Training methods, devices, equipment, and media for chromosome analysis models

Abstract

This invention discloses a training method, apparatus, electronic device, and storage medium for a chromosome analysis model. The method includes: acquiring a raw chromosome image to be labeled; wherein the raw chromosome image includes at least one raw individual chromosome; identifying each raw individual chromosome in the raw chromosome image based on a pre-stored individual chromosome template image to obtain a raw individual chromosome image, and determining the chromosome type corresponding to each raw individual chromosome; determining a raw mask image corresponding to the raw chromosome image based on the raw individual chromosome image; generating annotation information corresponding to the raw chromosome image based on the raw mask image and the chromosome type; generating training samples based on the annotation information and the raw chromosome image, and training a pre-established chromosome analysis model based on the training samples. The technical solution of this invention can improve annotation efficiency and the accuracy of training results.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of computer technology, and in particular to a training method, apparatus, device and medium for a chromosome analysis model. Background Technology

[0002] Currently, pre-trained neural network models are typically used as chromosome analysis models to analyze and identify chromosome karyotypes in images.

[0003] To improve the accuracy of chromosome analysis models, a large number of training samples are needed to train the neural network model. In existing technologies, when constructing the training sample set, annotators need to manually perform chromosome karyotype classification and masking operations on multiple actually acquired chromosome images to obtain annotation information. Based on the annotation information and chromosome images, training samples are constructed for training the neural network model.

[0004] However, in the process of realizing the present invention, it was found that the prior art has at least the following technical problems: due to the large number of chromosome images, the workload of markers increases, time is long, and annotation efficiency is low; and the manual annotation process is prone to errors, resulting in poor accuracy of the annotation information. Summary of the Invention

[0005] This invention provides a training method, apparatus, electronic device, and storage medium for a chromosome analysis model, aiming to improve annotation efficiency and accuracy, reduce time and resource waste, and improve the accuracy of training results.

[0006] According to one aspect of the present invention, a method for training a chromosome analysis model is provided, comprising:

[0007] Obtain the original chromosome image to be labeled; wherein the original chromosome image includes at least one original individual chromosome;

[0008] Based on pre-stored individual chromosome template images, each original individual chromosome in the original chromosome image is identified to obtain an original individual chromosome image, and the chromosome type corresponding to each original individual chromosome is determined.

[0009] Based on the original individual chromosome image, an original mask image corresponding to the original chromosome image is determined; based on the original mask image and the chromosome type, annotation information corresponding to the original chromosome image is generated.

[0010] Training samples are generated based on the annotation information and the original chromosome image, and the pre-established chromosome analysis model is trained based on the training samples.

[0011] According to another aspect of the present invention, a training apparatus for a chromosome analysis model is provided, the apparatus comprising:

[0012] The original chromosome image acquisition module is used to acquire the original chromosome image to be labeled; wherein, the original chromosome image includes at least one original individual chromosome;

[0013] The original individual chromosome identification module is used to identify each original individual chromosome in the original chromosome image based on a pre-stored individual chromosome template image, so as to obtain the original individual chromosome image and determine the chromosome type corresponding to each original individual chromosome;

[0014] The annotation information generation module is used to determine the original mask image corresponding to the original chromosome image based on the original individual chromosome image, and to generate annotation information corresponding to the original chromosome image based on the original mask image and the chromosome type;

[0015] The training sample generation module is used to generate training samples based on the annotation information and the original chromosome image, so as to train the pre-established chromosome analysis model based on the training samples.

[0016] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0017] At least one processor; and

[0018] A memory communicatively connected to the at least one processor; wherein,

[0019] The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the training method of the chromosome analysis model according to any embodiment of the present invention.

[0020] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the training method of the chromosome analysis model according to any embodiment of the present invention.

[0021] The technical solution of this invention involves acquiring a raw chromosome image to be labeled. The raw chromosome image includes at least one raw individual chromosome. Based on a pre-stored individual chromosome template image, each raw individual chromosome in the raw chromosome image is identified to obtain a raw individual chromosome image, and the chromosome type corresponding to each raw individual chromosome is determined. A raw mask image corresponding to the raw chromosome image is determined based on the raw individual chromosome image. Based on the raw mask image and the chromosome type, annotation information corresponding to the raw chromosome image is generated. This completes the annotation operation of the raw chromosome image, and training samples are generated based on the annotation information and the raw chromosome image. These training samples are then used to train a pre-established chromosome analysis model. This technical solution achieves automatic annotation of raw chromosome images using individual chromosome template images, eliminating the need for manual intervention, thus improving annotation efficiency and accuracy, reducing wasted time and resources, and improving the accuracy of training results.

