CN114219786B - Chromosome karyotype analysis method and system based on deep learning - Google Patents
Chromosome karyotype analysis method and system based on deep learning Download PDFInfo
- Publication number
- CN114219786B CN114219786B CN202111547786.5A CN202111547786A CN114219786B CN 114219786 B CN114219786 B CN 114219786B CN 202111547786 A CN202111547786 A CN 202111547786A CN 114219786 B CN114219786 B CN 114219786B
- Authority
- CN
- China
- Prior art keywords
- chromosome
- image
- mask
- classification
- deep learning
- 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 445
- 238000013135 deep learning Methods 0.000 title claims abstract description 62
- 238000004458 analytical method Methods 0.000 title claims abstract description 42
- 230000011218 segmentation Effects 0.000 claims abstract description 60
- 210000002230 centromere Anatomy 0.000 claims abstract description 40
- 238000007781 pre-processing Methods 0.000 claims abstract description 18
- 238000001514 detection method Methods 0.000 claims description 41
- 238000012549 training Methods 0.000 claims description 26
- 238000013145 classification model Methods 0.000 claims description 21
- 230000002159 abnormal effect Effects 0.000 claims description 19
- 238000000034 method Methods 0.000 claims description 19
- 230000005856 abnormality Effects 0.000 claims description 13
- 238000006243 chemical reaction Methods 0.000 claims description 13
- 230000002093 peripheral effect Effects 0.000 claims description 13
- 238000012545 processing Methods 0.000 claims description 9
- 238000013507 mapping Methods 0.000 claims description 8
- 238000010606 normalization Methods 0.000 claims description 8
- 238000012800 visualization Methods 0.000 claims description 8
- 230000000007 visual effect Effects 0.000 claims description 7
- 238000013528 artificial neural network Methods 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 5
- 230000008030 elimination Effects 0.000 claims description 3
- 238000003379 elimination reaction Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 2
- 238000004519 manufacturing process Methods 0.000 claims 1
- 238000002372 labelling Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 5
- 230000003287 optical effect Effects 0.000 description 3
- 238000004590 computer program Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 208000031404 Chromosome Aberrations Diseases 0.000 description 1
- 206010067477 Cytogenetic abnormality Diseases 0.000 description 1
- 102000017182 Ikaros Transcription Factor Human genes 0.000 description 1
- 108010013958 Ikaros Transcription Factor Proteins 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000003759 clinical diagnosis Methods 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 230000031864 metaphase Effects 0.000 description 1
- 238000000399 optical microscopy Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20092—Interactive image processing based on input by user
- G06T2207/20104—Interactive definition of region of interest [ROI]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
- Investigating Or Analysing Biological Materials (AREA)
Abstract
The invention relates to the technical field of chromosome karyotype analysis, and provides a chromosome karyotype analysis method and system based on deep learning, which comprises the following steps: obtaining an original image of a chromosome; preprocessing an original image of a chromosome to obtain a clear image; according to the semantic information of the clear image, segmenting the foreground semantic class into minimum chromosome unit clusters according to the number of contours to obtain a plurality of clustered images containing a single chromosome or a plurality of chromosomes; carrying out chromosome segmentation and classification on the clustered images by adopting a chromosome instance segmentation model based on deep learning to obtain single chromosome images with classification numbers and contour information; and finally, arranging the chromosomes according to the classification number, the polarity number, the structural variation number and the positions of the centromere to obtain a standard karyotype chart so as to complete karyotype analysis of the chromosomes.
Description
Technical Field
The invention relates to the technical field of chromosome karyotype analysis, in particular to a chromosome karyotype analysis method and system based on deep learning.
Background
Karyotyping is one of the important tools in genetic science research and in auxiliary clinical diagnosis. In the routine clinical workflow for diagnosing chromosome abnormality, a geneticist needs to visually check the chromosome number and form of the specimen by an optical microscope according to the bands of light and dark, so that the problems of high professional requirement and long time consumption exist.
Semi-automatic commercial karyotyping systems are currently in use to assist geneticists in performing analytical tasks under optical microscopy, such as CytoVision, Ikaros, ASI HiBand, etc. The system images chromosomes on a specimen slide by an optical microscope, and then performs operations such as chromosome segmentation and classification based on an image method. However, most of the above automatic operations are focused on a certain step of karyotype analysis, such as single chromosome type prediction, chromosome detection in metaphase maps, specific chromosome segmentation, etc., in actual operations, a user is required to manually adjust the definition of chromosomes in images, segment chromosomes, or drag chromosomes to the positions of corresponding types, and a full-flow full-automatic solution from a shot map to a karyotype map is lacking, so that the current chromosome karyotype analysis work is time-consuming, inefficient, and the accuracy cannot be guaranteed.
Disclosure of Invention
In order to overcome the defects that the chromosome karyotype analysis work is long in time consumption, low in efficiency and incapable of guaranteeing the accuracy rate due to the lack of full-automatic analysis operation from a shot image to a karyotype image in the prior art, the invention provides a chromosome karyotype analysis method based on deep learning and a chromosome karyotype analysis system based on deep learning.
