Peripheral blood cell number filtering and counting system
By designing a peripheral blood cell count filtering statistics system including acquisition unit and analysis unit, the error problem of edge cell recognition and statistics when non-continuously collecting cell microscopy images in the prior art is solved, and a more accurate count of peripheral blood cell count is achieved. , provides more reliable data support for clinical diagnosis.
Patent Information
- Application Number
- CN202411505287.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-28
- Publication Date
- 2025-02-18
AI Technical Summary
When existing automated cell recognition systems discontinuously collect cell microscopy images, it is difficult to accurately identify and count marginal cells, resulting in large errors in cell counting and increasing the risk of misdiagnosis in clinical diagnosis.
A peripheral blood cell count filtration statistics system is designed, including a collection unit and an analysis unit. The acquisition unit is used to collect cell microscope images under a microscope, and the analysis unit includes an identification module, a data capture module and a data analysis module. The identification module identifies and divides peripheral blood cells through deep learning models, the data capture module calculates the area of edge cells, and the data analysis module estimates the number of edge cells based on the area, and then accurately counts the number of peripheral blood cells.
Through the deep learning model of the identification module and the accurate calculation of the data analysis module, it can effectively filter out the repeated recognition edge cells, reduce count errors, improve the accuracy of peripheral blood cell count recognition, and provide a more accurate data basis for clinical diagnosis.
Smart Images
Figure CN119477830A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of cell microscopic image processing, and in particular relates to a peripheral blood cell number filtering and counting system. Background Art
[0002] Peripheral blood cell morphology examination is a key means of screening blood diseases, mainly including the observation of blood cell morphology and quantity, and microscopic examination is the "gold standard" of observation. At present, microscopic observation is to divide the microscopic image into different fields of view, and the automated equipment identifies and counts the number of cells in the field of view, and diagnoses the disease based on this. When the automated cell recognition system identifies edge cells in non-continuously collected cell microscopic images, it uses the "0" or "1" method. Cells that can be identified by the system are directly recorded as one cell, and cells that cannot be identified by the system are directly recorded as no cells. For example: when 10 identifiable cells in the same column are divided into two columns of identifiable edge cells, the above counting method will count 10 cells as 20 cells, which has a large error. This may cause a qualitative change in the clinical diagnosis results and increase the risk of misdiagnosis. Summary of the invention
[0003] In view of the above-mentioned deficiencies in the prior art, the technical problem to be solved by the present invention is to provide a peripheral blood cell number filtering and counting system.
[0004] In order to solve the above technical problems, the present invention provides the following technical solutions:
[0005] A peripheral blood cell number filtering and counting system comprises a collection unit and an analysis unit, wherein the collection unit is used to collect cell microscopic images of stained peripheral blood smears under a high-power microscope; the analysis unit comprises an identification module, a data capture module and a data analysis module, wherein the identification module is used to identify peripheral blood cells from the cell microscopic images, determine the types of peripheral blood cells, and divide each type of peripheral blood cells into complete cells and marginal cells; the data capture module is used to calculate the area of each marginal cell in the cell microscopic image; and the data analysis module is used to estimate the number of marginal cells in each type of peripheral blood cells according to the sum of the areas of the marginal cells in each type of peripheral blood cells, thereby obtaining the total number of peripheral blood cells in each type.
[0006] Furthermore, the high-power microscope is an "80×" or "100×" magnification microscope.
[0007] Furthermore, a peripheral blood cell type database is established in the recognition module, and the peripheral blood cell type database is obtained by training a deep learning model to perform deep learning on various properties of peripheral blood cells. The peripheral blood cell type database includes multiple types of abnormal cells such as white blood cells, red blood cells, and platelets.
[0008] Furthermore, the various properties of the peripheral blood cells include the morphology, size, inclusions, and nuclear-cytoplasmic ratio of the peripheral blood cells.
[0009] Furthermore, the peripheral blood cell type database is obtained by training a convolutional neural network.
[0010] Furthermore, the recognition module evaluates the prediction results through the values of parameters Precision, Recall and loU. The calculation formulas of parameters Precision, Recall and loU are as follows:
[0011]
[0012] Among them, Precision represents the precision rate, that is, the accuracy of the identified positive examples; Recall represents the recall rate; loU represents the intersection-over-union ratio between the evaluation prediction result and the true label; TN represents the number of true positives; FP represents the number of false positives; FN represents the number of false negatives; area of overlap is the intersection area of the two areas of the evaluation prediction result and the true label; area of union is the union area of the two areas of the evaluation prediction result and the true label.