[0022] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart of a training method for a chromosome analysis model according to an embodiment of the present invention;

[0025] Figure 2 This is a schematic diagram of an individual chromosome template image provided according to an embodiment of the present invention;

[0026] Figure 3 This is a schematic diagram of an original chromosome image provided according to an embodiment of the present invention;

[0027] Figure 4 This is a schematic diagram of an original mask image provided according to an embodiment of the present invention;

[0028] Figure 5 This is a schematic diagram of a group of images provided according to an embodiment of the present invention;

[0029] Figure 6 This is a schematic diagram of a set of individual template mask images provided according to an embodiment of the present invention;

[0030] Figure 7 This is a schematic diagram of the structure of a training device for a chromosome analysis model according to an embodiment of the present invention;

[0031] Figure 8 This is a schematic diagram of the structure of an electronic device that implements the training method of the chromosome analysis model in this embodiment of the invention. Detailed Implementation

[0032] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0033] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "etc.", and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0034] Figure 1 This is a flowchart of a training method for a chromosome analysis model according to an embodiment of the present invention. The method can be executed by a chromosome analysis model training device, which can be implemented in hardware and / or software.

[0035] like Figure 1 As shown, the method in this embodiment may specifically include:

[0036] S110. Obtain the original chromosome image to be labeled.

[0037] The original chromosome image includes at least one original individual chromosome. The original chromosome image can be a real chromosome image acquired under a microscope. In this embodiment, the real chromosome image acquired as a training sample can be identified as the original chromosome image. In specific implementations, multiple original chromosome images can be acquired simultaneously, or they can be acquired one by one.

[0038] S120. Based on the pre-stored individual chromosome template image, identify each original individual chromosome in the original chromosome image to obtain the original individual chromosome image and determine the chromosome type corresponding to each original individual chromosome.

[0039] The individual chromosome template image is an image of an individual's chromosomes with a standard karyotype, serving as the "gold standard" image for doctors. It should be noted that a healthy individual's normal cells have 23 pairs of chromosomes, including 22 pairs of autosomes named 1 through 22, and one pair of sex chromosomes, totaling 46 chromosomes. Therefore, the individual chromosome template image can include 23 pairs.

[0040] Figure 2 This is a schematic diagram of an individual chromosome template image provided according to an embodiment of the present invention. Figure 2 It includes multiple individual chromosome template images. Images of different shapes represent chromosomes of different types with different karyotypes. Two images with the same shape but different colors can be regarded as images corresponding to a pair of chromosomes.

[0041] In this embodiment, based on the image features of the individual chromosome template image, feature recognition can be performed on each original individual chromosome in the original chromosome image, thereby identifying at least one original individual chromosome contained in the original chromosome image, and obtaining the corresponding original individual chromosome image based on the original individual chromosome.

[0042] Furthermore, when storing an individual chromosome template image, the corresponding individual chromosome type can be obtained simultaneously. Therefore, when performing feature recognition on the original individual chromosomes based on the individual chromosome template image, the chromosome type corresponding to the original individual chromosomes can be directly determined. The chromosome type includes: the types corresponding to the 22 pairs of autosomes named from chromosome 1 to chromosome 22, and the type corresponding to one pair of sex chromosomes.

[0043] In specific implementation, the method of identifying each original individual chromosome in the original chromosome image based on the pre-stored individual chromosome template image includes: performing image segmentation operation on each original individual chromosome in the original chromosome image based on the threshold segmentation algorithm and the individual chromosome template image, and determining the original individual chromosome image in the original chromosome image based on the segmentation result.

[0044] Specifically, the original individual chromosomes with the same shape as the individual chromosome template image can be identified in the original chromosome image. A threshold segmentation algorithm is then used to segment the original individual chromosomes in the original chromosome image, and the segmented image is determined as the original individual chromosome image corresponding to the individual chromosome template image.

[0045] For example, based on the chromosome template image of individual chromosome 1, the original chromosome of individual chromosome 1 can be identified in the original chromosome image. The original chromosome of individual chromosome 1 can then be segmented in the original chromosome image using a threshold segmentation algorithm to obtain the original chromosome image of individual chromosome 1.

