Blood spherical red blood cell parameter, detection method and identification AI training method thereof

Abstract

In the blood spherical red blood cell parameter, the detection method and the identification AI training method thereof, a blood sample is preprocessed to obtain a microscopic examination sample, and the microscopic examination sample is tiled; a tiled microscopic examination sample image is shot, normal red blood cells and spherical red blood cells are recognized and labeled, labeled pictures are obtained and subjected to AI training, and an obtained AI feature data set A comprises normal red blood cell features and an obtained AI feature data set A comprises spherical red blood cell features. Identifying the microscopic examination sample image by using an AI identification algorithm, identifying spherical red blood cells in the blood sample in a selected area S1 of the image, and obtaining the total number NUMS1 of the spherical red blood cells in the blood sample in the selected image; the AI recognition algorithm recognizes the spherical red blood cells in the blood sample according to the feature data set including the spherical red blood cells. The AI training data collection efficiency is improved through diversified means, and the AI evolution efficiency is improved. Spherical red blood cell recognition is completed through advanced AI computing power, analysis is more efficient, and accuracy is higher.

Description

Technical Field

[0001] The present application belongs to the technical field of analysis of formed elements based on microscopically magnified images, and particularly relates to an AI training method for identifying blood spherical red blood cells and a computing and processing and storage device, as well as a blood spherical red blood cell parameter and detection method and a computing and processing and storage device. Background Art

[0002] Red blood cells (RBCs) are the most numerous type of blood cell and the primary means of transporting oxygen through the blood of vertebrates. Among vertebrates, mammalian RBCs are round, with a concave center. They are typically anucleate in circulating blood. Most mammalian RBCs are round, with convex edges and a concave center.

[0003] Human red blood cells are shaped like round pancakes with convex edges and a concave center. They are thicker at the edges and thinner in the middle, like a donut, but without the hole. This shape maximizes oxygen uptake from the surrounding area. It also provides flexibility, allowing it to release oxygen molecules through capillaries. The unique shape and relatively small size of red blood cells result in a large surface-to-volume ratio, allowing oxygen and carbon dioxide to quickly penetrate the cell. Abnormalities in red blood cell morphology are often indicative of certain diseases.

[0004] Spherocytes are a form of red blood cells. Spherocytes are small red blood cells with a diameter less than 2 / 3 of normal red blood cells. They are spherical rather than disc-shaped.

[0005] It is usually caused by genetic factors, immune factors, thalassemia, myelodysplastic syndrome, chronic anemia, etc. Spherocytes are commonly seen in hereditary spherocytosis and other hemolytic anemias with spherocytosis, such as autoimmune hemolytic anemia, hemolytic disease of the newborn, and hemolytic anemia caused by red blood cell enzyme defects.

[0006] Currently, the only conventional blood test suitable for whole-animal testing is blood smear microscopy. Blood smear microscopy is often affected by manual smearing techniques. Due to uneven front-to-back smears and the high workload involved in counting, full smear counts are not typically performed, resulting in poor reproducibility of data from multiple smears. This is significantly affected by the interpreter's ability. Spherocytes and normal red blood cells differ in morphology. In blood smear microscopy, these are flat, so the morphological differences between spherocytes and normal red blood cells are not obvious or prominent.

[0007] Existing hemocytometers utilize fluorescent staining, flow cytometry, or the Coulter impedance principle to count and classify mammalian red blood cells. These methods typically rely on the magnitude of electrical or optical signals, rather than image analysis, making it difficult to distinguish between spherocytes and normal red blood cells. Currently, no device is known that can identify and classify spherocytes.

[0008] The applicant filed a series of Chinese patents, such as

[0009] 1. CN2020112669290, “Cell analysis method and system and quantitative method and system”;

[0010] 2. CN2020112669182, “Cell Suspension Sample Imaging Method, System, and Kit”;

[0011] 3. CN2022104799126, “Fast focusing method for microscopic image acquisition device and microscopic image detection method”;

[0012] 4. CN2023100423151, “Blood sample imaging analysis system and method”;

[0013] 5. CN2023116834572, "AI Training Method and Computational Processing and Storage Device for Leukocyte Identification in Blood Samples," uses a novel AI recognition technology solution to measure blood samples and the content of formed elements in blood samples. Building on this foundation, the technical challenge remains to be solved: how to achieve spherocyte classification and counting.

[0014] If AI recognition technology is used to identify spherocytes, a large amount of AI training is required, and image samples of spherocytes are needed. The number of normal red blood cells is huge, and it is extremely difficult to find spherocytes manually. Without image samples, AI training cannot be carried out, and the AI ​​model cannot be verified and performance evaluated. Summary of the Invention

[0015] In this application, the inventors proposed to conduct AI training on images of blood sample suspensions containing spherical red blood cells. The obtained AI feature dataset A can effectively support AI detection of red blood cells in blood samples, and use advanced AI computing power to complete the identification of red blood cells, making the analysis of blood spherical red blood cells more efficient and more accurate.

