Chromosome image recognition method and device, computer device and storage medium

Abstract

The application relates to a chromosome image recognition method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: based on a chromosome original image and a chromosome banding image corresponding to the chromosome original image, the chromosome original image is converted into a chromosome band image, chromosome banding extraction is performed on the chromosome band image to obtain a band sequence feature, and chromosome recognition is performed on the chromosome original image according to the band sequence feature and a standard band sequence feature. According to the method, the chromosome original image is recognized according to the comparison between the band sequence feature and the standard band sequence feature, and the accuracy of chromosome recognition is improved from the dimension of the chromosome band sequence.

Description

Technical Field

[0001] This application relates to the field of intelligent medical technology, and in particular to a chromosome image recognition method, device, computer equipment, storage medium, and computer program product. Background Technology

[0002] Chromosomal structural abnormalities refer to morphological abnormalities of chromosomes, including translocations, inversions, and duplications, which lead to genomic instability. Chromosomal structural abnormalities are also associated with certain genetic diseases, such as translocation trisomy 21 syndrome and Cri-du-chat syndrome.

[0003] Currently, chromosomal karyotype analysis is used in clinical practice to analyze chromosomal structural abnormalities. For example, classification models for chromosomal structural deletions and inversions are used to predict chromosomal abnormalities.

[0004] Because there are many types of chromosomal structural abnormalities, corresponding abnormality classification models are used to identify different types of chromosomal structural abnormalities. However, abnormality classification models also have problems such as low accuracy and low efficiency in identifying chromosomal abnormalities. Summary of the Invention

[0005] Therefore, it is necessary to provide a chromosome image recognition method, apparatus, computer equipment, computer-readable storage medium, and computer program product that can improve recognition accuracy and efficiency in response to the above-mentioned technical problems.

[0006] Firstly, this application provides a chromosome image recognition method. The method includes:

[0007] Based on the original chromosome image and the corresponding chromosome banding map, the original chromosome image is converted into a chromosome banding image;

[0008] Chromosomal banding is extracted from the chromosome banding image to obtain banding sequence features;

[0009] Chromosome identification is performed on the original chromosome image based on the banded sequence features and the standard banded sequence features.

[0010] In one embodiment, the step of identifying chromosomes in the original chromosome image based on the banded sequence features and standard banded sequence features includes: determining a first feature similarity between the banded sequence features and the standard banded sequence features by comparing the consistency of each element in the banded sequence features and each element in the standard banded sequence features; if the first feature similarity is greater than or equal to a first preset value, then determining that the chromosomes in the original chromosome image are normal; if the first feature similarity is less than the first preset value, then determining a second feature similarity between the banded sequence features and the abnormal banded sequence features by comparing the consistency of each element in the banded sequence features of the chromosome image and each element in the abnormal banded sequence features; and identifying the chromosome abnormality type corresponding to the original chromosome image based on each of the second feature similarities.

[0011] In one embodiment, identifying the chromosome abnormality type corresponding to the original chromosome image based on each of the second feature similarities includes: if the second feature similarity is greater than or equal to a second preset value, then identifying the chromosome abnormality type corresponding to the original chromosome image based on the second feature similarity; if each of the second feature similarities is less than the second preset value, then sorting the second feature similarities in descending order; and generating chromosome abnormality prompt information based on the top N second feature similarities in the sorting.

[0012] In one embodiment, identifying the chromosome abnormality type corresponding to the original chromosome image based on each of the second feature similarities includes: sorting the second feature similarities in descending order;

[0013] If the second feature similarity score ranked first is greater than or equal to the second preset value, then the chromosome abnormality type corresponding to the original chromosome image is identified based on the second feature similarity score ranked first; if the second feature similarity score ranked first is less than the second preset value, then chromosome abnormality prompt information is generated based on the N second feature similarities ranked first.

[0014] In one embodiment, the elements in the banded sequence include elements in the length direction and elements in the width direction. Determining a first feature similarity between the banded sequence features and the standard banded sequence features by comparing the consistency of each element in the banded sequence features with each element in the standard banded sequence features includes: determining a third feature similarity between the banded sequence features and the standard banded sequence features by comparing each element in the length direction of the banded sequence features with each element in the length direction of the standard banded sequence features; determining a fourth feature similarity between the banded sequence features and the standard banded sequence features by comparing each element in the width direction of the banded sequence features with each element in the width direction of the standard banded sequence features; and fusing the third and fourth feature similarities to obtain the first feature similarity.

[0015] In one embodiment, before extracting chromosome banding patterns from the chromosome band images to obtain banding sequence features, the method includes: acquiring a first number of original chromosome images; expanding the first number of original chromosome images to obtain a second number of original chromosome images; transforming the second number of original chromosome images based on the original chromosome images and the corresponding chromosome banding maps to obtain a second number of chromosome band images; iteratively training a chromosome banding extraction model based on a training sample set constructed from the second number of chromosome band images, wherein the chromosome banding extraction model is used to extract chromosome banding patterns from the chromosome band images.

[0016] Secondly, this application also provides a chromosome image recognition device. The device includes: a conversion module, used to convert the original chromosome image into a chromosome banding image based on the original chromosome image and the corresponding chromosome banding map;

[0017] The extraction module is used to extract chromosome banding from the chromosome banding image to obtain banding sequence features;

[0018] The identification module is used to identify chromosomes in the original chromosome image based on the banded sequence features and the standard banded sequence features.

[0019] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:

[0020] Based on the original chromosome image and the corresponding chromosome banding map, the original chromosome image is converted into a chromosome banding image;

[0021] Chromosomal banding is extracted from the chromosome banding image to obtain banding sequence features;

[0022] Chromosome identification is performed on the original chromosome image based on the banded sequence features and the standard banded sequence features.

[0023] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:

[0024] Based on the original chromosome image and the corresponding chromosome banding map, the original chromosome image is converted into a chromosome banding image;

[0025] Chromosomal banding is extracted from the chromosome banding image to obtain banding sequence features;

[0026] Chromosome identification is performed on the original chromosome image based on the banded sequence features and the standard banded sequence features.

[0027] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, performs the following steps:

[0028] Based on the original chromosome image and the corresponding chromosome banding map, the original chromosome image is converted into a chromosome banding image;

[0029] Chromosomal banding is extracted from the chromosome banding image to obtain banding sequence features;

[0030] Chromosome identification is performed on the original chromosome image based on the banded sequence features and the standard banded sequence features.

