Sample analyzer and sample detection method

Abstract

The present application discloses a sample analyzer and a sample detection method, wherein the sample analyzer includes a detection module and a processing module, wherein the detection module is used to detect a sample to be detected and obtain first detection data and second detection data, wherein the first detection data and the second detection data correspond to different light receiving channels; the processing module is connected to the detection module and is used to obtain a two-dimensional image based on the first detection data and the second detection data, and to perform image segmentation processing on the two-dimensional image to determine the types of all particles in the two-dimensional image, calculate the fusion ratio between different types of particles, and compare the fusion ratio with a preset threshold; in response to the fusion ratio being greater than the preset threshold, it is determined that the sample to be detected corresponding to the two-dimensional image is an abnormal sample. By first acquiring a two-dimensional image and then segmenting and identifying anomalies in the two-dimensional image, the abnormal two-dimensional image can be excluded in advance to optimize the overall process, reduce resource consumption, and improve the accuracy of abnormal identification.

Description

Technical Field

[0001] The present application relates to the technical field of medical devices, in particular to a sample analyzer and a sample detection method. Background Art

[0002] In medical or experimental scenarios, it is often necessary to test and analyze various samples. For example, a blood cell analyzer is used to count and classify red blood cells, platelets, and white blood cells in the blood. Generally, the instrument pre-stores the cell value range when the organism is in a normal state. When the cell classification count result is not within the value range when the organism is in a normal state, an abnormal alarm can be issued.

[0003] In order to improve the judgment dimension of abnormal identification, the current cell analyzer introduces a scatter plot abnormal identification algorithm, that is, after the cell analyzer outputs the classification counting results, the scatter plot containing the classification results is image recognized by the scatter plot abnormal identification algorithm, and an abnormal alarm is issued according to the recognition results. However, this abnormal identification method has the problem of low recognition accuracy. Summary of the invention

[0004] The present application provides a sample analyzer and a sample detection method. By first acquiring a two-dimensional image and then segmenting and identifying abnormalities in the two-dimensional image, abnormal two-dimensional images can be excluded in advance to optimize the overall process and reduce resource consumption.

[0005] In order to solve the above technical problems, a technical solution adopted by the present application is: to provide a sample analyzer, which includes a detection module and a processing module. Among them, the detection module is used to detect the sample to be detected, and obtain the first detection data and the second detection data, and the first detection data and the second detection data correspond to different light receiving channels; the processing module is connected to the detection module, and is used to: obtain a two-dimensional image based on the first detection data and the second detection data; perform image segmentation processing on the two-dimensional image to determine the types of all particles in the two-dimensional image; calculate the fusion ratio between different types of particles, and compare the fusion ratio with a preset threshold; in response to the fusion ratio being greater than the preset threshold, determine that the sample to be detected corresponding to the two-dimensional image is an abnormal sample.

[0006] Among them, the processing module is also used to: perform semantic segmentation processing on the two-dimensional image to determine the first type of all particles in the two-dimensional image; determine an edge area image in which there are at least two different particle types based on the first type of all particles; and correct the type of particles in the edge area image to change the particle type in the edge area image from the first type to the second type to obtain the type of all particles in the two-dimensional image.

[0007] Among them, the processing module is also used to: split the two-dimensional image into several sub-images; determine the edge area image based on the several sub-images; wherein the edge area image contains at least two different types of particles; perform regional growth processing on the particles contained in the edge area image to determine the second type of particles contained in the edge area image; use the second type to correct the types of corresponding part of the particles in the first type to obtain the types of all particles in the two-dimensional image.

[0008] Among them, the processing module is also used for: after performing regional growth processing on the particles contained in the edge area image, if there are ungrown particles in the edge area image, calculating the center of gravity distance between the ungrown particles and each type of particles, and confirming the type with the smallest center of gravity distance as the second type of ungrown particles.

[0009] Among them, the processing module is also used to: determine any two different types of particles as a first particle and a second particle; define one of the first particle and the second particle as a particle of an unknown type, and define the other of the first particle and the second particle as a particle of a known type; perform regional growth processing on particles of known types to obtain a first growing particle corresponding to the first particle and a second growing particle corresponding to the second particle; determine a fusion ratio based on the first growing particle and the second growing particle; in response to the fusion ratio being not within a preset range, determine that the sample to be detected corresponding to the two-dimensional image is an abnormal sample.

[0010] The processing module is used to obtain a first total number by summing the first particle and the second particle; obtain a fusion number between the first particle and the second particle after merging the first growing particle and the second growing particle; and obtain a fusion ratio according to the ratio of the fusion number to the first total number.

[0011] Among them, the processing module is also used for: in response to the fusion ratio being less than or equal to a preset threshold, determining that the sample to be detected corresponding to the two-dimensional image is a normal sample; in response to the sample to be detected corresponding to the two-dimensional image being a normal sample, calculating the proportion of each type of particles in all particles, and comparing the proportion with a preset range; in response to the proportion not being within the preset range, determining that the sample to be detected corresponding to the two-dimensional image is an abnormal sample.

[0012] Among them, before performing image segmentation processing on the two-dimensional image to determine the types of all particles in the two-dimensional image, the processing module is also used to: perform binary classification judgment on the two-dimensional image based on a binary classification neural network; when the binary classification judgment result is abnormal, determine that the sample to be detected corresponding to the two-dimensional image is an abnormal sample; when the binary classification judgment result is normal, perform image segmentation processing on the two-dimensional image to determine the types of all particles in the two-dimensional image.