[0013] Furthermore, the formula for calculating the number of marginal cells in one type of peripheral blood cells by the data analysis module is as follows:
[0014]
[0015] In the formula, A represents the index of peripheral blood cell types, N A represents the estimated number of marginal cells in peripheral blood cells of type A; ceil() represents the upward rounding function; S A识别 represents the sum of the areas of the marginal cells identified in the peripheral blood cells of type A; S A标准 Represents the average area of type A peripheral blood cells.
[0016] Furthermore, the recognition module divides intact cells and edge cells in a method of: dividing peripheral blood cells located at the edge of the cell microscopic image into edge cells, and dividing the remaining peripheral blood cells into intact cells.
[0017] Furthermore, the acquisition unit acquires cell microscopic images of the peripheral blood smear under a high-power microscope in a non-continuous acquisition manner.
[0018] Furthermore, it also includes a storage unit, which is used to store the captured images of edge cells, the captured data of edge cells and the analysis results of edge cells.
[0019] In the present invention, the area size of the identified marginal cells is compared with the standard area size of the peripheral blood cells of this type to determine the number of marginal cells, and then the total number of peripheral blood cells of this type is determined. This method filters out the repeatedly identified marginal cells, greatly reduces the probability of repeated cells, reduces the peripheral blood cell counting error, increases the accuracy of identifying the total number of peripheral blood cells, and can provide a more accurate data basis for clinical diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:
[0021] Figure 1 It is a structural block diagram of an embodiment of a peripheral blood cell number filtering and counting system of the present invention.
[0022] Figures 2(a) and 2(b) are schematic diagrams of edge cells identified in cell microscopic images.
[0023] The accompanying drawings in the specification are as follows:
[0024] Acquisition unit-100; cell microscopic image-110; edge cell-111; analysis unit-200; identification module-210; data capture module-220; data analysis module-230; storage unit-300. DETAILED DESCRIPTION
[0025] The following describes the implementation methods of the present invention through specific examples. The illustrations provided in the following embodiments are only used to schematically illustrate the basic concept of the present invention. The following embodiments and features in the embodiments may be combined with each other without conflict.
[0026] See also Figure 1 , Figure 1 The structural block diagram of an embodiment of the peripheral blood cell number filtering and counting system of the present invention. The peripheral blood cell number filtering and counting system of this embodiment includes a collection unit 100 and an analysis unit 200. The collection unit 100 is used to collect cell microscopic images 110 under a high-power microscope of a peripheral blood smear stained with Wright-Giemsa; the high-power microscope is an "80×" microscope or a "100×" microscope. In this embodiment, the collection unit 100 collects cell microscopic images 110 under a high-power microscope of a peripheral blood smear in a non-continuous collection manner.
[0027] The analysis unit 200 includes an identification module 210, a data capture module 220 and a data analysis module 230. The identification module 210 is used to identify peripheral blood cells from the cell microscopic image 110, determine the type of peripheral blood cells, and divide each type of peripheral blood cells into intact cells and edge cells 111. In this embodiment, the algorithm of the identification module 210 adopts the YOLOv5s algorithm. The method for the identification module 210 to divide intact cells and edge cells 111 is: the peripheral blood cells located at the edge of the cell microscopic image 110 are divided into edge cells 111, and the remaining peripheral blood cells are divided into intact cells.
[0028] A peripheral blood cell type database is established in the identification module 210, and the peripheral blood cell type database is obtained by training a deep learning model to perform deep learning on the morphology, size, inclusions, nuclear-cytoplasmic ratio and other properties of peripheral blood cells. In this embodiment, the peripheral blood cell type database is obtained by training a convolutional neural network, and the peripheral blood cell type database includes white blood cells, red blood cells, platelets, and a total of 47 types of abnormal cells.
[0029] In this embodiment, the recognition module 210 evaluates the prediction result through the values of parameters Precision, Recall and loU. The calculation formulas of parameters Precision, Recall and loU are as follows:
[0030]
[0031]
[0032] Among them, Precision represents the precision rate, that is, the accuracy of the identified positive examples. In this embodiment, it is used to indicate how many correct targets are contained in the target box predicted by the model. Recall represents the recall rate, that is, the proportion of positive examples that are identified. In this embodiment, it is used to indicate how many targets are recognized by the model. loU represents the intersection-over-union ratio between the evaluation prediction result and the true label. TN represents the number of true positives, that is, the number of examples that are actually positive and are correctly judged as positive examples; FP represents the number of false positives, that is, the number of examples that are actually negative but are mistakenly judged as positive examples; FN represents the number of false negatives, that is, the number of examples that are actually positive but are mistakenly judged as negative examples. Area of overlap is the intersection area of the two areas of the evaluation prediction result and the true label; area of union is the union area of the two areas of the evaluation prediction result and the true label.
[0033] In this embodiment, TN, TP, and FN are used to define the precision and recall rates, and the recognition rate of peripheral blood cells is jointly evaluated by the recall rate and precision rate; the IoU parameter is used to evaluate the similarity between the predicted result and the true label; the higher the IoU value, the more accurate the predicted result.
