CN115619774B - Chromosome abnormality identification method, system and storage medium - Google Patents

Chromosome abnormality identification method, system and storage medium Download PDF

Info

Publication number
CN115619774B
CN115619774B CN202211465346.XA CN202211465346A CN115619774B CN 115619774 B CN115619774 B CN 115619774B CN 202211465346 A CN202211465346 A CN 202211465346A CN 115619774 B CN115619774 B CN 115619774B
Authority
CN
China
Prior art keywords
chromosome
abnormal
images
chromosomes
leukemia
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202211465346.XA
Other languages
Chinese (zh)
Other versions
CN115619774A (en
Inventor
卢沁阳
聂宇坤
徐思
刘丽珏
穆阳
彭伟雄
蔡昱峰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hunan Zixing Wisdom Medical Technology Co ltd
Original Assignee
Hunan Zixing Wisdom Medical Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hunan Zixing Wisdom Medical Technology Co ltd filed Critical Hunan Zixing Wisdom Medical Technology Co ltd
Priority to CN202211465346.XA priority Critical patent/CN115619774B/en
Publication of CN115619774A publication Critical patent/CN115619774A/en
Application granted granted Critical
Publication of CN115619774B publication Critical patent/CN115619774B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
    • G06V20/695—Preprocessing, e.g. image segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
    • G06V20/698—Matching; Classification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10056—Microscopic image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/03—Recognition of patterns in medical or anatomical images

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Medical Informatics (AREA)
  • Evolutionary Computation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Databases & Information Systems (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Quality & Reliability (AREA)
  • Radiology & Medical Imaging (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a method, a system and a storage medium for identifying chromosome abnormality, which are used for automatically identifying an abnormal chromosome image and relate to the technical field of molecular diagnostics; the method comprises the following steps: training the segmentation recognition model through the metaphase image of the leukemia type abnormal chromosome, and generating abnormal chromosomes in a large scale according to categories to obtain an automatic recognition abnormal chromosome model; then, selecting chromosomes with leukemia species specificity, inputting the chromosomes into an automatic abnormal chromosome recognition model for recognition, and outputting classification results corresponding to abnormal types; the invention selects three chromosome abnormality categories with leukemia category specificity, the three types of abnormalities have high relevance with Chronic Myelocytic Leukemia (CML), acute Myelocytic Leukemia (AML) and Acute Promyelocytic Leukemia (APL) respectively, and the three types of abnormal chromosomes can be identified more accurately, thereby bringing great value to diagnosis.