[0046] In this embodiment, by using a threshold segmentation algorithm and individual chromosome template images, each original individual chromosome can be quickly and accurately identified in the original chromosome image, and the original individual chromosome image can be segmented.

[0047] S130. Determine the original mask image corresponding to the original chromosome image based on the original individual chromosome image, and generate the annotation information corresponding to the original chromosome image based on the original mask image and chromosome type.

[0048] The original mask image is the mask image corresponding to the original chromosome image, and the annotation information includes information such as the chromosome type corresponding to each original individual chromosome image in the original mask image and the original chromosome image.

[0049] In this embodiment, when determining the original mask image, the original mask image can be determined by binarizing the original individual chromosome images in the original chromosome image and then determining the original mask image based on the image obtained after binarization.

[0050] Figure 3 This is a schematic diagram of an original chromosome image provided according to an embodiment of the present invention; as shown. Figure 3 As shown, the original individual chromosome images in the original chromosome image have the same shape as the individual chromosome template image, but differ in size and rotation direction. Figure 4 This is a schematic diagram of an original mask image provided according to an embodiment of the present invention; Figure 4 Each mask image in the image has the same shape and size as the original individual chromosome image. Different shades are used to represent the mask images corresponding to the binarized original individual chromosome images.

[0051] In this embodiment, determining the original mask image corresponding to the original individual chromosome image based on the original individual chromosome image includes: determining the individual chromosome template image corresponding to the original individual chromosome based on the chromosome type corresponding to each original individual chromosome in the original chromosome image; constructing a group image corresponding to the original chromosome image based on the determined individual chromosome template image; determining the similarity transformation matrix between the individual chromosome template image in the group image and the original individual chromosome image for each original individual chromosome image; generating an individual template mask image corresponding to each individual chromosome template image in the group image; and performing a similarity transformation on the individual template mask image corresponding to the original individual chromosome image based on the similarity transformation matrix corresponding to the original individual chromosome image to obtain the original mask image.

[0052] In practice, an individual chromosome template image corresponding to the original individual's chromosome can be determined. That is, based on the chromosome type of the original individual's chromosome, an individual chromosome template image of the same chromosome type is obtained. For example, Figure 3 The original individual chromosomes with a round karyotype should correspond to... Figure 2 A template image of an individual chromosome with a round karyotype.

[0053] Specifically, group images corresponding to the original chromosome images can be constructed based on the determined individual chromosome template images. Figure 5 This is a schematic diagram of a group of images provided according to an embodiment of the present invention; as shown. Figure 5 As shown, the group image includes at least one individual chromosome template image for each chromosome type contained in the original chromosome image. It should be noted that the individual chromosome template images in the group image can be sorted column-wise, and each original individual chromosome image in the original chromosome image can be considered as an image obtained after similarity transformation processing of the group image. In this embodiment, similarity transformation processing can be used to perform rotation, scaling, and translation operations on each individual chromosome template image in the group image to obtain the original chromosome image. Therefore, for each original individual chromosome image, a similarity transformation matrix can be determined between the individual chromosome template image in a group image and the original individual chromosome image.

[0054] In practice, an individual template mask image is generated corresponding to the chromosome template image of each individual in the group image. That is, the corresponding individual template mask image can be obtained by binarizing the chromosome template image of each individual in the group image. Figure 6 This is a schematic diagram of a set of individual template mask images provided according to an embodiment of the present invention; Figure 6The image is composed of individual template mask images, which are sorted column-wise according to their order in the group of images. Based on the above explanation, each original individual chromosome image in the original chromosome image can be considered as an image obtained after a similarity transformation of the group of images. Therefore, the original mask image can be considered as an image obtained after a similarity transformation of the image composed of individual template mask images. Thus, when determining the original mask image, a similarity transformation can be performed on the individual template mask images corresponding to the original individual chromosome images based on the similarity transformation matrix to obtain the original mask image.

[0055] For example, Figure 5 The first somatic chromosome template image, which is pentagonal in shape, can be transformed by similarity transformation matrix 1 to obtain... Figure 3 The first image of the original individual's chromosome, corresponding to the first row and second column, is pentagonal in shape. Therefore, after determining... Figure 6 After obtaining the pentagonal shape of the first somatic chromosome template mask image corresponding to the first somatic chromosome template image, a similarity transformation matrix 1 can be applied to the first somatic template mask image to obtain... Figure 4 The mask image corresponding to the individual chromosome with a pentagonal shape in the first row and second column of the original mask image. Therefore, the mask image corresponding to each individual chromosome in the original mask image can be obtained by performing a similarity transformation on the individual template mask image corresponding to the original individual chromosome image.