[0034] The data capture module 220 is used to calculate the area of each edge cell 111 in the cell microscopic image 110. In order to calculate the average area of each type of peripheral blood cell, in this embodiment, the data capture module 220 also calculates the area of each complete peripheral blood cell in the cell microscopic image 110.
[0035] The data analysis module 230 is used to estimate the number of marginal cells 111 in each type of peripheral blood cells according to the sum of the areas of the marginal cells 111 in each type of peripheral blood cells, and then obtain the total number of peripheral blood cells in each type. The formula for calculating the number of marginal cells 111 in one type of peripheral blood cells by the data analysis module 230 is as follows:
[0036]
[0037] In the formula, A represents the index of peripheral blood cell types, N A represents the estimated number of marginal cells 111 in peripheral blood cells of type A; ceil() represents the upward rounding function. S A识别 represents the sum of the areas of the marginal cells 111 identified in the peripheral blood cells of type A; S A标准 Represents the average area of type A peripheral blood cells.
[0038] S A识别 The calculation formula is as follows:
[0039] S A识别 =S A1 +S A2 +…+S An (2)
[0040] Where n represents the number of marginal cells 111 in peripheral blood cells of type A, S A1 , S A2 ,…,S An They respectively represent the areas of n border cells 111 in type A peripheral blood cells.
[0041] The data analysis module 230 calculates the average of the intact cell areas of each type of peripheral blood cells in the peripheral blood cell type database as the average area of the peripheral blood cells of that type, and the average area S of the peripheral blood cells of type A is A标准 The calculation formula is as follows:
[0042]
[0043] X represents the number of intact cells in peripheral blood cells of type A, d1, d2, ..., d X They represent the diameters of X intact cells in type A peripheral blood cells.
[0044] The method for the identification module 210 to determine whether a peripheral blood cell is a complete cell is to calculate the ratio of the area of the identified peripheral blood cell to the average area of this type of peripheral blood cells. If it is close to 1, the peripheral blood cell is determined to be a complete cell, otherwise, the peripheral blood cell is determined to be a marginal cell 111.
[0045] In order to store the collected image information and analyzed data, the peripheral blood cell number filtering and counting system may further include a storage unit 300 , which is used to store the captured images of the marginal cells 111 , the captured data of the marginal cells 111 , and the analysis results of the marginal cells 111 .
[0046] The working principle of this embodiment is as follows:
[0047] Please continue reading Figure 1 When the peripheral blood cells identified by the recognition module 210 are intact cells, their area is substantially the same as the average area of such peripheral blood cells recorded in the database, that is, Y Ai完整 =S Ai完整 / S A标准 =1; where Y Ai完整 represents the area coefficient of the i-th complete cell of type A; S Ai完整 represents the area of the ith complete cell of type A. When the peripheral blood cell identified by the recognition module 210 is a border cell 111, since the border cell 111 in the cell microscopic image 110 is generally not a complete cell and has a part missing, its area is smaller than the average area of the peripheral blood cells of this type recorded in the database, that is: Aj边缘 =S Aj边缘 / S A标准 <1; where Y Aj边缘 represents the area coefficient of the jth A-type edge cell 111; S Aj边缘 represents the area of the jth A-type edge cell 111. In the case where a complete cell is misidentified as an edge cell 111 because it is close to the edge of the cell microscopic image 110, its area is similar to S A标准 Therefore, it will be judged as one cell in the subsequent area calculation and will not affect the counting accuracy.
[0048] Assume that a complete peripheral blood cell of type A is divided into two parts located at the edges of two cell microscopic images 110 during image acquisition because it is located at the edge of the imaging field, and thus is identified as two edge cells 111. Since the two edge cells 111 are two parts of a peripheral blood cell of type A, the sum of their areas is just equal to the area of a complete cell of type A (i.e., S A标准 ), therefore, according to formula (1), only one type A peripheral blood cell will be counted instead of two, thus avoiding counting errors.
[0049] Assume that there are 10 type A peripheral blood cells located at the edge of the imaging field of view and are divided into two parts located at the edge of two cell microscopic images 110, respectively, and thus are identified as type A edge cells 111 in the two cell microscopic images 110. The type A edge cells 111 in one of the cell microscopic images 110 are shown in FIG2(a), and the type A edge cells 111 in the other cell microscopic image 110 are shown in FIG2(b). If the normal counting method is used, the type A edge cells 111 in the above two cell microscopic images 110 will be counted as 20 type A peripheral blood cells, which obviously has a large counting error. According to the method of this embodiment, according to formula (1), it can be obtained that:
[0050]
[0051] Among them, S a1 , S a2 ,…,S a10 They represent the areas of 10 A-type edge cells 111 in FIG2(a); S a11 , S a12 ,…,S a20 They represent the areas of 10 A-type edge cells 111 in Fig. 2(b) respectively. The counting result of the method of this embodiment is 10, which is the same as the actual situation, thus avoiding errors.