Description

Chromosome abnormality identification method, system and storage medium
Technical Field
The invention discloses a chromosome abnormality identification method, a system and a storage medium, which are used for automatic identification of an abnormal chromosome image and relate to the technical field of molecular diagnostics.
Background
Human somatic cells produce chromosomes in metaphase, normally 46 chromosomes (22 pairs of autosomes and a pair of sex chromosomes). Chromosomes act as carriers of genetic material, and abnormalities in their number or structure can lead to genetic disorders. The abnormal chromosome is caused by the exchange, deletion, repetition and the like of fragments on different types of normal chromosomes, finding the abnormal karyotype is an important work for analyzing the karyotype of the chromosome, and can help doctors to locate the abnormal type of patients, judge the genetic disease types of the patients and adopt a treatment scheme in time. Clinically, many different types of leukemia are associated with specific chromosomal abnormal karyotypes, e.g. abnormal karyotype t (9. If doctors can confirm the abnormal karyotype of the patients in time, the doctors can more quickly adopt proper treatment schemes to save the lives of the patients.
At present, there are many works in the industry to model chromosome abnormality recognition through a supervised deep neural network, for example, the publication number is CN115063412A, a chromosome image stitching method and a chromosome karyotype analysis method are disclosed, and the publication number is CN113989502A, and a chromosome segmentation recognition method based on a graph convolution neural network is provided, which provides convenience for chromosome analysis, but has many problems. On one hand, due to the fact that chromosome images of different sources and different band levels have large differences, abnormal chromosomes often appear on images which are difficult to analyze, such as large chromosome curvature and large overlapping degree, the problems that abnormal chromosome data are various and rare, and collection and labeling are difficult are caused, and therefore the neural network model is difficult to converge. On the other hand, the common deep neural network model is completely data-driven and lacks medical prior knowledge, which also leads to the deficiency of the prior art to some extent.
Disclosure of Invention
The invention aims to: in order to solve the problems of chromosome abnormality data deficiency and medical priori knowledge deficiency of a common deep neural network, the invention provides a deep neural network training method for chromosome abnormality recognition based on artificially generated abnormal chromosome data and combined with the banding information of chromosomes.
The technical scheme adopted by the invention is as follows: a method of identifying chromosomal abnormalities, comprising the steps of:
s1, training a segmentation recognition model through a metaphase image of the leukemia type abnormal chromosome, and generating abnormal chromosomes on a large scale according to types to obtain an automatic recognition abnormal chromosome model; the chromosome abnormality category in the automatic recognition abnormal chromosome model is provided with three branch structures of t (9, t (8), t (17;
s2, selecting chromosomes with leukemia species specificity, inputting the chromosomes into an automatic abnormal chromosome recognition model for recognition, and outputting classification results corresponding to abnormal types;
the specific steps of selecting chromosome with leukemia species specificity input are as follows:
s2.1, selecting two images of homologous chromosomes 9 and two images of homologous chromosomes 22 from each chromosome subgraph after the metaphase map segmentation and recognition, wherein the total number of the images is 4; or selecting two images of homologous chromosome 8 and two images of homologous chromosome 21, wherein the images are 4 images; or selecting images of two homologous chromosomes 15 and two homologous chromosomes 17, wherein the images are 4 images; the resolution of each subgraph is 224 multiplied by 224, and four graphs are stacked into a matrix of 224 multiplied by 4 to serve as input;
s2.2, reconstructing the image of each channel in the step S2.1 through a self-encoder of the image loss area, wherein the purpose is to enable the real image to be consistent with an artificially generated chromosome during training before entering the recognition network, and finally obtaining an input X _ in with the dimension of 224 multiplied by 4;
s2.3, respectively carrying out chromosome straightening, band-line extraction, band-line vectorization and probability normalization on chromosome images of four channels in X _ in through a traditional image processing algorithm;
s2.4, performing loss calculation on the probability feature vector and the V _ out output by the automatic identification abnormal chromosome model by using the relative entropy, and updating the automatic identification abnormal chromosome model by using a back propagation algorithm;
s2.5, after reasoning through the automatic identification abnormal chromosome model, three two-dimensional probability vectors are finally obtained, and distribution points to identification results of three abnormal types.
The chromosome abnormality categories with leukemia species specificity in the step S2 are t (9.
The specific steps of step S2.3 are as follows: after extracting the striae of the chromosome, using the average gray level of the striae as a value in a vector, and then performing probability normalization, wherein the probability normalization formula is as follows:
Figure 657215DEST_PATH_IMAGE001
the vector length is 50, the vector before probability normalization is recorded as f, and the target probability feature vector is recorded as V; f (x) represents the value of f at the x position, V (x) represents the value of V at the x position, and f (i) represents the value of f at the i position.
The formula of the relative entropy in step S2.4 is as follows:
Figure 535172DEST_PATH_IMAGE002
wherein KL represents the relative entropy, V represents the probability feature vector obtained in step 5, V _ out is the probability feature vector output by the model, and V (x) and V _ out (x) represent the values of the two vectors at the respective x positions, respectively.
A chromosome abnormality recognition system comprising:
at least one processor;
at least one memory for storing at least one program;
when the at least one program is executed by the at least one processor, the at least one program causes the at least one processor to implement a method of identifying a chromosome abnormality as described above.
A storage medium having stored therein processor-executable instructions for performing a method of identifying chromosome abnormalities as described above when executed by a processor.
In summary, due to the adoption of the technical scheme, the invention has the beneficial effects that:
the method extracts the chromosome banding information based on the traditional image algorithm, and skillfully integrates the chromosome banding information into the depth neural network; the model of the invention can automatically identify abnormal chromosomes with leukemia species specificity by integrating branch structures identified by various chromosome abnormal types and inputting chromosome images.
The deep neural network is trained by utilizing the generated scarce abnormal chromosome data, so that the model can accurately identify the abnormal chromosome, and the image recognition pressure of a doctor can be well relieved; the invention selects three types of chromosome abnormality types with leukemia species specificity, namely t (9, 22), t (8, 21) and t (15, 17), and the three types of abnormality are respectively highly associated with Chronic Myelogenous Leukemia (CML), acute Myelogenous Leukemia (AML) and Acute Promyelocytic Leukemia (APL) generally. If the three abnormal chromosomes can be accurately identified for doctors, great value is brought to diagnosis.
Drawings
The invention will now be described, by way of example, with reference to the accompanying drawings, in which:
FIG. 1 is a block diagram of the workflow of the present invention;
FIG. 2 is a diagram of a neural network model of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example 1
As shown in fig. 1-2, the method is based on artificially generated abnormal chromosome data, combines the banding information of chromosomes, and is used for a deep neural network training method for chromosome abnormality recognition, so as to solve the problems of chromosome abnormality data deficiency and medical priori knowledge deficiency of a common deep neural network.
A method for identifying a chromosome abnormality, as shown in fig. 2, taking a discrimination method of t (15:
s1, selecting two images of homologous chromosomes 15 and 17 from each chromosome subgraph subjected to metaphase map segmentation recognition by using the conventional segmentation recognition model, wherein the total number of the images is 4, the resolution of each subgraph is 224 multiplied by 224, and stacking the four images into a matrix of 224 multiplied by 4 to serve as input;
s2, respectively reconstructing the image of each channel in the S1 by using a self-encoder used for reconstructing an image loss area when a chromosome is generated in the traditional technology, wherein the purpose is to enable a real image to be consistent with an artificially generated chromosome during training before entering a recognition network, and finally obtaining an input X _ in with the dimension of 224 multiplied by 4;
s3, chromosome images of four channels in X _ in are subjected to chromosome straightening, band-line extraction, band-line vectorization and probability normalization in sequence through a traditional image processing algorithm respectively; specifically, after the banding of the chromosome is extracted, the average gray level of the banding is used as a value in a vector, and then probability normalization is performed, wherein the probability normalization formula is as follows:
Figure 137186DEST_PATH_IMAGE001
the vector length is 50, the vector before probability normalization is recorded as f, and the target probability feature vector is recorded as V; f (x) represents the value of f at the x position, V (x) represents the value of V at the x position, f (i) represents the value of f at the i position, and finally the vector V can be obtained from the vector f through the formula;
s4, because V obtained in S3 accords with discrete probability distribution, and the banding distribution of the chromosome has certain similarity with the probability distribution, namely banding disappearance at different positions has great difference on the banding characteristics of the chromosome, but the same banding disappearance in Euclidean space has the same distance (because each node of the vector is orthogonal); therefore, the relative entropy (K-L divergence) is used for carrying out loss calculation on V obtained in S3 and V _ out output by the neural network model, and the neural network model is updated by a back propagation algorithm; the formula for the relative entropy is as follows:
Figure 301451DEST_PATH_IMAGE003
wherein KL represents relative entropy, V represents probability feature vector, V _ out is probability feature vector output by the model, and V (x) and V _ out (x) respectively represent values of the two vectors at respective x positions;
s5, after reasoning of an automatic identification abnormal chromosome model, three two-dimensional probability vectors are finally obtained and respectively point to identification results of three abnormal types, and then output is extracted according to input chromosome types, in the example, the input is 15 and 17, and the example is a t (15) abnormal chromosome case, so that a third output vector representing a t (15) identification result is extracted, a first node value of the vector is 0.9, and a second node value of the vector is 0.1, so that the model output is't (15) abnormal'; during the training process, the gradient is also only propagated backwards from this. In addition, in the training process, a chromosome segmentation and recognition method is used for generating abnormal chromosomes in a large scale, and the abnormal chromosomes are combined with real normal chromosomes to form a training set for training of an automatic recognition abnormal chromosome model.
Example 2
A chromosome abnormality recognition method, taking a discrimination mode of t (9:
s1, selecting two images of homologous chromosomes 9 and 22 from chromosome subgraphs after a metaphase map is segmented and identified by using the conventional segmentation and identification model, wherein the total number of the images is 4, the resolution of each subgraph is 224 multiplied by 224, and stacking the four images into a matrix of 224 multiplied by 4 to serve as input;
s2, respectively reconstructing the image of each channel in the S1 by using a self-encoder used for reconstructing an image loss area when a chromosome is generated in the traditional technology, wherein the purpose is to enable a real image to be consistent with an artificially generated chromosome during training before entering a recognition network, and finally obtaining an input X _ in with the dimension of 224 multiplied by 4;
s3, chromosome images of four channels in X _ in are subjected to chromosome straightening, band-line extraction, band-line vectorization and probability normalization in sequence through a traditional image processing algorithm respectively; specifically, after the banding of the chromosome is extracted, the average gray level of the banding is used as a value in a vector, and then probability normalization is performed, wherein the probability normalization formula is as follows:
Figure 277497DEST_PATH_IMAGE001
the vector length is 50, the vector before probability normalization is recorded as f, and the target probability feature vector is recorded as V; f (x) represents the value of f at the x position, V (x) represents the value of V at the x position, f (i) represents the value of f at the i position, and finally the vector V can be obtained from the vector f through the formula;