[0056] In this embodiment, by determining the similarity transformation matrix corresponding to the original individual chromosome image and the individual template mask image, the original mask image can be quickly obtained through similarity transformation processing, thereby improving the effectiveness and accuracy of determining the original mask image.

[0057] In specific implementation, the method for determining the similarity transformation matrix between the individual chromosome template image in the group image and the original individual chromosome image can be as follows: Based on the feature detection algorithm, determine the first feature points in the individual chromosome template image in the group image and the matching second feature points in the original individual chromosome image; based on the correspondence between the first feature points and the second feature points, determine the feature matching relationship between the individual chromosome template image in the group image and the original individual chromosome image; based on the feature matching relationship, determine the similarity transformation matrix between the individual chromosome template image in the group image and the original individual chromosome image.

[0058] The feature detection algorithm can be the ORB (Oriented Fast and Rotated Brief) feature detection algorithm. ORB is an image local feature extraction algorithm that is scale- and rotation-invariant, and its speed and robustness can meet the real-time requirements of the system. This algorithm uses Oriented Fast to detect feature points and Rotation Brief to generate descriptors. After detecting feature points and generating descriptors, it uses brute-force matching to find the closest feature points between two images.

[0059] Specifically, a feature detection algorithm can be used to identify at least one feature point on the individual chromosome template image in each set of images, designated as a first feature point; for example, three or more feature points are identified on each individual chromosome template image as first feature points. Then, for each first feature point, a matching second feature point is identified in the corresponding original individual chromosome images. It should be noted that for each first feature point, one matching second feature point can be determined. Through the correspondence between each first and second feature point, the feature matching relationship between each individual chromosome template image and the original individual chromosome image in the set of images can be determined, thereby determining a similarity transformation matrix based on the feature matching relationship. For example, each original individual chromosome image corresponds to one similarity transformation matrix.

[0060] In this embodiment, a feature detection algorithm is used to quickly and accurately determine the matching relationship between the first feature point and the second feature point. Based on the matching relationship between the points, the similarity transformation matrix is ​​determined, which helps to improve the accuracy and effectiveness of determining the similarity transformation matrix.

[0061] Optionally, before determining the feature matching relationship between the individual chromosome template image and the original individual chromosome image, the method further includes: determining the first mismatch point in each first feature point and the second mismatch point corresponding to the first mismatch point in the second feature points based on the random sampling consensus algorithm; filtering out the first mismatch point in the first feature points and updating the first feature points based on the first remaining points after filtering out the first mismatch point; filtering out the second mismatch point in the second feature points and updating the second feature points based on the second remaining points after filtering out the second mismatch point.

[0062] To improve the accuracy of matching relationships between feature points, a filtering operation can be performed before determining the feature matching relationship to reduce the impact of mismatches on the accuracy of the results. Specifically, the Random Sample Consensus (RANSAC) algorithm can be used to filter out the first mismatches in the first feature points and the second mismatches on the original individual chromosome image. To avoid the impact of mismatches on matching accuracy, the first mismatches are removed from the first feature points, and the remaining feature points are determined as the first residual points, and the first feature points are updated based on the first residual points. Similarly, the second mismatches are removed from the second feature points, and the remaining feature points are determined as the second residual points, and the second feature points are updated based on the second residual points.

[0063] In this embodiment, by filtering out the first and second mismatch points, the accuracy of the determined matching results is improved, and the stability of the feature detection results of the feature detection algorithm is enhanced.

[0064] In specific implementation, after determining the similarity transformation matrix between the individual chromosome template image and the original individual chromosome image in the group image, the process also includes: determining each scaling parameter in each similarity transformation matrix; determining the abnormal parameters in each scaling parameter based on the absolute median difference algorithm; determining the abnormal matching relationship between the individual chromosome template image and the original individual chromosome image in the group image based on the abnormal matching relationship; performing a removal operation on the individual chromosome template image and the original individual chromosome image in the group image based on the removal result; and updating the similarity transformation matrix based on the removal result.

[0065] It should be noted that the similarity transformation matrix may include scaling parameters to reflect the scaling relationship between the individual chromosome template images in the group images and the original individual chromosome images.