[0013] The sample analyzer further includes: an interactive module connected to the processing module, for displaying abnormal indication information of the sample to be detected in response to the sample to be detected corresponding to the two-dimensional image being an abnormal sample; wherein the abnormal indication information includes at least one of the particle type distribution after the two-dimensional image is processed by image segmentation, the particle type with a fusion ratio greater than a preset threshold, and the fusion ratio corresponding to the particle type with a fusion ratio greater than the preset threshold.

[0014] In order to solve the above technical problems, another technical solution adopted in the present application is: providing a sample detection method, which includes: detecting a sample to be detected to obtain first detection data and second detection data, the first detection data and the second detection data corresponding to different light receiving channels; obtaining a two-dimensional image based on the first detection data and the second detection data; performing image segmentation processing on the two-dimensional image to determine the types of all particles in the two-dimensional image; calculating the fusion ratio between different types of particles, and comparing the fusion ratio with a preset threshold; in response to the fusion ratio being greater than the preset threshold, determining that the sample to be detected corresponding to the corresponding two-dimensional image is an abnormal sample.

[0015] The beneficial effects of the present application are as follows: Different from the prior art, the sample analyzer provided by the present application includes a detection module and a processing module coupled to each other, wherein the detection module is used to detect the sample to be detected, obtain the first detection data and the second detection data, the first detection data and the second detection data correspond to different light receiving channels, and the processing module is used to obtain a two-dimensional image based on the first detection data and the second detection data; perform image segmentation processing on the two-dimensional image to determine the types of all particles in the two-dimensional image; calculate the fusion ratio between different types of particles, and compare the fusion ratio with a preset threshold; in response to the fusion ratio being greater than the preset threshold, determine that the sample to be detected corresponding to the two-dimensional image is an abnormal sample. The above-mentioned method of first acquiring a two-dimensional image, and then segmenting and identifying anomalies on the two-dimensional image can not only exclude the abnormal two-dimensional image in advance, improve the data reliability of the data used in the abnormal identification process, but also optimize the overall process, reduce resource consumption, and perform abnormal identification by calculating the fusion ratio, which can avoid the problem of inaccurate segmentation in particle segmentation, thereby improving the accuracy of overall abnormal identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0017] Figure 1is a structural schematic diagram of a first embodiment of a sample analyzer provided by the present application;

[0018] Figure 2 is a schematic diagram of an embodiment of a 2-classification neural network provided by the present application;

[0019] Figure 3 is a schematic diagram of an embodiment of a semantic segmentation network provided by the present application;

[0020] Figure 4 is a schematic diagram of an embodiment of a semantic segmentation process provided by the present application;

[0021] Figure 5 is a structural schematic diagram of a second embodiment of a sample analyzer provided by the present application;

[0022] Figure 6 It is a flow chart of the first embodiment of the sample detection method provided by the present application;

[0023] Figure 7 is a flowchart of an embodiment of step 63 provided by the present application;

[0024] Figure 8 It is a flow chart of the second embodiment of the sample detection method provided by the present application;

[0025] Fig. 9 It is a structural schematic diagram of an embodiment of a computer-readable storage medium provided by the present application. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0027] See also Figure 1 , Figure 1It is a structural schematic diagram of the first embodiment of the sample analyzer provided in the present application, and the sample analyzer 10 includes a detection module 101 and a processing module 102 coupled to each other. The detection module 101 is used to detect the sample to be detected and obtain the first detection data and the second detection data, and the first detection data and the second detection data correspond to different light receiving channels; the processing module 102 is used to: obtain a two-dimensional image based on the first detection data and the second detection data; perform image segmentation processing on the two-dimensional image to determine the types of all particles in the two-dimensional image; calculate the fusion ratio between different types of particles, and compare the fusion ratio with a preset threshold; in response to the fusion ratio being greater than the preset threshold, determine that the sample to be detected corresponding to the two-dimensional image is an abnormal sample. Among them, the size of the preset threshold can be determined according to the actual situation, such as 0.5.

[0028] It is understandable that the samples to be tested may be different for different testing items. The samples to be tested may be blood samples, saliva samples, urine samples or others.

[0029] When the sample analyzer 10 detects the samples to be detected, it is necessary to add corresponding reagents to each sample to be detected, and place the sample with the added reagents in the detection area. When the detection area is irradiated with a light source (such as a laser) in the detection area, the light source irradiates each cell in the sample to be detected to generate two-dimensional signal data and / or three-dimensional signal data.

[0030] In some embodiments, the detection module 101 may directly detect the sample to be detected to obtain two-dimensional signal data, wherein the two-dimensional signal data includes first detection data and second detection data.

[0031] In one application scenario, the two-dimensional signal data includes side scattered light (SSC) signal data and side fluorescence (SFL) signal data. Specifically, the first detection data is the peak height value of the side scattered light pulse signal, and the second detection data is the peak height value of the side fluorescence pulse signal. In other embodiments, the two-dimensional signal data may be forward scattered light signal data and side scattered light signal data, specifically, the first detection data is the peak height value of the forward scattered light pulse signal, and the second detection data is the peak height value of the side scattered light pulse signal.

[0032] Among them, the side scattered light signal data / side scattered light pulse signal is used to characterize the information of the internal particles and cell nucleus of the cell; the side fluorescence signal data / side fluorescence pulse signal is used to characterize the content of deoxyribonucleic acid (DNA) and ribonucleic acid (RNA) in the cell; the forward scattered light signal data / forward scattered light pulse signal is used to characterize the size of the cell.