s4, because V obtained in S3 accords with discrete probability distribution, and the banding distribution of the chromosome has certain similarity with the probability distribution, namely banding disappearance at different positions has great difference on the banding characteristics of the chromosome, but the same banding disappearance in Euclidean space has the same distance (because each node of the vector is orthogonal); therefore, the relative entropy (K-L divergence) is used to perform a loss calculation on V obtained at S3 and V _ out output by the neural network model, and the neural network model is updated by a back propagation algorithm. The formula for the relative entropy is as follows:
Figure 959146DEST_PATH_IMAGE003
wherein KL represents relative entropy, V represents probability feature vector, V _ out is probability feature vector output by the model, and V (x) and V _ out (x) respectively represent values of the two vectors at respective x positions;
s5, after reasoning through an automatic identification abnormal chromosome model, three two-dimensional probability vectors are finally obtained and respectively point to identification results of three abnormal types, and then output is extracted according to input chromosome types, in the example, the input is 9 and 22, and the example is a normal chromosome case, so that a first output vector representing the identification result of t (9) is extracted, the first node value of the vector is 0.2, the second node value of the vector is 0.8, and the output of the model is not t (9) abnormal; during the training process, the gradient is also only propagated backwards from this. In addition, in the training process, a chromosome segmentation recognition method is used for generating abnormal chromosomes in a large scale, and then the abnormal chromosomes and the real normal chromosomes are combined into a training set for training of an automatic recognition abnormal chromosome model.
Example 3
A method for identifying chromosome abnormality, taking a discrimination mode of t (8:
s1, selecting two images of homologous chromosomes 8 and 21 from each chromosome subgraph subjected to metaphase map segmentation recognition by using the conventional segmentation recognition model, wherein the total number of the images is 4, the resolution of each subgraph is 224 multiplied by 224, and stacking the four images into a matrix of 224 multiplied by 4 to serve as input;
s2, respectively reconstructing the image of each channel in the S1 by using a self-encoder used for reconstructing an image loss area when a chromosome is generated in the traditional technology, wherein the purpose is to enable a real image to be consistent with an artificially generated chromosome during training before entering a recognition network, and finally obtaining an input X _ in with the dimension of 224 multiplied by 4;
s3, chromosome images of four channels in the X _ in are subjected to chromosome straightening, band-line extraction, band-line vectorization and probability normalization in sequence through a traditional image processing algorithm respectively; specifically, after the banding of the chromosome is extracted, the average gray level of the banding is used as a value in a vector, and then probability normalization is performed, wherein the probability normalization formula is as follows:
Figure 399354DEST_PATH_IMAGE001
the vector length is 50, the vector before probability normalization is recorded as f, and the target probability feature vector is recorded as V; f (x) represents the value of f at the x position, V (x) represents the value of V at the x position, f (i) represents the value of f at the i position, and finally the vector V can be obtained from the vector f through the formula;
s4, because V obtained in S3 accords with discrete probability distribution, and the banding distribution of the chromosome has certain similarity with the probability distribution, namely banding disappearance at different positions has great difference on the banding characteristics of the chromosome, but the same banding disappearance in Euclidean space has the same distance (because each node of the vector is orthogonal); therefore, the relative entropy (K-L divergence) is used to perform the loss calculation on V obtained in step 3 and V _ out output by the neural network model, and the neural network model is updated by the back propagation algorithm. The formula for the relative entropy is as follows:
Figure 141045DEST_PATH_IMAGE003
wherein KL represents relative entropy, V represents the probability feature vector obtained in step 5, V _ out is the probability feature vector output by the model, and V (x) and V _ out (x) represent the values of the two vectors at respective x positions, respectively;
s5, after reasoning through an automatic identification abnormal chromosome model, three two-dimensional probability vectors are finally obtained and respectively point to identification results of three abnormal types, and then output is extracted according to input chromosome types, in the example, the input is No. 8 and No. 21, and the example is a t (8) abnormal chromosome case, so that a second output vector representing the identification result of t (8) 21 is extracted, the first node value of the vector is 0.9, the second node value is 0.1, and the model output is't (8) abnormal'; during the training process, the gradient is also only propagated backwards from this. In addition, in the training process, a chromosome segmentation recognition method is used for generating abnormal chromosomes in a large scale, and then the abnormal chromosomes and the real normal chromosomes are combined into a training set for training of an automatic recognition abnormal chromosome model.
Example 4
A chromosome abnormality recognition system comprising:
at least one processor;
at least one memory for storing at least one program;
when the at least one program is executed by the at least one processor, the at least one processor is caused to implement one of the chromosome abnormality recognition methods described above.
Example 5
A storage medium having stored therein processor-executable instructions for performing a method of identifying chromosome abnormalities as described above when executed by a processor.
Verification example
The test set used in the present invention was from 2693 real cases, a total of 11842 metaphase images, where there were 290 in t (9; finally, the sensitivity of the model trained by the method is 96% on t (9; sensitivity at t (8; sensitivity at t (15; the above indexes fully demonstrate the beneficial effects of the invention.
The above description is only a preferred embodiment of the present invention, and not intended to limit the present invention, the scope of the present invention is defined by the appended claims, and all structural changes that can be made by using the contents of the description and the drawings of the present invention are intended to be embraced therein.

