CN115147661B - Chromosome classification method, device, equipment and readable storage medium - Google Patents
Chromosome classification method, device, equipment and readable storage medium Download PDFInfo
- Publication number
- CN115147661B CN115147661B CN202210879228.7A CN202210879228A CN115147661B CN 115147661 B CN115147661 B CN 115147661B CN 202210879228 A CN202210879228 A CN 202210879228A CN 115147661 B CN115147661 B CN 115147661B
- Authority
- CN
- China
- Prior art keywords
- chromosome
- classifier
- network
- metaphase
- images
- 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
Links
- 210000000349 chromosome Anatomy 0.000 title claims abstract description 187
- 238000000034 method Methods 0.000 title claims abstract description 34
- 230000031864 metaphase Effects 0.000 claims abstract description 59
- 238000012549 training Methods 0.000 claims abstract description 38
- 210000002230 centromere Anatomy 0.000 claims abstract description 16
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 10
- 230000006870 function Effects 0.000 claims description 31
- 238000012545 processing Methods 0.000 claims description 11
- 238000010606 normalization Methods 0.000 claims description 8
- 238000004590 computer program Methods 0.000 claims description 7
- 238000010276 construction Methods 0.000 claims description 5
- 238000013528 artificial neural network Methods 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims 1
- 238000013135 deep learning Methods 0.000 abstract description 4
- 238000004891 communication Methods 0.000 description 7
- 230000005236 sound signal Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000012216 screening Methods 0.000 description 3
- 230000005856 abnormality Effects 0.000 description 2
- 210000004027 cell Anatomy 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- KLDZYURQCUYZBL-UHFFFAOYSA-N 2-[3-[(2-hydroxyphenyl)methylideneamino]propyliminomethyl]phenol Chemical compound OC1=CC=CC=C1C=NCCCN=CC1=CC=CC=C1O KLDZYURQCUYZBL-UHFFFAOYSA-N 0.000 description 1
- 241000282472 Canis lupus familiaris Species 0.000 description 1
- 241000282326 Felis catus Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 201000001098 delayed sleep phase syndrome Diseases 0.000 description 1
- 208000033921 delayed sleep phase type circadian rhythm sleep disease Diseases 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- PCHJSUWPFVWCPO-UHFFFAOYSA-N gold Chemical compound [Au] PCHJSUWPFVWCPO-UHFFFAOYSA-N 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000000394 mitotic effect Effects 0.000 description 1
- 210000003765 sex chromosome Anatomy 0.000 description 1
Classifications
-
- 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
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a chromosome classification method, a chromosome classification device, chromosome classification equipment and a readable storage medium, wherein the chromosome classification method comprises the following steps: acquiring a plurality of metaphase chromosome images; constructing a chromosome classifier model based on a twin network strategy, wherein the chromosome classifier model comprises two network models, the two network models are identical, and each network model comprises a convolutional neural network, a Transformer encoder and a softmax classifier; training the chromosome classifier model by using the image set to obtain a trained chromosome classifier model; and classifying the metaphase chromosome images to be classified by using the trained chromosome classifier model to obtain classification results. The chromosome classifier model constructed by the invention is a deep learning classifier capable of simultaneously fusing the texture characteristics, the length and the centromere position information of the chromosome, and the accuracy of classification result prediction can be improved through the classifier.
Description
Technical Field
The invention relates to the technical field of chromosomes, in particular to a chromosome classification method, a chromosome classification device, a chromosome classification equipment and a readable storage medium.
Background
Chromosome karyotyping is a gold standard for clinical screening for defects. Chromosome screening in China is not popular, and the main reasons are that related equipment is monopoly abroad, the equipment is expensive and clumsy, and the karyotype analysis process is strongly dependent on doctor experience to process, so that more than 2 hours are required for completing diagnosis of a single patient. The process of chromosome karyotyping mainly comprises 1) screening high-quality images from hundreds of metaphase mitotic cell microscopic images on patients; 2) Extracting each chromosome from the cell image; 3) And classifying the chromosomes to construct a karyotype map, and diagnosing the structural abnormality or the quantitative abnormality of the chromosomes based on the karyotype map. Obviously, chromosome classification is a key ring, and currently, by manual classification, the workload is huge and complex, so that the realization of automatic classification has important clinical significance.
Disclosure of Invention
The object of the present invention is to provide a chromosome classification method, apparatus, device and readable storage medium, which improve the above problems.
