storage process, and this repetition is performed until 100 fields of view are reached.
[Control calculation process / output process] The result obtained from the image processing / storage process is output to a display or printer. Table 1 shows an example of the output (printing), and the result in the high magnification field of view (HPF) is shown by symbols.

  Table 2 shows the coincidence ratio of the inspection results obtained by the image processing in this embodiment. That is, each captured image was visually observed to determine whether or not it matches the components (red blood cells, white blood cells, etc.) classified and analyzed by image processing. Further, as a control comparison, the same measurement was performed using a single-point focusing device without a focus adjustment function. The results are summarized in Table 2.

Example 3
FIG. 2 is a diagram showing another example of the urine particle analyzer of the present invention. A required amount is dispensed into the reaction tube 11 by a probe from a sample container containing a urine specimen. At this time, in order to collect the sample more uniformly, sampling may be performed after the stirring operation. When using a sample container with a barcode attached in advance, the barcode is read in advance by a barcode reading device. By using the bar code, the presence / absence, type, and specimen number of the sample can be read, and can be used for subsequent processing data and sample identification. Next, a required amount of the staining solution is dispensed into the reaction tube 11, and the urine formed portion is stained. About 2 minutes later, a necessary amount is taken from the reaction tube 11 and dispensed onto the slide glass 12. The dyeing reaction time and reaction temperature may be set arbitrarily.

  In FIG. 2, the slide glass 12 is placed on an XY table 13 as an imaging stage, and the XY table 13 can be driven so that the slide glass 12 can freely move and stop in the X coordinate axis direction and the Y coordinate direction. . The light emitted from the lamp 14 as the light source travels on the optical axis, passes through the condenser lens 15 and is condensed on the test solution on the slide glass 12. The formed image in the test solution is enlarged by the objective lens 16 and adjusted to the image forming position by the automatic focusing mechanism. The image at this imaging position is projected as an image on the imaging surface of the CCD camera 17 and subjected to photoelectric conversion. The lens driving circuit 19 instructs the lens driving unit to move the objective lens 16 up and down by a very small distance every predetermined period until the in-focus state is detected by the focus calculation circuit 18. Then, the variance value of the brightness level obtained by the focus calculation circuit 18 in each cycle is determined as an evaluation value, and when the evaluation value becomes the minimum value, the movement position of the objective lens 16 when the minimum value is reached is detected. Is in the in-focus state, and the lens drive circuit 19 is commanded to set the objective lens 16 to the in-focus position, thereby completing the automatic in-focus adjustment. The table drive circuit 21 for driving the XY table 13 is controlled by the image processing control circuit 20, and the image processing control circuit 20 controls the movement and stop of the XY table 13, and when the XY table 13 stops. The CCD camera 17 is controlled to take an image. An image obtained by the CCD camera 17 is digitized by an A / D converter in the image storage circuit 22 and stored in the image storage circuit 22. After the storage in the image storage circuit 22 is completed, the XY table 13 is moved to change the field of view, and the image is captured, A / D converted, and stored in the same process until the number of fields set in advance is reached. This operation is repeated. The movement of the XY table 13 for changing the visual field may be controlled to be performed after the image analysis for one visual field is completed. The writing to the image storage circuit 22 and the subsequent image processing control are performed by the image processing control circuit 20 to control the input of the image data stored in the image storage circuit 22 to the identification calculation circuit 23. The discriminating operation circuit 23 calculates the image feature amount (formation area, circularity coefficient, equivalent circle diameter, circumference length, absolute maximum length, ferret diameter X / Y ratio, maximum chord length X / Y ratio, short axis length. (Such as length / major axis length ratio) is extracted as a primary parameter. The image processing control circuit 20 inputs these primary parameters and secondary parameters generated by the combination calculation to the identification calculation circuit 23. The classification operation circuit 23 classifies the formed component using a neural network, but the neural network performs learning using a large amount of data in advance based on expert judgment, and the coupling coefficient between each neuron is optimal. Therefore, it is possible to perform a neural network calculation using the input feature parameter and to perform automatic classification of the target formed component. Moreover, you may perform the automatic classification | category of a formation using a statistical learning recognition method, without using a neural network. Further, the identification calculation circuit 23 stores the classification result and the image data in the output device 24. In this embodiment, a magneto-optical disk having a large storage capacity is used as the output device.