[0016] In this application, the inventors proposed a blood sample preprocessing process that can accelerate the collection efficiency of spherocyte samples and improve the AI ​​training efficiency of identifying spherocytes.

[0017] The technical solution of the present application to solve the above-mentioned technical problems is an AI training method for identifying blood spherical red blood cells. The blood sample is preprocessed to obtain a microscopic sample; the microscopic sample is flattened; the red blood cells are allowed to settle to the bottom; the flattened microscopic sample is photographed to obtain a microscopic sample image; normal red blood cells and spherical red blood cells in the microscopic sample image are identified and labeled to obtain a labeled image, and the labeled image is used for AI training to obtain an AI feature data set A; the AI ​​feature data set A includes normal red blood cell features, and the AI ​​feature data set A includes spherical red blood cell features.

[0018] The pretreatment of blood samples includes any one of the following technical features: TA1: adding a dye to the blood sample, staining the blood sample, and diluting the blood sample after staining to obtain a microscopic sample; TA2: adding a liquid dye to the blood sample, staining and diluting the blood sample to obtain a microscopic sample; TA3: preparing a diluted staining solution with dry powder dye and water, mixing the diluted staining solution and the blood sample to obtain a microscopic sample; TA4: adding a hypotonic solution to dilute the blood sample, leaving it for N seconds, adding a protective agent, and then adding the dye to obtain a microscopic sample, 10≤N≤300; TA5: mixing the hypotonic solution and the protective agent to obtain an intermediate dilution solution, adding the blood sample to the intermediate dilution solution, and then adding the dye to obtain a microscopic sample.

[0019] Preprocessing of blood samples includes any of the following technical features: TC1: After the microscopic sample is prepared, it is photographed after waiting for a period of time TW; the time TW ranges from 0.5 minutes to 5 minutes; TC2: the above blood sample comes from multiple animal individuals; TC3: multiple blood samples are mixed to obtain a microscopic sample; TC4: the above blood sample comes from one or more individuals with spherocytosis.

[0020] Use AI feature dataset A as the feature dataset, use AI software to identify one or more input images, and obtain recognition output images. The above-mentioned recognition output images include recognized normal red blood cell images and spherical red blood cell images. The recognized spherical red blood cell images are manually reviewed to form manual review mark images. The manual review mark images include marked normal red blood cell images and marked spherical red blood cell images. Use the manual review mark images for AI training to obtain AI feature dataset B. AI feature dataset B includes normal red blood cell features, and AI feature dataset B includes spherical red blood cell features.

[0021] A computing and processing device, comprising any one of the following technical features: TD1: used to run all or part of the above-mentioned AI training method for identifying blood spherical red blood cells; TD2: the memory of the above-mentioned computing and processing device includes the above-mentioned AI feature data set A; TD3: the memory of the above-mentioned computing and processing device includes the above-mentioned AI feature data set B.

[0022] A data storage device includes any one of the following technical features: TE1: storing program code for executing all or part of the aforementioned AI training method for identifying spherical red blood cells; TE2: storing the aforementioned AI feature dataset A; storing the aforementioned AI feature dataset B. A detection device is configured to execute part or all of the aforementioned AI training method for identifying spherical red blood cells.

[0023] A method for detecting blood spherocyte parameters comprises the following steps: pre-processing a liquid sample to obtain a microscopic sample; flattening the microscopic sample; photographing the flattened microscopic sample to obtain a microscopic sample image; using an AI recognition algorithm to recognize the microscopic sample image, identifying spherocytes in the blood sample in a selected area S1 of the image, and obtaining the total number NUMS1 of spherocytes in the blood sample in the selected image; and the AI ​​recognition algorithm identifying spherocytes in the blood sample based on a feature data set including spherocytes.

[0024] The spherical red blood cell feature dataset in the above blood samples is obtained by manually annotating the spherical red blood cell images in the blood samples and then training.

[0025] The above-mentioned blood spherocyte parameter detection method includes any one of the following technical features: TG1: the above-mentioned blood spherocyte parameter is the number of spherocytes in a unit volume of blood sample, and the above-mentioned blood sample has been pre-treated by dilution and diluted N times; the tile height of the above-mentioned microscopic sample is H; the volume V corresponding to the total number of spherocytes NUMS1 in the above-mentioned blood sample = the selected area S1×H; the number of spherocytes in a unit volume of blood sample = N×the total number of spherocytes NUMS1 in the blood sample / the volume of the blood sample V; TG2: the above-mentioned blood spherocyte parameter is the number of spherocytes in a unit volume of blood sample, and the above-mentioned blood sample has been pre-treated by concentration and concentrated N times, and the tile height of the above-mentioned microscopic sample is H; the volume V corresponding to the total number of spherocytes NUMS1 in the above-mentioned blood sample = the selected area S1×H; the number of spherocytes in a unit volume of blood sample = the total number of spherocytes NUMS1 in the blood sample / (the volume of the blood sample V×N).