[0031] The aforementioned chromosome image recognition method, apparatus, computer equipment, storage medium, and computer program product, based on the original chromosome image and the corresponding chromosome banding map, converts the original chromosome image into a chromosome banding image. The original chromosome image is then converted based on the chromosome banding image to obtain chromosome banding images reflecting key colorimetric regions, providing a basis for improving the prediction accuracy of subsequent models. Chromosomal banding is extracted from the chromosome banding image to obtain banding sequence features. Based on these banding sequence features and standard banding sequence features, chromosome recognition is performed on the original chromosome image. This method, on the one hand, improves the accuracy of chromosome recognition by comparing the banding sequence features with the standard banding sequence features, starting from the dimension of chromosome banding sequences. On the other hand, it directly identifies chromosomes from the original chromosome image based on the chromosome's own banding sequence features and standard banding sequence features, eliminating the need to establish corresponding recognition models for each type of chromosomal abnormality, thus improving recognition efficiency. Attached Figure Description

[0032] Figure 1 This is an application environment diagram of the chromosome image recognition method in one embodiment;

[0033] Figure 2 This is a flowchart illustrating a chromosome image recognition method in one embodiment;

[0034] Figure 3 This is a schematic diagram of the original chromosome image in one embodiment;

[0035] Figure 4 Here is a standard banding diagram of chromosomes in one embodiment;

[0036] Figure 5 This is a schematic diagram of chromosome banding images in one embodiment;

[0037] Figure 6 This is a flowchart illustrating a chromosome identification method in one embodiment;

[0038] Figure 7 This is a flowchart illustrating a method for assisting in the identification of chromosome structural abnormalities in one embodiment;

[0039] Figure 8 This is a schematic diagram of the structure of a chromosome banding extraction model in one embodiment;

[0040] Figure 9 This is a structural block diagram of a chromosome image recognition device in one embodiment;

[0041] Figure 10 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0043] The chromosome image recognition method provided in this application can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104, or it can be located in the cloud or on another network server.

[0044] Server 104 converts the original chromosome image into a chromosome banding image based on the original chromosome image and the corresponding chromosome banding image. Server 104 extracts chromosome banding from the chromosome banding image to obtain banding sequence features. Server 104 then performs chromosome identification on the original chromosome image based on the banding sequence features and standard banding sequence features.

[0045] The terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, and smart in-vehicle systems. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted devices. The terminal 102 can also be an electron microscope or an electronic camera. Specifically, the terminal 102 can acquire raw chromosome images in real time. The server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0046] In one embodiment, such as Figure 2 As shown, a chromosome image recognition method is provided, which can be applied to... Figure 1 Taking the server in the example of this, the explanation includes:

[0047] S202, based on the original chromosome image and the corresponding chromosome banding image, converts the original chromosome image into a chromosome banding image.

[0048] Chromosomes are the specific form of DNA present in eukaryotic cells during mitosis or meiosis. They are formed by the spiral winding of chromatin threads, which gradually shorten and thicken. Specifically, chromosomes are the chromatin structures in metaphase of cell division. After staining, they can be clearly displayed. The original image of a chromosome can be an image corresponding to different stages of the cell in which the chromosome is located. For example, the original image of a chromosome can be an image of metaphase, prophase, anaphase, and anaphase.

[0049] Abnormalities in the original chromosome image refer to morphological abnormalities of chromosomes, including translocations, inversions, and duplications, which lead to genomic instability. Chromosomal structural abnormalities are also associated with certain genetic diseases, such as translocational trisomy 21 syndrome and Cri-du-chat syndrome.

[0050] Currently, in clinical practice, karyotype analysis is mainly used to analyze chromosomal structural abnormalities. Karyotype analysis refers to the analysis and research of a cell's chromosome number, morphology, length, banding pattern, and centromere position.

[0051] This system can acquire raw chromosome images in real time or extract them from an image database for analysis. Specifically, raw chromosome images can be acquired in real time using an electronic camera or electron microscope, or retrieved from an image database.

[0052] Among them, the chromosome banding image can be a G-banding image of chromosomes. G-banding is achieved by processing chromosome specimens and then staining them with dye, resulting in distinct alternating light and dark bands on the chromosomes. Specifically, the chromosome banding image can be obtained by converting the original chromosome image based on the human normal chromosome G-banding pattern diagram and the original chromosome image in the International Society for Human Genetics (ISCN) nomenclature system.

[0053] Specifically, such as Figure 3 The diagram shown illustrates the transformation of the original chromosome image into a chromosome banding image based on the original chromosome image and its corresponding chromosome banding diagram. For example, ... Figure 4 The chromosome standard banding diagram shown is a standard image of ISCN (International System of Nomenclature for Human Cell Genetics). The chromosome banding diagram includes a chromatographic banding pattern.

[0054] Among them, chromosome banding images can be composed of multiple chromosome images reflecting key chromogenic regions, such as Figure 5In the schematic diagram of a chromosome banding image, black areas can represent key chromosomal regions, while the areas between black areas can represent non-key chromosomal regions. By analyzing the key and non-key chromosomal regions in the chromosome banding image, it is possible to determine whether the chromosome image is abnormal and the type of abnormality.

[0055] S204, Chromosomal banding is extracted from the chromosome banding image to obtain banding sequence features.

[0056] Chromosomal banding can include key and non-key coloring regions in a chromosome banding image.

[0057] This involves extracting chromosome banding patterns from chromosome band images to obtain chromosome banding features. Chromosome banding features include banding type and banding region. The banding type can correspond to either key colored regions or non-key colored regions; for example, the banding type can be black or white. The banding region can refer to the size and distribution of the region corresponding to each band.

[0058] After obtaining the chromosome banding features, further feature extraction is performed from the chromosome banding features to obtain the banding sequence features.

[0059] The band pattern sequence feature includes multiple elements. For example, a certain band pattern sequence feature expression is "00110..." where element 0 represents non-critical color regions and 1 represents critical color regions. It should be noted that... Figure 5 As shown, the key color areas are black, and the non-key color areas are white.

[0060] Banding sequence features can be sequence features in different directions. For example, banding sequence features can include banding sequence features along the length of the chromosome and banding sequence features along the width of the chromosome. Banding sequence features along the length of the chromosome can characterize the distribution of key chromogenic regions along the length of the chromosome. Banding sequence features along the width of the chromosome can characterize the distribution of key chromogenic regions along the width of the chromosome.

[0061] S206, chromosome identification is performed on the original chromosome image based on banded sequence characteristics and standard banded sequence characteristics.

