Chromosome image analysis method, system, computer device and storage medium

Abstract

The present invention discloses a chromosome image analysis method, system, computer device, and storage medium, comprising the following steps: S01: constructing a primary classification input set; S02: constructing a secondary classification input set using one of two methods; Method 1: using a neural network instance segmentation model to segment chromatid outlines and obtain a probability matrix for the chromatids; the probability matrix and image feature information constitute the secondary classification input set; Method 2: obtaining an outline thumbnail based on all chromatid outlines obtained in Method 1, and inferring the probability of the chromatid thumbnail belonging to 24 chromosome categories based on a classifier; the probability matrix and image feature information constitute the secondary classification input set; S03: obtaining final chromatid category information using a pairing model. This sequential two-step classification and identification of chromatids reduces the impact of chromatid crossing, overlap, and impurity interference in microscopic images on the final classification accuracy, significantly improving chromosome classification accuracy.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a chromosome image analysis method, system, computer equipment and storage medium. Background Art

[0002] Human somatic cells have a total of 23 pairs of chromosomes, including 22 pairs of autosomes (1-22) and one pair of sex chromosomes (XX in females and XY in males), for a total of 24 types. Chromosome image recognition is an important part of chromosome karyotype analysis.

[0003] In the existing technology, in the "A chromosome image recognition method, device, computer equipment and storage medium" disclosed in application number 202310058038.3, the chromosome microscopic image is first subjected to chromosome instance segmentation processing to obtain multiple chromatid images, and then classified using a neural network chromosome classifier to obtain chromatid categories.

[0004] Patent application number 202111547786.5 discloses a "chromosome karyotype analysis method and system based on deep learning" which states that "a deep learning-based instance segmentation model is used to segment and classify clustered images to obtain single chromosome images with classification numbers and contour position information." "S9. According to the classification number, polarity number, structural variation number, and centromere position of the single chromosome image, the single chromosomes are arranged to obtain a standard karyotype map, and the chromosome numbers with abnormal numbers and single chromosomes with abnormal structures are marked on the karyotype map to complete the chromosome karyotype analysis." That is, the chromosome instance segmentation model is used to segment and classify chromosomes in microscopic images.

[0005] The two aforementioned patents utilize a deep learning-based neural network model to derive the probability of chromosome classification into 24 categories, selecting the category with the highest probability as the final chromatid category. In actual microscopic images, factors such as chromatid crossing and overlapping, as well as interference from image impurities, can affect the accuracy of chromatid classification probabilities. Directly selecting the category with the highest probability as the corresponding chromatid category will yield illogical chromosome classification results and reduce chromosome classification accuracy. For cases where the probability of two chromatids being classified into the same category is high and the probability values ​​are not significantly different, further chromosome pairing is necessary by combining features such as banding consistency, area consistency, and size consistency across multiple chromatids, along with the natural laws governing the deterministic distribution of human chromosome categories and numbers.

[0006] Patent application number 202011352831.7 discloses "a chromosome karyotype analysis system" that "designs an identification algorithm and a correction algorithm to identify and pair the extracted chromosomes to generate a karyotype map." Its correction algorithm uses a probability-based chromosome correction algorithm, and its correction algorithm is only for the case where a chromosome metaphase microscopic image contains 46 chromatids.

[0007] In actual engineering applications, it is very common and objective that in the images of chromosome metaphase division phases in the microscopic images taken by the camera, there are fewer than 46 chromatids (for example, some chromosomes are scattered far in space and are not captured. Or there are missing chromosomes in the division phase), or there are more than 46 chromatids (for example, multiple chromosome division phases are captured in the same image. Or only one division phase is captured, but there is an extra chromosome of a certain type). In most of the multiple microscopic images on the same test slide of the same patient, the number of chromosomes is not equal to 46, which may be a manifestation of abnormal chromosome number, which has important diagnostic value in clinical practice. Therefore, the identification and pairing of microscopic images of chromosome metaphase division with a chromosome number not equal to 46 chromosomes is a technical problem that needs to be solved and has practical application value.

[0008] Furthermore, patent application number 202011352831.7 uses a traditional algorithm ensemble learning model to determine chromosome classification probabilities. Traditional algorithms are inferior to deep learning algorithms in extracting chromosome features and achieving classification results. Furthermore, their correction algorithms fail to account for the similarities in height and area between chromatids of the same type. These factors reduce the effectiveness of chromosome pairing. Designing a chromosome classification and pairing method based on these two perspectives can improve the accuracy of chromosome classification and produce more reasonable karyotype results.