[0033] In some embodiments, the detection module 101 is also used to detect the sample to be detected, obtain three-dimensional signal data, and then obtain any two signal data from the three-dimensional signal data as two-dimensional signal data, namely, first detection data and second detection data. In other words, the two-dimensional signal data can also avoid the detection of the sample to be detected and be directly obtained from the three-dimensional signal data, that is, the three-dimensional signal data can contain the two-dimensional signal data.

[0034] In one application scenario, the three-dimensional signal data includes side scattered light signal data, forward scattered light signal data and side fluorescence (SFL) signal data, and the two-dimensional signal data is directly obtained from the three-dimensional signal data, and the two-dimensional signal data includes side scattered light signal data and side fluorescence signal data.

[0035] Specifically, the detection module 101 detects the samples to be detected to obtain the peak height value of the side scattered light pulse signal, the peak height value of the side fluorescence pulse signal and the peak height value of the forward scattered light pulse signal of the particles in each sample to be detected, so as to obtain two-dimensional signal data based on these three signal data, and the two-dimensional signal data is specifically the peak height value of the side scattered light pulse signal and the peak height value of the side fluorescence pulse signal, so as to obtain a two-dimensional image based on the peak height value of the side scattered light pulse signal and the peak height value of the side fluorescence pulse signal.

[0036] In other application scenarios, the two-dimensional signal data includes side scattered light signal data, forward scattered light signal data, or side scattered light signal data and side fluorescence signal data. In other words, the two-dimensional signal data can be any two of the three-dimensional signal data.

[0037] And, in other embodiments, the three-dimensional signal data can be a low-angle scattered light signal, a medium-angle scattered light signal, and a high-angle scattered light signal, wherein the low-angle scattered light signal is used to characterize the volume size of particles (cells); the medium-angle scattered light signal is used to characterize the complexity information inside the particles; and the high-angle scattered light signal is used to characterize the granularity information inside the particles.

[0038] Taking the blood sample as an example, when the blood sample is subjected to white blood cell detection, the particle types involved in the processing by the processing module 102 in the sample analyzer 10 include blood ghosts, lymphocytes, monocytes, eosinophils, basophils, and neutrophils, etc. For another example, when the blood sample is subjected to reticulocyte detection, the particle types involved in the processing by the processing module 102 in the sample analyzer 10 include red blood cells, reticulocytes, platelets, white blood cells, etc. Based on the above content, after acquiring the two-dimensional signal data and the three-dimensional signal data, a corresponding scatter image can be established, and the scatter points in the scatter image correspond one-to-one to the particles in the sample to be detected, and the position of each scatter point in the scatter image is determined by the multi-dimensional signal data of the corresponding particle.

[0039] For example, a two-dimensional scatter image is established based on the two-dimensional signal data: the two-dimensional signal data is the side scattered light signal data and the side fluorescence signal data, and a coordinate system is established with the side scattered light signal data as the X-axis and the side fluorescence signal data as the Y-axis. The two-dimensional signal data of the particles detected from the sample to be detected is mapped to the coordinate system, and a two-dimensional scatter image can be established.

[0040] For another example, a three-dimensional scatter image is established based on three-dimensional signal data: the three-dimensional signal data is side scattered light signal data, forward scattered light signal data and side fluorescence signal data, a coordinate system is established with the side scattered light signal data as the X-axis, the forward scattered light signal data as the Y-axis and the side fluorescence signal data as the Z-axis, and the three-dimensional signal data of the particles detected in the sample to be detected is mapped to the coordinate system, so that a three-dimensional scatter image can be established.

[0041] It is worth noting that the two-dimensional image processed by the processing module 102 of the present application is a two-dimensional scattered point image, and the two-dimensional image is a grayscale image, not an RGB (red-green-blue, three primary colors) image, and will not be affected by color when performing subsequent image segmentation and other processing on the two-dimensional image. In addition, the two-dimensional image does not carry the classification results of the particle type, and then in the subsequent process, the processing module 102 in the sample analyzer 10 performs image segmentation processing on the two-dimensional image, which can reduce the resources consumed by the image segmentation processing.

[0042] Different from the prior art, the sample analyzer 10 provided in the present application includes a detection module 101 and a processing module 102 coupled to each other, wherein the detection module 101 is used to detect the sample to be detected, obtain the first detection data and the second detection data, and the first detection data and the second detection data correspond to different light receiving channels; the processing module 102 is used to obtain a two-dimensional image based on the first detection data and the second detection data, perform image segmentation processing on the two-dimensional image to determine the types of all particles in the two-dimensional image, calculate the fusion ratio between different types of particles, compare the fusion ratio with a preset threshold, and in response to the fusion ratio being greater than the preset threshold, determine that the sample to be detected corresponding to the two-dimensional image is an abnormal sample. The above-mentioned method of first acquiring a two-dimensional image, then segmenting and identifying anomalies on the two-dimensional image can not only exclude abnormal two-dimensional images in advance, improve the data reliability of the data used in the abnormal identification process, but also optimize the overall process, reduce resource consumption, and perform abnormal identification by calculating the fusion ratio, which can avoid the problem of inaccurate segmentation in particle segmentation, thereby improving the accuracy of overall abnormal identification.

[0043] It is worth noting that after the processing module 102 acquires the two-dimensional image and before performing image segmentation processing on the two-dimensional image, the processing module 102 needs to perform a binary classification anomaly analysis on the acquired two-dimensional image so as to eliminate abnormal two-dimensional images before the image segmentation processing step to reduce resource waste.