Claims (3)

1. A method for identifying a chromosomal abnormality, comprising the steps of:
s1, training a segmentation recognition model through a leukemia type abnormal chromosome metaphase image, and generating abnormal chromosomes in a large scale according to categories to obtain an automatic recognition abnormal chromosome model; the chromosome abnormality category in the automatic identification abnormal chromosome model is provided with three branch structures of t (9; the types of chromosome abnormalities with leukemia species specificity are t (9, 22), t (8), t (15, 17), and the three types of abnormalities are generally highly correlated with Chronic Myelogenous Leukemia (CML), acute Myelogenous Leukemia (AML), and Acute Promyelocytic Leukemia (APL), respectively;
s2, selecting chromosomes with leukemia species specificity, inputting the chromosomes into an automatic abnormal chromosome recognition model for recognition, and outputting classification results corresponding to abnormal types;
the specific steps of selecting chromosome with leukemia species specificity input are as follows:
s2.1, selecting two images of homologous chromosomes 9 and 22 from each chromosome subgraph after the metaphase map is segmented and identified, wherein the images are 4 images; or selecting two images of homologous chromosome 8 and two images of homologous chromosome 21, wherein the images are 4 images; or selecting images of two homologous chromosomes 15 and 17, and taking 4 images; the resolution of each subgraph is 224 multiplied by 224, and four graphs are stacked into a matrix of 224 multiplied by 4 to serve as input;
s2.2, reconstructing the image of each channel in the step S2.1 through a self-encoder of the image loss area, wherein the purpose is to enable the real image to be consistent with an artificially generated chromosome during training before entering the recognition network, and finally obtaining an input X _ in with the dimension of 224 multiplied by 4;
s2.3, respectively carrying out chromosome straightening, band-line extraction, band-line vectorization and probability normalization on chromosome images of four channels in the X _ in through a traditional image processing algorithm; the method comprises the following specific steps: after extracting the striae of the chromosome, using the average gray level of the striae as a value in a vector, and then performing probability normalization, wherein the probability normalization formula is as follows:
Figure QLYQS_1
the vector length is 50, the vector before probability normalization is recorded as f, and the target probability feature vector is recorded as V; f (x) represents the value of f at the x position, V (x) represents the value of V at the x position, f (i) represents the value of f at the i position, and finally the vector V can be obtained from the vector f through the formula;
s2.4, performing loss calculation on the probability feature vector and the V _ out output by the automatic identification abnormal chromosome model by using the relative entropy, and updating the automatic identification abnormal chromosome model by using a back propagation algorithm; the formula of the relative entropy is as follows:
Figure QLYQS_2
wherein KL represents relative entropy, V represents probability feature vector, V _ out is probability feature vector of model output, and V (x) and V _ out (x) represent values of two vectors at respective x positions, respectively;
s2.5, after reasoning through the automatic identification abnormal chromosome model, three two-dimensional probability vectors are finally obtained, and distribution points to identification results of three abnormal types.
2. A system for identifying chromosomal abnormalities, comprising:
at least one processor;
at least one memory for storing at least one program;
when executed by the at least one processor, cause the at least one processor to implement a method of identifying chromosomal abnormalities as recited in claim 1.
3. A storage medium having stored therein instructions executable by a processor, the storage medium comprising: the processor-executable instructions, when executed by a processor, are for performing a method of chromosome abnormality identification as recited in claim 1.
CN202211465346.XA 2022-11-22 2022-11-22 Chromosome abnormality identification method, system and storage medium Active CN115619774B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211465346.XA CN115619774B (en) 2022-11-22 2022-11-22 Chromosome abnormality identification method, system and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211465346.XA CN115619774B (en) 2022-11-22 2022-11-22 Chromosome abnormality identification method, system and storage medium