In order to achieve the above purpose, the embodiment of the present application provides the following technical solutions:
in one aspect, embodiments of the present application provide a method of chromosome classification, the method comprising:
acquiring an image set, the image set comprising a plurality of metaphase chromosome images;
constructing a chromosome classifier model based on a twin network strategy, wherein the chromosome classifier model comprises two network models, the two network models are identical, and each network model comprises a convolutional neural network, a Transformer encoder and a softmax classifier;
training the chromosome classifier model by using the image set to obtain a trained chromosome classifier model;
and classifying the metaphase chromosome images to be classified by using the trained chromosome classifier model to obtain classification results.
In a second aspect, embodiments of the present application provide a chromosome classification device that includes an acquisition module, a construction module, a training module, and a classification module.
The acquisition module is used for acquiring an image set, wherein the image set comprises a plurality of metaphase chromosome images;
the construction module is used for constructing a chromosome classifier model based on a twin network strategy, wherein the chromosome classifier model comprises two network models, the two network models are identical, and each network model comprises a convolutional neural network, a Transformer encoder and a softmax classifier;
the training module is used for training the chromosome classifier model by utilizing the image set to obtain a trained chromosome classifier model;
and the classification module is used for classifying the metaphase chromosome images to be classified by using the trained chromosome classifier model to obtain classification results.
In a third aspect, embodiments of the present application provide a chromosome classification device comprising a memory and a processor. The memory is used for storing a computer program; the processor is configured to implement the steps of the chromosome classification method described above when executing the computer program.
In a fourth aspect, embodiments of the present application provide a readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the chromosome classification method described above.
The beneficial effects of the invention are as follows:
1. in order to alleviate the problems of similarity between classes and intra-class variability, the chromosome classifier model is constructed on the basis of a twin network strategy, and meanwhile, the chromosome classifier model constructed by the method is a deep learning classifier capable of simultaneously fusing the texture characteristics, the length and the centromere position information of the chromosomes, and the accuracy of classification result prediction can be improved through the classifier.
2. In the invention, considering that the centromere position information is important information for distinguishing chromosome types, the information and the coding features are input into a softmax classifier to predict the chromosome type information, and the classification performance can be improved in this way.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be apparent from the description, or may be learned by practice of the embodiments of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims thereof as well as the appended drawings.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the embodiments will be briefly described below, it being understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and other related drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic flow chart of a chromosome classification method according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a chromosome classification apparatus according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a chromosome classification apparatus according to an embodiment of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. The components of the 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 invention, as 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 made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It should be noted that: like reference numerals or letters denote like items in the following figures, and thus once an item is defined in one figure, no further definition or explanation thereof is necessary in the following figures. Meanwhile, in the description of the present invention, the terms "first", "second", and the like are used only to distinguish the description, and are not to be construed as indicating or implying relative importance.
Example 1
As shown in fig. 1, the present embodiment provides a chromosome classification method, which includes step S1, step S2, step S3, and step S4.
S1, acquiring an image set, wherein the image set comprises a plurality of metaphase chromosome images;
s2, constructing a chromosome classifier model based on a twin network strategy, wherein the chromosome classifier model comprises two network models, the two network models are identical, and each network model comprises a convolutional neural network, a transducer encoder and a softmax classifier;
the human cytogenomics International naming system divides chromosomes into 22 pairs of autosomes (designated by numbers 1-22) and 1 pair of sex chromosomes (designated by X and Y, respectively) according to the length of the chromosome (long, medium, short), the location of the centromere (medium, sub-medium and proximal), and the zonal distribution. Therefore, it is clinically classified mainly based on the length of chromosome, the position of centromere, and the distribution of streaks. It is apparent that this challenge for automatic classification is great, and in particular, the imaging features (e.g., zoned texture features) of different numbers of chromosomes, which are not as intuitive as the categories in natural images (e.g., cats and dogs), are subject to inter-category similarity and intra-category variability. Thus, it is difficult to extract reliable imaging features to construct a chromosome classifier. The deep learning classifier represented by the Convolutional Neural Network (CNN) at present can avoid the extraction of artificial features and is expected to obtain satisfactory effects. However, CNN generally encodes only the texture features of a chromosome image, and cannot encode the length of the chromosome and centromere position information, which affects classification performance to some extent. For this reason, this embodiment proposes a deep learning classifier capable of simultaneously fusing chromosome texture features and length and centromere position information. In addition, in order to alleviate the problems of similarity between classes and intra-class variability, it is proposed to construct a chromosome classifier model based on a twin network strategy;