[0066] In this embodiment, to further improve the accuracy and stability of the determined similarity transformation matrix, the scaling parameters in the similarity transformation matrix can be constrained. Specifically, for each original individual chromosome, the similarity transformation matrix corresponding to the original individual chromosome image and the scaling parameters in the similarity transformation matrix can be determined. For each determined scaling parameter, parameter detection is performed based on the Median Absolute Deciation (MAD) algorithm to identify outlier parameters in the scaling parameters.

[0067] Specifically, the similarity transformation matrix corresponding to the abnormal parameters can be regarded as an abnormal matching relationship; the individual chromosome template image and the original individual chromosome image corresponding to the abnormal matching relationship are removed, and the matching is re-performed based on the removed individual chromosome template image and the original individual chromosome image, and the similarity transformation matrix is ​​updated based on the feature matching relationship obtained by the re-matching.

[0068] In this embodiment, the absolute median difference algorithm can quickly identify abnormal parameters, thereby removing the individual chromosome template image and the original individual chromosome image corresponding to the abnormal parameters, so as to further improve the accuracy of the determined similarity transformation matrix.

[0069] In a specific implementation, after determining each first feature point in the individual chromosome template image in the group image and the matching second feature point in the original individual chromosome image, the process includes: determining at least two overlapping individual chromosome images in the original chromosome image; determining the number of feature points of the matching second feature point in each overlapping individual chromosome image; determining the maximum value among the number of feature points, and deleting overlapping individual chromosome images with a feature point count less than the maximum value in the original chromosome image; and updating the first and second feature points based on the original chromosome image after deleting the overlapping individual chromosome images.

[0070] To improve the effectiveness of training samples, at least two overlapping individual chromosome images in the original chromosome images can be deduplicated. Overlapping individual chromosome images are those with the same karyotype and size.

[0071] Specifically, the number of feature points for the second feature point corresponding to each overlapping individual chromosome image can be determined. It should be noted that the larger the number of feature points, the higher the confidence level; during the deduplication process, overlapping individual chromosome images with lower confidence levels can be deleted.

[0072] In this embodiment, the maximum value of the number of feature points can be determined, and any overlapping individual chromosome image with a feature point count less than the maximum value can be deleted. Alternatively, each overlapping individual chromosome image with a feature point count less than the maximum value can be deleted from the original chromosome image, thereby completing the deduplication operation on the original individual chromosome images. Based on the original chromosome image after deduplication of overlapping individual chromosome images, new first feature points and new second feature points are re-determined. Therefore, this embodiment, by performing deduplication on overlapping individual chromosome images, helps improve the effectiveness and diversity of training samples, and contributes to improving training accuracy.

[0073] S140. Generate training samples based on annotation information and original chromosome images, and train the pre-established chromosome analysis model based on the training samples.

[0074] Specifically, at least one original chromosome image and its corresponding annotation information can be used to generate training samples. The annotation information includes the original mask image corresponding to the original chromosome image, and the chromosome type corresponding to each original individual chromosome image within the original chromosome image. In this embodiment, a pre-established chromosome analysis model can be trained based on the training samples; wherein the chromosome analysis model can be a pre-established convolutional neural network model.

[0075] The technical solution of this invention involves acquiring a raw chromosome image to be labeled. The raw chromosome image includes at least one raw individual chromosome. Based on a pre-stored individual chromosome template image, each raw individual chromosome in the raw chromosome image is identified to obtain a raw individual chromosome image, and the chromosome type corresponding to each raw individual chromosome is determined. A raw mask image corresponding to the raw chromosome image is determined based on the raw individual chromosome image. Based on the raw mask image and the chromosome type, annotation information corresponding to the raw chromosome image is generated. This completes the labeling operation of the raw chromosome image, and training samples are generated based on the annotation information and the raw chromosome image. These training samples are then used to train a pre-established chromosome analysis model. The automatic labeling process of the raw chromosome image is achieved using individual chromosome template images, eliminating the need for manual intervention, which helps improve labeling efficiency and accuracy, reduces wasted time and resources, and improves the accuracy of training results.