[0052] Of course, the above is just an example of continuous acquisition. Based on the same principle, the above method can also reduce the counting error in the discontinuous acquisition mode. A1边缘 and Y A2边缘 Meet Y A1边缘 +Y A2边缘 =1, that is, the sum of the areas of its two A-type marginal cells 111 is equal to the standard area of A-type peripheral blood cells. It can be considered that the above two A-type marginal cells A1 and A2 can be spliced into a complete A-type peripheral blood cell.
[0053] In this embodiment, the area size of the identified edge cells 111 is compared with the standard area size of the peripheral blood cells of this type to determine the number of edge cells, and then the total number of peripheral blood cells of this type is determined. This method filters out the repeatedly identified edge cells 111, greatly reduces the probability of duplicate cells, reduces the error in peripheral blood cell counting, increases the accuracy of identifying the total number of peripheral blood cells, and can provide a more accurate data basis for clinical diagnosis.
[0054] The above embodiments only express the preferred implementation modes of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for those of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.
Claims
1. A peripheral blood cell number filtering and counting system, characterized in that: It comprises a collection unit and an analysis unit, wherein the collection unit is used to collect cell microscopic images of stained peripheral blood smears under a high-power microscope; the analysis unit comprises an identification module, a data capture module and a data analysis module, wherein the identification module is used to identify peripheral blood cells from the cell microscopic images, determine the types of peripheral blood cells, and divide each type of peripheral blood cells into complete cells and marginal cells; the data capture module is used to calculate the area of each marginal cell in the cell microscopic image; and the data analysis module is used to estimate the number of marginal cells in each type of peripheral blood cells according to the sum of the areas of the marginal cells in each type of peripheral blood cells, thereby obtaining the total number of peripheral blood cells in each type.
2. The peripheral blood cell number filtering and counting system according to claim 1, characterized in that: The high-power microscope is an "80×" or "100×" microscope.
3. The peripheral blood cell number filtering and counting system according to claim 1, characterized in that: A peripheral blood cell type database is established in the recognition module. The peripheral blood cell type database is obtained by training a deep learning model to perform deep learning on multiple properties of peripheral blood cells. The peripheral blood cell type database includes multiple types of abnormal cells such as white blood cells, red blood cells, and platelets.
4. The peripheral blood cell number filtering and counting system as claimed in claim 3, characterized in that: The various properties of the peripheral blood cells include the morphology, size, contents, and nuclear-cytoplasmic ratio of the peripheral blood cells.
5. The peripheral blood cell number filtering and counting system as claimed in claim 3, characterized in that: The peripheral blood cell type database is obtained by training a convolutional neural network.
6. The peripheral blood cell number filtering and counting system according to claim 1, characterized in that: The recognition module evaluates the prediction results through the values of parameters Precision, Recall and loU. The calculation formulas of parameters Precision, Recall and loU are as follows: Among them, Precision is the precision rate, that is, the accuracy of the identified positive examples; Recall is the recall rate; loU is the intersection-over-union ratio between the evaluation prediction result and the true label; TN is the number of true positives; FP is the number of false positives; FN is the number of false negatives; area of overlap is the intersection area of the two areas of the evaluation prediction result and the true label; area of union is the union area of the two areas of the evaluation prediction result and the true label.
7. The peripheral blood cell number filtering and counting system according to claim 1, characterized in that: The formula for calculating the number of marginal cells in one type of peripheral blood cells by the data analysis module is as follows: In the formula, A represents the index of peripheral blood cell types, N A represents the estimated number of marginal cells in peripheral blood cells of type A; ceil() represents the upward rounding function; S A识别 represents the sum of the areas of the marginal cells identified in the peripheral blood cells of type A; S A标准 Represents the average area of type A peripheral blood cells.
8. The peripheral blood cell number filtering and counting system according to claim 7, characterized in that: The method for the recognition module to divide intact cells and edge cells is: the peripheral blood cells located at the edge of the cell microscopic image are divided into edge cells, and the remaining peripheral blood cells are divided into intact cells.
9. The peripheral blood cell number filtering and counting system according to any one of claims 1 to 8, characterized in that: The acquisition unit acquires cell microscopic images of the peripheral blood smear under a high-power microscope in a non-continuous acquisition manner.
10. The peripheral blood cell number filtering and counting system according to any one of claims 1 to 8, characterized in that: The device also includes a storage unit, which is used to store the captured images of the edge cells, the captured data of the edge cells and the analysis results of the edge cells.