[0062] Among them, the standard banding sequence features can be the banding sequence features corresponding to the standard chromosome image. The standard banding sequence features contain the distribution of key display areas of the standard chromosome. The standard banding sequence features can be analyzed to obtain the structural information of each part of the standard chromosome.

[0063] Multiple standard chromosomes can be selected, and the original images corresponding to each standard chromosome can be processed to obtain standard banding images.

[0064] The method of chromosome identification based on banded pattern sequence features and standard banded pattern sequence features includes: matching the banded pattern sequence features with the standard banded pattern sequence features to generate a similarity matching result; if the similarity matching result is greater than or equal to a preset threshold, the original chromosome image corresponding to the banded pattern sequence features is determined to be normal; if the similarity matching result is less than the preset threshold, the original chromosome image corresponding to the banded pattern sequence features is determined to be abnormal.

[0065] In addition, the region corresponding to the key color region in the banded sequence feature can be compared with the region corresponding to the key color region in the standard banded sequence feature. If the overlap is greater than or equal to a preset threshold, the original chromosome image corresponding to the banded sequence feature is determined to be normal. If the overlap is less than the preset threshold, the original chromosome image corresponding to the banded sequence feature is determined to be abnormal.

[0066] Specifically, the similarity between the banded sequence features and the standard banded sequence features can be compared to generate a similarity score. In particular, each element in the banded sequence features is compared with each element in the standard banded sequence features. If the similarity score of each element is greater than or equal to a preset value, the original chromosome image is determined to be normal. If the similarity score of each element is less than the preset value, the original chromosome image is determined to be abnormal.

[0067] Banding sequence features can include banding sequence features along chromosome length and banding sequence features along chromosome width.

[0068] Specifically, the distribution of key chromogenic regions along the length direction can be compared with the key chromogenic regions of a standard chromosome along the length direction to obtain chromosome-assisted identification results. Then, the distribution of key chromogenic regions along the width direction can be compared with the key chromogenic regions of a standard chromosome along the width direction to obtain chromosome-assisted identification results. Finally, the banding sequence features along the length direction and the banding sequence features along the width direction of the chromosome are fused, and the fused features are compared with the key chromogenic regions of a standard chromosome to obtain chromosome-assisted identification results. The chromosome-assisted identification results can include whether the chromosome structure is abnormal and the type of chromosome structural abnormality.

[0069] In the aforementioned chromosome image recognition method, based on the original chromosome image and its corresponding chromosome banding map, the original chromosome image is transformed into a chromosome banding image. The original chromosome image is then transformed based on the chromosome banding image to obtain chromosome banding images reflecting key coloring regions, providing a basis for improving the prediction accuracy of the subsequent model. Chromosomal banding is extracted from the chromosome banding image to obtain banding sequence features. Based on these banding sequence features and standard banding sequence features, chromosome recognition is performed on the original chromosome image. This method, on the one hand, improves the accuracy of chromosome recognition by comparing the banding sequence features with the standard banding sequence features, starting from the dimension of chromosome banding sequences. On the other hand, it directly identifies chromosomes from the original chromosome image based on the chromosome's own banding sequence features and standard banding sequence features, eliminating the need to establish corresponding recognition models for each type of chromosomal abnormality, thus improving recognition efficiency.

[0070] In one embodiment, such as Figure 6 The flowchart shown illustrates the chromosome identification method. Based on banding sequence features and standard banding sequence features, chromosome identification is performed on the original chromosome image, including:

[0071] S602, by comparing the consistency of each element in the banded sequence feature with each element in the standard banded sequence feature, the first feature similarity between the banded sequence feature and the standard banded sequence feature is determined.

[0072] Among them, the elements in the banding sequence features refer to the key and non-key colored regions of the chromosome band image corresponding to the banding sequence features. The elements in the banding sequence features can be determined according to the position, size and type of the colored region. For example, the banding sequence feature corresponding to a key colored region of one unit size can be "1", or the banding sequence feature corresponding to three consecutive key colored regions of one unit size can be "111", or the banding sequence feature corresponding to a combination of a key colored region of one unit size and a non-key colored region of one unit size can be "10" or "01".

[0073] This process involves first partitioning each element in the banded sequence feature and each element in the standard banded sequence feature into separate partitions. Then, a consistency comparison is performed based on the elements within each partition to obtain the first feature similarity between the banded sequence feature and the standard banded sequence feature. By partitioning each element, the computational load for matching the banded sequence feature and the standard banded sequence feature is reduced, the computation time is decreased, and the efficiency of chromosome identification is improved.

[0074] Furthermore, the elements in the banded sequence feature and the elements in the standard banded sequence feature can be compared for consistency according to their order of arrangement. For example, starting from the chromosome length direction, the elements in the banded sequence feature and the elements in the standard banded sequence feature can be compared sequentially to obtain the first feature similarity between the banded sequence feature and the standard banded sequence feature. It should be noted that the elements in the chromosome width direction can also be compared for consistency to obtain the corresponding feature similarity.

[0075] S604, if the similarity of the first feature is greater than or equal to the first preset value, then the chromosome in the original chromosome image is determined to be normal.

[0076] The first feature similarity represents the degree of similarity between the banded sequence feature and the standard banded sequence feature. If the first similarity is greater than or equal to the first preset value, it means that the original chromosome image corresponding to the banded sequence feature does not have obvious abnormalities. If the first feature similarity is less than or equal to the first preset value, it means that the original chromosome image corresponding to the banded sequence feature may have abnormalities in chromosome structure.

[0077] S606, if the first feature similarity is less than the first preset value, then the second feature similarity between the banded sequence features and the abnormal banded sequence features is determined by comparing the consistency of each element in the banded sequence features of the chromosome image with each element in the abnormal banded sequence features.

[0078] It should be noted that the original chromosome image may contain various abnormal structural types. It is necessary to perform consistency comparison between each element in the banding sequence feature of the chromosome image and each element in each abnormal banding sequence feature to obtain the second feature similarity between the banding sequence feature and the abnormal banding sequence feature.

[0079] If the similarity of the first feature is less than the first preset value, it means that the original chromosome image corresponding to the banded sequence feature may have abnormalities in chromosome structure and needs further verification.

[0080] Among them, the abnormal banding sequence feature is the banding sequence feature corresponding to the abnormal chromosome image. The abnormal chromosome image contains at least one type of chromosome abnormality, including: deletion, inversion, reciprocal translocation, insertion, circular chromosome and other abnormality types.