[0076] Figure 7 This is a schematic diagram of a training device for a chromosome analysis model according to an embodiment of the present invention. This device is used to execute the training method for the chromosome analysis model provided in any of the above embodiments. This device and the training methods for the chromosome analysis models in the above embodiments belong to the same inventive concept. Details not described in detail in the embodiments of the chromosome analysis model training device can be found in the embodiments of the chromosome analysis model training methods described above. Figure 7 As shown, the device includes:

[0077] The original chromosome image acquisition module 10 is used to acquire the original chromosome image to be labeled; wherein, the original chromosome image includes at least one original individual chromosome;

[0078] The original individual chromosome identification module 11 is used to identify each original individual chromosome in the original chromosome image based on the pre-stored individual chromosome template image, so as to obtain the original individual chromosome image and determine the chromosome type corresponding to each original individual chromosome.

[0079] The annotation information generation module 12 is used to determine the original mask image corresponding to the original chromosome image based on the original individual chromosome image, and to generate annotation information corresponding to the original chromosome image based on the original mask image and chromosome type;

[0080] The training sample generation module 13 is used to generate training samples based on annotation information and original chromosome images, so as to train the pre-established chromosome analysis model based on the training samples.

[0081] Based on any optional technical solution in the embodiments of the present invention, the original individual chromosome identification module 11 optionally includes:

[0082] The image segmentation unit is used to perform image segmentation on each original individual chromosome in the original chromosome image based on the threshold segmentation algorithm and the individual chromosome template image, and to determine the original individual chromosome image based on the segmentation results.

[0083] Based on any optional technical solution in the embodiments of the present invention, optionally, the annotation information generation module 12 includes:

[0084] The individual chromosome template image determination unit is used to determine the individual chromosome template image corresponding to the original individual chromosome based on the chromosome type corresponding to each original individual chromosome in the original chromosome image, and to construct a group image corresponding to the original chromosome image based on the determined individual chromosome template image;

[0085] The similarity transformation matrix determination unit is used to determine the similarity transformation matrix between the individual chromosome template image in the group image and the original individual chromosome image for each original individual chromosome image;

[0086] The individual template mask image generation unit is used to generate an individual template mask image corresponding to the individual chromosome template image of each individual in the group image. Based on the similarity transformation matrix corresponding to the original individual chromosome image, the unit performs a similarity transformation on the individual template mask image corresponding to the original individual chromosome image to obtain the original mask image.

[0087] Based on any optional technical solution in the embodiments of the present invention, the optional similarity transformation matrix determination unit includes:

[0088] The second feature point determination subunit is used to determine, based on the feature detection algorithm, each first feature point in the individual chromosome template image in the group image and the second feature point that matches it in the original individual chromosome image.

[0089] The feature matching relationship determination subunit is used to determine the feature matching relationship between the individual chromosome template image and the original individual chromosome image in the group image based on the correspondence between the first feature point and the second feature point;

[0090] The similarity transformation matrix determines the sub-unit, which is used to determine the similarity transformation matrix between the individual chromosome template image and the original individual chromosome image in the group image based on the feature matching relationship.

[0091] Based on any optional technical solution in the embodiments of the present invention, the similarity transformation matrix determination unit may optionally further include:

[0092] The second mismatch point determination subunit is used to determine the first mismatch point in each first feature point and the second mismatch point in the second feature points corresponding to the first mismatch point before determining the feature matching relationship between the individual chromosome template image and the original individual chromosome image, based on the random sampling consensus algorithm.

[0093] The second feature point update subunit is used to filter out the first mismatched point from the first feature points, update the first feature point based on the first remaining point after filtering, filter out the second mismatched point from the second feature points, and update the second feature point based on the second remaining point after filtering.

[0094] Based on any optional technical solution in the embodiments of the present invention, the similarity transformation matrix determination unit may optionally further include:

[0095] The scaling parameter determination subunit is used to determine each scaling parameter in each similarity transformation matrix after determining the similarity transformation matrix between the individual chromosome template image and the original individual chromosome image in the group image;

[0096] The abnormal parameter determination subunit is used to determine the abnormal parameters among the scaling parameters based on the absolute median difference algorithm.

[0097] The abnormal matching relationship determination subunit is used to determine the abnormal matching relationship between the individual chromosome template image and the original individual chromosome image in the group image based on the abnormal parameters;

[0098] The image removal subunit is used to remove individual chromosome template images and original individual chromosome images in the group images based on abnormal matching relationships, and to update the similarity transformation matrix based on the removal results.

[0099] Based on any optional technical solution in the embodiments of the present invention, the similarity transformation matrix determination unit may optionally further include:

[0100] The overlapping individual chromosome image determination subunit is used to determine at least two overlapping individual chromosome images in the original chromosome image after each first feature point in the individual chromosome template image in the determination group image and a second feature point that matches in the original individual chromosome image.