In order to solve the technical problems, the technical scheme of the invention is as follows:
a chromosome karyotype analysis method based on deep learning comprises the following steps:
s1, obtaining an original chromosome image;
s2, preprocessing the original image of the chromosome to obtain a clear image;
s3, according to the semantic information of the clear image, segmenting the foreground semantic class into the smallest chromosome unit cluster according to the contour number to obtain a plurality of clustered images containing a single chromosome or a plurality of chromosomes;
s4, chromosome segmentation and classification are carried out on the clustered images by adopting an example segmentation model based on deep learning, and a single chromosome image with a classification number and contour position information is obtained;
s5, backtracking the single chromosome image with the classification number and the contour position information to a chromosome image with clear stripes for visual display;
s6, performing polarity prediction and structural variation classification on the single chromosome image by adopting a deep learning-based classification model to obtain a single chromosome image with a polarity number and a structural variation number;
s7, counting the total number of the single chromosomes, and respectively counting the number of the single chromosomes with each number to obtain the number abnormality prompt of the chromosomes with each number;
s8, carrying out centromere detection on the single chromosome by adopting a feature point detection model based on deep learning to obtain the centromere position of the single chromosome;
and S9, arranging the single chromosomes according to the classification number, the polarity number and the structural variation number of the single chromosome image and the positions of the centromere to obtain a standard karyotype chart, and marking the chromosome numbers with abnormal quantity and the single chromosomes with abnormal structure on the karyotype chart to finish the karyotype analysis of the chromosomes.
Furthermore, the invention also provides a chromosome karyotype analysis system based on deep learning, which comprises:
the image acquisition module is used for acquiring an original chromosome image;
the image preprocessing module is used for preprocessing the original image of the chromosome to obtain a clear image;
the clustering module is used for segmenting the foreground semantic class into a minimum chromosome unit cluster according to the outline number according to the semantic class information of the clear image to obtain a plurality of clustered images containing a single chromosome or a plurality of chromosomes;
the segmentation and classification module is used for carrying out chromosome segmentation and classification on the clustering images based on a chromosome instance segmentation model of deep learning to obtain single chromosome images with classification numbers; the single chromosome image processing system is used for carrying out polarity prediction and structural variation classification on the single chromosome image by adopting a classification model based on deep learning to obtain a single chromosome image with a polarity number and a structural variation number;
the system comprises a centromere detection module, a characteristic point detection module and a characteristic point detection module, wherein the centromere detection module is used for carrying out centromere detection on a single chromosome by adopting a characteristic point detection model based on deep learning to obtain the centromere position of the single chromosome;
and the visualization module is used for backtracking the single chromosome image with the classification number to the chromosome image with clear stripes, arranging the single chromosomes according to the classification number, the polarity number and the structural variation number of the single chromosome image and the positions of the centromere to obtain a standard karyotype chart, marking the chromosome number with abnormal quantity and the single chromosome with abnormal structure on the karyotype chart, and visually displaying the karyotype chart.
Compared with the prior art, the technical scheme of the invention has the beneficial effects that: the method is suitable for tasks such as chromosome segmentation, chromosome classification, chromosome arrangement, chromosome counting and the like, and the denoised chromosome image with clear stripes is obtained by preprocessing the image, so that the subsequent operation of the operation tasks such as chromosome clustering, chromosome segmentation, chromosome classification and the like is facilitated; the invention also realizes the automatic operation of the whole process from the shooting end to the analysis end, and can effectively improve the working efficiency of the user on the chromosome karyotype analysis.
Drawings
FIG. 1 is a flowchart of a method for deep learning-based karyotyping according to example 1.
Fig. 2 is a chromosome original image.
FIG. 3 is a clearly striped chromosome image.
Fig. 4 is a clear image.
Fig. 5 is a clustered image.
Fig. 6 is a single chromosome image.
Fig. 7 is a chromosome image displayed visually.
FIG. 8 is a karyotype image of chromosomes arranged in centromeres.
FIG. 9 is an image of karyotypes arranged in bottom alignment.
Fig. 10 is an architecture diagram of the chromosome karyotype analysis system based on deep learning of example 2.
Detailed Description
The drawings are for illustrative purposes only and are not to be construed as limiting the patent;
it will be understood by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.
The technical solution of the present invention is further described below with reference to the accompanying drawings and examples.
Example 1
The present embodiment provides a method for analyzing a karyotype based on deep learning, and as shown in fig. 1, the method is a flowchart of the method for analyzing a karyotype based on deep learning of the present embodiment.
The method for analyzing the karyotype based on deep learning provided by the embodiment comprises the following steps:
step 1: and acquiring a chromosome original image.
In the embodiment, the chromosome image is shot and collected through equipment to obtain the original chromosome image. As shown in fig. 2, this is the original image of the chromosome of this embodiment.
Step 2: and preprocessing the original image of the chromosome to obtain a clear image.