[0081] Specifically, multiple abnormal chromosome images and their corresponding abnormality types are acquired beforehand. The elements of the banding sequence features of the chromosome images and the elements of the abnormal banding sequence features are then compared for consistency. This comparison determines the second feature similarity between the banding sequence features and the abnormal banding sequence features, which aids in subsequent processes. The abnormality type of the chromosome image is determined based on this second feature similarity. By comparing the elements of the banding sequence features of the chromosome images with the elements of the abnormal banding sequence features, it is unnecessary to build a corresponding model for each chromosome abnormality type. This simplifies the chromosome identification process and improves identification efficiency to some extent while maintaining the accuracy of chromosome structural abnormality identification.

[0082] S608, based on the similarity of each second feature, identifies the type of chromosome abnormality corresponding to the original chromosome image.

[0083] The second feature similarity characterizes the degree of similarity between the original chromosome image and the abnormal chromosome image. The higher the second feature similarity, the higher the probability that the original chromosome image contains the corresponding abnormal type structure.

[0084] Specifically, if at least one of the second feature similarities is greater than or equal to a second preset value, the type of chromosomal abnormality corresponding to the original chromosome image is identified based on that second feature similarity. If all the second feature similarities are less than the second preset value, it indicates that the original chromosome image has a risk of structural abnormality, and chromosomal abnormality warning information needs to be generated based on that second feature similarity.

[0085] It should be noted that the first preset value and the second preset value can be flexibly selected according to the actual situation, and are not limited here. For example, the first preset value can be 0.98 and the second preset value can be 0.99.

[0086] In this embodiment, on the one hand, by comparing the consistency of each element in the banded sequence features of the chromosome image with each element in the abnormal banded sequence features, it is not necessary to establish a corresponding model for each chromosome abnormality type. While ensuring the recognition accuracy of chromosome structural abnormalities, the chromosome recognition process is simplified to a certain extent and the recognition efficiency is improved. On the other hand, the chromosome abnormality type corresponding to the original chromosome image is generated based on the second feature similarity, or chromosome abnormality prompt information is generated based on the second feature similarity, which improves the recognition effect.

[0087] In one embodiment, identifying the chromosome abnormality type corresponding to the original chromosome image based on the similarity of each second feature includes: if the similarity of the second feature is greater than or equal to a second preset value, then identifying the chromosome abnormality type corresponding to the original chromosome image based on the similarity of the second feature; if the similarity of each second feature is less than the second preset value, then sorting the second feature similarities in descending order, and generating chromosome abnormality prompt information based on the top N similarities of the sorted second feature.

[0088] The second feature similarity represents the similarity between the banded pattern sequence features and the abnormal banded pattern sequence features, that is, the similarity between the original chromosome image and the original images of the abnormal chromosomes. If the second feature similarity is greater than or equal to a second preset value, it means that the original chromosome image is similar to the original image of the abnormal chromosome, and that there is an abnormal structure in the original chromosome image. Specifically, the presence of an abnormal structure in the original chromosome image is determined based on the abnormal structure of the original image of the abnormal chromosome corresponding to the abnormal banded pattern sequence features.

[0089] If the similarity of one of the second features is less than the second preset value, it means that the original chromosome image corresponding to the second feature similarity may have an anomaly. In order to improve the chromosome recognition effect, the original chromosome image needs to be further recognized.

[0090] Specifically, the similarity of the second feature is sorted in descending order, and the similarity between the original chromosome image and the original images of each abnormal chromosome is ranked. Then, an anomaly recognition structure is generated based on the ranking results. Specifically, the similarity of the second feature is sorted in descending order, and chromosome abnormality prompt information is generated based on the top N similarity scores of the second feature.

[0091] The anomaly alert information is used to indicate the types of abnormalities present in the original chromosome image to medical personnel. Specifically, it can be achieved by packaging the top N chromosome abnormality types corresponding to the second feature similarity in a ranking list to obtain the chromosome abnormality types associated with the original chromosome image.

[0092] In this embodiment, by further identifying based on the similarity of each second feature, the similarity of the second features is sorted in descending order, and chromosomal abnormality prompt information is generated based on the top N similarity of the second features, thereby improving the identification effect of chromosomal abnormality types.

[0093] In one embodiment, identifying the chromosome abnormality type corresponding to the original chromosome image based on the similarity of each second feature includes: sorting the second feature similarities in descending order; if the second feature similarity at the top of the sort is greater than or equal to a second preset value, then identifying the chromosome abnormality type corresponding to the original chromosome image based on the second feature similarity at the top of the sort; if the second feature similarity at the top of the sort is less than the second preset value, then generating chromosome abnormality prompt information based on the top N second feature similarities.

[0094] The method involves first sorting the second feature similarity in descending order, then selecting the second feature similarity with the highest similarity, and prioritizing the determination of the most likely abnormal type of the original chromosome image from multiple abnormal types. If the second feature similarity ranked first is greater than or equal to a second preset value, then the chromosome abnormality type corresponding to the original chromosome image is identified based on the second feature similarity ranked first, thus improving the recognition efficiency.

[0095] If the similarity of the second feature ranked first is less than the second preset value, chromosome abnormality information can be obtained based on the similarity between the original chromosome image and the abnormal chromosome image, or based on the similarity between the banded sequence features and the abnormal banded sequence features.

[0096] Specifically, if the similarity of each second feature is less than the second preset value, the similarity of the second features is sorted in descending order, and chromosome abnormality prompt information is generated based on the top N similarity values ​​of the second features.

[0097] In this embodiment, if the similarity of the second feature ranked first is greater than or equal to the second preset value, the chromosome abnormality type corresponding to the original chromosome image is identified based on the similarity of the second feature ranked first, thereby improving the identification efficiency.

[0098] In one embodiment, the elements in the striped sequence include elements in the length direction and elements in the width direction. A first feature similarity between the striped sequence features and the standard striped sequence features is determined by comparing the consistency of each element in the striped sequence features with each element in the standard striped sequence features. This includes: determining a third feature similarity between the striped sequence features and the standard striped sequence features by comparing the elements in the length direction of the striped sequence features with each element in the length direction of the standard striped sequence features; determining a fourth feature similarity between the striped sequence features and the standard striped sequence features by comparing the elements in the width direction of the striped sequence features with each element in the width direction of the standard striped sequence features; and fusing the third and fourth feature similarities to obtain the first feature similarity.