[0026] An AI recognition algorithm is used to identify the above-mentioned microscopic sample image, identify normal red blood cells in the blood sample in the selected area S1 of the image, and obtain the total number NUMS2 of normal red blood cells in the blood sample in the selected image; the above-mentioned AI recognition algorithm identifies normal red blood cells in the blood sample based on a characteristic data set of normal red blood cells in the blood sample; the ratio of spherical red blood cells to normal red blood cells = NUM1 / NUM2.

[0027] The blood sample pretreatment includes diluting the blood sample with a liquid dye to obtain a microscopic specimen, with the dilution factor being 50 ≤ M ≤ 800. The blood spherocyte parameter detection method includes any of the following technical features: TH1: microscopic examination and photography are completed within time T15 after the microscopic specimen is prepared as a hypotonic suspension; time T15 ranges from 10 minutes to 30 minutes; TH2: microscopic specimen preparation and microscopic photography are completed at a temperature greater than 18 degrees Celsius and less than 45 degrees Celsius.

[0028] A computing and processing device is used to execute all or part of the above-mentioned blood spherocyte parameter detection method.

[0029] A data storage device storing all or part of the program code for executing the above-mentioned blood spherocyte parameter detection method. A detection device for executing part or all of the above-mentioned blood spherocyte parameter detection method.

[0030] A blood spherocyte parameter used to evaluate the number of spherocytes in a blood sample, including any one of the following technical features: TJ1: the number of spherocytes in a unit volume of blood sample = the number of spherocytes in the blood sample per unit volume of blood sample / the volume of blood sample; TJ2: the proportion of the number of spherocytes in a blood sample = the number of spherocytes in the same volume of blood sample / the number of red blood cells.

[0031] The technical effects of the above technical solution include: improving the efficiency of AI training data collection through diversified means and improving the efficiency of AI evolution.

[0032] The technical effects of the above technical solution include: using advanced AI computing power to complete the identification of spherical red blood cells, making the analysis of blood spherical red blood cells more efficient and more accurate.

[0033] The technical effects of the above technical solution include: encapsulating the trained AI feature data set A including spherical red blood cell features into independent data, which can be sold separately, reducing the engineering difficulty of the entire industry and allowing new technologies to be promoted and implemented as soon as possible.

[0034] The technical effects of the above technical solution include: improving the image collection efficiency of spherical red blood cells through a variety of preprocessing methods, which is conducive to accelerating the AI ​​training iteration process and improving efficiency.

[0035] The technical effects of the above technical solution include: the purpose of blood sample staining is to detect other cells in the blood sample, such as white blood cells. When training the AI ​​model, using the same staining can improve the accuracy of simultaneous identification of multiple cells and further improve system efficiency.

[0036] The technical effects of the above technical solution include: direct addition of liquid dye, which can achieve staining and dilution at the same time, and high sample preparation efficiency.

[0037] The technical benefits of the above-described technical solution include: after preparing the microscopic sample, imaging can be performed after a waiting time TW. After a sufficiently long waiting time, normal red blood cells or already spherical red blood cells can be transformed into spherical red blood cells, thereby improving the efficiency of obtaining spherical red blood cell samples. When the microscopic sample is prepared as a hypotonic suspension, this process is accelerated, and the waiting time TW is shortened.

[0038] The technical benefits of the above-mentioned technical solution include: by mixing blood samples from multiple individual animals to obtain microscopic samples, the efficiency of obtaining spherocyte samples is improved. Since not all samples contain spherocytes, the above-mentioned method can improve the efficiency of obtaining characteristic data.

[0039] The technical effects of the above technical solution include: the blood sample is derived from one or more individuals with spherocytosis, which can further improve the efficiency of obtaining spherocytosis samples.

[0040] The technical effects of the above technical solution include: the feature data set B evolved based on the AI ​​feature data set A is encapsulated into independent data, so that the AI ​​recognition capability of spherical red blood cells can be iterated and improved.

[0041] The technical effects of the above technical solution include: manually reviewing and marking images, including marking normal red blood cell images and marking spherical red blood cell images, which can reduce the probability of misidentifying normal red blood cells as spherical red blood cells and improve the accuracy of recognition.

[0042] The technical effects of the above technical solution include: the total number of spherical red blood cells NUMS1 in the blood sample in the selected image is given, the basic parameters for quantitative analysis of blood spherical red blood cells are given, and the AI ​​recognition algorithm is used, with high efficiency and accuracy.

[0043] The technical effects of the above technical solution include: the characteristic data set of spherical red blood cells in blood samples is obtained by training after manually annotating the images of spherical red blood cells in blood samples, and the accuracy of AI recognition is more guaranteed.