[0044] Based on this, binary classification anomaly analysis can include the following process: perform binary classification judgment on the two-dimensional image based on the binary classification neural network; when the binary classification judgment result is abnormal, determine that the sample to be detected corresponding to the two-dimensional image is an abnormal sample; when the binary classification judgment result is normal, perform image segmentation processing on the two-dimensional image to determine the types of all particles in the two-dimensional image.

[0045] Specifically, a trained binary classification neural network is used to perform binary classification anomaly analysis on a two-dimensional image. The binary classification neural network can be a basic network of ResNet18 and MobileNet. Resnet18 is a residual-based network, and MobileNet is a neural network suitable for mobile devices. Among them, the binary classification neural network is obtained by training the position information, aggregation degree, quantity information, etc. of different types of scattered points in the two-dimensional image. When using the binary classification neural network to perform binary classification anomaly analysis on a two-dimensional image, it is necessary to compare the algorithm result output by the binary classification neural network with the set threshold to determine whether the sample to be detected corresponding to the two-dimensional image is a normal sample or an abnormal sample. When the algorithm result is less than the set threshold, a feedback signal can be output to indicate that the binary classification judgment result is abnormal and the sample to be detected is an abnormal sample; when the algorithm result is greater than the set threshold, another feedback signal can be output to indicate that the binary classification judgment result is normal and the sample to be detected is a normal sample. Among them, the set threshold is usually 0.5, or it can be other values, which are not limited here.

[0046] In the embodiments of the present application, the binary classification neural network used can be as follows: Figure 2 As shown in the figure, the binary classification neural network includes 3 convolutional layers, 3 pooling layers, 1 flatten layer and 1 fully connected layer, and then the classification result is obtained by activation with the sigmoid (activation) function. The classification result will show whether the sample to be detected corresponding to the corresponding two-dimensional image is an abnormal sample. If the output "1" indicates that the binary classification result is abnormal and the sample to be detected is an abnormal sample, the output "0" indicates that the binary classification result is normal and the sample to be detected is a normal sample.

[0047] In some embodiments, the processing module 102 is used to perform image segmentation processing on the two-dimensional image to determine the types of all particles in the two-dimensional image, specifically including: the processing module 102 is used to perform semantic segmentation processing on the two-dimensional image to determine the first type of all particles in the two-dimensional image; determine an edge area where there are at least two different particle types based on the first types of all particles; and correct the type of particles in the edge area image to change the particle type in the edge area image from the first type to the second type to obtain the type of all particles in the two-dimensional image.

[0048] Specifically, the processing module 102 is used to perform semantic segmentation processing on the two-dimensional image to determine the first type of all particles in the two-dimensional image, including: the processing module 102 uses the semantic segmentation network to identify the types of all particles in the two-dimensional image based on the pixels of each particle in the two-dimensional image. It can be understood that when the target in the two-dimensional image is an irregular image, the image formed by the types of particles contained in the segmentation map output by the semantic segmentation network is also irregular, that is, each particle in the segmentation image output by the semantic segmentation network has a corresponding category.

[0049] The segmentation algorithm used by the semantic segmentation network may include but is not limited to a 2D segmentation algorithm (Unet, U2net) and a 3D segmentation algorithm (Pointnet, 3DUnet). Figure 3 As shown in Figure 1, the semantic segmentation network includes 2 convolutional layers, 2 deconvolutional layers, 2 pooling layers and 2 upsampling layers.

[0050] It can be understood that the image input to the semantic segmentation network is a grayscale image, and the output image of the semantic segmentation network is an RGB image. Figure 4 As shown in the figure, the grayscale image ① is input into the semantic segmentation network to obtain the RGB image ②, in which different types of particles are marked with different colors in the RGB (red-green-blue, three primary colors) image ②. The type of particles can be determined by observing the color of the particles, such as blue represents basophils (Baso) and purple represents lymphocytes (Lyn).

[0051] Specifically, a region growing process is performed on the two-dimensional image after the semantic segmentation process. Before performing the region growing process, an edge region image needs to be determined. The process of acquiring the edge region image includes: performing image segmentation process on the two-dimensional image to split it into a number of sub-images; based on the several sub-images, determining the edge region image; wherein the edge region image contains at least two different types of particles.

[0052] It can be understood that since there are at least two different types of particles in the edge region, it is easy to have inaccurate identification of particles in the edge region, such as identifying particle A as particle B. In other words, the step of correcting the particle type in the two-dimensional image is only for the edge region, not for all particles. Targeted correction can reduce the time required for the overall process and avoid the phenomenon of incorrect particle type classification.

[0053] Based on this, after acquiring the edge area image, the type of particles in the edge area image can be corrected to change the type of particles in the edge area image from the first type to the second type, so as to obtain the types of all particles in the two-dimensional image. Specifically, the particles contained in the edge area image are subjected to regional growth processing to determine the second type of particles contained in the edge area image; the types of the corresponding part of the particles in the first type are corrected using the second type to obtain the types of all particles in the two-dimensional image. For example, if the first type of particles is eosinophils and the second type is basophils, the second type can be used as the final particle type, and the first type can be modified, that is, the type of the target particles is modified from eosinophils to basophils; for another example, if the first type of particles is abnormal lymphocytes and the second type is atypical lymphocytes, the second type can be used as the final particle type, and the first type can be modified, that is, the type of the target particles is modified from abnormal lymphocytes to atypical lymphocytes.