[0009] The above information disclosed in this Background section is only for enhancement of understanding of the background of the application and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0010] The purpose of the present invention is to solve the technical problems of low chromosome classification accuracy in the prior art and chromosome identification and pairing when the number of chromosomes in the microscopic image is not equal to 46, and propose a chromosome image analysis method, system, computer equipment and storage medium.

[0011] In a first aspect, a technical solution provided in an embodiment of the present invention is: a chromosome image analysis method, comprising the following steps:

[0012] S01: Take microscopic images of chromosomes and construct a classification input set;

[0013] S02: Based on the primary classification input set, a secondary classification input set is constructed using one of the following two methods;

[0014] Method 1: Use a deep learning-based neural network instance segmentation model to segment and classify the microscopic images in the primary classification input set, and obtain the probabilities of all chromatid outlines and corresponding chromatids belonging to the 24 chromosome categories (1, 2, 3, .. 21, 22, X, Y) in the microscopic images; calculate the image feature information of each chromatid based on the chromatid outlines; the probabilities of all chromatids belonging to the 24 chromosome categories and the image feature information of all chromatids constitute the secondary classification input set;

[0015] Method 2: Based on all chromatid outlines obtained in Method 1, a small chromatid outline image is cut out from the microscopic image and the chromatid image is vertically positioned at 90 degrees. A deep neural network-based classifier infers the probability that the chromatid image belongs to one of the 24 chromosome categories. The probabilities of all chromatids belonging to the 24 chromosome categories output by the classifier, together with the image feature information of all chromatids, form the secondary classification input set.

[0016] S03: Order Indicates rounding up to an integer; input the secondary classification input set, call the assign_num chromatid pairing model, correct the chromatid category, and obtain the final chromatid category information.

[0017] Preferably, the microscopic image is a microscopic image of chromosome metaphase.

[0018] Preferably, the deep learning-based neural network instance segmentation model is a deep learning-based neural network model obtained by using a microscopic image dataset of chromosome metaphase to annotate the chromosomes in the image and train it, which can segment the outlines of different chromatids and classify different chromatid categories.

[0019] Preferably, the deep learning-based neural network classifier refers to a deep learning-based neural network model that uses a chromatid image dataset, labels different chromatid images with their categories, and is trained to be able to classify chromatid categories.

[0020] Preferably, the image feature information of the microscopic image includes the contour area and height information of the chromatid.

[0021] Preferably, the chromatid image pairing model has the function of pairing the paired chromatids, as well as the function of reclassifying and correcting the chromatid categories. This patent uniformly adopts the pairing model for identification.

[0022] Preferably, the objective function F of the chromatid pairing model is:

[0023]

[0024] Among them, score i,j is the probability that the i-th chromatid is classified as the j-th category; the variable w i,j Indicates whether the i-th chromatid is finally classified into category j; when w i,j = 0, indicating that the i-th chromatid is not classified as category j, w i,j =1, indicating that the i-th chromatid is classified as category j.

[0025] Preferably, the chromatid pairing model further includes the following constraints:

[0026] Chromatid classification value constraint: w i,j =0,orw i,j =1;

[0027] Constraint on the total number of chromatids participating in pairing each time: Chromatid classification category constraints: Constraints on the number of autosomes in a mitotic phase: Constraint on the number of sex chromosomes in a mitotic phase: Constraints on the number of Y chromosomes in a mitotic phase: Where N represents the total number of chromatids that remain unsorted and paired during the pairing process.

[0028] Preferably, the chromatid pairing model further includes image feature information constraints:

[0029]

[0030] Among them, area i represents the outline area of ​​the i-th chromatid; h i Represents the height information of the i-th chromatid.

[0031] In a second aspect, a technical solution further provided in an embodiment of the present invention is: a chromosome image analysis system, comprising: a chromosome microscopic image acquisition module, a primary segmentation and classification module, and a chromatid pairing module;

[0032] The chromosome microscopic image acquisition module is used to capture microscopic images of chromatids and construct a classification input set;

[0033] The primary segmentation and classification module performs a segmentation and classification on each microscopic image in the primary classification set based on the instance segmentation model to obtain the outline of the chromatid, the probability that the chromatid belongs to the chromosome category, and the characteristic information of the chromatid; optionally, the probability that the chromatid small image belongs to the 24 chromosome categories can be obtained in the following way: based on all the chromatid outlines obtained by the instance segmentation model, the chromatid outline small image is cut out from the microscopic image, and the chromatid small image is erected at 90 degrees. The classifier based on the deep neural network infers the probability that the chromatid small image belongs to the 24 chromosome categories. The above-mentioned probability of the chromatid small image belonging to the 24 chromosome categories and the characteristic information of the chromatid constitute the secondary classification input set.