[0044] The technical effects of the above technical solution include: the number of spherocytes in a unit volume of blood sample is based on classification recognition, combined with image-based sample quantification, to achieve accurate spherocyte counting, which takes the accurate quantitative analysis of formed elements in blood samples a step further, and can perform precise quantitative analysis of spherocytes, making the formed elements of blood samples that can be accurately quantitatively analyzed richer and more diverse.

[0045] The technical benefits of the above-mentioned technical solution include: calculating the proportion of spherocytes in a blood sample by the ratio of the number of spherocytes in the same volume of blood sample to the number of red blood cells, taking the analysis of formed elements in blood samples a step further and enabling accurate quantitative analysis of the proportion of spherocytes. This provides a more dimensional quantitative analysis basis for determining the severity of disease, and provides clinicians with comprehensive red blood cell analysis data, enriching the clinical data reference set. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is one of the steps in the AI ​​training method for identifying blood spherical red blood cells;

[0047] Figure 2 This is the second step of the AI ​​training method for identifying blood spherical red blood cells;

[0048] Figure 3 This is one of the pictures of blood cells, including normal red blood cells and spherocytes;

[0049] Figure 4 yes Figure 3 Some screenshots of Figure 4 The dotted box in the middle is the image of the identified spherical red blood cells;

[0050] Figure 5 This is one of the schematic diagrams of the blood sample pretreatment method;

[0051] Figure 6 This is the second diagram of the blood sample pretreatment method;

[0052] Figure 7 This is the third diagram of the blood sample pretreatment method;

[0053] Figure 8 This is the fourth diagram of the blood sample pretreatment method;

[0054] Figure 9 This is the fifth diagram of the blood sample pretreatment method;

[0055] Figure 10 This is the sixth diagram of the blood sample pretreatment method;

[0056] Figure 11 This is the seventh diagram of the blood sample pretreatment method;

[0057] Figure 12 This is the eighth diagram of the method for pre-processing blood samples;

[0058] Figure 13 This is the second picture of blood cells; the framed cells in the picture are identified spherical red blood cells;

[0059] Figure 14It is one of the schematic diagrams of the method for detecting blood spherocyte parameters;

[0060] Figure 15 This is the second schematic diagram of the method for detecting blood spherocyte parameters;

[0061] Figure 16 This is the third picture of blood cells;

[0062] Figure 17 yes Figure 16 Some screenshots show Figure 17 The image of spherical red blood cells identified in the figure;

[0063] Figure 18 This is the ninth diagram of the method for pre-processing a blood sample;

[0064] Figure 19 This is the tenth diagram of the blood sample pretreatment method. DETAILED DESCRIPTION

[0065] The contents of this application are further described in detail below in conjunction with the accompanying drawings. It should be noted that the following is a description of the preferred embodiments of the present invention and does not constitute any limitation to the present invention. The description of the preferred embodiments of the present invention is only an illustration of the general principles of the present invention. The numbers such as "first", "second" and "A" and "B" involved in the present invention are only for the convenience of explanation and do not represent the order relationship in time or space. The combination of letters and numbers "TA", "TB" and "H" involved in the present invention are only for the convenience of explanation, and the specific meaning is determined by the specific words referred to.

[0066] The presence of a large number of spherocytes in the blood is considered to be caused by immune-mediated hemolysis. Other diseases can also lead to an increase in spherocytes, such as Heinz body hemolytic anemia, blood parasites, toxin poisoning, microvascular diseases and tumor diseases. Such as lymphoma and hemangioma. Spherocyte classification and counting has clinical value as a reference for disease diagnosis. The existing technology lacks separate classification and counting of spherocytes. Whether it is a smear or a conventional cell analyzer that classifies and identifies cells based on photoelectric signals, it is difficult to accurately classify, identify and count spherocytes. That is, a general blood routine test cannot interpret spherocytes, nor can it indicate their abnormalities.

[0067] For mammalian blood analysis, fully automated spherocyte counting has not yet been achieved, making it difficult to more accurately assess the animal's condition through large-scale counting. This application can use AI+ red blood cell quantification to perform large-scale spherocyte counting in mammals, overcoming the difficulty and inaccuracy of spherocyte counting on blood smears.

[0068] Existing technologies lack spherocyte calibration standards, making it difficult to accurately determine the accuracy of spherocyte identification. Spherocyte identification and classification remain unrealistic. The clinical probability of an abnormal spherocyte is much lower than that of normal red blood cells. Collecting large numbers of spherocytes for AI training is unrealistic. Improving the efficiency of AI training, increasing the efficiency of collecting abnormal training samples, and increasing the rate of AI evolution remain unresolved technical challenges.

[0069] In this application, spherical red blood cell samples are collected through multiple methods to improve the efficiency of training data acquisition, provide richer spherical red blood cell samples for AI training, improve the efficiency of collecting and acquiring feature data, and improve the efficiency of AI training methods.