s3, training the chromosome classifier model by using the image set to obtain a trained chromosome classifier model;
the specific implementation steps of the step comprise a step S31, a step S32 and a step S33;
step S31, randomly extracting two metaphase chromosome images from the image set during each training to form a group of image pairs; inputting an extracted metaphase chromosome image into each network model respectively, wherein in each network model, a convolution neural network is utilized to extract multi-scale convolution characteristics of the input metaphase chromosome image; normalizing the length of the chromosome in the input metaphase chromosome image to obtain a normalization processing result; inputting the multi-scale convolution characteristics and the normalization processing result into a transform encoder, and outputting coding characteristics; inputting the coding features into a first softmax classifier, predicting to obtain centromere position information of a chromosome, inputting the centromere position information of the chromosome and the coding features into a second softmax classifier, and predicting to obtain chromosome category information, wherein the first softmax classifier is the same as the second softmax classifier;
in the step, considering that the centromere position information is important information for distinguishing chromosome types, inputting the information and the coding features into a softmax classifier to predict and obtain chromosome type information, and in this way, the classification performance can be improved;
step S32, judging whether the two metaphase chromosome images extracted randomly are of the same category or not based on the two coding features obtained through the two network models, and minimizing the distance between the similar metaphase chromosome images and maximizing the distance between the heterogeneous metaphase chromosome images by using a contrast loss function based on the judging result;
the specific implementation steps of the step comprise a step S321 and a step S322;
s321, inputting two coding features obtained through two network models into a fully-connected network or a multi-layer perceptron, inputting the output of the fully-connected network or the multi-layer perceptron into a sigmoid function, and converting the output into a probability value;
and S322, comparing and analyzing the probability value with a preset probability threshold, wherein if the probability value is larger than the probability threshold, two randomly extracted metaphase chromosome images are of the same kind, and otherwise, are of different kinds.
And step S33, judging whether a training stopping condition is met, stopping training if the training stopping condition is met, obtaining the trained chromosome classifier model, otherwise, returning to the step of randomly extracting two metaphase chromosome images from the image set.
The specific implementation steps of the step comprise a step S331 and a step S322;
step S331, constructing total loss, wherein the total loss is as follows:
in the formula (1),a cross entropy loss function corresponding to the first softmax classifier in a first network model; />A cross entropy loss function corresponding to the first softmax classifier in a second network model; />A cross entropy loss function corresponding to the second softmax classifier in the first network model; />A cross entropy loss function corresponding to the second softmax classifier in a second network model; />Cross loss entropy corresponding to the sigmoid function; />For the contrast loss function, the two network models include the first network model and the second network model;
in the formula (2), d is the Euclidean distance of two coding features; y is a label of whether two randomly extracted metaphase chromosome images are matched or not, wherein the same class y=1, and the different class y=0; margin is a set threshold, 1.25 is taken in this step; n is the sequence number of the image pair; n is the number of image pairs in the image set;
step S332, judging whether the total loss is smaller than a preset loss threshold, if so, reaching the training stop condition, otherwise, not reaching the training stop condition.
In this step, the preset loss threshold value may be set in a user-defined manner according to the user's requirement, and 0.01 is taken in this step;
and S4, classifying the metaphase chromosome images to be classified by using the trained chromosome classifier model to obtain a classification result.
In practical application, the metaphase chromosome image to be classified is input into a trained chromosome classifier model, and a chromosome classification result can be output.
Example 2
As shown in fig. 2, the present embodiment provides a chromosome classification apparatus, which includes an acquisition module 701, a construction module 702, a training module 703, and a classification module 704.
An acquisition module 701 for acquiring a set of images, the set of images comprising a plurality of metaphase chromosome images;
a construction module 702, configured to construct a chromosome classifier model based on a twin network policy, where the chromosome classifier model includes two network models, and the two network models are identical, and each network model includes a convolutional neural network, a Transformer encoder, and a softmax classifier;
the training module 703 is configured to train the chromosome classifier model by using the image set, so as to obtain a trained chromosome classifier model;
and the classification module 704 is configured to perform classification processing on the metaphase chromosome image to be classified by using the trained chromosome classifier model, so as to obtain a classification result.