[0099] Specifically, based on chromosome length direction, each element in the length direction of the banded sequence feature is compared with each element in the length direction of the standard banded sequence feature to obtain the third feature similarity between the banded sequence feature and the standard banded sequence feature. In particular, the third feature similarity is obtained by sequentially comparing each element in the length direction of the banded sequence feature with each element in the length direction of the standard banded sequence feature.

[0100] For example, taking the striped sequence feature as 001 and the standard striped sequence feature as 011 as an example, the elements in the length direction of the striped sequence feature and the elements in the length direction of the standard striped sequence feature are compared sequentially, and the similarity of the third feature is approximately 0.67.

[0101] Specifically, based on the chromosome width direction, each element in the width direction of the banded sequence feature is compared with each element in the width direction of the standard banded sequence feature to obtain the fourth feature similarity between the banded sequence feature and the standard banded sequence feature. In particular, the fourth feature similarity is obtained by sequentially comparing each element in the width direction of the banded sequence feature with each element in the width direction of the standard banded sequence feature.

[0102] Specifically, the third feature similarity and the fourth feature similarity can be fused according to the feature weights to obtain the first feature similarity. The feature weights can be determined based on the number of elements contained in the feature similarity. For example, if there are 95 elements in the length direction of the chromosome and 5 elements in the width direction of the chromosome, the feature weight corresponding to the third feature similarity is 0.95, and the feature weight corresponding to the fourth feature similarity is 0.05.

[0103] Alternatively, the first feature similarity can be obtained by averaging the third feature similarity and the fourth feature similarity.

[0104] In this embodiment, by fusing the third feature similarity in the chromosome length direction and the fourth feature similarity in the chromosome width direction, a more accurate first feature similarity is obtained, thereby improving the accuracy of chromosome identification.

[0105] Existing chromosome banding extraction models suffer from low extraction accuracy and precision. Therefore, in one embodiment, before extracting chromosome banding patterns from chromosome band images to obtain banding sequence features, the method includes: acquiring a first number of original chromosome images; expanding the first number of original chromosome images to obtain a second number of original chromosome images; transforming the second number of original chromosome images based on the original chromosome images and their corresponding chromosome banding maps to obtain a second number of chromosome banding images; and iteratively training the chromosome banding extraction model using a training sample set constructed from the second number of chromosome banding images. The chromosome banding extraction model is used to extract chromosome banding patterns from the chromosome banding images.

[0106] Among them, various image processing methods can be used to process the original chromosome images, and a second number of original chromosome images can be obtained based on the processed original chromosome images and the first number of original chromosome images.

[0107] The second number is greater than or equal to the first number. The size of the second number depends on the type of image processing method used and the number of original chromosome images actually processed.

[0108] Specifically, the image processing methods include: flipping the original chromosome image horizontally or vertically, rotating the original chromosome image at random angles, and randomly adjusting the contrast and brightness of the original chromosome image, and then standardizing the processed original chromosome image. Image standardization first involves normalizing pixel values ​​to 0-1, followed by mean normalization and standard deviation normalization.

[0109] The second number of original chromosome images are converted to obtain a second number of chromosome band images, including: converting the original chromosome images into chromosome band images based on the second number of original chromosome images and the corresponding chromosome banding diagrams to obtain a second number of chromosome band images.

[0110] The chromosome banding extraction model is iteratively trained based on a training sample set constructed from a second set of chromosome banding images.

[0111] In this embodiment, by expanding the first number of chromosome band images to obtain a second number of chromosome band images, the size of the training sample set is increased, thereby improving the chromosome banding extraction model and its extraction effect.

[0112] In one embodiment, such as Figure 7As shown, a method for assisting in the identification of chromosome structural abnormalities is provided, including a training part of the chromosome banding extraction model and a usage part of the chromosome banding extraction model.

[0113] The training component of the chromosome banding extraction model includes:

[0114] S702, acquire the first number of raw chromosome images.

[0115] S704, expand the first number of original chromosome images to obtain a second number of original chromosome images.

[0116] S706, based on the original chromosome image and the corresponding chromosome banding image, the second number of original chromosome images are transformed to obtain the second number of chromosome banding images.

[0117] S708, based on the training sample set constructed from the second number of chromosome band images, iteratively train the chromosome banding extraction model.

[0118] Among them, the chromosome banding extraction model is used to extract chromosome banding from chromosome banding images.

[0119] Specifically, the ADAM algorithm is used to optimize the error gradient of the neural network model by steepest descent using the training sample set, and a chromosome banding extraction model is constructed through offline training.

[0120] Among them, such as Figure 8 The schematic diagram of the chromosome banding extraction model shown can be generated based on the ResNet101 network. In this model, conv represents a convolutional layer, stride represents the stride, stride1 represents a stride of 1 unit, padding represents the padding operation, padding1 represents padding of 1 unit, max pool represents the max pooling layer, average represents the average pooling layer, Ls-d fc represents a fully connected layer of length Ls, softmax represents the normalization function and normalization operation, and 7x7, 3x3, and 1x1 are the kernel sizes in the convolutional layer conv and the pooling window sizes in the pooling layer.

[0121] Specifically, the final fully connected layer of the model is changed to an output of length Ls*1*1. The number of parameters in the fully connected layer is proportional to the size of the input feature map, while the number of parameters in the Ls*1*1 convolutional kernel is only related to the number of channels in the input feature map and the number of output channels. For large input feature map sizes and large number of output channels, this can significantly reduce the number of model parameters.

[0122] The usage of the chromosome banding extraction model includes:

[0123] S710 converts the original chromosome image into a chromosome banding image based on the original chromosome image and the corresponding chromosome banding map.

[0124] S712, chromosome banding is extracted from chromosome band images to obtain banding sequence features.

[0125] S714, by comparing each element in the length direction of the banded sequence feature with each element in the length direction of the standard banded sequence feature, the third feature similarity between the banded sequence feature and the standard banded sequence feature is determined.

[0126] S716, by comparing each element in the width direction of the banded sequence feature with each element in the width direction of the standard banded sequence feature, the fourth feature similarity between the banded sequence feature and the standard banded sequence feature is determined.

[0127] S718, the third feature similarity and the fourth feature similarity are fused to obtain the first feature similarity.

[0128] The first similarity between the features of the banded sequence and the features of the standard banded sequence can be obtained based on the fitting degree between the banded sequence and the standard banded sequence.