[0070] In this application, images of blood cells in a living state are obtained from suspended blood microscopy samples. The AI-based feature data can be used to identify spherical red blood cells. AI-based recognition enables accurate classification and counting of these cells. In mammals, this method can clearly alert clinicians to abnormalities in this indicator through optical imaging.

[0071] like Figure 1 and Figure 2 , an AI training method for identifying blood spherical red blood cells, the blood sample is pre-processed to obtain a microscopic sample; the microscopic sample is flattened; the red blood cells are allowed to settle to the bottom; the flattened microscopic sample is photographed to obtain a microscopic sample image; the normal red blood cells and spherical red blood cells in the microscopic sample image are identified and labeled to obtain a labeled image, and the labeled image is used for AI training to obtain an AI feature data set A; the AI ​​feature data set A includes normal red blood cell features, and the AI ​​feature data set A includes spherical red blood cell features. Figure 1 During image processing, image segmentation is performed first to obtain red blood cell sub-images. These sub-images are then labeled, with normal red blood cells labeled as normal red blood cells and spherocytes labeled as spherocytes. The labeled image data is used to train the AI ​​model, outputting AI feature dataset A, which includes both normal red blood cells and spherocytes. AI feature dataset A can be used for subsequent cell recognition.

[0072] like Figure 2, an AI training method for identifying blood spherical red blood cells, using AI feature data set A as a feature data set, using AI software to identify one or more input images, obtaining a recognition output image, the above-mentioned recognition output image includes the recognized normal red blood cell image and the spherical red blood cell image, manually reviewing the recognized spherical red blood cell image to form a manual review mark image, the manual review mark image includes a marked normal red blood cell image and a marked spherical red blood cell image, using the manual review mark image for AI training, obtaining an AI feature data set B, AI feature data set B includes normal red blood cell features, and AI feature data set B includes spherical red blood cell features.

[0073] Figure 1 In the , only one round of training was performed. Figure 2 The training can be repeated multiple times. Figure 2 In this example, AI feature dataset A is used for AI recognition verification, identifying and outputting normal red blood cells and spherocytes. The identified and output spherocytes are manually reviewed, and normal and spherocytes are labeled. A second round of AI training is then conducted on this basis, outputting the iterative AI feature dataset B. This enriches the dataset and reduces the probability of misidentification of spherocytes. The difference in projection between spherocytes and normal red blood cells is not significant from a morphological perspective, making misidentification more likely in the absence of sufficient training data. By incorporating manual review and secondary AI training, and through repeated iterations, the feature dataset can be continuously optimized, improving recognition accuracy.

[0074] Figure 3 This is one of the pictures of blood cells, including normal red blood cells and spherocytes; Figure 4 yes Figure 3 Some screenshots of Figure 4 The dotted box in the middle is the image of the identified spherical red blood cells.

[0075] Pretreatment of blood samples: adding dye to the blood sample, staining the blood sample, and diluting the blood sample after staining to obtain a microscopic sample. Staining can be done first and then diluted. Alternatively, liquid dye can be added to the blood sample, staining and diluting the blood sample to obtain a microscopic sample. Alternatively, dry powder dye can be mixed with water to make a diluted staining solution, and the diluted staining solution and blood sample can be mixed to obtain a microscopic sample. Figure 19 , add dye to the blood sample, stain the blood sample, and dilute it after staining to obtain a microscopic examination sample.

[0076] like Figure 5, pretreatment of blood samples, adding liquid dyes to blood samples, staining and diluting the blood samples to obtain microscopic samples. Liquid dyes include diluted dye solutions, which are used to stain and dilute blood samples. Liquid dyes can be high-purity dyes, which are then diluted after staining to obtain microscopic samples. Whether it is a dry powder dye or a liquid dye, it can be added at the appropriate link. Staining is to take into account other components in blood cells, such as white blood cells, which can be identified and classified based on the same process. Therefore, during AI training, it is also based on the same staining background to improve the consistency of AI training and actual application scenarios, thereby improving recognition accuracy.

[0077] like Figure 6 , including mixing water and dry powder stain to form a diluted staining solution, and mixing the diluted staining solution with blood to prepare a microscopic specimen.

[0078] like Figure 7 The blood sample and the diluted staining solution are mixed to form an intermediate sample, which is then left to stand for N seconds before being used as a microscopic examination sample. The value of N ranges from 10 to 300 seconds; it can be 20, 30, 40, or 50 seconds. N can be any time between 10 and 300 seconds.

[0079] like Figure 8 The blood sample and the hypotonic solution are mixed and diluted to form an intermediate sample. During the N seconds of standing, a protective agent is added to form a microscopic sample. Figure 18 After mixing the hypotonic solution and protective agent to obtain an intermediate dilution, the blood sample is added to the intermediate dilution, followed by the staining agent to obtain a microscopic specimen. The protective agent is added to maintain the spherical erythrocyte morphology. The staining agent is then added to the microscopic specimen for staining. Figure 9 The hypotonic solution in the microscope can be a simple diluent, without a stain. Of course, the hypotonic solution can also be a diluting stain, performing both dilution and staining simultaneously. Mixing blood with the hypotonic solution causes some normal red blood cells to absorb the solution and gradually transform into spherocytes. Adding a protective agent during the static process stabilizes the morphology of the spherocytes, facilitating the capture of as many spherocytes as possible during subsequent imaging. Dyes can also be added to the microscopic sample to stain other blood cells that need to be identified simultaneously, improving the efficiency of the system.

[0080] like Figure 9 After preparing the microscopic sample, wait for TW before taking the image. In some cases, this waiting period is necessary to allow more red blood cells to absorb extracellular fluid and become spherocytes, increasing the probability of spherocytes appearing and improving the efficiency of subsequent AI training and recognition. The TW period ranges from 0.5 minutes to 5 minutes.

[0081] like Figure 10After the microscopic sample is prepared, the shooting is completed within time T15; the time T15 ranges from 10 minutes to 30 minutes. In the microscopic sample, the maintenance time of spherical red blood cells is certain, especially in the hypotonic suspension environment. The spherical red blood cells will eventually rupture, and hemolysis will make it impossible to subsequently identify and classify the spherical red blood cells. Under ideal conditions, the cell morphology in the isotonic suspension environment can maintain its original state. However, in actual applications, the hypotonic suspension environment is usually close to isotonic, and most red blood cells will eventually become spherical red blood cells.

[0082] Maintaining an osmotic pressure close to that of red blood cells by diluting the solution can ensure cell morphology. Adding a hypotonic solution swells the red blood cells, and then adding a protective agent. This can cause the red blood cells to undergo a spheroidization reaction, producing spherical red blood cells. The osmotic pressure range of a hypotonic suspension is a solution less than 280mmol / L. An actual reagent formula: 1. Hypotonic solution (0.45% NaCl) 2. Protective agent is one or more of formaldehyde, paraformaldehyde, acetaldehyde, propionaldehyde, butyraldehyde, valeraldehyde, glutaraldehyde, or phenol; the hypotonic solution swells the red blood cells, while the protective agent fixes the swollen red blood cells to prevent them from swelling and hemolysis. An actual sample pretreatment step: 1. Use a pipette to add 10ul of fresh blood to the preservation solution containing the isotonic solution, and gently invert it up and down 10 times to mix. 2. Take 150ul of the mixture and add it to the detection chip for rapid detection on the instrument. 3. Obtain data on spherical red blood cells. Figure 13 is a picture of blood cells; the boxed cells are identified as spherocytes. Figure 16 Also a picture of blood cells; Figure 17 yes Figure 16 Some screenshots show Figure 17 Image of spherical red blood cells identified in .

[0083] like Figure 11 and Figure 12 , multiple blood samples are mixed to obtain microscopic samples. Figure 11 , the above blood samples come from multiple animal individuals, increasing the probability of obtaining characteristic data.

[0084] like Figure 12 The above blood samples are derived from one or more individuals with spherocytosis. When multiple blood samples are mixed, some pretreatment and screening steps are omitted. These pretreatment and screening steps can avoid aggregation of blood cells from different individuals. For example, blood samples with the same blood type are selected. Figure 12 , the multiple blood samples can be from the same individual or from different individuals who have spherocytosis.

[0085] A computing and processing device is used to run all or part of the above-mentioned AI training method for identifying blood spherical red blood cells; the memory of the computing and processing device includes the above-mentioned AI feature data set A; the memory of the computing and processing device includes the above-mentioned AI feature data set B.

[0086] A data storage device storing program code for executing all or part of the aforementioned AI training method for identifying spherical red blood cells; storing the aforementioned AI feature dataset A; storing the aforementioned AI feature dataset B. A detection device for executing part or all of the aforementioned AI training method for identifying spherical red blood cells.

[0087] like Figure 14 A method for detecting blood spherocyte parameters comprises preprocessing a blood sample to obtain a microscopic sample; flattening the microscopic sample; photographing the flattened microscopic sample to obtain a microscopic sample image; and using an AI recognition algorithm to identify the microscopic sample image, identifying spherocytes in the blood sample within a selected area S1 of the image, and obtaining the total number NUMS1 of spherocytes in the blood sample in the selected image. The AI ​​recognition algorithm identifies spherocytes in the blood sample based on a feature dataset that includes spherocytes. The feature dataset for spherocytes in the blood sample is obtained by training after manually annotating spherocyte images of the blood sample.

[0088] A blood spherocyte parameter is defined for evaluating the number of spherocytes in a blood sample. The number of spherocytes in a unit volume of a blood sample = the number of spherocytes in the blood sample per unit volume of the blood sample / the volume of the blood sample.