In a specific embodiment of the disclosure, the training module 703 further includes an extracting unit 7031, a calculating unit 7032, and a judging unit 7033.
An extraction unit 7031 for randomly extracting two metaphase chromosome images from the image set at each training time, to form a group of image pairs; inputting an extracted metaphase chromosome image into each network model respectively, wherein in each network model, a convolution neural network is utilized to extract multi-scale convolution characteristics of the input metaphase chromosome image; normalizing the length of the chromosome in the input metaphase chromosome image to obtain a normalization processing result; inputting the multi-scale convolution characteristics and the normalization processing result into a transform encoder, and outputting coding characteristics; inputting the coding features into a first softmax classifier, predicting to obtain centromere position information of a chromosome, inputting the centromere position information of the chromosome and the coding features into a second softmax classifier, and predicting to obtain chromosome category information, wherein the first softmax classifier is the same as the second softmax classifier;
a calculating unit 7032, configured to determine, based on two coding features obtained by two network models, whether two metaphase chromosome images extracted randomly are of the same class, and based on a determination result, minimize a distance between similar metaphase chromosome images and maximize a distance between heterogeneous metaphase chromosome images using a contrast loss function;
and a judging unit 7033, configured to judge whether a training stopping condition is reached, and if so, stop training to obtain the trained chromosome classifier model, otherwise, return to a step of randomly extracting two metaphase chromosome images from the image set.
In one embodiment of the present disclosure, the computing unit 7032 further includes a transformant subunit 70321 and an analyzer subunit 70322.
A converter unit 70321, configured to input the two coding features obtained by the two network models into a fully-connected network or a multi-layer perceptron, input the output of the fully-connected network or the multi-layer perceptron into a sigmoid function, and convert the output into a probability value;
and the analysis subunit 70322 is configured to compare and analyze the probability value with a preset probability threshold, and if the probability value is greater than the probability threshold, the two metaphase chromosome images extracted randomly are similar, otherwise, are heterogeneous.
In one embodiment of the disclosure, the determining unit 7033 further includes a constructing subunit 70331 and a determining subunit 70332.
A building subunit 70331 for building a total loss of:
in the formula (1),a cross entropy loss function corresponding to the first softmax classifier in a first network model; />A cross entropy loss function corresponding to the first softmax classifier in a second network model; />A cross entropy loss function corresponding to the second softmax classifier in the first network model; />A cross entropy loss function corresponding to the second softmax classifier in a second network model; />Cross loss corresponding to the sigmoid functionEntropy loss; />For the contrast loss function, the two network models include the first network model and the second network model;
in the formula (2), d is the Euclidean distance of two coding features; y is a label of whether two randomly extracted metaphase chromosome images are matched or not, wherein the same class y=1, and the different class y=0; margin is a set threshold, n is the sequence number of the image pair; n is the number of image pairs in the image set;
and the judging subunit 70332 is configured to judge whether the total loss is smaller than a preset loss threshold, if so, reach the training stop condition, otherwise, not reach the training stop condition.
It should be noted that, regarding the apparatus in the above embodiments, the specific manner in which the respective modules perform the operations has been described in detail in the embodiments regarding the method, and will not be described in detail herein.
Example 3
Corresponding to the above method embodiments, the present disclosure further provides a chromosome classification apparatus, which is described below and the chromosome classification method described above may be referred to correspondingly to each other.
Fig. 3 is a block diagram illustrating a chromosome classification device 800, according to an example embodiment. As shown in fig. 3, the chromosome classification apparatus 800 may include: a processor 801, a memory 802. The chromosome classification device 800 may also include one or more of a multimedia component 803, an i/O interface 804, and a communication component 805.