[0129] Specifically, the elements at the same positions in the banded sequence are compared with those in the standard banded sequence. If an element at a certain position is the same, the counter is incremented by one. The value of the counter is then divided by the default length to obtain the goodness of fit, as shown in the following formula:

[0130]

[0131] Where S represents the goodness of fit, L s Sf represents the length of a chromosome along its length or width. i Sn represents the length of the banded sequence, typically 512. If it is less than 512, it is automatically padded. i This represents the length of the standard banded sequence. Specifically, if an element at a certain position is the same, the counter is incremented by one; if all elements at the same position in the banded sequence are the same as those in the standard banded sequence, s = 1.

[0132] S720, if the similarity of the first feature is greater than or equal to the first preset value, then the chromosome in the original chromosome image is determined to be normal.

[0133] S722, if the first feature similarity is less than the first preset value, then the second feature similarity between the banded sequence features and the abnormal banded sequence features is determined by comparing the consistency of each element in the banded sequence features of the chromosome image with each element in the abnormal banded sequence features.

[0134] S724, if the second feature similarity is greater than or equal to the second preset value, then the chromosome abnormality type corresponding to the original chromosome image is identified based on the second feature similarity.

[0135] S726, if the similarity of each second feature is less than the second preset value, then sort the second feature similarity in descending order.

[0136] S728 generates chromosome abnormality warning information based on the similarity of the top N second features.

[0137] S730, sort the similarity of the second feature in descending order.

[0138] S732, if the similarity of the second feature ranked first is greater than or equal to the second preset value, then the chromosome abnormality type corresponding to the original chromosome image is identified based on the similarity of the second feature ranked first.

[0139] S734, if the similarity of the second feature ranked first is less than the second preset value, then generate chromosome abnormality prompt information based on the similarity of the top N second features ranked first.

[0140] In this embodiment, based on the original chromosome image and the corresponding chromosome banding map, the original chromosome image is converted into a chromosome banding image. The original chromosome image is then converted based on the chromosome banding image to obtain chromosome banding images reflecting key chromogenic regions, providing a basis for improving the prediction accuracy of the subsequent model. Chromosomal banding is extracted from the chromosome banding image to obtain banding sequence features. Based on these banding sequence features and standard banding sequence features, chromosome identification is performed on the original chromosome image. This method, on the one hand, improves the accuracy of chromosome identification by comparing the banding sequence features with the standard banding sequence features, starting from the dimension of chromosome banding sequences. On the other hand, it directly identifies chromosomes from the original chromosome image based on the chromosome's own banding sequence features and the standard banding sequence features, eliminating the need to establish corresponding identification models for each type of chromosomal abnormality, thus improving identification efficiency.

[0141] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0142] Based on the same inventive concept, this application also provides a chromosome image recognition device for implementing the chromosome image recognition method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more chromosome image recognition device embodiments provided below can be found in the limitations of the chromosome image recognition method described above, and will not be repeated here.

[0143] In one embodiment, such as Figure 9 As shown, a chromosome image recognition device is provided, comprising: a conversion module 902, an extraction module 904, and a recognition module 906, wherein:

[0144] The conversion module 902 is used to convert the original chromosome image into a chromosome banding image based on the original chromosome image and the chromosome banding map corresponding to the original chromosome image.

[0145] Extraction module 904 is used to extract chromosome banding from the chromosome banding image to obtain banding sequence features;

[0146] The identification module 906 is used to identify chromosomes in the original chromosome image based on the banded sequence features and the standard banded sequence features.

[0147] In one embodiment, the identification module 906 is further configured to determine a first feature similarity between the banded sequence features and the standard banded sequence features by comparing the consistency of each element in the banded sequence features with each element in the standard banded sequence features; if the first feature similarity is greater than or equal to a first preset value, then the chromosome in the original chromosome image is determined to be normal; if the first feature similarity is less than the first preset value, then a second feature similarity between the banded sequence features and the abnormal banded sequence features is determined by comparing the consistency of each element in the banded sequence features of the chromosome image with each element in the abnormal banded sequence features; and based on each second feature similarity, the chromosome abnormality type corresponding to the original chromosome image is identified.

[0148] In one embodiment, the identification module 906 is further configured to: if the second feature similarity is greater than or equal to a second preset value, identify the chromosome abnormality type corresponding to the original chromosome image based on the second feature similarity; if all second feature similarities are less than the second preset value, sort the second feature similarities in descending order; and generate chromosome abnormality prompt information based on the top N second feature similarities in the sorting.

[0149] In one embodiment, the identification module 906 is further configured to sort the second feature similarity in descending order; if the second feature similarity ranked first is greater than or equal to a second preset value, then the chromosome abnormality type corresponding to the original chromosome image is identified based on the second feature similarity ranked first; if the second feature similarity ranked first is less than the second preset value, then chromosome abnormality prompt information is generated based on the N second feature similarities ranked first.

[0150] In one embodiment, the elements in the stripe sequence include elements in the length direction and elements in the width direction. The recognition module 906 is further configured to determine a third feature similarity between the stripe sequence feature and the standard stripe sequence feature by comparing each element in the length direction of the stripe sequence feature with each element in the length direction of the standard stripe sequence feature; determine a fourth feature similarity between the stripe sequence feature and the standard stripe sequence feature by comparing each element in the width direction of the stripe sequence feature with each element in the width direction of the standard stripe sequence feature; and fuse the third feature similarity and the fourth feature similarity to obtain a first feature similarity.

[0151] In one embodiment, the chromosome image recognition device further includes: an expansion module for acquiring a first number of original chromosome images; expanding the first number of original chromosome images to obtain a second number of original chromosome images; converting the second number of original chromosome images based on the original chromosome images and the corresponding chromosome banding images to obtain a second number of chromosome banding images; and iteratively training a chromosome banding extraction model based on a training sample set constructed from the second number of chromosome banding images, wherein the chromosome banding extraction model is used to extract chromosome banding from the chromosome banding images.

[0152] Each module in the aforementioned chromosome image recognition device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0153] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 10 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The database stores banding sequence feature data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a chromosome image recognition method.

[0154] Those skilled in the art will understand that Figure 10 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0155] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0156] Based on the original chromosome image and the corresponding chromosome banding map, the original chromosome image is converted into a chromosome banding image; chromosome banding is extracted from the chromosome banding image to obtain banding sequence features; chromosome identification is performed on the original chromosome image based on the banding sequence features and standard banding sequence features.

[0157] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0158] By comparing the consistency of each element in the banded sequence feature with each element in the standard banded sequence feature, the first feature similarity between the banded sequence feature and the standard banded sequence feature is determined. If the first feature similarity is greater than or equal to a first preset value, the chromosome in the original chromosome image is determined to be normal. If the first feature similarity is less than the first preset value, by comparing the consistency of each element in the banded sequence feature of the chromosome image with each element in each abnormal banded sequence feature, the second feature similarity between the banded sequence feature and the abnormal banded sequence feature is determined. Based on each second feature similarity, the chromosome abnormality type corresponding to the original chromosome image is identified.