In this embodiment, the step of preprocessing the chromosome image includes: carrying out style conversion on the acquired chromosome original image by adopting an image conversion model based on a countermeasure generation network to obtain a clear-striped chromosome image shown in FIG. 3; and (3) eliminating the background of the chromosome image with clear stripes by adopting a semantic segmentation model based on a neural network to obtain a clean and clear image as shown in figure 4.
The image conversion model is a model designed end to end and used for converting a source domain image into a target domain image, wherein the source domain provides image content, the target domain provides the attribute or style of the image, and the stylization of the target domain is realized under the image content provided by the source domain, so that the conversion from the source domain image to the target domain image is realized.
In the embodiment, original chromosome images acquired by different devices or shooting devices are converted into chromosome images with clear stripes, so that subsequent chromosome segmentation and classification operations are facilitated, the difference of the images is reduced, and segmentation and classification errors caused by chromosome images of different shooting styles are reduced.
In one embodiment, the image conversion model includes a first generator, a second generator, a first discriminator, and a second discriminator; when the training set is input into a chromosome image conversion model for training, a first generator generates a corresponding image G with clear stripes according to an input chromosome original image x; the first discriminator discriminates the image G generated by the first generator according to the target domain image set Y to obtain the mapping of the source domain image set X in the target domain image set Y, and an optimal mapping exists to enable the image G and the target domain image set Y to have the same distribution; selecting a chromosome image with clear stripes from the target domain image set Y, inputting the chromosome image into a second generator, and generating an image G' by the second generator; the second discriminator discriminates the image G 'generated by the second generator according to the source domain image set X, so as to obtain the mapping of the target domain image set Y in the source domain image set X, and an optimal mapping exists so that the image G' and the source domain image set X have the same distribution.
In another embodiment, pix2pix, cycleGAN, DuaIGAN, DiscoGAN, UNIT, MUNIT, starGAN, APADE, etc. models are employed as image conversion models, and are not limited to the image conversion models.
In one embodiment, the step of background elimination of the clearly striped chromosome image by using the semantic segmentation model based on the neural network comprises the following steps: the semantic segmentation model takes a chromosome region in an image as a foreground semantic class, takes a background region in the image as a background semantic class, predicts a semantic class to which each pixel point of an input image belongs, generates a mask according to the semantic class of each pixel point to obtain a foreground semantic mask and a background semantic mask, deletes the background semantic mask and retains the foreground semantic mask, so that the segmentation of chromosome and background content is realized, and a clear image without background noise is obtained.
The semantic segmentation model adopts one or more of U-Net, FCN, SegNet, PSPNet and deep Lab, and is not limited to the semantic segmentation model.
And step 3: and according to the semantic information of the clear image, segmenting the foreground semantic class into the minimum chromosome unit cluster according to the contour number to obtain a plurality of clustered images containing a single chromosome or a plurality of chromosomes.
The method comprises the following steps of carrying out segmentation on the foreground semantic class by the minimum chromosome unit cluster according to the number of contours, wherein the segmentation comprises the following steps:
s3.1, extracting the outlines of the foreground semantic masks in the clear images, respectively calculating the minimum peripheral rectangular frames of the extracted n outlines, and respectively storing the coordinates [ pointx, pointy ] at the upper left corners of the n minimum peripheral rectangular frames, the length value of the rectangular frame and the width value of the rectangular frame;
further, storing the coordinates [ pointx, pointy ] at the upper left corner of the n minimum peripheral rectangular frames, the length value length and the width value width of the rectangular frames in a list form;
s3.2, creating n pure-color background pictures according to the size of the minimum peripheral rectangular frame, and then drawing the extracted n outlines in the corresponding pure-color background pictures one by one to obtain n outline masks;
and S3.3, displaying an ROI area image in the clear image according to the outline mask to obtain a chromosome cluster corresponding to the outline mask, storing the ROI area as a cluster image, and carrying out image size normalization on the cluster image to obtain the cluster image shown in the figure 5.
Further, the step of normalizing the image size in this embodiment includes:
if the length value of the size length of the clustered image is greater than the preset length L or the width of the clustered image is greater than the preset width W, namely if the length value length is greater than the preset length L, the long edge of the clustered image is taken as the length of the picture after filling, the two sides of the wide edge are filled at equal intervals according to the proportion of L/W, so that the length-width ratio of the filled picture is equal to L/W, and the size of the picture is reset to L W after the picture is filled into a rectangular picture; if the width value width is larger than W, taking the wide edge of the clustering graph as the length of the filled picture, carrying out equidistant filling on two edges of the long edge according to the proportion of L/W, enabling the length-width ratio of the filled image to be equal to L/W, and resetting the size of the filled image to be L x W after filling the image into a rectangular image;
if the length value length of the clustering image is smaller than the preset length L and the width value width of the clustering image is smaller than the preset width W, edge filling is carried out on the clustering image until the size of the clustering image is L x W, the chromosome image is placed in the middle of the clustering image during filling, and the equal-distance filling of the upper side, the lower side, the left side and the right side is carried out on the clustering image. In this embodiment, the value of the preset length L ranges from 100 to 1000, and the value of the preset width W ranges from 100 to 1000.
And 4, step 4: and (3) carrying out chromosome segmentation and classification on the clustered images by adopting a chromosome example segmentation model based on deep learning to obtain single chromosome images with classification numbers.
In this embodiment, the step of performing chromosome segmentation and classification on the clustered image includes: constructing a chromosome instance segmentation model based on deep learning, and inputting the clustering image into the chromosome instance segmentation model to obtain a chromosome number and a chromosome mask corresponding to the chromosome number in the clustering image; and (3) cutting the corresponding clustered images according to the chromosome mask to obtain a single chromosome image shown in figure 6, and labeling the chromosome numbers on the cut images to obtain a series of single chromosome images with classification numbers.
In the process of classifying the clustered images, the example category in the foreground semantics is distinguished through the foreground semantics and the background semantics of each pixel point in the clustered images, namely the serial number of each chromosome in the clustered images is predicted. And obtaining the serial number of each chromosome in the clustering image and the corresponding chromosome mask after chromosome segmentation and classification.
In one embodiment, the chromosome instance segmentation model employs Mask RCNN, BlendMask, SOLO, SOLOV2, and the like.
Further, the step of constructing a training data set for training the chromosome instance segmentation model is as follows:
and (4-a) acquiring a clinically existing karyotype graph, and cutting the karyotype graph according to regions to obtain each numbered chromosome image and a corresponding label.
And (4-b) performing feature matching on each numbered chromosome image on a pre-stored clear image by using an image registration mode to obtain a chromosome mask matched with each numbered chromosome image in the clear image, and assigning the number of the chromosome image to the matched chromosome in the clear image.
And (4-c) creating a background picture of a background semantic mask for each chromosome mask subjected to feature matching by taking the size of the clear image as a target size.
(4-d) drawing each feature-matched chromosome mask on the background picture in the form of filled-in inner contours.
(4-e) judging whether the chromosome mask completing the feature matching on the clear image has a coincidence region, if so, merging the chromosome masks with the coincidence region to generate a clustering map mask containing a plurality of chromosomes, such as a new cross mask, an adhesion mask, an overlapping mask, a cross overlapping mask or a cross adhesion mask, and simultaneously storing the chromosome masks and the serial numbers thereof on the clustering map mask.
And (4-f) taking the chromosome mask and the clustering map mask as regions of interest, and displaying the ROI region image of the chromosome mask subjected to feature matching.
(4-g) extracting a minimum rectangular area according to the chromosome mask and the clustering map mask, cutting the minimum rectangular area of the image after the chromosome mask, the clustering map mask and the ROI corresponding to the chromosome mask and the clustering map mask are displayed, and then carrying out image size normalization processing on the cut image to obtain a single chromosome image and a clustering map containing a plurality of chromosomes.
The image size normalization processing step in this step is the same as the image size normalization processing performed on the clustered image in the step S3.3 described above.
(4-h) writing chromosome labels and chromosome mask outline information corresponding to the single chromosome image and the clustering graph containing the plurality of chromosomes into a labeling file, making example segmentation labeling information, obtaining a large number of single chromosome images and the plurality of chromosome images and the corresponding labeling files, and forming a training data set for training the chromosome example segmentation model.
Through the steps, a large number of single or multiple chromosome images and corresponding annotation files stored in a json file form can be automatically manufactured.
And 5: and backtracking the single chromosome image with the classification number and the contour position information to a chromosome image with clear stripes for visual display.
The method comprises the following steps of tracing back a single chromosome image with a classification number to a chromosome image with clear stripes for visual display, wherein the steps comprise:
(5-a) calculating the minimum peripheral rectangle of the chromosome mask of the single chromosome image with the classification number to obtain a minimum peripheral rectangle list, and the length and width of the image when the clustering graph where the single chromosome is located is clustered;
(5-b) carrying out image size normalization processing on the single chromosome images according to the length and width of each single chromosome image cluster; the method comprises the following specific steps:
if the length and the width of the single chromosome image cluster are judged, and the image size is adjusted: when the length is greater than the preset length L or the width is greater than the preset width W, the chromosome mask of the single chromosome image is resized: when the length value length is larger than a preset length L, taking the long edge of the chromosome mask of the single chromosome image as the length of the filled image, filling the wide edge at equal intervals at two sides according to the L/W ratio to enable the length-width ratio of the filled image to be equal to L/W, and resetting the size of the filled image to be L x W; when the width value width is larger than the preset width W, taking the wide edge of the single chromosome image as the length of the filled image, filling the long edge of the filled image at equal intervals according to the proportion of L/W, enabling the length-width ratio of the filled image to be equal to L/W, and resetting the size of the filled image to be L x W;
if the length value length of the single chromosome image is smaller than the preset length L and the width value width is smaller than the preset width W, edge filling is carried out on the chromosome mask of the single chromosome image until the image size is L x W, the chromosome image is placed in the middle of the image during filling, and then equidistant filling is carried out on the upper side, the lower side, the left side and the right side of the single chromosome image. In the embodiment, the value range of L is 100-1000, and the value range of W is 100-1000.
(5-c) extracting the contour of the chromosome mask in the single chromosome image after the image size adjustment is finished, then mapping the starting point of the minimum peripheral matrix of the clustering graph to the clear image as the starting point of the contour, and mapping the contour to the clear image;
and (5-d) after traversing the chromosome mask in all the single chromosome images of the basic unit, backtracking the single chromosome images with the classification numbers which are segmented and classified into the clear images for visual display.
In the actual visual display, all the single chromosomes in the chromosome image are segmented by color contours, and the classification numbers of the single chromosome images are distinguished by contour colors among different chromosomes, so that the visually displayed chromosome image shown in fig. 7 is obtained.
Step 6: and performing polarity prediction and structural variation classification on the single chromosome image by adopting a deep learning-based classification model to obtain the single chromosome image with the polarity number and the structural variation number.
In the step, a chromosome classification model based on deep learning is constructed, and a single chromosome image is input into the chromosome classification model to obtain chromosome polarity numbers and abnormal numbers in the single chromosome image. The specific steps for constructing the chromosome classification model are as follows:
(6-a) acquiring a clinically existing karyotype graph, and cutting the karyotype graph according to regions to obtain each numbered chromosome image and a corresponding label;
(6-b) making a polarity label for each numbered chromosome image, vertically inverting the randomly selected chromosome image by a certain amount, marking the partial image with a downward direction (indicated by the number a), and marking the rest numbered images with an upward direction (indicated by the number b);
(6-c) creating a structural abnormality label for the image labeled with the polarity label, labeling a chromosome image with structural abnormality as abnormal (denoted by the number AN), and labeling a chromosome image without structural abnormality as normal (denoted by the number N);
(6-d) writing chromosome numbers, polarity labels and structural abnormality labels corresponding to the chromosome images into an annotation file, making single chromosome image annotation information, obtaining a large number of single chromosome images and corresponding annotation files thereof, and forming a training data set for training the chromosome classification model;
and (6-e) inputting the training data set obtained in the step (6-d) into a chromosome classification model for training, adjusting parameters of the chromosome classification model according to a classification result output by the model, and completing construction of the chromosome classification model based on deep learning.
The deep learning-based chromosome classification model adopted in the embodiment comprises one or more of AlexNet, ZFNet, VGG, inclusion, ResNet, widereset (inclusion-ResNet _ v1/v2), DenseNet, resext, DPN, SENet and the like.
And 7: and counting the total number of the single chromosomes, and respectively counting the number of the single chromosomes with each number to obtain the number abnormality prompt of the chromosomes with each number.
And 8: and (3) carrying out the centromere detection on the single chromosome by adopting a characteristic point detection model based on deep learning to obtain the centromere position of the single chromosome.
In the step, the positions (X, Y) of the centromere in the single chromosome image are obtained by constructing a feature point detection model based on deep learning and inputting the single chromosome image into the chromosome feature point detection model. The steps for constructing the chromosome feature point detection model for training are as follows:
(8-a) collecting a clinically existing karyotype graph, and cutting the karyotype graph according to regions to obtain chromosome images of each number;
(8-b) making a centromere position label for each numbered chromosome image, and labeling coordinates (X, Y) of the centromere position of the chromosome in the image;
(8-c) writing the corresponding chromosome number and the corresponding centromere position into a label file, making single chromosome image label information, obtaining a large number of single chromosome images and corresponding label files thereof, and forming a training data set for training the chromosome feature point detection model;
and (8-d) inputting the training data set obtained in the step (8-c) into the chromosome feature point detection model, and adjusting parameters in the chromosome feature point detection model according to a detection result to complete the construction of the chromosome feature point detection model.
The chromosome feature point detection model in the present embodiment includes one or more of MTCNN, TCDCN, and the like.
And step 9: and arranging the single chromosomes according to the classification numbers, the polarity labels, the structural variation labels and the positions of the centromere of the single chromosome images to obtain a standard karyotype chart, and marking the chromosome numbers with abnormal numbers and the single chromosomes with abnormal structures on the karyotype chart to finish the karyotype analysis of the chromosomes.
In another embodiment, in order to observe the karyotype chart more intuitively, the following steps are also provided: and arranging the single chromosome images with the classification numbers, the polarity labels, the structural variation labels and the positions of the centromere according to the order of the chromosome numbers to generate a chromosome karyotype chart. The method comprises the following specific steps:
1) creating a white background image of the chromosome arrangement image in pixel values of 255;
2) for each clustered image, taking a single chromosome mask output by chromosome example segmentation as an ROI, and displaying an interested area on the clustered image to obtain a segmented single chromosome image;
3) cutting the minimal rectangular area of the obtained chromosome image, rotating the chromosome image by a corresponding angle, and correcting the chromosome;
4) and sequentially placing the single chromosome images after the rotation and the arrangement in the corresponding positions in the arrangement diagram according to the chromosome number sequence, printing the corresponding chromosome numbers, the classification numbers, the polarity numbers, the structural variation numbers and the filament-forming point positions at the proper positions below the chromosomes, and marking the chromosome numbers with abnormal quantity and the single chromosomes with abnormal structure on the karyotype diagram to obtain the chromosome karyotype arrangement diagram.
The arrangement mode of the single chromosome image in the chromosome karyotype arrangement chart comprises centromere arrangement and/or bottom alignment arrangement, and the chromosome karyotype arrangement charts shown in fig. 8 and 9 are respectively obtained by selection according to the use requirements of users. The chromosome karyotype analysis method provided by the embodiment can be applied to tasks such as chromosome segmentation, chromosome classification, chromosome arrangement, chromosome counting, karyotype graph output and the like, and realizes the automatic operation of the whole process from shooting to analysis of chromosome karyotype analysis, namely, the full-automatic karyotype analysis of the chromosome original image shot by an optical microscope can be effectively realized.
Example 2
The embodiment provides a chromosome karyotype analysis system based on deep learning, which is applied to the chromosome karyotype analysis method based on deep learning provided in embodiment 1. Fig. 10 is an architecture diagram of the chromosome karyotype analysis system based on deep learning according to this embodiment.
The deep learning-based karyotype analysis system provided by the embodiment includes:
the image acquisition module 1 is used for acquiring a chromosome original image;
the image preprocessing module 2 is used for preprocessing the original image of the chromosome to obtain a clear image;
the clustering module 3 is used for segmenting the foreground semantic class into a minimum chromosome unit cluster according to the outline number according to the semantic class information of the clear image to obtain a plurality of clustered images containing a single chromosome or a plurality of chromosomes;
the segmentation and classification module 4 is used for carrying out chromosome segmentation and classification on the clustering images based on a chromosome example segmentation model of deep learning to obtain single chromosome images with classification numbers; the single chromosome image processing system is used for carrying out polarity prediction and structural variation classification on the single chromosome image by adopting a classification model based on deep learning to obtain a single chromosome image with a polarity number and a structural variation number;
the centromere detection module 5 is used for carrying out centromere detection on the single chromosome by adopting a feature point detection model based on deep learning to obtain the centromere position of the single chromosome;
and the visualization module 6 is used for backtracking the single chromosome image with the classification number to the chromosome image with clear stripes, arranging the single chromosomes according to the classification number, the polarity number and the structural variation number of the single chromosome image and the positions of the centromere to obtain a standard karyotype chart, marking the chromosome number with abnormal quantity and the single chromosome with abnormal structure on the karyotype chart, and visually displaying the karyotype chart.
In the specific implementation process, the image acquisition module 1 is connected with external equipment to acquire the original chromosome image and transmit the acquired original chromosome image to the image preprocessing module 2.
The image preprocessing module 2 performs style conversion on the input chromosome original image by adopting an image conversion model based on the countermeasure generation network to obtain a chromosome image with clear stripes, and performs background elimination on the chromosome image with clear stripes by adopting a semantic segmentation model based on a neural network to obtain a clear image. The image preprocessing module 2 transmits the clear image obtained by preprocessing to the clustering module 3.
The clustering module 3 performs segmentation of the foreground semantic class according to the contour number on the basis of the semantic class information of the clear image to obtain a plurality of clustered images containing a single chromosome or a plurality of chromosomes, and then transmits the clustered images to the segmentation and classification module 4.
The segmentation and classification module 4 is loaded with a chromosome instance segmentation model based on deep learning and used for completing training, and the segmentation and classification module 4 inputs the clustering images into the chromosome instance segmentation model to obtain chromosome numbers and chromosome masks corresponding to the chromosome numbers in the clustering images; and cutting the corresponding clustered images according to the chromosome mask, marking the chromosome numbers on the cut images to obtain a series of single chromosome images with classification numbers, and transmitting the series of single chromosome images with classification numbers to the visualization module 6.
Meanwhile, the segmentation and classification module 4 performs polarity prediction and structural variation classification on the single chromosome image by using a preset deep learning-based classification model to obtain a single chromosome image with a polarity number and a structural variation number, and then transmits the single chromosome image with the polarity number and the structural variation number to the visualization module 5.
The centromere detection module 5 adopts a preset feature point detection model based on deep learning to perform centromere detection on the single chromosome to obtain the centromere position of the single chromosome, and then the centromere position of the single chromosome is transmitted to the visualization module 5.
The visualization module 6 takes the clear image as a unit, backtracks the single chromosome image with the classification number to the single chromosome image with clear stripes according to the classification number, the polarity number and the structural variation number of the single chromosome image, the position of the centromere, and arranges the single chromosomes to obtain a standard karyotype chart, and marks the chromosome number with abnormal quantity and the single chromosome with abnormal structure on the karyotype chart to complete the karyotype analysis of the chromosomes.
In the embodiment, a full-flow full-automatic chromosome karyotype analysis system capable of realizing the conversion from a shot image to a karyotype image is constructed, a large amount of background noise in the shot image is removed through the image preprocessing module 2, and the communication of tasks such as chromosome segmentation, chromosome classification, chromosome arrangement, chromosome counting and the like is realized by matching with the functional modules such as the clustering module 3, the segmentation and classification module 4, the centromere detection module 5 and the like, so that the workload of a geneticist is reduced, and the working efficiency of chromosome karyotype analysis is improved.
Example 3
The present embodiment provides a chromosome karyotype analysis system based on deep learning, which is applied to the chromosome karyotype analysis method based on deep learning provided in embodiment 1.
The deep learning based karyotyping system according to this embodiment includes a processor and a memory, where the memory stores a computer program, and the processor implements the steps of the deep learning based karyotyping method according to embodiment 1 when executing the computer program in the memory.
The same or similar reference numerals correspond to the same or similar parts;
it should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.
Claims (10)
1. A chromosome karyotype analysis method based on deep learning is characterized by comprising the following steps:
s1, obtaining an original chromosome image;
s2, preprocessing the original image of the chromosome to obtain a clear image;
s3, according to the semantic information of the clear image, segmenting the foreground semantic class into the smallest chromosome unit cluster according to the contour number to obtain a plurality of clustered images containing a single chromosome or a plurality of chromosomes;
s4, chromosome segmentation and classification are carried out on the clustered images by adopting an example segmentation model based on deep learning, and a single chromosome image with a classification number and contour position information is obtained;
s5, backtracking the single chromosome image with the classification number and the contour position information to a chromosome image with clear stripes for visual display;
s6, performing polarity prediction and structural variation classification on the single chromosome image by adopting a deep learning-based classification model to obtain a single chromosome image with a polarity number and a structural variation number;
s7, counting the total number of the single chromosomes, and respectively counting the number of the single chromosomes with each number to obtain the number abnormality prompt of the chromosomes with each number;
s8, carrying out centromere detection on the single chromosome by adopting a feature point detection model based on deep learning to obtain the centromere position of the single chromosome;
and S9, arranging the single chromosomes according to the classification number, the polarity number and the structural variation number of the single chromosome image and the positions of the centromere to obtain a standard karyotype chart, and marking the chromosome numbers with abnormal quantity and the single chromosomes with abnormal structure on the karyotype chart to finish the karyotype analysis of the chromosomes.
2. The deep learning based karyotype analysis method according to claim 1, wherein the step of preprocessing the chromosome image in the step of S2 includes: carrying out style conversion on the acquired chromosome original image by adopting an image conversion model based on a countermeasure generation network to obtain a chromosome image with clear stripes; and (3) eliminating the background of the chromosome image with clear stripes by adopting a semantic segmentation model based on a neural network to obtain a clear image.
3. The deep learning-based chromosome karyotyping analysis method according to claim 2, wherein the step of background elimination for the clearly striped chromosome image using the neural network-based semantic segmentation model comprises: the semantic segmentation model takes a chromosome region in an image as a foreground semantic class and a background region in the image as a background semantic class, predicts the semantic class to which each pixel point of an input image belongs, generates a mask according to the semantic class of each pixel point to obtain a foreground semantic mask and a background semantic mask, deletes the background semantic mask, and retains the foreground semantic mask to obtain a clear image with background noise eliminated.
4. The deep learning based karyotype analysis method according to claim 3, wherein the step of dividing the foreground semantic class by the number of contours for the minimum chromosome unit cluster in the step of S3 includes:
s3.1, extracting the contours of the foreground semantic mask in the clear image, respectively calculating the minimum peripheral rectangular frames of the extracted n contours, and respectively storing the coordinates of the upper left corners of the n minimum peripheral rectangular frames and the length and width values of the rectangular frames;
s3.2, creating n pure-color background pictures according to the size of the minimum peripheral rectangular frame, and then drawing the extracted n outlines in the corresponding pure-color background pictures one by one to obtain n outline masks;
and S3.3, displaying an ROI area image in the clear image according to the outline mask to obtain a chromosome cluster corresponding to the outline mask, storing the ROI area as a clustering image, and carrying out image size normalization on the clustering image.
5. The deep learning-based karyotype analysis method according to claim 3, wherein the step of performing chromosome segmentation and classification on the clustered images comprises:
constructing a chromosome instance segmentation model based on deep learning, and inputting the clustering image into the chromosome instance segmentation model to obtain a chromosome number and a chromosome mask corresponding to the chromosome number in the clustering image;
and cutting the corresponding clustered image according to the chromosome mask, and marking the chromosome number on the cut image to obtain a single chromosome image with a classification number.