[0089] A method for detecting a blood spherocyte parameter, wherein the blood spherocyte parameter is the number of spherocytes per unit volume of a blood sample, the blood sample being diluted N times after dilution pretreatment; the tile height of the microscopic sample being H; the volume V corresponding to the total number of spherocytes NUMS1 in the blood sample being equal to the selected area S1×H; and the number of spherocytes per unit volume of the blood sample being equal to N×the total number of spherocytes NUMS1 in the blood sample being equal to / the volume V of the blood sample being V.

[0090] A method for detecting a blood spherocyte parameter, wherein the blood spherocyte parameter is the number of spherocytes per unit volume of a blood sample, the blood sample undergoes concentration pretreatment and is concentrated N-fold, and the microscopic sample is tiled to a height of H; the total number of spherocytes NUMS1 in the blood sample corresponds to a volume V = a selected area S1 × H; and the number of spherocytes per unit volume of the blood sample = the total number of spherocytes NUMS1 in the blood sample / (blood sample volume V × N).

[0091] A blood spherocyte parameter is defined to evaluate the proportion of spherocytes in a blood sample. The proportion of spherocytes in a blood sample = the number of spherocytes in the same volume of blood sample / the number of red blood cells. The number of red blood cells in the denominator can be as follows: Figure 15 The number of normal red blood cells shown. The number of red blood cells in the denominator can also be the number of all red blood cells, and the number of all red blood cells can be the sum of the number of normal red blood cells and the number of spherocytes.

[0092] like Figure 15 , use the AI ​​recognition algorithm to identify the above-mentioned microscopic sample image, identify the normal red blood cells in the blood sample in the selected area S1 of the image, and obtain the total number NUMS2 of normal red blood cells in the blood sample in the selected image; the above-mentioned AI recognition algorithm identifies the normal red blood cells in the blood sample based on the normal red blood cell feature data set in the blood sample; the ratio of spherical red blood cells to normal red blood cells = NUM1 / NUM2.

[0093] The blood sample pretreatment described above includes diluting the blood sample with a liquid stain to obtain a microscopic specimen, with the dilution factor being 50 ≤ M ≤ 800. The dilution factor is appropriate to ensure that the number of red blood cells in the resulting red blood cell flattening image is suitable for identification. Microscopic examination and photography should be completed within T15 of the time the microscopic specimen is prepared as a hypotonic suspension; and the preparation and microscopic examination of the specimen should be completed at a temperature greater than 18°C ​​and less than 45°C.

[0094] A computing device for executing all or part of the above-mentioned method for detecting spherocyte parameters. A data storage device for storing all or part of the program code for executing the above-mentioned method for detecting spherocyte parameters. A detection device for executing part or all of the above-mentioned method for detecting spherocyte parameters.

[0095] Although the present invention is illustrated and described based on the preferred embodiment and several alternatives, the invention is not limited by the specific description in this specification. Other additional replacement or equivalent components can also be used to practice the present invention.

Claims

1. A method for AI training to identify spherical red blood cells, characterized in that: Blood samples were pre-processed to obtain microscopic specimens; Spread the sample for microscopic examination flatly; let the red blood cells settle to the bottom; photographing the flattened microscopic sample to obtain an image of the microscopic sample; Identify and label normal red blood cells and spherocytes in microscopic sample images to obtain labeled images. Use the labeled images for AI training to obtain AI feature dataset A. The AI ​​feature dataset A includes normal red blood cell features, and the AI ​​feature dataset A includes spherical red blood cell features.

2. The AI ​​training method for identifying spherical red blood cells according to claim 1, characterized in that: Pretreatment of blood samples, including any of the following technical features: TA1: adding a dye to the blood sample, staining the blood sample, and diluting the blood sample after staining to obtain a microscopic examination sample; TA2: Add liquid dye to the blood sample, stain and dilute the blood sample to obtain a microscopic sample; TA3: Dry powder stain is mixed with water to make a diluted staining solution, which is then mixed with the blood sample to obtain a sample for microscopic examination. TA4: Dilute the blood sample with hypotonic solution, let it sit for N seconds, add a protective agent, and then add a stain to obtain a microscopic sample, 10≤N≤300; TA5: After mixing the hypotonic solution and the protective agent, an intermediate dilution is obtained. The blood sample is added to the intermediate dilution, and then a stain is added to obtain a microscopic specimen.

3. The AI ​​training method for identifying spherical red blood cells according to claim 1, characterized in that: Pretreatment of blood samples, including any of the following technical features: TC1: After the microscopic sample is prepared, wait for TW before taking the picture; the time TW ranges from 0.5 minutes to 5 minutes; TC2: The blood samples come from multiple animal individuals; TC3: Mix multiple blood samples to obtain microscopic samples; TC4: The blood sample is derived from one or more individuals with spherocytosis.