[0159] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0160] If the second feature similarity is greater than or equal to the second preset value, the chromosome abnormality type corresponding to the original chromosome image is identified based on the second feature similarity; if all second feature similarities are less than the second preset value, the second feature similarities are sorted in descending order; and chromosome abnormality prompt information is generated based on the top N second feature similarities.

[0161] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0162] The second feature similarity scores are sorted in descending order. If the second feature similarity score at the top of the sort is greater than or equal to a second preset value, the chromosome abnormality type corresponding to the original chromosome image is identified based on the second feature similarity score at the top of the sort. If the second feature similarity score at the top of the sort is less than the second preset value, chromosome abnormality warning information is generated based on the top N second feature similarity scores.

[0163] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0164] The third feature similarity between the striped sequence feature and the standard striped sequence feature is determined by comparing each element in the length direction of the striped sequence feature with each element in the length direction of the standard striped sequence feature. The fourth feature similarity between the striped sequence feature and the standard striped sequence feature is determined by comparing each element in the width direction of the striped sequence feature with each element in the width direction of the standard striped sequence feature. The third feature similarity and the fourth feature similarity are then fused to obtain the first feature similarity.

[0165] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0166] A first number of original chromosome images are obtained; the first number of original chromosome images are expanded to obtain a second number of original chromosome images; based on the original chromosome images and the corresponding chromosome banding images, the second number of original chromosome images are transformed to obtain a second number of chromosome banding images; based on the training sample set constructed from the second number of chromosome banding images, the chromosome banding extraction model is iteratively trained, wherein the chromosome banding extraction model is used to extract chromosome banding from the chromosome banding images.

[0167] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0168] Based on the original chromosome image and the corresponding chromosome banding map, the original chromosome image is converted into a chromosome banding image; chromosome banding is extracted from the chromosome banding image to obtain banding sequence features; chromosome identification is performed on the original chromosome image based on the banding sequence features and standard banding sequence features.

[0169] In one embodiment, when the computer program is executed by a processor, it further performs the following steps:

[0170] By comparing the consistency of each element in the banded sequence feature with each element in the standard banded sequence feature, the first feature similarity between the banded sequence feature and the standard banded sequence feature is determined. If the first feature similarity is greater than or equal to a first preset value, the chromosome in the original chromosome image is determined to be normal. If the first feature similarity is less than the first preset value, by comparing the consistency of each element in the banded sequence feature of the chromosome image with each element in each abnormal banded sequence feature, the second feature similarity between the banded sequence feature and the abnormal banded sequence feature is determined. Based on each second feature similarity, the chromosome abnormality type corresponding to the original chromosome image is identified.

[0171] In one embodiment, when the computer program is executed by a processor, it further performs the following steps:

[0172] If the second feature similarity is greater than or equal to the second preset value, the chromosome abnormality type corresponding to the original chromosome image is identified based on the second feature similarity; if all second feature similarities are less than the second preset value, the second feature similarities are sorted in descending order; and chromosome abnormality prompt information is generated based on the top N second feature similarities.

[0173] In one embodiment, when the computer program is executed by a processor, it further performs the following steps:

[0174] The second feature similarity scores are sorted in descending order. If the second feature similarity score at the top of the sort is greater than or equal to a second preset value, the chromosome abnormality type corresponding to the original chromosome image is identified based on the second feature similarity score at the top of the sort. If the second feature similarity score at the top of the sort is less than the second preset value, chromosome abnormality warning information is generated based on the top N second feature similarity scores.

[0175] In one embodiment, when the computer program is executed by a processor, it further performs the following steps:

[0176] The third feature similarity between the striped sequence feature and the standard striped sequence feature is determined by comparing each element in the length direction of the striped sequence feature with each element in the length direction of the standard striped sequence feature. The fourth feature similarity between the striped sequence feature and the standard striped sequence feature is determined by comparing each element in the width direction of the striped sequence feature with each element in the width direction of the standard striped sequence feature. The third feature similarity and the fourth feature similarity are then fused to obtain the first feature similarity.

[0177] In one embodiment, when the computer program is executed by a processor, it further performs the following steps:

[0178] A first number of original chromosome images are obtained; the first number of original chromosome images are expanded to obtain a second number of original chromosome images; based on the original chromosome images and the corresponding chromosome banding images, the second number of original chromosome images are transformed to obtain a second number of chromosome banding images; based on the training sample set constructed from the second number of chromosome banding images, the chromosome banding extraction model is iteratively trained, wherein the chromosome banding extraction model is used to extract chromosome banding from the chromosome banding images.

[0179] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:

[0180] Based on the original chromosome image and the corresponding chromosome banding map, the original chromosome image is converted into a chromosome banding image; chromosome banding is extracted from the chromosome banding image to obtain banding sequence features; chromosome identification is performed on the original chromosome image based on the banding sequence features and standard banding sequence features.

[0181] In one embodiment, when the computer program is executed by a processor, it further performs the following steps:

[0182] By comparing the consistency of each element in the banded sequence feature with each element in the standard banded sequence feature, the first feature similarity between the banded sequence feature and the standard banded sequence feature is determined. If the first feature similarity is greater than or equal to a first preset value, the chromosome in the original chromosome image is determined to be normal. If the first feature similarity is less than the first preset value, by comparing the consistency of each element in the banded sequence feature of the chromosome image with each element in each abnormal banded sequence feature, the second feature similarity between the banded sequence feature and the abnormal banded sequence feature is determined. Based on each second feature similarity, the chromosome abnormality type corresponding to the original chromosome image is identified.

[0183] In one embodiment, when the computer program is executed by a processor, it further performs the following steps:

[0184] If the second feature similarity is greater than or equal to the second preset value, the chromosome abnormality type corresponding to the original chromosome image is identified based on the second feature similarity; if all second feature similarities are less than the second preset value, the second feature similarities are sorted in descending order; and chromosome abnormality prompt information is generated based on the top N second feature similarities.

[0185] In one embodiment, when the computer program is executed by a processor, it further performs the following steps:

[0186] The second feature similarity scores are sorted in descending order. If the second feature similarity score at the top of the sort is greater than or equal to a second preset value, the chromosome abnormality type corresponding to the original chromosome image is identified based on the second feature similarity score at the top of the sort. If the second feature similarity score at the top of the sort is less than the second preset value, chromosome abnormality warning information is generated based on the top N second feature similarity scores.

