Search

CN-121999482-A - AI training and detection method for identifying cell-derived species

CN121999482ACN 121999482 ACN121999482 ACN 121999482ACN-121999482-A

Abstract

In the AI training and detecting method for identifying the cell source species, a microscopic examination sample image A comprises a cell type A image of the species A, a microscopic examination sample image B comprises a cell type B image of the species B, the cell type A image is marked to obtain a cell type marking image A of the species A, the cell type B image is marked to obtain a cell type marking image B of the species B, AI training is carried out by using the cell marking image A and the cell marking image B to obtain an AI species cell characteristic data set, and the AI species cell characteristic data set is combined with a corresponding AI algorithm module to have the capability of identifying and distinguishing the cells of the species A and the cells of the species B. And the AI species cell characteristic data set is obtained by training the marked cell pictures, and the cell pictures are derived from two or more species.

Inventors

  • FANG XIANGFEI
  • WANG ZHIPING
  • LU JINGJIANG

Assignees

  • 深圳安侣医学科技有限公司

Dates

Publication Date
20260508
Application Date
20241108

Claims (13)

  1. 1. An AI training method for identifying a cell-derived species, characterized in that, The microscopic sample image a includes a cell type a image of species a; the microscopic sample image B comprises a cell type B image of the species B; Labeling the cell type A image to obtain a cell type labeling image A of the species A; labeling the cell type B image to obtain a cell type labeling image B of the species B; carrying out AI training by using the cell labeling image A and the cell labeling image B to obtain an AI species cell characteristic data set; the AI species cell characteristic dataset, in combination with the corresponding AI algorithm module, has the ability to identify cells of species a from cells of species B.
  2. 2. The AI training method as defined in claim 1, wherein the microscopic sample image A or the microscopic sample image B is derived from detection data, The detection data are detection images of microscopic imaging of a suspension microscopic sample.
  3. 3. The AI training method of claim 1, wherein the sample comprising cells of the selected species is pre-treated to obtain a microscopic sample; tiling the microscopic sample, and suspending the species cells to the bottom; shooting a tiled microscopic examination sample to obtain a microscopic examination sample image; Identifying and labeling selected species cells in the microscopic examination sample image to obtain labeling pictures, and performing AI training by using the labeling pictures to obtain an AI species cell characteristic data set; the AI species cell characteristic dataset comprises sample cell characteristics of the selected species; The AI species cell characteristic dataset, in combination with the corresponding AI algorithm module, has the ability to identify cells of the selected species.
  4. 4. The AI training method for identifying a cell-derived species of claim 1, comprising any one of the following technical features: Characterized by TC1, cell type A is nucleated red blood cell of fish, and cell type B is mammalian non-nucleated red blood cell; characterized by TC2, cell type A is avian nucleated red blood cells and cell type B is mammalian non-nucleated red blood cells; characterized by TC3, cell type A is reptile nucleated red blood cell and cell type B is mammalian non-nucleated red blood cell; Characterized in that TC4, cell type A is mammalian nucleated red blood cells and cell type B is mammalian non-nucleated red blood cells.
  5. 5. The AI training method for identifying a cell-derived species of claim 1, comprising any one of the following technical features: the characteristic TD1 is that the cell type A is the white blood cell of fish, and the cell type B is the mammalian white blood cell; characterized by TD2, cell type A is avian leucocyte and cell type B is mammalian leucocyte; Characterized in that the cell type A is a reptile leukocyte and the cell type B is a mammalian leukocyte.
  6. 6. The AI training method of claim 5, comprising any of the following features: Characterized in that the cell type A is thrombocyte or thrombocyte cluster of fish, and the cell type B is mammalian platelet or agglutinated platelet; Characterized in that the cell type A is thrombocyte or thrombocyte cluster of birds and the cell type B is mammalian platelet or agglutinated platelet; Characterized in that the cell type A is thrombocyte or thrombocyte cluster of reptile and the cell type B is mammalian platelet or agglutinated platelet.
  7. 7. A detection method for identifying cell-derived species, characterized in that, Preprocessing a blood sample to obtain a microscopic sample; Tiling the microscopic sample, and suspending the cells to the bottom; shooting a tiled microscopic examination sample to obtain a microscopic examination sample image; Identifying the microscopic sample image by using an AI identification algorithm to identify the cell source species in the sample in the selected area S1 of the image, wherein the AI identification algorithm identifies the cell source species in the sample according to an AI species cell characteristic data set; the AI species cell characteristic data set is obtained by training labeled cell pictures, wherein the cell pictures are derived from two or more species.
  8. 8. The method for detecting a cell-derived species according to claim 7, wherein, Including any two or more of birds, reptiles, fish, mammals.
  9. 9. The method for detecting a cell-derived species according to claim 8, wherein, The cells include any one of white blood cells, red blood cells, thrombocytes or thrombocytes clusters, mammalian platelets or aggregated platelets.
  10. 10. The method for detecting the presence of a cell-derived species according to claim 8, comprising any one of the following technical features: TF1, identifying the microscopic sample image by using an AI identification algorithm, and identifying the total number NUMA of corresponding cells of the species A in a sample in a selected area S1 of the image, the total number NUMB of corresponding cells of the species B, wherein the volume of the microscopic sample corresponding to the selected area S1 is V1, the unit volume content of the corresponding cells of the species A in the microscopic sample is NUMA/V1, and the unit volume content of the corresponding cells of the species B in the microscopic sample is NUMB/V1; TF2, identifying the microscopic sample image by using an AI identification algorithm, and identifying the total number NUMA of the corresponding cells of the species A, the total number NUMB of the corresponding cells of the species B and the ratio of the corresponding cells of the species A to the corresponding cells of the species B=NUMA/NUMB in the sample in the selected area S1 of the image.
  11. 11. A computing processing device is characterized in that, The method is characterized by comprising any one of the following technical characteristics: TH1, all or part of a method for operating any one of claims 1 to 10; TH2, the memory of the computing device comprises all or part of the method of any one of claims 1 to 10.
  12. 12. A data storage device, characterized in that, The method is characterized by comprising any one of the following technical characteristics: TG1 storing all or part of the program code for performing the method of any of claims 1 to 10; TG2 data storing all or part of the method according to any of claims 1 to 10.
  13. 13. A detection device is characterized in that, Part or all of a method for performing any one of claims 1 to 10 Or store all or part of the data of the method of any one of claims 1 to 10.

Description

AI training and detection method for identifying cell-derived species Technical Field The application belongs to the technical field of analysis of visible components based on microscopic amplified images, and particularly relates to an AI training and detecting method for identifying cell source species, and a computing processing and storing device. Background Blood cell characteristics vary from species to species. For example, erythrocytes are the most abundant type of blood cells in blood, and most of mammalian erythrocytes in vertebrates are in the shape of a round cake with a concave center, and human erythrocytes are also in the shape of a round cake with a convex edge and a concave center. When an abnormality occurs in erythrocytes, it is often an indicator of some diseases. However, for some species of different evolutionary hierarchy, the morphology and characteristics of the erythrocytes are different, e.g., most non-mammalian erythrocytes are nucleated erythrocytes. The applicant has proposed a series of Chinese patents, such as 1. CN2022104799126, "microscopic image acquisition device quick focusing method and microscopic image detection method"; 2. CN2023100423151, "sample imaging analysis system and method"; 3. CN2023116834572, a sample white blood cell identification AI training method, a calculation processing device and a storage device, and a brand new AI identification technical scheme is used for measuring the sample and the content of the tangible components in the sample. How to perform different types of species identification based on the above-described techniques is a technical problem to be solved. In the prior art, animal types are known for detection, and the animal types need to be input for analysis, so that the device has the technical problem of automatically identifying the source of sample species. In the field of food safety, for example, daily detection or sampling detection is required for food source species sold, the traditional detection technology is difficult to realize, and long detection time is required. In the scientific research or reconnaissance field, many scenes need to identify and identify the types of the substances according to the blood samples of small samples, and the prior art does not have an efficient means, so that the DNA detection time is long, the cost is high, and the efficiency is low. Disclosure of Invention In the application, the inventor proposes to use the cell type labeling images of two different species to perform AI training, and the obtained AI species cell characteristic data set is combined with the corresponding AI algorithm module, so that the AI species cell characteristic data set has the capability of identifying and distinguishing the cells of the species A from the cells of the species B, and the different species identification is completed by using advanced AI algorithm, so that the species identification based on the cell characteristics is possible, and the species identification is more efficient and accurate. The technical scheme for solving the technical problems is that the AI training method for identifying the cell source species is characterized in that an microscopic sample image A comprises a cell type A image of a species A, a microscopic sample image B comprises a cell type B image of a species B, the cell type A image is marked to obtain a cell type marked image A of the species A, the cell type B image is marked to obtain a cell type marked image B of the species B, AI training is carried out by using the cell marked image A and the cell marked image B to obtain an AI species cell characteristic data set, and the AI species cell characteristic data set is combined with a corresponding AI algorithm module to have the capability of identifying and distinguishing cells of the species A and cells of the species B. The microscopic sample image A or the microscopic sample image B is derived from detection data, and the detection data is a detection image of microscopic imaging of the suspension microscopic sample. The method comprises the steps of preprocessing a sample containing selected species cells to obtain a microscopic examination sample, tiling the microscopic examination sample, suspending the species cells to sink, photographing the tiled microscopic examination sample to obtain a microscopic examination sample image, identifying and labeling the selected species cells in the microscopic examination sample image to obtain a labeling picture, carrying out AI training on the labeling picture to obtain an AI species cell characteristic data set, wherein the AI species cell characteristic data set comprises the selected species sample cell characteristics, and the AI species cell characteristic data set is combined with a corresponding AI algorithm module to have the capability of identifying the selected species cells. The AI training method for identifying the cell-derived species is characterized by