4. The AI ​​training method for identifying spherical red blood cells according to any one of claims 1 to 3, characterized in that: Using AI feature dataset A as the feature dataset, AI software is used to identify one or more input images to obtain a recognition output image, wherein the recognition output image includes an identified normal red blood cell image and a spherical red blood cell image. The identified spherical red blood cell image is manually reviewed to form a manually reviewed marked image, wherein the manually reviewed marked image includes a marked normal red blood cell image and a marked spherical red blood cell image. AI training is performed using the manually reviewed marked image to obtain AI feature dataset B, wherein AI feature dataset B includes normal red blood cell features, and AI feature dataset B includes spherical red blood cell features.

5. A computing and processing device, characterized in that: Include any one of the following technical features: TD1: used to run all or part of the AI ​​training method for identifying blood spherical red blood cells according to any one of claims 1 to 7; TD2: The memory of the computing and processing device includes the AI ​​feature data set A according to any one of claims 1 to 7; TD3: The memory of the computing and processing device includes the AI ​​feature data set B described in any one of claims 6 to 7.

6. A data storage device, characterized in that Include any one of the following technical features: TE1: stores program codes for executing all or part of the AI ​​training method for identifying spherical red blood cells according to any one of claims 1 to 7; TE2: stores the AI ​​feature dataset A described in any one of claims 1 to 7; TE3: stores the AI ​​feature dataset B described in any one of claims 6 to 7.

7. A detection device, characterized in that: Used to execute part or all of the AI ​​training method for identifying blood spherical red blood cells as described in any one of claims 1 to 7.

8. A method for detecting blood spherical red blood cell parameters, characterized in that: Blood samples were pre-processed to obtain microscopic specimens; Tiling of samples for microscopic examination; photographing the flattened microscopic sample to obtain an image of the microscopic sample; An AI recognition algorithm is used to identify the microscopic sample image, identify the spherocytes in the blood sample in the selected area S1 of the image, and obtain the total number NUMS1 of spherocytes in the blood sample in the selected image; the AI ​​recognition algorithm identifies the spherocytes in the blood sample based on a feature data set including spherocytes.

9. The method for detecting blood spherocyte parameters according to claim 8, characterized in that: The spherocyte feature dataset in the blood sample is obtained by manually annotating the spherocyte images in the blood sample and then training.

10. The method for detecting blood spherocyte parameters according to claim 8, characterized in that: Include any one of the following technical features: TG1: The blood spherocyte parameter is the number of spherocytes per unit volume of a blood sample. The blood sample is diluted N-fold after dilution pretreatment. The microscopic sample tile height is H. The total number of spherocytes NUMS1 in the blood sample corresponds to a volume V = selected area S1 × H. The number of spherocytes per unit volume of a blood sample = N × the total number of spherocytes NUMS1 in the blood sample / blood sample volume V. TG2: The blood spherocyte parameter is the number of spherocytes per unit volume of a blood sample. The blood sample is concentrated N-fold after concentration pretreatment, and the microscopic sample is tiled at a height of H. The total number of spherocytes NUMS1 in the blood sample corresponds to a volume V = selected area S1 × H. The number of spherocytes per unit volume of a blood sample = the total number of spherocytes NUMS1 in the blood sample / (blood sample volume V × N).

11. The method for detecting blood spherocyte parameters according to claim 8, characterized in that: using an AI recognition algorithm to identify the microscopic sample image, identify normal red blood cells in the blood sample in a selected area S1 of the image, and obtain a total number NUMS2 of normal red blood cells in the blood sample in the selected image; the AI ​​recognition algorithm identifies normal red blood cells in the blood sample based on a feature dataset of normal red blood cells in the blood sample; The ratio of spherocytes to normal red blood cells = NUM1 / NUM2.

12. The method for detecting blood spherocyte parameters according to claim 8, characterized in that: The blood sample pretreatment includes diluting the blood sample with a liquid dye to obtain a microscopic examination sample, with the dilution multiple being 50≤M≤800.

13. The method for detecting blood spherocyte parameters according to claim 12, characterized in that: Include any one of the following technical features: TH1: Complete microscopic examination and photography within T15, when the sample is prepared into a hypotonic suspension; the range of T15 is 10 minutes to 30 minutes; TH2: Complete the preparation and microscopic photography of samples at a temperature greater than 18 degrees Celsius and less than 45 degrees Celsius.

14. A computing and processing device, characterized in that: Used to perform all or part of the blood spheroidal red blood cell parameter detection method according to any one of claims 8 to 13.

15. A data storage device, characterized in that: The device stores all or part of the program codes for executing the method for detecting blood spheroidal red blood cell parameters according to any one of claims 8 to 13.

16. A detection device, characterized in that: Used to run part or all of the blood spherocyte parameter detection method according to any one of claims 8 to 13.

17. A blood spherocyte parameter for evaluating the number of spherocytes in a blood sample, characterized in that: Include any one of the following technical features: TJ1: Number of spherocytes per unit volume of blood sample = number of spherocytes per unit volume of blood sample / volume of blood sample; TJ2: The proportion of spherocytes in a blood sample = the number of spherocytes in the same volume of blood sample / the number of red blood cells.

Images