Wherein the processor 801 is configured to control the overall operation of the chromosome classification apparatus 800 to perform all or part of the steps of the chromosome classification method described above. Memory 802 is used to store various types of data to support the operation of the chromosome classification device 800, which may include, for example, instructions for any application or method operating on the chromosome classification device 800, as well as application-related data, such as contact data, transceived messages, pictures, audio, video, and the like. The Memory 802 may be implemented by any type or combination of volatile or non-volatile Memory devices, such as static random access Memory (Static Random Access Memory, SRAM for short), electrically erasable programmable Read-Only Memory (Electrically Erasable Programmable Read-Only Memory, EEPROM for short), erasable programmable Read-Only Memory (Erasable Programmable Read-Only Memory, EPROM for short), programmable Read-Only Memory (Programmable Read-Only Memory, PROM for short), read-Only Memory (ROM for short), magnetic Memory, flash Memory, magnetic disk, or optical disk. The multimedia component 803 may include a screen and an audio component. Wherein the screen may be, for example, a touch screen, the audio component being for outputting and/or inputting audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in the memory 802 or transmitted through the communication component 805. The audio assembly further comprises at least one speaker for outputting audio signals. The I/O interface 804 provides an interface between the processor 801 and other interface modules, which may be a keyboard, mouse, buttons, etc. These buttons may be virtual buttons or physical buttons. The communication component 805 is configured to perform wired or wireless communication between the chromosome classification device 800 and other devices. Wireless communication, such as Wi-Fi, bluetooth, near field communication (Near FieldCommunication, NFC for short), 2G, 3G or 4G, or a combination of one or more thereof, the respective communication component 805 may thus comprise: wi-Fi module, bluetooth module, NFC module.
In an exemplary embodiment, the chromosome classification device 800 may be implemented by one or more application specific integrated circuits (Application Specific Integrated Circuit, ASIC), digital signal processor (DigitalSignal Processor, DSP), digital signal processing device (Digital Signal Processing Device, DSPD), programmable logic device (Programmable Logic Device, PLD), field programmable gate array (Field Programmable Gate Array, FPGA), controller, microcontroller, microprocessor, or other electronic components for performing the chromosome classification method described above.
In another exemplary embodiment, a computer readable storage medium comprising program instructions which, when executed by a processor, implement the steps of the chromosome classification method described above is also provided. For example, the computer readable storage medium may be the memory 802 described above including program instructions executable by the processor 801 of the chromosome classification device 800 to perform the chromosome classification method described above.
Example 4
Corresponding to the above method embodiments, the present disclosure further provides a readable storage medium, and a readable storage medium described below and a chromosome classification method described above may be referred to correspondingly to each other.
A readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the chromosome classification method of the above method embodiment.
The readable storage medium may be a usb disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a magnetic disk, or an optical disk, and the like.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (6)
1. A method of chromosome classification, comprising:
acquiring an image set, the image set comprising a plurality of metaphase chromosome images;
constructing a chromosome classifier model based on a twin network strategy, wherein the chromosome classifier model comprises two network models, the two network models are identical, and each network model comprises a convolutional neural network, a Transformer encoder and a softmax classifier;
training the chromosome classifier model by using the image set to obtain a trained chromosome classifier model;
classifying the metaphase chromosome images to be classified by using the trained chromosome classifier model to obtain classification results;
training the chromosome classifier model by using the image set to obtain a trained chromosome classifier model, wherein the training method comprises the following steps of:
randomly extracting two metaphase chromosome images from the image set during each training to form a group of image pairs; inputting an extracted metaphase chromosome image into each network model respectively, wherein in each network model, a convolution neural network is utilized to extract multi-scale convolution characteristics of the input metaphase chromosome image; normalizing the length of the chromosome in the input metaphase chromosome image to obtain a normalization processing result; inputting the multi-scale convolution characteristics and the normalization processing result into a transform encoder, and outputting coding characteristics; inputting the coding features into a first softmax classifier, predicting to obtain centromere position information of a chromosome, inputting the centromere position information of the chromosome and the coding features into a second softmax classifier, and predicting to obtain chromosome category information, wherein the first softmax classifier is the same as the second softmax classifier;
judging whether the two metaphase chromosome images extracted randomly are of the same class or not based on the two coding features obtained through the two network models, and minimizing the distance between similar metaphase chromosome images and maximizing the distance between heterogeneous metaphase chromosome images by using a contrast loss function based on a judging result;
judging whether a training stopping condition is reached, stopping training if the training stopping condition is reached, obtaining the trained chromosome classifier model, otherwise, returning to the step of randomly extracting two metaphase chromosome images from the image set.
2. The chromosome classification method according to claim 1, wherein determining whether two of the metaphase chromosome images extracted at random are of the same class based on the two coding features obtained by the two network models, comprises:
inputting the two coding features obtained through the two network models into a fully-connected network or a multi-layer perceptron, inputting the output of the fully-connected network or the multi-layer perceptron into a sigmoid function, and converting the output into a probability value;
and comparing and analyzing the probability value with a preset probability threshold, wherein if the probability value is larger than the probability threshold, two chromosome images in metaphase are randomly extracted to be of the same kind, and otherwise, the two chromosome images in metaphase are of different kinds.