[0054] In an application scenario, a two-dimensional image is segmented to obtain several sub-images of the same size; it is determined whether there are different types of pixels in each sub-image. If so, it means that the sub-image is an edge region image, and the type of particles in the edge region image is set to an unknown type; a target sub-image with only one type of particles (i.e., only one type of pixels) is used as a seed region, and the seed region is adjacent to the edge region image. At this time, based on the continuous growth of the seed region to the surroundings, the particles around are added to the target sub-image according to a certain rule to combine into a new sub-image (or stack). The types of all particles contained in the new sub-image are consistent, and the actual type of the unknown type of particles in the edge region image can be determined. For particles that have not grown in the edge region, the distance of the center of gravity of the new sub-image is used to determine the type of particles in the sub-image, that is, the distance between the center of gravity of the ungrown particles and each type of particles is calculated, and the type with the smallest center of gravity distance is confirmed as the second type of the ungrown particles. Among them, the center of gravity refers to a point among all particles of the same type, and the sum of the distances from this point to all other particles is the smallest. At this time, the distance from the center of gravity of the unknown type of particles to each known type of particles is the smallest.

[0055] Through the above method, the second type of particles in the two-dimensional image can be determined again. At this time, the second type can be compared with the first type to reset / correct the type of particles that distinguish the first type from the second type, and the type of the corresponding particles can be changed from the first type to the second type, thereby determining the final type of all particles on the two-dimensional image.

[0056] Based on the above content, the semantic segmentation network is first used to segment the two-dimensional image. Since the semantic segmentation network is divided according to the category of pixels, the types contained in the segmentation map output by the semantic segmentation network are arranged by region, that is, the output segmentation map is divided into several regions of equal size, each region displays one type, but not all regions contain only one type of pixel, that is, not all regions contain only one type. Based on this, when there are multiple pixels in the region, there will be controversy in the way of determining the particle type based on pixels, that is, when there are multiple different types of particles in the region, there may be controversy in determining the particle type in the region. Therefore, after the semantic segmentation process is performed, the present application can use the regional growing algorithm to correct / reset the type of particles in the controversial region obtained after the semantic segmentation. In addition, because the regional growing algorithm is also an image segmentation algorithm based on pixel division, through two image segmentation processes, the accuracy of particle type determination is guaranteed, and the failure rate of subsequent operations of the sample analyzer 10 is reduced.

[0057] In some embodiments, see Figure 1 , the processing module 102 of the present application can also be used to: determine any two different types of particles as a first particle and a second particle; define one of the first particle and the second particle as a particle of an unknown type, and define the other of the first particle and the second particle as a particle of a known type; perform regional growth processing on the particles of known types to obtain a first growth particle corresponding to the first particle and a second growth particle corresponding to the second particle; determine a fusion ratio based on the first growth particle and the second growth particle; in response to the fusion ratio not being within a preset range, determine that the sample to be detected corresponding to the two-dimensional image is an abnormal sample. Among them, the step of determining the fusion ratio may include: summing the first particle and the second particle to obtain a first total number; merging the first growth particle and the second growth particle to obtain the fusion number between the first particle and the second particle; and obtaining the fusion ratio according to the ratio of the fusion number to the first total number.

[0058] Taking the inter-particle fusion ratio of lymphatic mononuclear particles as an example: the particle types corresponding to the particles other than lymphatic mononuclear particles are set to unknown types; the particle types corresponding to mononuclear type particles are set to unknown types, lymphatic particles are used as seeds, and the particles grow in all directions, and different colors are used to mark the particles in the new growth area and the area to which the seeds belong, and the number of particles before the growth of lymphatic particles and the number of newly grown particles (corresponding to the first growing particles) are counted; the particle types corresponding to lymphatic type particles are set to unknown types, mononuclear particles are used as seeds, and the particles grow in all directions, and different colors are used to mark the particles in the growth area and the area described by the seeds, and the number of particles before the growth of mononuclear particles and the number of newly grown particles (corresponding to the second growing particles) are counted; the sum of the numbers of lymphatic particles and mononuclear particles before growth is determined as the first total number, and the sum of the numbers of newly grown lymphatic particles and newly grown mononuclear particles is determined as the fusion number; the fusion number is divided by the first total number to obtain the fusion ratio of lymphatic mononuclear particles.

[0059] In some embodiments, see Figure 1 , the processing module 102 of the present application can also be used to: in response to the fusion ratio being less than or equal to a preset threshold, determine that the sample to be detected corresponding to the two-dimensional image is a normal sample; in response to the sample to be detected corresponding to the two-dimensional image being a normal sample, determine the proportion of each type of particles in all particles; in response to the proportion value not being within a preset range, determine that the sample to be detected corresponding to the two-dimensional image is an abnormal sample, wherein the preset range is determined according to actual conditions, such as [0.8, 0.9]. For example, determine the proportion of the number of type A particles in the total number of particles, and then determine whether the proportion meets the proportion range of normal samples. If so, it means that the sample to be detected corresponding to the two-dimensional image is a normal sample, otherwise it is considered that the sample to be detected corresponding to the two-dimensional image is an abnormal sample.

[0060] Among them, each type of particle in the two-dimensional image has a preset range such as 0.2 to 0.5. The preset ranges of different types of particles can be the same or different. When making a judgment, all types of particles need to be judged. As long as the number of one type of particles is not within the preset range, the sample to be detected corresponding to the two-dimensional image is judged to be an abnormal sample.