Example 4
[Urine Specimen Collection Step] After stirring 200 urine stock solutions without centrifuging, 0.75 ml of each was dispensed into a predetermined reaction tube.
[Staining Step] 0.25 ml of S (Sternheimer) staining solution is added to a reaction tube into which a urine sample has been dispensed, and stirred.
[Specimen Preparation Step] 0.015 ml is taken from the solution to which the staining solution has been added, and dispensed into the space between the slide glasses to prepare a sample.
[Sample moving step] The prepared sample is set on the XY stage. The specimen moves to the imaging position by the XY stage.
[Imaging Step] With a CCD camera set on the imaging stage, 100 magnified images corresponding to a 10-magnification internal lens and a 400-magnification objective lens are imaged (strongly magnified field of view: HPF).
[Image processing / storage process] The captured image is stored in the magneto-optical disk. The captured image is classified and counted into various components using the learning recognition function, and the result is transmitted to the control calculation step. Further, after storing the image, the XY stage moves to the next visual field, and again proceeds to the focus process, the imaging process, and the storage process, and this repetition is performed until 100 visual fields are reached.
[Control calculation step / output step] The result obtained from the image processing / storage step and the result obtained from the microorganism amount analysis step are integrated and output to a display or printer. Table 2 summarizes the coincidence rate of the inspection results by the image processing in this example. That is, each captured image was visually observed to determine whether or not it coincided with the white blood cell fine classification classified and analyzed by image processing. Moreover, the same measurement was performed using a one-point focusing device without a focus adjustment function as an object comparison. The results are summarized in Table 3.

The present invention relates to an apparatus for classifying tangible components in urine using computer image processing and recognition technology, and is applied to the field of clinical examination.
As described above, by using the urinary component classification device of the present invention, the work related to the component classification in the test solution is simplified, reducing the burden on the laboratory technician, and more precise, An analysis result with high accuracy can be provided.

It is a figure which shows an example of the urine substance analyzer of this invention. It is a figure which shows an example of the urine substance analyzer of this invention.

Explanation of symbols

1 Flow Cell 2 Light Source 3 Condenser Lens 4 Objective Lens 5 CCD Camera 6 Arithmetic Circuit 7 Drive Control Circuit 8 Image Processing / Storage Circuit 9 Control Calculation Circuit 10 Output Device 11 Reaction Tube 12 Glass Slide 13 XY Table 14 Light Source 15 Condenser Lens 16 Objective Lens 17 CCD camera 18 focus calculation circuit 19 lens drive circuit 20 image processing control circuit 21 table drive circuit 22 drawing storage circuit 23 identification calculation circuit 24 output device

Claims (1)

Urinary formation characterized in that it has means for imaging the test liquid in the flow cell and is equipped with an automatic focusing function that automatically focuses on the formation in the test liquid in the means A classification device,
In the urine having an automatic focusing function capable of converting the light intensity of the formed component into an electric signal with respect to the formed component in the urine in the flow cell and automatically adjusting the focus using the change of the electric signal In the component classification device,
Performs phase difference detection for detecting a focus state for urinary particles in the flow cell, the focus have a autofocus function capable of automatically adjusting the utilizing phase difference analysis,
Furthermore, the automatic focusing function includes means for completing the automatic focusing adjustment through the following steps (1) and (2):
(1) Step of moving the objective lens left and right by a very small distance every predetermined period until the in-focus state is detected
(2) The variance value of the luminance level obtained at each predetermined period is determined as an evaluation value, and when the evaluation value becomes the minimum value, the movement position of the objective lens when the minimum value is reached is in focus. The process of determining that there is an objective and setting the objective lens to the in-focus position

Images