3. The chromosome classification method according to claim 1, wherein determining whether a training stop condition is reached comprises:
constructing a total loss, wherein the total loss is:
in the formula (1),a cross entropy loss function corresponding to the first softmax classifier in a first network model;a cross entropy loss function corresponding to the first softmax classifier in a second network model; />A cross entropy loss function corresponding to the second softmax classifier in the first network model; />A cross entropy loss function corresponding to the second softmax classifier in a second network model; />Cross loss entropy corresponding to the sigmoid function; />For the contrast loss function, the two network models include the first network model and the second network model;
in the formula (2), d is the Euclidean distance of two coding features; y is a label of whether two randomly extracted metaphase chromosome images are matched or not, wherein the same class y=1, and the different class y=0; margin is a set threshold, n is the sequence number of the image pair; n is the number of image pairs in the image set;
and judging whether the total loss is smaller than a preset loss threshold value, if so, reaching a training stop condition, otherwise, not reaching the training stop condition.
4. A chromosome classification apparatus, comprising:
the acquisition module is used for acquiring an image set, wherein the image set comprises a plurality of metaphase chromosome images;
the construction module is used for constructing a chromosome classifier model based on a twin network strategy, wherein the chromosome classifier model comprises two network models, the two network models are identical, and each network model comprises a convolutional neural network, a Transformer encoder and a softmax classifier;
the training module is used for training the chromosome classifier model by utilizing the image set to obtain a trained chromosome classifier model;
the classification module is used for classifying the metaphase chromosome images to be classified by using the trained chromosome classifier model to obtain classification results;
a training module, comprising:
the extraction unit is used for randomly extracting two metaphase chromosome images from the image set during each training to form a group of image pairs; inputting an extracted metaphase chromosome image into each network model respectively, wherein in each network model, a convolution neural network is utilized to extract multi-scale convolution characteristics of the input metaphase chromosome image; normalizing the length of the chromosome in the input metaphase chromosome image to obtain a normalization processing result; inputting the multi-scale convolution characteristics and the normalization processing result into a transform encoder, and outputting coding characteristics; inputting the coding features into a first softmax classifier, predicting to obtain centromere position information of a chromosome, inputting the centromere position information of the chromosome and the coding features into a second softmax classifier, and predicting to obtain chromosome category information, wherein the first softmax classifier is the same as the second softmax classifier;
the computing unit is used for judging whether the two metaphase chromosome images extracted randomly are of the same category or not based on the two coding features obtained through the two network models, and minimizing the distance between the similar metaphase chromosome images and maximizing the distance between the heterogeneous metaphase chromosome images by using a contrast loss function based on a judging result;
the judging unit is used for judging whether the training stopping condition is met, if so, stopping training to obtain the trained chromosome classifier model, otherwise, returning to the step of randomly extracting two metaphase chromosome images from the image set;
a computing unit comprising:
the conversion subunit is used for inputting the two coding features obtained through the two network models into a fully-connected network or a multi-layer perceptron, inputting the output of the fully-connected network or the multi-layer perceptron into a sigmoid function, and converting the output into a probability value;
the analysis subunit is used for comparing and analyzing the probability value with a preset probability threshold, and if the probability value is larger than the probability threshold, the two randomly extracted metaphase chromosome images are of the same type, otherwise, the two randomly extracted metaphase chromosome images are of different types;
a judgment unit including:
a building subunit for building a total loss, the total loss being:
in the formula (1),a cross entropy loss function corresponding to the first softmax classifier in a first network model;a cross entropy loss function corresponding to the first softmax classifier in a second network model; />A cross entropy loss function corresponding to the second softmax classifier in the first network model; />A cross entropy loss function corresponding to the second softmax classifier in a second network model; />Cross loss entropy corresponding to the sigmoid function;for the contrast loss function, the two network models include the first network model and the second network model;
in the formula (2), d is the Euclidean distance of two coding features; y is a label of whether two randomly extracted metaphase chromosome images are matched or not, wherein the same class y=1, and the different class y=0; margin is a set threshold, n is the sequence number of the image pair; n is the number of image pairs in the image set;
and the judging subunit is used for judging whether the total loss is smaller than a preset loss threshold value, if so, reaching the training stop condition, otherwise, not reaching the training stop condition.