Publications (2)

Publication Number Publication Date
CN115619774A CN115619774A (en) 2023-01-17
CN115619774B true CN115619774B (en) 2023-03-14

Family

ID=84879088

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211465346.XA Active CN115619774B (en) 2022-11-22 2022-11-22 Chromosome abnormality identification method, system and storage medium

Country Status (1)

Country Link
CN (1) CN115619774B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116777904B (en) * 2023-08-16 2023-10-24 笑纳科技(苏州)有限公司 Different resolution abnormal chromosome generation network and method based on stripe characteristics

Also Published As

Publication number Publication date
CN115619774A (en) 2023-01-17

Similar Documents

Publication Publication Date Title
US20220375242A1 (en) Deep learning automated dermatopathology
CN112101451B (en) Breast cancer tissue pathological type classification method based on generation of antagonism network screening image block
CN110853022B (en) Pathological section image processing method, device and system and storage medium
CN109271958B (en) Face age identification method and device
CN106202329A (en) Sample data process, data identification method and device, computer equipment
CN109829478B (en) Problem classification method and device based on variation self-encoder
CN115619774B (en) Chromosome abnormality identification method, system and storage medium
CN109390053B (en) Fundus image processing method, fundus image processing apparatus, computer device, and storage medium
EP1711919A1 (en) Example-based diagnosis decision support
CN114841294B (en) Classifier model training method and device for detecting chromosome structure abnormality
CN109271957B (en) Face gender identification method and device
CN111899259A (en) Prostate cancer tissue microarray classification method based on convolutional neural network
CN111028923A (en) Digital pathological image dyeing normalization method, electronic device and storage medium
Sulam et al. Maximizing AUC with Deep Learning for Classification of Imbalanced Mammogram Datasets.
CN114330499A (en) Method, device, equipment, storage medium and program product for training classification model
CN113628297A (en) COVID-19 deep learning diagnosis system based on attention mechanism and transfer learning
CN112614573A (en) Deep learning model training method and device based on pathological image labeling tool
Chen et al. A novel graphical model approach to segmenting cell images
CN113012249B (en) Method, device and storage medium for generating focus on CT image
CN111180011A (en) Method and device for detecting gene mutation of focus
CN114708264B (en) Light spot quality judging method, device, equipment and storage medium
CN110970100A (en) Method, device and equipment for detecting item coding and computer readable storage medium
JP2011248680A (en) Image processing system
JP2024508095A (en) Graph construction and visualization of multiplex immunofluorescence images
Abubacker et al. Correlation-based feature selection for association rule mining in semantic annotation of mammographic medical images

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant