CN114972796A - Chromosome extraction method, device, equipment and readable storage medium - Google Patents
Chromosome extraction method, device, equipment and readable storage medium Download PDFInfo
- Publication number
- CN114972796A CN114972796A CN202210624344.4A CN202210624344A CN114972796A CN 114972796 A CN114972796 A CN 114972796A CN 202210624344 A CN202210624344 A CN 202210624344A CN 114972796 A CN114972796 A CN 114972796A
- Authority
- CN
- China
- Prior art keywords
- chromosome
- image
- anchor frame
- training
- loss function
- 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.)
- Pending
Links
Images
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
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
-
- 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
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16B—BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY
- G16B15/00—ICT specially adapted for analysing two-dimensional or three-dimensional molecular structures, e.g. structural or functional relations or structure alignment
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Multimedia (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Databases & Information Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- Biotechnology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Crystallography & Structural Chemistry (AREA)
- Bioinformatics & Computational Biology (AREA)
- Chemical & Material Sciences (AREA)
- Evolutionary Biology (AREA)
- Investigating Or Analysing Biological Materials (AREA)
Abstract
The invention provides a chromosome extraction method, a chromosome extraction device, chromosome extraction equipment and a readable storage medium, wherein the method comprises the following steps: acquiring a first image, wherein the first image comprises a metaphase cell image of a chromosome to be extracted; training a characteristic multi-scale network to obtain a trained characteristic multi-scale network model, wherein the trained characteristic multi-scale network model comprises a saliency map prediction part, an anchor frame classification part and an anchor frame regression part; and processing the first image by using the trained feature multi-scale network model to obtain a positioning frame of each chromosome in the metaphase cell image of the chromosome to be extracted, and extracting the chromosome according to the positioning frame. The method of the invention can be used for detecting the chromosome rapidly and accurately.
Description
Technical Field
The invention relates to the technical field of chromosomes, in particular to a chromosome extraction method, a chromosome extraction device, chromosome extraction equipment and a readable storage medium.
Background
Chromosome abnormality can cause abortion, dysnoesia, congenital malformation and other birth defects, seriously affect family happiness and bring huge burden to society. Karyotyping is one of the important clinical screening tools for chromosomal abnormalities, and the procedure generally includes the following steps: (1) selecting a high quality image from hundreds of metaphase cell microscopic images of the patient; (2) extracting and classifying each chromosome from the cell image to generate a chromosome karyotype image; (3) and finally screening the chromosome structural and quantitative abnormality based on the karyotype chart. The above steps are complicated and require multiple experienced genetics physicians to coordinate and consume a great deal of time and effort to complete the diagnostic process for a single patient. The extraction of each chromosome from an image is a time-consuming and labor-consuming main link, so that an efficient and reliable chromosome extraction method is urgently needed to reduce the workload of medical staff.
Disclosure of Invention
It is an object of the present invention to provide a chromosome extraction method, apparatus, device and readable storage medium to improve the above-mentioned problems.
In order to achieve the above object, the embodiments of the present application provide the following technical solutions:
in one aspect, an embodiment of the present application provides a chromosome extraction method, including:
acquiring a first image, wherein the first image comprises a metaphase cell image of a chromosome to be extracted;
training a characteristic multi-scale network to obtain a trained characteristic multi-scale network model, wherein the trained characteristic multi-scale network model comprises a saliency map prediction part, an anchor frame classification part and an anchor frame regression part;
and processing the first image by using the trained feature multi-scale network model to obtain a positioning frame of each chromosome in the metaphase cell image of the chromosome to be extracted, and extracting the chromosome according to the positioning frame.
In a second aspect, an embodiment of the present application provides a chromosome extraction apparatus, which includes an acquisition module, a training module, and an extraction module.
The acquisition module is used for acquiring a first image, wherein the first image comprises a metaphase cell image of the chromosome to be extracted;
the training module is used for training the characteristic multi-scale network to obtain a trained characteristic multi-scale network model, and the trained characteristic multi-scale network model comprises a saliency map prediction part, an anchor frame classification part and an anchor frame regression part;
and the extraction module is used for processing the first image by using the trained characteristic multi-scale network model to obtain a positioning frame of each chromosome in the metaphase cell image of the chromosome to be extracted, and extracting the chromosome according to the positioning frame.
In a third aspect, embodiments of the present application provide a chromosome extraction device, which includes a memory and a processor. The memory is used for storing a computer program; the processor is used for realizing the steps of the chromosome extraction method when executing the computer program.
In a fourth aspect, the present application provides a readable storage medium, on which a computer program is stored, and the computer program, when executed by a processor, implements the steps of the chromosome extraction method.
The invention has the beneficial effects that:
1. in the traditional detector, the anchor frame can be arranged in the whole image and contains a large number of useless anchor frames with backgrounds, and the number of redundant anchor frames can be greatly reduced by arranging the anchor frame by using the skeleton image in the saliency map; meanwhile, in the invention, because the skeleton and the edge can better enable each chromosome to be independently opened, and the cross points can enable the model to focus more on the cross region, the saliency map is used as the supplementary information of the anchor frame classification part and the anchor frame regression part, and the performance of the model for identifying the chromosomes can be improved.
2. The existing mainstream target detector is basically based on the detection of a positive frame, the positive frame detection is not suitable for the characteristics of any angle and dense distribution of chromosomes, but the invention adopts a rotating frame for detection, a plurality of anchor frames with different sizes, different length-width ratios and different rotating angles are arranged at the arrangement position of each anchor frame, and a series of anchor frames with different rotating angles are arranged to be more suitable for the inherent characteristics of chromosomes, namely dense distribution, any direction and the like, so that the chromosome examples can be more accurately positioned.
3. In the invention, the problem of unbalanced anchor frames and local similarity is possibly more serious by arranging a plurality of anchor frames at the setting position of each anchor frame, so that all the anchor frames are screened, the total number of the anchor frames can be obviously reduced by the method, and the problems of unbalanced anchor frames and local similarity can be effectively solved; meanwhile, the number of the anchor frames is reduced, so that the reasoning process can be accelerated and the detection performance can be improved.
4. The invention can carry out center point offset, size scaling and angle offset on the reserved anchor frame through the anchor frame regression part, and can enable the anchor frame to position the target more accurately through the mode.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by the 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 hereof 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 needed to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained according to the drawings without inventive efforts.
FIG. 1 is a schematic flow chart of a chromosome extraction method according to an embodiment of the present invention;
FIG. 2 is a schematic structural diagram of a chromosome extraction apparatus according to an embodiment of the present invention;
FIG. 3 is a schematic structural diagram of a chromosome extraction apparatus according to an embodiment 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.
It should be noted that: like reference numbers or letters refer to like items in the following figures, and thus, once an item is defined in one figure, it need not be further defined or explained in subsequent figures. Meanwhile, in the description of the present invention, the terms "first", "second", and the like are used only for distinguishing 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 extraction method including step S1, step S2, and step S3.
Step S1, acquiring a first image, wherein the first image comprises a metaphase cell image of the chromosome to be extracted;
step S2, training a characteristic multi-scale network to obtain a trained characteristic multi-scale network model, wherein the trained characteristic multi-scale network model comprises a saliency map prediction part, an anchor frame classification part and an anchor frame regression part;
the specific implementation manner of the step is step S21 and step S22;
step S21, constructing a training set, wherein the training set comprises a plurality of samples, each sample comprises first data, second data and third data, the first data comprises a metaphase cell image of chromosomes, the second data comprises a metaphase cell image of chromosomes with skeletons, edges and intersections of overlapped chromosomes marked in an artificial mode, and the third data comprises a standard frame obtained by marking the chromosomes in the metaphase cell image of chromosomes in an artificial mode;
in the step, the second data is used for marking the metaphase cell image of the chromosome by a doctor, and comprises marking a skeleton, an edge and a cross point of overlapped chromosomes; the third data is that the standard frame marking is carried out on the chromosome in the chromosome metaphase cell image by a doctor, and the chromosome can be extracted and classified through the standard frame;
and step S22, training the feature multi-scale network model by using the training set based on a random gradient descent method, judging whether a training stopping condition is met according to a loss function, and stopping training if the training stopping condition is met to obtain the trained feature multi-scale network model.
In this step, a loss function needs to be constructed, and the specific steps of constructing the loss function are step S221 and step S222.
Step S221, constructing a loss function corresponding to the prediction part of the saliency map, constructing a loss function corresponding to the classification part of the anchor frame and constructing a loss function corresponding to the regression part of the anchor frame;
the corresponding loss function of the prediction part of the saliency map is:
in the formula (1), M sei Is a predicted saliency map;
the method comprises the following steps of (1) being a predicted target, wherein the predicted target is a chromosome metaphase cell image of a skeleton, an edge and an intersection point of overlapped chromosomes which are marked in an artificial mode;
the corresponding loss function of the anchor frame classification part is as follows:
in formula (2), FL is a focal loss function; s - Is a negative anchor frame set, and is provided with a negative anchor frame set,
for the prediction probability that an anchor frame in the negative anchor frame set contains chromosomes, Y i A tag for the anchor frame, which has a value of 0; s. the + Is a set of positive anchor frames and is,
is the predicted probability that a given anchor in the set of positive anchors contains a chromosome, Y j A tag for the anchor frame, having a value of 1; as the focus Loss function, a conventional focus Loss function, such as that mentioned in Tsung-YiLin, Priya Goyal, Ross Girshick, Kaiming He, Pittr Doll a, Focal local for depth Object Detection, can be used in the present embodiment;
the method for identifying the negative anchor frame and the positive anchor frame comprises the following steps:
setting a third threshold, where the third threshold may be set in a user-defined manner according to user requirements, and in this embodiment, the third threshold is 0.7; setting a fourth threshold, where the fourth threshold may be set in a user-defined manner according to user requirements, and in this embodiment, the third threshold is 0.3; when the intersection ratio between the anchor frame and any one standard frame is larger than a third threshold value, the anchor frame is a positive anchor frame; when the intersection ratio between the anchor frame and any one standard frame is smaller than a fourth threshold value, the anchor frame is a negative anchor frame;
the corresponding loss function of the regression part of the anchor frame is as follows:
in equation (3), the regression loss is calculated only for the positive anchor box, where D j The Gaussian distribution of a certain anchor frame and the Gaussian score of the corresponding standard frameKullback-Leibler divergence between cloths;
and step S222, constructing a total loss function according to the loss function corresponding to the prediction part of the saliency map, the loss function corresponding to the anchor frame classification part and the loss function corresponding to the anchor frame regression part.
In the formula (4),
predicting a partial corresponding loss function for the saliency map;
a loss function corresponding to the classification part of the anchor frame;
a loss function corresponding to the regression part of the anchor frame; lambda [ alpha ] 1 Weights for the classified parts of the anchor frame; lambda [ alpha ] 2 Weights for the regression portion of the anchor frame;
and S3, processing the first image by using the trained feature multi-scale network model to obtain a positioning frame of each chromosome in the metaphase cell image of the chromosome to be extracted, and extracting the chromosomes according to the positioning frame.
The specific implementation manner of the step is step S31, step S32 and step S33;
step S31, inputting the metaphase cell image of the chromosome to be detected into the trained feature multi-scale network model for feature extraction to obtain a feature map, and processing the feature map by using the prediction part of the feature map to obtain a significant map, wherein the significant map comprises three channel images which are a skeleton image of the chromosome, an edge image of the chromosome and a cross point image of an overlapped chromosome;
in the traditional detector, an anchor frame is arranged in the whole image and contains a large number of useless anchor frames of backgrounds, and in the step, the anchor frame is arranged by using a skeleton image in a saliency map, so that the number of redundant anchor frames can be greatly reduced;
in the step, the skeleton and the edge can better enable each chromosome to be independently opened, and meanwhile, the cross point can enable the model to focus more on the cross region. Therefore, the saliency map is used as supplementary information of the anchor frame classification part and the anchor frame regression part, and the performance of the model for recognizing the chromosome can be improved.
Step S32, analyzing the skeleton image of the chromosome, taking pixel points with pixel values larger than a first threshold value on the skeleton image of the chromosome as anchor frame setting positions, and setting a preset number of anchor frames at each anchor frame setting position, wherein the specifications of the anchor frames are different;
in this step, the first threshold and the preset number may be set by user according to the requirement of the user, in this embodiment, the first threshold is 0.3, and the preset number is 36, that is, 36 anchor frames with different sizes, different length-width ratios, and different rotation angles are set at each anchor frame setting position; a series of anchor frames with different rotation angles are arranged, so that the intrinsic characteristics of chromosomes, namely dense distribution, any direction and the like, can be better adapted, and chromosome examples can be positioned more accurately;
and S33, performing feature splicing processing on the feature map and the saliency map to obtain a spliced map, and obtaining the positioning frame based on the spliced map, the anchor frame classification part and the anchor frame regression part.
The specific implementation manner of the step is step S331, step S332, step S333 and step S334;
step S331, inputting the splicing map into the anchor frame classification part, firstly processing the splicing map by using a 1 × 1 convolutional layer to obtain a first processing result, and then processing the first processing result by using a 9 × 9 void convolutional layer to obtain a second processing result, wherein the second processing result comprises the probability of each anchor frame corresponding to the chromosome;
step S332, inputting the mosaic into the anchor frame regression part, processing the mosaic by using a 1 × 1 convolutional layer to obtain a third processing result, and processing the first processing result by using a 9 × 9 void convolutional layer to obtain a fourth processing result, wherein the fourth processing result comprises offset data of each anchor frame, and the offset data comprises an offset coefficient of an x-axis coordinate of a center point of the anchor frame, an offset coefficient of a y-axis coordinate of the center point of the anchor frame, a scaling coefficient of a width of the anchor frame, a scaling coefficient of a height of the anchor frame and a rotation coefficient of an angle of the anchor frame;
in this step, the offset coefficient of the x-axis coordinate of the center point of the anchor frame, the offset coefficient of the y-axis coordinate of the center point of the anchor frame, the scaling coefficient of the width of the anchor frame, the scaling coefficient of the height of the anchor frame, and the rotation coefficient of the angle of the anchor frame are sequentially recorded as
S333, screening all the anchor frames according to the second processing result, wherein if the probability of containing chromosomes corresponding to the anchor frames is greater than a second threshold value, the anchor frames are reserved to obtain reserved anchor frames;
considering that the problem of unbalanced anchor frames and local similarity may become more serious when a plurality of anchor frames are arranged at each anchor frame arrangement position, the method in step S333 is adopted to screen all the anchor frames, and by adopting the method, the total number of the anchor frames can be significantly reduced, so that the problem of unbalanced anchor frames and local similarity can be effectively solved; meanwhile, the number of the anchor frames is reduced, so that the reasoning process can be accelerated and the detection performance can be improved.
In this step, the second threshold may be set by user according to the user's requirement, and in this embodiment, the second threshold is 0.5;
step S334, processing each of the reserved anchor frames according to the fourth processing result to obtain a final positioning frame.
The step can be used for carrying out center point offset, size scaling and angle offset on the reserved anchor frame, and the anchor frame can be used for accurately positioning the target in such a way.
After the positioning frame is obtained, extracting and classifying the chromosome according to the positioning frame to generate a chromosome karyotype image, and analyzing and processing the generated chromosome karyotype image, for example, screening chromosome structure and quantity abnormality according to the karyotype image;
the specific implementation steps of the step are as follows:
each anchor box is denoted as (x) a ,y a ,w a ,h a ,θ a ) Wherein x is a Is the x-axis coordinate of the center point of the anchor frame, y a Is the y-axis coordinate of the center point of the anchor frame, w a Is the width of the anchor frame, h a Is the height, theta, of the anchor frame a Is the rotation angle of the anchor frame;
to obtain
Then, the calculation is carried out according to the formula (5), and the final positioning frame (x) can be obtained p ,y p ,w p ,h p ,θ p )；
Example 2
As shown in fig. 2, the present embodiment provides a chromosome extraction apparatus, which includes an acquisition module 701, a training module 702, and an extraction module 703.
An obtaining module 701, configured to obtain a first image, where the first image includes a metaphase cell image of a chromosome to be extracted;
a training module 702, configured to train a feature multi-scale network to obtain a trained feature multi-scale network model, where the trained feature multi-scale network model includes a saliency map prediction portion, an anchor frame classification portion, and an anchor frame regression portion;
an extracting module 703, configured to process the first image by using the trained feature multi-scale network model, to obtain a location frame of each chromosome in the metaphase cell image of the chromosome to be extracted, and extract the chromosome according to the location frame.
In a specific embodiment of the present disclosure, the training module 702 further includes a building unit 7021 and a training unit 7022.
A constructing unit 7021, configured to construct a training set, where the training set includes a plurality of samples, each sample includes first data, second data, and third data, the first data includes a metaphase cell image of chromosomes, the second data includes a metaphase cell image of chromosomes with intersections of a skeleton, an edge, and overlapping chromosomes labeled in an artificial manner, and the third data includes a standard frame obtained by labeling chromosomes in the metaphase cell image of chromosomes in an artificial manner;
a training unit 7022, configured to train the feature multi-scale network model by using the training set based on a stochastic gradient descent method, and determine whether a training stop condition is met according to a loss function, if so, stop training, and obtain the trained feature multi-scale network model.
In a specific embodiment of the present disclosure, the training unit 7022 further includes a first building subunit 70221 and a second building subunit 70222.
A first constructing subunit 70221, configured to construct a loss function corresponding to the saliency map prediction portion, a loss function corresponding to the anchor frame classification portion, and a loss function corresponding to the anchor frame regression portion;
and a second constructing subunit 70222, configured to construct a total loss function according to the loss function corresponding to the saliency map prediction portion, the loss function corresponding to the anchor frame classification portion, and the loss function corresponding to the anchor frame regression portion.
In a specific embodiment of the present disclosure, the extracting module 703 further includes a processing unit 7031, an analyzing unit 7032, and a splicing unit 7033.
A processing unit 7031, configured to input the metaphase cell image of the chromosome to be detected into the trained feature multi-scale network model for feature extraction to obtain a feature map, and process the feature map by using the saliency map prediction part to obtain a saliency map, where the saliency map includes three channel images, which are a skeleton image of the chromosome, an edge image of the chromosome, and a cross point image of an overlapping chromosome;
an analyzing unit 7032, configured to analyze the skeleton image of the chromosome, use a pixel point on the skeleton image of the chromosome, where a pixel value of the pixel point is greater than a first threshold, as an anchor frame setting position, and set a preset number of anchor frames at each anchor frame setting position, where specifications of each anchor frame are different;
a splicing unit 7033, configured to perform feature splicing processing on the feature map and the saliency map to obtain a spliced map, and obtain the location frame based on the spliced map, the anchor frame classification portion, and the anchor frame regression portion.
In a specific embodiment of the present disclosure, the splicing unit 7033 further includes a first processing subunit 70331, a second processing subunit 70332, a screening subunit 70333, and a third processing subunit 70334.
A first processing subunit 70331, configured to input the mosaic into the anchor frame classification part, process the mosaic by using a 1 × 1 convolutional layer to obtain a first processing result, and process the first processing result by using a 9 × 9 void convolutional layer to obtain a second processing result, where the second processing result includes a probability that each anchor frame corresponds to and includes a chromosome;
a second processing subunit 70332, configured to input the mosaic into the anchor frame regression portion, process the mosaic using a 1 × 1 convolutional layer to obtain a third processing result, and process the first processing result using a 9 × 9 void convolutional layer to obtain a fourth processing result, where the fourth processing result includes offset data of each anchor frame, and the offset data includes an offset coefficient of an anchor frame center point x-axis coordinate, an offset coefficient of an anchor frame center point y-axis coordinate, a scaling coefficient of an anchor frame width, a scaling coefficient of an anchor frame height, and a rotation coefficient of an anchor frame angle;
a screening subunit 70333, configured to screen all the anchor frames according to the second processing result, where if the probability that the chromosome is included corresponding to the anchor frame is greater than a second threshold, the anchor frame is retained to obtain a retained anchor frame;
a third processing subunit 70334, configured to process each of the reserved anchor frames according to the fourth processing result, so as to obtain a final location frame.
It should be noted that, regarding the apparatus in the above embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment related to the method, and will not be elaborated herein.
Example 3
Corresponding to the above method embodiments, the embodiments of the present disclosure also provide a chromosome extraction apparatus, and the chromosome extraction apparatus described below and the chromosome extraction method described above may be referred to correspondingly to each other.
Fig. 3 is a block diagram illustrating a chromosome extraction apparatus 800 according to an exemplary embodiment. As shown in fig. 3, the chromosome extraction apparatus 800 may include: a processor 801, a memory 802. The chromosome extraction device 800 can also include one or more of a multimedia component 803, an I/O interface 804, and a communication component 805.
The processor 801 is configured to control the overall operation of the chromosome extraction apparatus 800, so as to complete all or part of the steps in the chromosome extraction method. The memory 802 is used to store various types of data to support the operation of the chromosome extraction device 800, such data can include, for example, instructions for any application or method operating on the chromosome extraction device 800, as well as application-related data, such as contact data, messaging, pictures, audio, video, and so forth. The Memory 802 may be implemented by any type of volatile or non-volatile Memory device or combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic Memory, flash Memory, magnetic disk or optical disk. The multimedia components 803 may include screen and audio components. Wherein the screen may be, for example, a touch screen and the audio component is used for outputting and/or inputting audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signal may further be stored in the memory 802 or transmitted through the communication component 805. The audio assembly also includes at least one speaker for outputting audio signals. The I/O interface 804 provides an interface between the processor 801 and other interface modules, such as a keyboard, mouse, buttons, etc. These buttons may be virtual buttons or physical buttons. The communication component 805 is used for wired or wireless communication between the chromosome extraction device 800 and other devices. Wireless communication, such as Wi-Fi, bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination of one or more of them, so that the corresponding communication component 805 may include: Wi-Fi module, bluetooth module, NFC module.
In an exemplary embodiment, the chromosome extraction Device 800 may be implemented by one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components for performing the above-described chromosome extraction method.
In another exemplary embodiment, there is also provided a computer-readable storage medium including program instructions which, when executed by a processor, implement the steps of the chromosome extraction method described above. For example, the computer-readable storage medium may be the above-described memory 802 including program instructions that are executable by the processor 801 of the chromosome extraction device 800 to perform the above-described chromosome extraction method.
Example 4
Corresponding to the above method embodiment, the embodiment of the present disclosure further provides a readable storage medium, and a readable storage medium described below and the above chromosome extraction method can be correspondingly referred to each other.
A readable storage medium, on which a computer program is stored, which, when executed by a processor, implements the steps of the chromosome extraction method of the above-described method embodiments.
The readable storage medium may be a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and various other readable storage media capable of storing program codes.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A chromosome extraction method, comprising:
acquiring a first image, wherein the first image comprises a metaphase cell image of a chromosome to be extracted;
training a characteristic multi-scale network to obtain a trained characteristic multi-scale network model, wherein the trained characteristic multi-scale network model comprises a saliency map prediction part, an anchor frame classification part and an anchor frame regression part;
and processing the first image by using the trained feature multi-scale network model to obtain a positioning frame of each chromosome in the metaphase cell image of the chromosome to be extracted, and extracting the chromosome according to the positioning frame.
2. The chromosome extraction method according to claim 1, wherein training the feature multi-scale network model to obtain the trained feature multi-scale network model comprises:
constructing a training set, wherein the training set comprises a plurality of samples, each sample comprises first data, second data and third data, the first data comprises a metaphase cell image of chromosomes, the second data comprises a metaphase cell image of chromosomes with frameworks, edges and intersections of overlapped chromosomes labeled in a manual mode, and the third data comprises a standard frame obtained by labeling the chromosomes in the metaphase cell image of chromosomes in a manual mode;
and training the characteristic multi-scale network model by using the training set based on a random gradient descent method, judging whether a training stopping condition is met or not according to a loss function, and stopping training if the training stopping condition is met to obtain the trained characteristic multi-scale network model.
3. The chromosome extraction method according to claim 2, characterized in that the method for constructing the loss function comprises:
constructing a first loss function corresponding to the prediction part of the saliency map, a second loss function corresponding to the classification part of the anchor frame and a third loss function corresponding to the regression part of the anchor frame;
and constructing the loss function according to the first loss function corresponding to the saliency map prediction part, the second loss function corresponding to the anchor frame classification part and the third loss function corresponding to the anchor frame regression part.
4. The chromosome extraction method according to claim 1, wherein the step of processing the first image by using the trained feature multi-scale network model to obtain a location frame of each chromosome in the metaphase cell image of the chromosome to be detected comprises:
inputting the metaphase cell image of the chromosome to be detected into the trained feature multi-scale network model for feature extraction to obtain a feature map, and processing the feature map by using the prediction part of the feature map to obtain a significant map, wherein the significant map comprises three channel images which are a skeleton image of the chromosome, an edge image of the chromosome and a cross point image of an overlapped chromosome;
analyzing the skeleton image of the chromosome, taking pixel points with pixel values larger than a first preset threshold value on the skeleton image of the chromosome as anchor frame setting positions, and setting a preset number of anchor frames at each anchor frame setting position, wherein the specifications of the anchor frames are different;
and performing feature splicing processing on the feature map and the saliency map to obtain a spliced map, and obtaining the positioning frame based on the spliced map, the anchor frame classification part and the anchor frame regression part.
5. A chromosome extraction apparatus, characterized by comprising:
the acquisition module is used for acquiring a first image, wherein the first image comprises a metaphase cell image of the chromosome to be extracted;
the training module is used for training the characteristic multi-scale network to obtain a trained characteristic multi-scale network model, and the trained characteristic multi-scale network model comprises a saliency map prediction part, an anchor frame classification part and an anchor frame regression part;
and the extraction module is used for processing the first image by using the trained characteristic multi-scale network model to obtain a positioning frame of each chromosome in the metaphase cell image of the chromosome to be extracted, and extracting the chromosome according to the positioning frame.
6. The chromosome extraction apparatus according to claim 5, wherein the training module includes:
the system comprises a construction unit, a data acquisition unit and a data processing unit, wherein the construction unit is used for constructing a training set, the training set comprises a plurality of samples, each sample comprises first data, second data and third data, the first data comprises a metaphase chromosome cell image, the second data comprises a metaphase chromosome cell image of which a skeleton, an edge and a cross point of overlapped chromosomes are marked in a man-made mode, and the third data comprises a standard frame obtained by marking the chromosomes in the metaphase chromosome cell image in the man-made mode;
and the training unit is used for training the characteristic multi-scale network model by using the training set based on a random gradient descent method, judging whether a training stopping condition is met or not according to a loss function, and stopping training if the training stopping condition is met to obtain the trained characteristic multi-scale network model.
7. The chromosome extraction apparatus according to claim 6, wherein the training unit includes:
the first construction subunit is used for constructing a loss function corresponding to the prediction part of the saliency map, a loss function corresponding to the classification part of the anchor frame and a loss function corresponding to the regression part of the anchor frame;
and the second construction subunit is used for constructing a total loss function according to the loss function corresponding to the prediction part of the saliency map, the loss function corresponding to the anchor frame classification part and the loss function corresponding to the anchor frame regression part.
8. The chromosome extraction apparatus according to claim 5, wherein the extraction module includes:
the processing unit is used for inputting the metaphase cell image of the chromosome to be detected into the trained feature multi-scale network model for feature extraction to obtain a feature map, and processing the feature map by using the prediction part of the feature map to obtain a significant map, wherein the significant map comprises three channel images which are a skeleton image of the chromosome, an edge image of the chromosome and a cross point image of an overlapped chromosome;
the analysis unit is used for analyzing the skeleton image of the chromosome, taking pixel points with pixel values larger than a first threshold value on the skeleton image of the chromosome as anchor frame setting positions, and setting a preset number of anchor frames at each anchor frame setting position, wherein the specifications of the anchor frames are different;
and the splicing unit is used for performing characteristic splicing processing on the characteristic diagram and the saliency map to obtain a splicing diagram, and obtaining the positioning frame based on the splicing diagram, the anchor frame classification part and the anchor frame regression part.
9. A chromosome extraction device characterized by comprising:
a memory for storing a computer program;
a processor for implementing the steps of the chromosome extraction method according to any one of claims 1 to 4 when executing the computer program.
10. 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 extraction method according to any one of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210624344.4A CN114972796A (en) | 2022-06-02 | 2022-06-02 | Chromosome extraction method, device, equipment and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210624344.4A CN114972796A (en) | 2022-06-02 | 2022-06-02 | Chromosome extraction method, device, equipment and readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114972796A true CN114972796A (en) | 2022-08-30 |
Family
ID=82960230
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210624344.4A Pending CN114972796A (en) | 2022-06-02 | 2022-06-02 | Chromosome extraction method, device, equipment and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114972796A (en) |
-
2022
- 2022-06-02 CN CN202210624344.4A patent/CN114972796A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10817717B2 (en) | Method and device for parsing table in document image | |
| CN111160335B (en) | Image watermark processing method and device based on artificial intelligence and electronic equipment | |
| CN111476284A (en) | Image recognition model training method, image recognition model training device, image recognition method, image recognition device and electronic equipment | |
| WO2021217857A1 (en) | Slice defect detection method and apparatus, and electronic device and readable storage medium | |
| EP1586897B1 (en) | Image analysis supporting method, image analysis supporting program, and image analysis supporting device | |
| US20200242398A1 (en) | Information processing method and information processing system | |
| CN113095434A (en) | Target detection method and device, electronic equipment and storage medium | |
| Li et al. | Automatic comic page segmentation based on polygon detection | |
| CN110827236A (en) | Neural network-based brain tissue layering method and device, and computer equipment | |
| JP2022548160A (en) | Preparing training datasets using machine learning algorithms | |
| CN115409069A (en) | Village and town building identification method, classification method, device, electronic equipment and medium | |
| CN111310826A (en) | Method and device for detecting labeling abnormity of sample set and electronic equipment | |
| CN114549993A (en) | Method, system and device for scoring line segment image in experiment and readable storage medium | |
| Xue et al. | Table analysis and information extraction for medical laboratory reports | |
| CN115862113A (en) | Stranger abnormity identification method, device, equipment and storage medium | |
| Lee et al. | Image analysis using machine learning for automated detection of hemoglobin H inclusions in blood smears-a method for morphologic detection of rare cells | |
| CN112102250A (en) | Method for establishing and detecting pathological image detection model with training data as missing label | |
| CN114694130A (en) | Method and device for detecting telegraph poles and pole numbers along railway based on deep learning | |
| CN106682669A (en) | Image processing method and mobile terminal | |
| CN116152576B (en) | Image processing method, device, equipment and storage medium | |
| US8755606B2 (en) | Systems and methods for efficient feature extraction accuracy using imperfect extractors | |
| CN114972796A (en) | Chromosome extraction method, device, equipment and readable storage medium | |
| CN111539390A (en) | Small target image identification method, equipment and system based on Yolov3 | |
| CN108665769B (en) | Network teaching method and device based on convolutional neural network | |
| US20220207724A1 (en) | Method of determining a distribution of stem cells in a cell image, electronic device, and storage medium |
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 |