[0101] The feature point number determination subunit is used to determine the number of feature points of the second feature point matched in each overlapping individual chromosome image;

[0102] The feature point update subunit is used to determine the maximum value among the number of each feature point, and to delete overlapping individual chromosome images with fewer than the maximum value in the original chromosome image; based on the original chromosome image after deleting the overlapping individual chromosome images, the first feature point and the second feature point are updated.

[0103] The technical solution of this invention involves acquiring a raw chromosome image to be labeled. The raw chromosome image includes at least one raw individual chromosome. Based on a pre-stored individual chromosome template image, each raw individual chromosome in the raw chromosome image is identified to obtain a raw individual chromosome image, and the chromosome type corresponding to each raw individual chromosome is determined. A raw mask image corresponding to the raw chromosome image is determined based on the raw individual chromosome image. Based on the raw mask image and the chromosome type, annotation information corresponding to the raw chromosome image is generated. This completes the labeling operation of the raw chromosome image, and training samples are generated based on the annotation information and the raw chromosome image. These training samples are then used to train a pre-established chromosome analysis model. The automatic labeling process of the raw chromosome image is achieved using individual chromosome template images, eliminating the need for manual intervention, which helps improve labeling efficiency and accuracy, reduces wasted time and resources, and improves the accuracy of training results.

[0104] It is worth noting that in the embodiments of the training device for the above chromosome analysis model, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy differentiation and are not used to limit the scope of protection of the present invention.

[0105] Figure 8This is a schematic diagram of the structure of an electronic device that implements the training method of the chromosome analysis model according to embodiments of the present invention. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0106] like Figure 8 As shown, the electronic device 20 includes at least one processor 21 and a memory, such as a read-only memory (ROM) 22 or a random access memory (RAM) 23, communicatively connected to the at least one processor 21. The memory stores computer programs executable by the at least one processor. The processor 21 can perform various appropriate actions and processes based on the computer program stored in the ROM 22 or loaded from storage unit 28 into the RAM 23. The RAM 23 can also store various programs and data required for the operation of the electronic device 20. The processor 21, ROM 22, and RAM 23 are interconnected via a bus 24. An input / output (I / O) interface 25 is also connected to the bus 24.

[0107] Multiple components in electronic device 20 are connected to I / O interface 25, including: input unit 26, such as keyboard, mouse, etc.; output unit 27, such as various types of monitors, speakers, etc.; storage unit 28, such as disk, optical disk, etc.; and communication unit 29, such as network card, modem, wireless transceiver, etc. Communication unit 29 allows electronic device 20 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0108] Processor 21 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 21 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 21 performs the various methods and processes described above, such as the training methods for chromosome analysis models.

[0109] In some embodiments, the training method for the chromosome analysis model can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 28. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 20 via ROM 22 and / or communication unit 29. When the computer program is loaded into RAM 23 and executed by processor 21, one or more steps of the training method for the chromosome analysis model described above can be performed. Alternatively, in other embodiments, processor 21 can be configured to execute the training method for the chromosome analysis model by any other suitable means (e.g., by means of firmware).

[0110] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0111] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0112] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0113] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0114] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0115] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0116] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0117] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A training method for a chromosome analysis model, characterized in that, include: Obtain the original chromosome image to be labeled; wherein the original chromosome image includes at least one original individual chromosome; Based on pre-stored individual chromosome template images, each original individual chromosome in the original chromosome image is identified to obtain an original individual chromosome image, and the chromosome type corresponding to each original individual chromosome is determined. Based on the original individual chromosome image, an original mask image corresponding to the original chromosome image is determined; based on the original mask image and the chromosome type, annotation information corresponding to the original chromosome image is generated. Training samples are generated based on the annotation information and the original chromosome image, and the pre-established chromosome analysis model is trained based on the training samples. The step of determining the original mask image corresponding to the original chromosome image based on the original individual chromosome image includes: Based on the chromosome type corresponding to each original individual chromosome in the original chromosome image, an individual chromosome template image corresponding to the original individual chromosome is determined, and a group image corresponding to the original chromosome image is constructed based on the determined individual chromosome template image; For each of the original individual chromosome images, a similarity transformation matrix is ​​determined between the individual chromosome template image in the group of images and the original individual chromosome image; Generate an individual template mask image corresponding to each individual chromosome template image in the group of images. Based on the similarity transformation matrix corresponding to the original individual chromosome image, perform a similarity transformation on the individual template mask image corresponding to the original individual chromosome image to obtain the original mask image.

2. The method according to claim 1, characterized in that, The process of identifying each original individual chromosome in the original chromosome image based on a pre-stored individual chromosome template image includes: Based on the threshold segmentation algorithm and the individual chromosome template image, image segmentation operation is performed on each of the original individual chromosomes in the original chromosome image, and the original individual chromosome images in the original chromosome image are determined based on the segmentation results.

3. The method according to claim 1, characterized in that, Determining the similarity transformation matrix between the individual chromosome template image in the group of images and the original individual chromosome image includes: Based on the feature detection algorithm, each first feature point in the individual chromosome template image in the group of images is determined, and the second feature point that matches it in the original individual chromosome image is determined. Based on the correspondence between the first feature point and the second feature point, the feature matching relationship between the individual chromosome template image and the original individual chromosome image in the group of images is determined; Based on the feature matching relationship, a similarity transformation matrix is ​​determined between the individual chromosome template image in the group of images and the original individual chromosome image.

4. The method according to claim 3, characterized in that, Before determining the feature matching relationship between the individual chromosome template image and the original individual chromosome image, the process also includes: Based on the random sampling consensus algorithm, the first mismatch point in each of the first feature points and the second mismatch point in the second feature points corresponding to the first mismatch point are determined; The first mismatched point is removed from the first feature point, and the first feature point is updated based on the first remaining point after removal. The second mismatched point is removed from the second feature point, and the second feature point is updated based on the second remaining point after removal.

5. The method according to claim 3, characterized in that, After determining the similarity transformation matrix between the individual chromosome template image in the group of images and the original individual chromosome image, the method further includes: Determine each scaling parameter in each of the aforementioned similarity transformation matrices; Based on the absolute median difference algorithm, abnormal parameters among the scaling parameters are determined. Based on the abnormal parameters, the abnormal matching relationship between the individual chromosome template image and the original individual chromosome image in the group of images is determined; Based on the abnormal matching relationship, the individual chromosome template image and the original individual chromosome image in the group of images are removed, and the similarity transformation matrix is ​​updated based on the removal result.

6. The method according to claim 3, characterized in that, After determining each first feature point in the individual chromosome template image in the group of images, and the second feature point that matches it in the original individual chromosome image, the process includes: Identify at least two overlapping individual chromosome images in the original chromosome image; Determine the number of feature points of the second feature point matched in the chromosome image of each of the overlapping individuals; Determine the maximum value among the number of each feature point, and delete overlapping individual chromosome images in the original chromosome image whose number of feature points is less than the maximum value; Based on the original chromosome image after deleting the overlapping individual chromosome images, the first feature point and the second feature point are updated.

7. A training device for a chromosome analysis model, characterized in that, include: The original chromosome image acquisition module is used to acquire the original chromosome image to be labeled; wherein, the original chromosome image includes at least one original individual chromosome; The original individual chromosome identification module is used to identify each original individual chromosome in the original chromosome image based on a pre-stored individual chromosome template image, so as to obtain the original individual chromosome image and determine the chromosome type corresponding to each original individual chromosome; The annotation information generation module is used to determine the original mask image corresponding to the original chromosome image based on the original individual chromosome image, and to generate annotation information corresponding to the original chromosome image based on the original mask image and the chromosome type; The training sample generation module is used to generate training samples based on the annotation information and the original chromosome image, so as to train the pre-established chromosome analysis model based on the training samples; The annotation information generation module includes: The individual chromosome template image determination unit is used to determine an individual chromosome template image corresponding to the original individual chromosome based on the chromosome type corresponding to each original individual chromosome in the original chromosome image, and to construct a group image corresponding to the original chromosome image based on the determined individual chromosome template image; The similarity transformation matrix determination unit is used to determine, for each of the original individual chromosome images, the similarity transformation matrix between the individual chromosome template image in the group of images and the original individual chromosome image; An individual template mask image generation unit is used to generate an individual template mask image corresponding to each individual chromosome template image in the group of images, and to perform a similarity transformation on the individual template mask image corresponding to the original individual chromosome image based on the similarity transformation matrix corresponding to the original individual chromosome image to obtain the original mask image.

8. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor to enable the at least one processor to perform the training method of the chromosome analysis model according to any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a processor, implement the training method for the chromosome analysis model according to any one of claims 1-6.

Images