[0187] In one embodiment, when the computer program is executed by a processor, it further performs the following steps:

[0188] The third feature similarity between the striped sequence feature and the standard striped sequence feature is determined by comparing each element in the length direction of the striped sequence feature with each element in the length direction of the standard striped sequence feature. The fourth feature similarity between the striped sequence feature and the standard striped sequence feature is determined by comparing each element in the width direction of the striped sequence feature with each element in the width direction of the standard striped sequence feature. The third feature similarity and the fourth feature similarity are then fused to obtain the first feature similarity.

[0189] In one embodiment, when the computer program is executed by a processor, it further performs the following steps:

[0190] A first number of original chromosome images are obtained; the first number of original chromosome images are expanded to obtain a second number of original chromosome images; based on the original chromosome images and the corresponding chromosome banding images, the second number of original chromosome images are transformed to obtain a second number of chromosome banding images; based on the training sample set constructed from the second number of chromosome banding images, the chromosome banding extraction model is iteratively trained, wherein the chromosome banding extraction model is used to extract chromosome banding from the chromosome banding images.

[0191] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0192] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0193] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0194] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A chromosome image recognition method, characterized in that, The method includes: Based on the original chromosome image and the corresponding chromosome banding map, the original chromosome image is converted into a chromosome banding image; Chromosomal banding is extracted from the chromosome banding image to obtain banding sequence features; Chromosome identification is performed on the original chromosome image based on the banded sequence features and the standard banded sequence features; The step of performing chromosome identification on the original chromosome image based on the banding sequence features and the standard banding sequence features includes: By comparing the consistency of each element in the banded sequence feature with each element in the standard banded sequence feature, the first feature similarity between the banded sequence feature and the standard banded sequence feature is determined. If the first feature similarity is greater than or equal to the first preset value, then the chromosome in the original chromosome image is determined to be normal. If the first feature similarity is less than the first preset value, then the second feature similarity between the banded sequence features and the abnormal banded sequence features is determined by comparing the consistency of each element in the banded sequence features of the chromosome image and each element in the abnormal banded sequence features. Based on the similarity of each of the second features, the chromosome abnormality type corresponding to the original chromosome image is identified, including: If the second feature similarity is greater than or equal to the second preset value, then the chromosome abnormality type corresponding to the original chromosome image is identified based on the second feature similarity. If the similarity of each of the second features is less than the second preset value, then the similarity of the second features is sorted in descending order. Generate chromosome abnormality alert information based on the similarity of the top N second features; Sort the similarity of the second feature in descending order; If the second feature similarity of the highest sorted value is greater than or equal to the second preset value, then the chromosome abnormality type corresponding to the original chromosome image is identified based on the second feature similarity of the highest sorted value. If the similarity of the second feature ranked first is less than the second preset value, then a chromosome abnormality warning message is generated based on the similarity of the top N second features. The elements in the banded sequence include elements in the length direction and elements in the width direction. The step of determining the first feature similarity between the banded sequence features and the standard banded sequence features by comparing the consistency of each element in the banded sequence features with each element in the standard banded sequence features includes: The third feature similarity between the banded sequence feature and the standard banded sequence feature is determined by comparing each element in the length direction of the banded sequence feature with each element in the length direction of the standard banded sequence feature. The fourth feature similarity between the banded sequence feature and the standard banded sequence feature is determined by comparing each element in the width direction of the banded sequence feature with each element in the width direction of the standard banded sequence feature. The third feature similarity and the fourth feature similarity are fused to obtain the first feature similarity.

2. The method according to claim 1, characterized in that, Before extracting chromosome banding patterns from the chromosome banding image to obtain banding sequence features, the method includes: Obtain the first number of raw chromosome images; The first number of original chromosome images are expanded to obtain a second number of original chromosome images; Based on the original chromosome image and the corresponding chromosome banding image, the second number of original chromosome images are transformed to obtain the second number of chromosome banding images; The chromosome banding extraction model is iteratively trained based on the training sample set constructed from the second number of chromosome banding images, wherein the chromosome banding extraction model is used to extract chromosome banding from the chromosome banding images.

3. A chromosome image recognition device, characterized in that, The device includes: The conversion module is used to convert the original chromosome image into a chromosome banding image based on the original chromosome image and the chromosome banding map corresponding to the original chromosome image. The extraction module is used to extract chromosome banding from the chromosome banding image to obtain banding sequence features; The identification module is used to identify chromosomes in the original chromosome image based on the banded sequence features and the standard banded sequence features; and to determine the first feature similarity between the banded sequence features and the standard banded sequence features by comparing the consistency of each element in the banded sequence features and each element in the standard banded sequence features. If the first feature similarity is greater than or equal to the first preset value, then the chromosome in the original chromosome image is determined to be normal. If the first feature similarity is less than the first preset value, then the second feature similarity between the banded sequence features and the abnormal banded sequence features is determined by comparing the consistency of each element in the banded sequence features of the chromosome image and each element in the abnormal banded sequence features. If the second feature similarity is greater than or equal to the second preset value, then the chromosome abnormality type corresponding to the original chromosome image is identified based on the second feature similarity. If the similarity of each of the second features is less than the second preset value, then the similarity of the second features is sorted in descending order. Generate chromosome abnormality alert information based on the similarity of the top N second features; Sort the similarity of the second feature in descending order; If the second feature similarity of the highest sorted value is greater than or equal to the second preset value, then the chromosome abnormality type corresponding to the original chromosome image is identified based on the second feature similarity of the highest sorted value. If the similarity of the second feature ranked first is less than the second preset value, then a chromosome abnormality warning message is generated based on the similarity of the top N second features. The elements in the banded sequence include elements in the length direction and elements in the width direction. The step of determining the first feature similarity between the banded sequence features and the standard banded sequence features by comparing the consistency of each element in the banded sequence features with each element in the standard banded sequence features includes: The third feature similarity between the banded sequence feature and the standard banded sequence feature is determined by comparing each element in the length direction of the banded sequence feature with each element in the length direction of the standard banded sequence feature. The fourth feature similarity between the banded sequence feature and the standard banded sequence feature is determined by comparing each element in the width direction of the banded sequence feature with each element in the width direction of the standard banded sequence feature. The third feature similarity and the fourth feature similarity are fused to obtain the first feature similarity.

4. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 3.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 3.

6. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 3.

Images