[0061] Different from the prior art, the sample analyzer 10 provided in the present application first acquires a two-dimensional image and then segments and identifies anomalies in the two-dimensional image. This method can not only eliminate abnormal two-dimensional images in advance and improve the data reliability of the data used in the abnormality identification process, but also optimize the overall process and reduce resource consumption. By calculating the fusion ratio for anomaly identification, the problem of inaccurate segmentation in particle segmentation can be avoided, thereby improving the accuracy of overall abnormality identification.

[0062] In other embodiments, the sample analyzer 10 may further include a control module, an interaction module, a configuration module, and other modules, which may be configured according to actual needs. Figure 5 As shown, the interactive module 103 in the sample analyzer 10 is connected to the processing module 102, and is used to display abnormal indication information of the sample to be detected in response to the sample to be detected corresponding to the two-dimensional image being an abnormal sample. The abnormal indication information includes at least one of the particle type distribution after the two-dimensional image is processed by image segmentation, the particle type with a fusion ratio greater than a preset threshold, and the fusion ratio corresponding to the particle type with a fusion ratio greater than the preset threshold.

[0063] Different from the prior art, the sample analyzer 10 provided in the present application can not only exclude abnormal two-dimensional images in advance and improve the data reliability of the data used in the abnormality identification process, but also optimize the overall process and reduce resource consumption. By calculating the fusion ratio for abnormality identification, the problem of inaccurate segmentation of particle segmentation can be avoided, thereby improving the overall accuracy of abnormality identification.

[0064] In summary, see Figure 6 , Figure 6 1 is a flow chart of a first embodiment of a sample detection method provided by the present application. The sample detection method is applied to a sample analyzer, which is the same as or similar to the sample analyzer 10 of the above embodiment. The sample detection method includes:

[0065] Step 61: Detect the sample to be detected to obtain first detection data and second detection data, where the first detection data and the second detection data correspond to different light receiving channels.

[0066] The sample to be tested may be a blood sample, a saliva sample, a urine sample or others.

[0067] When the sample analyzer 10 detects the samples to be detected, it is necessary to add corresponding reagents to each sample to be detected, and place the sample to be detected with the added reagents in the detection area. When the detection area is irradiated with a light source (such as a laser) in the detection area, the light source irradiates each cell in the sample to be detected to generate two-dimensional signal data and / or three-dimensional signal data.

[0068] If two-dimensional signal data is obtained by detecting the sample to be detected, the two-dimensional signal data corresponds to the first detection data and the second detection data; if three-dimensional signal data is obtained by detecting the sample to be detected, any two signal data can be directly determined from the three-dimensional signal data, and then the two signal data are used as the two-dimensional signal data, namely the first detection data and the second detection data.

[0069] Step 62: Obtain a two-dimensional image based on the first detection data and the second detection data.

[0070] In some embodiments, the two-dimensional image is determined based on the three-dimensional image, the three-dimensional image is obtained based on the three-dimensional signal data, and the two-dimensional image is determined based on two signal data in the three-dimensional signal data.

[0071] The three-dimensional signal data includes side scattered light signal data, forward scattered light signal data and side fluorescent signal data; the two-dimensional signal data includes side scattered light signal data and side fluorescent signal data.

[0072] After acquiring the two-dimensional signal, a scattered point image corresponding to the sample to be detected can be established, that is, the two-dimensional image corresponding to the two-dimensional signal can be detected. The scattered points in the two-dimensional image correspond one-to-one to the particles in the sample to be detected, and the position of each scattered point in the two-dimensional image can be determined by the multi-dimensional signal data of the corresponding particle.

[0073] Specifically, the process of establishing a two-dimensional particle image based on two-dimensional signal data may include: the first detection data is side scattered light signal data, the second detection data is side fluorescent signal data, a coordinate system is established with the side scattered light signal data as the X-axis and the side fluorescent signal data as the Y-axis, and the two-dimensional signal data of the particles detected from the sample to be detected is mapped to the coordinate system, so that a two-dimensional particle image can be established. Of course, a coordinate system can also be established with the side fluorescent signal data as the X-axis and the side scattered light signal data as the Y-axis.

[0074] It can be understood that the two-dimensional image is a two-dimensional particle image, and the two-dimensional image is a grayscale image, not an RGB image. At this time, the sample analyzer 10 can perform image segmentation processing on the two-dimensional image, and can determine the types of all scattered points in the two-dimensional image. The type of scattered points is related to the sample to be detected. The sample to be detected is a blood sample. If the blood sample is subjected to a white blood cell test, the types of scattered points involved include blood ghosts, lymphocytes, monocytes, eosinophils, basophils, and neutrophils; if the blood sample is subjected to a reticulocyte test, the types of particles involved include red blood cells, reticulocytes, platelets, white blood cells, etc.

[0075] In addition, after acquiring the two-dimensional image, it is necessary to perform abnormal analysis on the two-dimensional image. Specifically, the two-dimensional image is subjected to binary classification processing to eliminate the abnormal detection samples corresponding to the two-dimensional image. At this time, the following method can be used: Figure 2 The 2-class classification neural network shown is used for processing.

[0076] Step 63: Perform image segmentation processing on the two-dimensional image to determine the types of all particles in the two-dimensional image.

[0077] In some embodiments, the two-dimensional image is subjected to two image segmentation processes, including a semantic segmentation process and a region growing process. Figure 7, step 63 may include the following steps:

[0078] Step 71: Perform semantic segmentation processing on the two-dimensional image to determine the first type of all particles in the two-dimensional image.

[0079] In some embodiments, using Figure 3 The semantic segmentation network shown performs image segmentation on the two-dimensional image to obtain the first type of all particles in the two-dimensional image.

[0080] Step 72: Determine, based on the first type of all particles, an edge region image in which at least two different particle types exist.

[0081] In some embodiments, step 72 may include the following process (not shown):

[0082] S11: split the two-dimensional image into several sub-images.

[0083] S12: Determine an edge region image based on the plurality of sub-images; wherein the edge region image contains at least two different types of particles.

[0084] Step 73: Correct the type of particles in the edge region image to change the type of particles in the edge region image from the first type to the second type, so as to obtain the types of all particles in the two-dimensional image.

[0085] In some embodiments, step 73 may include the following process (not shown):

[0086] S21: performing region growing processing on the particles contained in the edge region image to determine the second type of the particles contained in the edge region image.

[0087] S22: using the second type to correct the types of some particles corresponding to the first type, to obtain the types of all particles in the two-dimensional image.

[0088] Step 64: Calculate the fusion ratio between different types of particles, and compare the fusion ratio with a preset threshold.

[0089] In some embodiments, step 64 may include the following steps (not shown):

[0090] S31: Determine any two different types of particles as first particles and second particles.

[0091] S32: defining one of the first particles and the second particles as a particle of an unknown type, and defining the other of the first particles and the second particles as a particle of a known type.

[0092] S33: performing a region growth process on particles of known type, performing a region growth process on particles of known type, and obtaining first grown particles corresponding to the first particles and second grown particles corresponding to the second particles.

[0093] S34: Determine a fusion ratio based on the first growing particles and the second growing particles.

[0094] S35: Compare the fusion ratio with a preset threshold.

[0095] The preset threshold may be 0.5, 1.5 or others, which is not limited here.

[0096] Step 65: In response to the fusion ratio being greater than a preset threshold, determining that the sample to be detected corresponding to the corresponding two-dimensional image is an abnormal sample.

[0097] In addition, in response to the fusion ratio being less than or equal to a preset threshold, it is determined that the sample to be detected corresponding to the two-dimensional image is a normal sample.

[0098] Further, after executing step 65, the ratio of each type of particles in all particles can also be determined; in response to the ratio value not being within the preset range, the sample to be detected corresponding to the two-dimensional image is determined to be an abnormal sample. The preset range can be 0.8 to 0.9 or other, which is not limited here. It is worth noting that each type of particle has a corresponding preset range.

[0099] Different from the prior art, the sample detection method provided in the present application can exclude abnormal two-dimensional images to optimize the overall process and reduce resource consumption, and uses two-dimensional images to process the two-dimensional images using the image segmentation technology twice, which can improve the accuracy of identifying the type of particles in the two-dimensional images.

[0100] In an application scenario, see Figure 8 , the sample detection method applied to the sample analyzer 10 may include the following steps:

[0101] Step 81: Detect the sample to be detected to obtain first detection data and second detection data.

[0102] Step 82: Obtain a two-dimensional image based on the first detection data and the second detection data.

[0103] Step 83: Perform binary classification judgment on the two-dimensional image based on the binary classification neural network.

[0104] When the binary classification result is abnormal, the sample to be detected corresponding to the two-dimensional image is determined to be an abnormal sample, and detection continues on another two-dimensional image; otherwise, step 84 is executed.

[0105] Step 84: Perform semantic segmentation processing on the two-dimensional image to determine the first type of all particles in the two-dimensional image.

[0106] Step 85: Determine, based on the first type of all particles, an edge region image in which at least two different particle types exist.

[0107] Step 86: Correct the type of particles in the edge region image to change the type of particles in the edge region image from the first type to the second type, so as to obtain the types of all particles in the two-dimensional image.

[0108] Step 87: Calculate the fusion ratio between different types of particles.

[0109] Step 88: Determine whether the fusion ratio is greater than a preset threshold.

[0110] If so, it is determined that the sample to be detected corresponding to the two-dimensional image is an abnormal sample, and the subsequent process is stopped; otherwise, step 89 is continued.

[0111] Step 89: Calculate the proportion of each type of particles in all the particles.

[0112] Step 810: Determine whether the ratio value is within a preset range.

[0113] If not, it is determined that the sample to be detected corresponding to the two-dimensional image is an abnormal sample; if so, it is determined that the sample to be detected corresponding to the two-dimensional image is a normal sample.

[0114] Different from the prior art, the sample detection method provided by the present application can perform an abnormality detection on the two-dimensional image before image segmentation to eliminate abnormal samples to be detected, avoid waste of resources, and then perform image segmentation processing on the two-dimensional image twice, which can reduce the error rate of particle type recognition, and after identifying the particle type, two abnormality recognitions are required, which can reduce the error rate of abnormal recognition of samples to be detected. In other words, the sample detection method provided by the present application can greatly improve the success rate of abnormal recognition of samples to be detected.

[0115] See also Fig. 9 , Fig. 9 It is a structural diagram of an embodiment of a computer-readable storage medium provided in the present application. The computer-readable storage medium 90 stores a computer program 901. When the computer program 901 is executed, it is used to implement the sample detection method of any of the above embodiments.

[0116] The storage media used in this application include various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), or a CD.