5. A chromosome classification apparatus, comprising:
a memory for storing a computer program;
a processor for implementing the steps of the chromosome classification method as claimed in any one of claims 1 to 3 when executing the computer program.
6. A readable storage medium, characterized by: the readable storage medium has stored thereon a computer program which, when executed by a processor, implements the steps of the chromosome classification method as defined in any one of claims 1 to 3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210879228.7A CN115147661B (en) | 2022-07-25 | 2022-07-25 | Chromosome classification method, device, equipment and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210879228.7A CN115147661B (en) | 2022-07-25 | 2022-07-25 | Chromosome classification method, device, equipment and readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115147661A CN115147661A (en) | 2022-10-04 |
| CN115147661B true CN115147661B (en) | 2023-07-25 |
Family
ID=83414525
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210879228.7A Active CN115147661B (en) | 2022-07-25 | 2022-07-25 | Chromosome classification method, device, equipment and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115147661B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118429731B (en) * | 2024-07-04 | 2024-09-06 | 湘江实验室 | Chromosome classification method |
| CN118430650B (en) * | 2024-07-05 | 2024-09-27 | 湘江实验室 | Chromosome key point sequence prediction method and device |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109300111B (en) * | 2018-08-27 | 2020-05-12 | 杭州德适生物科技有限公司 | Chromosome recognition method based on deep learning |
| CN110516745B (en) * | 2019-08-28 | 2022-05-24 | 北京达佳互联信息技术有限公司 | Training method and device of image recognition model and electronic equipment |
| CN112598643B (en) * | 2020-12-22 | 2023-06-23 | 百度在线网络技术（北京）有限公司 | Depth fake image detection and model training method, device, equipment and medium |
| CN112784130B (en) * | 2021-01-27 | 2022-05-27 | 杭州网易云音乐科技有限公司 | Twin network model training and measuring method, device, medium and equipment |
-
2022
- 2022-07-25 CN CN202210879228.7A patent/CN115147661B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN115147661A (en) | 2022-10-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115147661B (en) | Chromosome classification method, device, equipment and readable storage medium | |
| CN109919928B (en) | Medical image detection method and device and storage medium | |
| CN109993102B (en) | Similar face retrieval method, device and storage medium | |
| CN111582342B (en) | Image identification method, device, equipment and readable storage medium | |
| CN110796199B (en) | Image processing method and device and electronic medical equipment | |
| CN110147878B (en) | Data processing method, device and equipment | |
| CN113469236A (en) | Deep clustering image recognition system and method for self-label learning | |
| CN110827236B (en) | Brain tissue layering method, device and computer equipment based on neural network | |
| CN115223715A (en) | Cancer prediction method and system based on multi-modal information fusion | |
| CN111340213B (en) | Neural network training method, electronic device, and storage medium | |
| CN112101162A (en) | Image recognition model generation method and device, storage medium and electronic equipment | |
| CN112233798B (en) | Interpretable disease risk analysis system based on pathological mode and attention mechanism | |
| CN112465231B (en) | Method, apparatus and readable storage medium for predicting regional population health status | |
| CN111737479B (en) | Data acquisition method and device, electronic equipment and storage medium | |
| CN108877928A (en) | Patient information acquisition method, device, computer equipment and storage medium | |
| CN110244854A (en) | A kind of artificial intelligence approach of multi-class eeg data identification | |
| CN111931616A (en) | Emotion recognition method and system based on mobile intelligent terminal sensor equipment | |
| CN117438087A (en) | Health management method and system based on health monitoring data analysis | |
| CN113626591A (en) | Electronic medical record data quality evaluation method based on text classification | |
| CN116304042A (en) | False news detection method based on multi-modal feature self-adaptive fusion | |
| CN113705595A (en) | Method, device and storage medium for predicting degree of abnormal cell metastasis | |
| CN113160987B (en) | Health state prediction method, apparatus, computer device and storage medium | |
| CN115375954B (en) | Chemical experiment solution identification method, device, equipment and readable storage medium | |
| CN112183336A (en) | Expression recognition model training method and device, terminal equipment and storage medium | |
| CN114387524B (en) | Image identification method and system for small sample learning based on multilevel second-order representation |
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 |