[0034] Based on the secondary classification input set, the chromatid pairing module is used to perform secondary classification on the secondary classification input set to obtain the chromatid thumbnail classification results.

[0035] In a third aspect, a technical solution provided in an embodiment of the present invention is: a computer device comprising at least one processor and at least one memory, wherein the at least one memory is used to store at least one computer program; when the at least one computer program is executed by the at least one processor, the at least one processor implements the above-mentioned chromosome image analysis method.

[0036] In a fourth aspect, a technical solution provided in an embodiment of the present invention is: a storage medium, wherein the storage medium stores processor executable instructions, and the executable instructions are used to execute the above-mentioned chromosome image analysis method when executed by the processor.

[0037] Beneficial effects of the present invention: A chromosome image analysis method, system, computer device, and storage medium designed by the present invention perform a primary chromatid identification based on the chromatid outlines, classification probabilities, and chromatid image feature information output by a deep learning neural network model. Chromatids are then secondary classified using a chromatid pairing model. Specifically, the deep learning-based neural network model can better extract chromosome microscopic image features, resulting in higher chromatid classification accuracy. Chromatid classification probabilities output by the deep learning neural network model are used for preliminary chromatid identification, followed by secondary classification using the chromatid pairing model. These two classification methods, as a whole, mitigate the impact of chromatid crossing, overlap, and interference from impurities in the microscopic image on the final classification accuracy. Furthermore, the chromatid pairing model considers the image features of similar heights and outline areas between chromatids of the same category, as well as the chromosome category and number characteristics, eliminating chromatid pairings with excessively different heights and outline areas, and eliminating classification results that do not conform to the human chromosome category and number patterns, thereby improving chromosome pairing accuracy. The design of the chromatid pairing model itself includes the processing of chromosome pairing when the number of chromosomes is not equal to 46. The present invention can solve the technical problem of chromosome pairing when the number of chromosomes in the metaphase microscopic image of chromosome division is not equal to 46, which has practical application value.

[0038] The above content of the invention is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Other features, objects, and advantages of the present invention will become more apparent upon reading the detailed description of the non-limiting embodiments made with reference to the following drawings. The drawings are for the purpose of illustrating preferred embodiments only and are not to be construed as limiting the present invention. Like reference characters are used throughout the drawings to designate like parts.

[0040] Figure 1 The figure is a flow chart of an embodiment of a chromosome image analysis method of the present invention.

[0041] Figure 2 This is a flow chart of another embodiment of a chromosome image analysis method of the present invention.

[0042] Figure 3 This is a schematic structural diagram of a chromosome image analysis system of the present invention. DETAILED DESCRIPTION

[0043] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific implementation method described herein is only an optimal embodiment of the present invention, which is only used to explain the present invention and does not limit the scope of protection of the present invention. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0044] Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flow charts. Although the flow charts describe the operations (or steps) as sequential processes, many of the operations (or steps) therein can be performed in parallel, concurrently, or simultaneously. In addition, the order of the operations can be rearranged. The process can be terminated when its operations are completed, but can also have additional steps not included in the figures; the process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0045] Example 1

[0046] like Figure 1 As shown, the first aspect of this embodiment provides a chromosome image analysis method, comprising the following steps:

[0047] S1. Obtain microscopic images of chromatids to construct a classification input set.

[0048] Specifically, the microscopic image is a microscopic image of chromosome metaphase.

[0049] In practical applications, the general method for obtaining a microscopic image of chromosome metaphase is: first locate and screen the target image with a good division phase under a low-power microscope, and then switch to a high-power microscope to capture a microscopic image of chromosome metaphase.

[0050] S2. Segment and classify the microscopic images in the primary classification input set using a deep learning-based neural network instance segmentation model to obtain the probabilities of all chromatid outlines and corresponding chromatids belonging to the 24 chromosome categories (1, 2, 3, .. 21, 22, X, Y). Image feature information for each chromatid can be calculated based on the chromatid outlines. The probabilities of all chromatids belonging to the 24 chromosome categories, together with the image feature information for all chromatids, constitute the secondary classification input set.

[0051] Specifically, the deep learning-based neural network instance segmentation model is a deep learning-based neural network model that uses a microscopic image dataset of chromosome metaphase to annotate the chromosomes in the image and is trained to segment the outlines of different chromatids and classify different chromatid categories.

[0052] Specifically, the image feature information of the microscopic image includes the contour area and height information of the chromatid.

[0053] S3, order Indicates rounding up to an integer; input the secondary classification input set, call the assign_num chromatid pairing model, correct the chromatid category, and obtain the final chromatid category information.

[0054] Specifically, the objective function F of the chromatid pairing model is:

[0055]

[0056] The objective function F represents the solution that can maximize the probability of classification of all chromatid categories, where score i,j is the probability that the i-th chromatid is classified as the j-th category; the variable w i,j Indicates whether the i-th chromatid is finally classified into category j; when w i,j = 0, indicating that the i-th chromatid is not classified as category j, w i,j =1, indicating that the i-th chromatid is classified as category j.

[0057] Specifically, the chromatid pairing model also includes the following constraints:

[0058] Chromatid classification value constraint: w i,j =0,orw i,j =1;

[0059] Constraint on the total number of chromatids participating in pairing each time: Chromatid classification category constraints: Constraints on the number of autosomes in a mitotic phase:

[0060] Constraint on the number of sex chromosomes in a mitotic phase:

[0061] Constraints on the number of Y chromosomes in a mitotic phase:

[0062] Where N represents the total number of chromatids that remain unsorted and paired during the pairing process.

[0063] It is understandable that the chromatid classification value constraint is: i,j It can only take one of the two values ​​0 or 1.

[0064] It is understandable that the total number of chromatids involved in each classification is constrained: in one microscopic image, chromosomes are classified multiple times, and the maximum number of chromatids involved in each chromosome classification is 46. This constraint is intended to limit the total number of chromatids, making it easier to use background knowledge about the number of human chromosomes.

[0065] It can be understood that the chromatid classification category constraint is: for each chromosome, at most one category can be assigned.

[0066] It can be understood that there is a constraint on the number of autosomes in a mitotic phase: for autosomes, in a mitotic phase, there are at most two chromosomes of each category.

[0067] It can be understood that there is a constraint on the number of sex chromosomes in a cleavage phase: for sex chromosomes X and Y, there are at most two sex chromosomes in a cleavage phase.

[0068] It can be understood that the number of Y chromosomes in one division phase is constrained: for the sex chromosome Y, there can be at most one Y chromosome in one division phase.

[0069] Specifically, the chromatid pairing model also includes image feature information constraints:

[0070]

[0071] Among them, area i represents the outline area of ​​the i-th chromatid; h i Represents the height information of the i-th chromatid.

[0072] It can be understood that the image feature information ratio constraint represents that when the ratio of the contour areas between chromatids is less than or equal to 0.8 and the ratio of the heights between chromatids is less than or equal to 0.7, the two chromosomes cannot be simultaneously classified as the same chromosome category.

[0073] As a preferred embodiment, for the probability that all chromatids in the secondary classification input set belong to the 24 chromosome categories, when the probability value is lower than a certain threshold (for example, 0.005), the corresponding element value is set to 0. This method can improve the efficiency of classification calculation.

[0074] Example 2

[0075] like Figure 2As shown, Example 2 is an alternative to Example 1. The remaining features and steps are generally the same, except that S2 can also be replaced by: using a deep learning-based neural network instance segmentation model to segment and classify the microscopic images in the primary classification input set to obtain all chromatid outlines in the image; extracting a chromatid outline thumbnail from the microscopic image and vertically positioning the chromatid thumbnail at 90 degrees. Using a deep neural network-based classifier, the chromatid thumbnail is inferred to have a probability of belonging to one of the 24 categories. The probabilities of all chromatids belonging to the 24 chromosome categories output by the classifier model, together with the image feature information of all chromatids, constitute the secondary classification input set.

[0076] The deep learning-based neural network classifier refers to a deep learning-based neural network model that uses a chromatid image dataset, labels different chromatid images with their categories, and is trained to be able to classify chromatid categories.

[0077] Example 3

[0078] like Figure 3 As shown, a technical solution provided in the second aspect of the embodiment of the present invention is: a chromosome image analysis system, which is composed of a chromosome microscopic image acquisition module, a primary segmentation and classification module, and a secondary pairing module in sequence.

[0079] The chromosome microscopic image acquisition module is used to capture microscopic images of chromatids and construct a classification input set.

[0080] The primary segmentation and classification input module performs a segmentation and classification on each microscopic image in the primary classification set based on the instance segmentation model to obtain the outline of the chromatid, the probability that the chromatid belongs to the chromosome category, and the characteristic information of the chromatid; optionally, the probability that the chromatid small image belongs to the 24 chromosome categories can be obtained in the following way: based on all the chromatid outlines obtained by the instance segmentation model, the chromatid outline small image is cut out from the microscopic image, and the chromatid small image is erected at 90 degrees. The classifier based on the deep neural network infers the probability that the chromatid small image belongs to the 24 chromosome categories. The above-mentioned probability that the chromatid small image belongs to the 24 chromosome categories and the characteristic information of the chromatid constitute the secondary classification input set.

[0081] The pairing module performs secondary classification on the secondary classification input set using a chromatid pairing model based on the secondary classification input set to obtain a chromatid thumbnail classification result.

[0082] Example 4

[0083] A third aspect of an embodiment of the present invention provides a technical solution: a computer device comprising at least one processor and at least one memory, wherein the at least one memory is configured to store at least one computer program; when the at least one computer program is executed by the at least one processor, the at least one processor implements the chromosome image analysis method described above. Specifically, the memory may include, but is not limited to, random-access memory (RAM), read-only memory (ROM), flash memory, first-in-first-out memory (FIFO), and / or first-in-last-out memory (FILO), etc.

[0084] Example 5

[0085] A fourth aspect of the present invention provides a technical solution comprising: a storage medium storing processor-executable instructions, wherein the processor, when executing the instructions, is used to perform the chromosome image analysis method described above. The storage medium refers to a data storage medium and may include, but is not limited to, computer-readable storage media such as a floppy disk, an optical disk, a hard disk, a flash memory, a USB flash drive, and / or a memory stick. The computer may be a general-purpose computer, a dedicated computer, a computer network, or other programmable device.

[0086] Example 6

[0087] A fifth aspect of the present invention provides a technical solution comprising a computer program product comprising instructions that, when executed on a computer, cause the computer to perform the chromosome image analysis method described in the first aspect. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device.

[0088] Verification Example

[0089] After actual testing, it was found that the image analysis method using the instance segmentation model combined with the chromatid pairing model designed in this embodiment was compared with the image classification method using the instance segmentation model alone, and the accuracy of chromosome classification and pairing in the final karyotype diagram was compared to obtain a comparison result of the karyotype diagram results. In 48.6% of the test images, the chromosome karyotype diagram results were consistent with those of the scheme using the image classification method using the instance segmentation model alone; in 42.9% of the test microscopic images, the chromosome karyotype diagram results were better than those of the scheme using the image classification method using the instance segmentation model alone; and in 8.5% of the test images, the chromosome karyotype diagram results were worse than those of the scheme using the image classification method using the instance segmentation model alone. Overall, the image analysis method using the instance segmentation model combined with the chromatid pairing model designed in this embodiment will improve the results of chromosome classification and pairing. The above test data demonstrates the beneficial effects of the present invention.

[0090] The specific embodiments described above are preferred embodiments of the chromosome image analysis method, system, computer device, and storage medium of the present invention, and are not intended to limit the specific implementation scope of the present invention. The scope of the present invention includes but is not limited to this specific embodiment. Any equivalent changes made in accordance with the shape and structure of the present invention are within the scope of protection of the present invention.

Claims

1. A chromosome image analysis method, characterized in that: The steps include: S01: Take microscopic images of chromosomes and construct a classification input set; S02: Based on the primary classification input set, a secondary classification input set is constructed using one of the following two methods; Method 1: Use a deep learning-based neural network instance segmentation model to segment and classify the microscopic images in the primary classification input set, obtaining the outlines of all chromatids in the microscopic images and the probabilities that the corresponding chromatids belong to the 24 chromosome categories (1, 2, 3, .. 21, 22, X, Y). Image feature information of each chromatid is calculated based on the chromatid outlines. The probabilities that all chromatids belong to the 24 chromosome categories and the image feature information of all chromatids constitute the secondary classification input set. Method 2: Use a deep learning-based neural network instance segmentation model to segment and classify the microscopic images in the primary classification input set, obtain all chromatid outlines in the microscopic image, cut out the chromatid outline thumbnails from the microscopic image, and stand the chromatid thumbnails upright at 90 degrees; a deep neural network-based classifier infers the probability that the chromatid thumbnails belong to the 24 chromosome categories; the probabilities of all chromatid thumbnails belonging to the 24 chromosome categories output by the classifier and the image feature information of all chromatids form a secondary classification input set, wherein the image feature information of each chromatid is calculated based on the chromatid outline; S03: Let assign_num = Total number of chromatids in one microscopic image / 46 , Indicates rounding up to an integer; input the secondary classification input set, call the assign_num chromatid pairing model, correct the chromatid category, and obtain the final chromatid category information.

2. A chromosome image analysis method according to claim 1, characterized in that: The microscopic image is a microscopic image of chromosome metaphase.

3. A chromosome image analysis method according to claim 1, characterized in that: The deep learning-based neural network instance segmentation model is a deep learning-based neural network model that uses a microscopic image dataset of chromosome metaphase to annotate chromosomes in microscopic images and is trained to segment the outlines of different chromatids and classify different chromatid categories.

4. The chromosome image analysis method according to claim 1, characterized in that: The deep neural network classifier refers to a neural network model based on deep learning that is able to classify chromatid categories and is obtained by using a chromatid image dataset to label different chromatid images with corresponding category labels and training them.

5. The chromosome image analysis method according to claim 1, characterized in that: The image feature information of the microscopic image includes contour area and height information of chromatids.

6. A chromosome image analysis method according to claim 1, characterized in that: The objective function of the chromatid pairing model is: ； in, is the probability that the i-th chromatid is classified as the j-th category; N represents the total number of unclassified chromatids remaining in the pairing process, and the variable Indicates whether the i-th chromatid is finally classified into category j; when , indicating that the i-th chromatid is not classified as category j, , indicating that the i-th chromatid is classified as category j.

7. A chromosome image analysis method according to claim 6, characterized in that: The chromatid pairing model also includes the following constraints: Chromatid classification value constraints: ; Constraint on the total number of chromatids participating in each classification: ; Chromatid classification category constraints: ; Constraints on the number of autosomes in a mitotic phase: ; Constraint on the number of sex chromosomes in a mitotic phase: ; Constraints on the number of Y chromosomes in a mitotic phase: ; Where N represents the total number of chromatids that remain unpaired during the pairing process.

8. A chromosome image analysis method according to claim 6 or 7, characterized in that: The image pairing model also includes image feature information constraints: in, represents the outline area of ​​the i-th chromatid; Represents the height information of the i-th chromatid.

9. A chromosome image analysis system, characterized in that: include: Chromosome microscopic image acquisition module, one-time segmentation and classification module, and chromatid pairing module; The chromosome microscopic image acquisition module is used to capture the microscopic image of the chromosome and construct a classification input set; The primary segmentation and classification module is configured to segment and classify the microscopic images in the primary classification input set based on a deep learning neural network instance segmentation model, obtain the outlines of all chromatids in the microscopic image, extract a chromatid outline thumbnail from the microscopic image, and vertically position the chromatid thumbnail at 90 degrees; a deep neural network-based classifier infers the probability that the chromatid thumbnail belongs to one of the 24 chromosome categories; the probability of the chromatid thumbnail belonging to the 24 chromosome categories and the image feature information of all chromatids constitute a secondary classification input set; wherein the image feature information of each chromatid is calculated based on the chromatid outline; The chromatid pairing module is used to perform secondary classification on the secondary classification input set using an image pairing model to obtain a chromatid thumbnail classification result based on a secondary classification input set, wherein the chromatid thumbnail classification result obtained by performing secondary classification on the secondary classification input set using an image pairing model based on the secondary classification input set is specifically: Let assign_num = Total number of chromatids in one microscopic image / 46 , Indicates rounding up to an integer; input the secondary classification input set, call the assign_num chromatid pairing model, correct the chromatid category, and obtain the final chromatid category information.

10. A computer device, characterized in that: The apparatus comprises at least one processor and at least one memory, wherein the at least one memory is used to store at least one computer program; when the at least one computer program is executed by the at least one processor, the at least one processor implements a chromosome image analysis method according to any one of claims 1 to 8.

11. A storage medium, characterized in that: The storage medium stores processor-executable instructions, which, when executed by the processor, are used to execute the chromosome image analysis method according to any one of claims 1 to 8.

Images