[0117] In summary, the sample analyzer 10 and the corresponding sample detection method provided by the present application can perform abnormal detection on the sample to be detected to screen out abnormal samples, and at least three abnormal analyses need to be performed during the abnormal detection process, which can improve the success rate of abnormal detection. In the field of medical pathology, the cause of the patient's disease can be analyzed in multiple dimensions, which is conducive to determining the cause of the patient's disease and can also play a preventive role to a certain extent.

[0118] In addition, two-dimensional images, two-dimensional images after image segmentation processing, etc. can also be displayed on the display interface, and the type of particles can be selected by checking boxes or double-clicking, and the type of particles can be corrected when the particle type is identified incorrectly.

[0119] The above description is only an implementation method of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly used in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A sample analyzer, characterized in that: The sample analyzer comprises: A detection module, used to detect a sample to be detected to obtain first detection data and second detection data, wherein the first detection data and the second detection data correspond to different light receiving channels; A processing module is connected to the detection module, and the processing module is used to: Obtaining a two-dimensional image based on the first detection data and the second detection data; Performing image segmentation processing on the two-dimensional image to determine the types of all particles in the two-dimensional image; calculating a fusion ratio between particles of different types, and comparing the fusion ratio with a preset threshold; In response to the fusion ratio being greater than the preset threshold, it is determined that the sample to be detected corresponding to the two-dimensional image is an abnormal sample.

2. The sample analyzer according to claim 1, characterized in that: The processing module is also used for: Performing semantic segmentation processing on the two-dimensional image to determine the first type of all particles in the two-dimensional image; Determine an edge region image where at least two different particle types exist based on the first type of all particles; The type of particles in the edge area image is corrected to change the type of particles in the edge area image from the first type to the second type, so as to obtain the types of all particles in the two-dimensional image.

3. The sample analyzer according to claim 2, characterized in that: The processing module is also used for: Splitting the two-dimensional image into a plurality of sub-images; Based on the plurality of sub-images, an edge region image is determined; wherein the edge region image contains at least two different types of particles; Performing region growing processing on the particles contained in the edge region image to determine a second type of the particles contained in the edge region image; The types of some particles corresponding to the first type are corrected using the second type to obtain the types of all particles in the two-dimensional image.

4. The sample analyzer according to claim 3, characterized in that: The processing module is also used for: After performing region growing processing on the particles contained in the edge area image, if there are ungrown particles in the edge area image, the center of gravity distance between the ungrown particles and each type of particles is calculated, and the type with the smallest center of gravity distance is confirmed as the second type of the ungrown particles.

5. The sample analyzer according to claim 1, characterized in that: The processing module is also used for: Determine any two different types of particles as first particles and second particles; defining one of the first particle and the second particle as a particle of an unknown type, and defining the other of the first particle and the second particle as a particle of a known type; Performing a regional growth process on the particles of the known type to obtain first grown particles corresponding to the first particles and second grown particles corresponding to the second particles; determining the fusion ratio based on the first growing particles and the second growing particles; In response to the fusion ratio being not within a preset range, it is determined that the sample to be detected corresponding to the two-dimensional image is an abnormal sample.

6. The sample analyzer according to claim 5, characterized in that: The processing module is also used for: a first total number obtained by summing the first particles and the second particles; Merging the first growing particle and the second growing particle to obtain a fusion quantity between the first particle and the second particle; The fusion ratio is obtained according to the ratio of the fusion quantity to the first total quantity.

7. The sample analyzer according to claim 1, characterized in that: The processing module is also used for: In response to the fusion ratio being less than or equal to the preset threshold, determining that the sample to be detected corresponding to the two-dimensional image is a normal sample; In response to the sample to be detected corresponding to the two-dimensional image being a normal sample, calculating a ratio value of each type of particles in all the particles, and comparing the ratio value with a preset range; In response to the ratio value not being within the preset range, it is determined that the sample to be detected corresponding to the two-dimensional image is an abnormal sample.

8. The sample analyzer according to claim 1, characterized in that: Before performing image segmentation processing on the two-dimensional image to determine the types of all particles in the two-dimensional image, the processing module is further used to: Performing binary classification judgment on the two-dimensional image based on a binary classification neural network; When the binary classification judgment result is abnormal, determining that the sample to be detected corresponding to the two-dimensional image is an abnormal sample; When the binary classification result is normal, the step of performing image segmentation processing on the two-dimensional image to determine the types of all particles in the two-dimensional image is performed.

9. The sample analyzer according to claim 1, characterized in that: The sample analyzer also includes: An interactive module is connected to the processing module and is used to display abnormal indication information of the sample to be detected in response to the sample to be detected corresponding to the two-dimensional image being an abnormal sample; wherein the abnormal indication information includes at least one of the particle type distribution of the two-dimensional image after image segmentation processing, the particle type whose fusion ratio is greater than the preset threshold, and the fusion ratio corresponding to the particle type whose fusion ratio is greater than the preset threshold.

10. A sample detection method, characterized in that: Applications in sample analyzers include: Detecting a sample to be detected to obtain first detection data and second detection data, wherein the first detection data and the second detection data correspond to different light receiving channels; Obtaining a two-dimensional image based on the first detection data and the second detection data; Performing image segmentation processing on the two-dimensional image to determine the types of all particles in the two-dimensional image; calculating a fusion ratio between particles of different types, and comparing the fusion ratio with a preset threshold; In response to the fusion ratio being greater than the preset threshold, it is determined that the sample to be detected corresponding to the corresponding two-dimensional image is an abnormal sample.

Images