6. The deep learning based karyotyping method according to claim 5, further comprising the steps of: constructing a training data set for training the chromosome instance segmentation model; the method comprises the following specific steps:
collecting the clinically existing karyotype chart, and cutting the karyotype chart according to regions to obtain each numbered chromosome image and a corresponding label;
carrying out feature matching on each numbered chromosome image on a pre-stored clear image by using an image registration mode to obtain a chromosome mask matched with the numbered chromosome image in the clear image, and assigning the number of the chromosome image to the matched chromosome in the clear image;
establishing a background picture of a background semantic mask for each chromosome mask which is subjected to feature matching by taking the size of the clear image as a target size;
drawing each chromosome mask with completed feature matching on the background picture in the form of filled inner contour;
judging whether the chromosome mask completing the feature matching on the clear image has a coincidence region, if so, merging the chromosome masks with the coincidence region to generate a clustering chart mask containing a plurality of chromosomes, and simultaneously storing the chromosome masks and the serial numbers thereof on the clustering chart mask;
taking the chromosome mask and the clustering graph mask as regions of interest, and displaying ROI area images of the chromosome mask subjected to feature matching;
extracting a minimum rectangular area according to the chromosome mask and the clustering chart mask, cutting the minimum rectangular area of the image after the chromosome mask, the clustering chart mask and the ROI corresponding to the chromosome mask and the clustering chart mask are displayed, and then carrying out image size normalization processing on the cut image to obtain a single chromosome image and a clustering chart containing a plurality of chromosomes;
and writing the chromosome labels and the chromosome mask outline information corresponding to the single chromosome image and the clustering graph containing the plurality of chromosomes into an annotation file, making example segmentation annotation information, obtaining a large number of single and plurality of chromosome images and corresponding annotation files, and forming a training data set for training the chromosome example segmentation model.
7. The deep learning-based karyotyping method according to claim 5, wherein in the step S5, the step of tracing back the single chromosome image with the classification number to the clearly-striped chromosome image for visualization includes:
calculating the minimum peripheral rectangle of the chromosome mask of the single chromosome image with the classification number to obtain a minimum peripheral rectangle list, and the length l and the width w of the image when the clustering graph where the single chromosome is located is clustered; carrying out image size normalization processing on the single chromosome images according to the length l and the width w of each single chromosome image cluster;
extracting the contour of the chromosome mask in the single chromosome image after the image size adjustment is finished, then mapping the starting point of the minimum peripheral matrix of the clustering graph to the clear image as the starting point of the contour, and mapping the contour to the clear image;
and after traversing the chromosome mask in all the single chromosome images of the basic unit, backtracking the single chromosome images with the classification numbers which are segmented and classified into the clear image for visual display.
8. The deep learning based karyotyping analysis method according to claim 1, wherein in the step S6, the constructing of the deep learning based classification model includes the steps of:
collecting the clinically existing karyotype chart, and cutting the karyotype chart according to regions to obtain each numbered chromosome image and a corresponding label;
respectively manufacturing polarity labels for each chromosome image, randomly selecting a certain number of chromosome images for vertical turning operation, marking the polarity labels of the partial images as the direction is downward, and marking the polarity labels of the rest chromosome images as the direction is upward;
making a structural abnormality label for the chromosome image marked with the polarity label, marking the structural abnormality label of the chromosome image with structural abnormality as abnormal, and marking the structural abnormality label of the chromosome image without structural abnormality as normal;
writing chromosome numbers, polarity labels and structural abnormality labels corresponding to the chromosome images into an annotation file, making single chromosome image annotation information, obtaining a large number of single chromosome images and corresponding annotation files thereof, and forming a training data set for training the chromosome classification model;
and inputting the training data set into the chromosome classification model for training, and adjusting parameters of the chromosome classification model according to the classification result output by the chromosome classification model to complete the construction of the chromosome classification model based on deep learning.
9. The deep-learning-based karyotype analysis method according to claim 1, wherein in the step S8, the construction of the deep-learning-based feature point detection model includes the steps of:
acquiring a clinically existing karyotype graph, and cutting the karyotype graph according to regions to obtain chromosome images with various numbers;
making a centromere position label for each numbered chromosome image, and marking coordinates (X, Y) of the centromere position of the chromosome in the chromosome image;
writing the corresponding chromosome number and the position of the centromere into a label file, making single chromosome image label information, obtaining a large number of single chromosome images and corresponding label files thereof, and forming a training data set for training the chromosome feature point detection model;
and inputting the training data set into the chromosome feature point detection model, and adjusting parameters in the chromosome feature point detection model according to the detection result to complete the construction of the chromosome feature point detection model.
10. A deep learning based karyotype analysis system, comprising:
the image acquisition module is used for acquiring an original chromosome image;
the image preprocessing module is used for preprocessing the original image of the chromosome to obtain a clear image;
the clustering module is used for segmenting the foreground semantic class into a minimum chromosome unit cluster according to the outline number according to the semantic class information of the clear image to obtain a plurality of clustered images containing a single chromosome or a plurality of chromosomes;
the segmentation and classification module is used for carrying out chromosome segmentation and classification on the clustering images based on a chromosome instance segmentation model of deep learning to obtain single chromosome images with classification numbers; the single chromosome image processing system is used for carrying out polarity prediction and structural variation classification on the single chromosome image by adopting a classification model based on deep learning to obtain a single chromosome image with a polarity number and a structural variation number;
the system comprises a centromere detection module, a characteristic point detection module and a characteristic point detection module, wherein the centromere detection module is used for carrying out centromere detection on a single chromosome by adopting a characteristic point detection model based on deep learning to obtain the centromere position of the single chromosome;
and the visualization module is used for backtracking the single chromosome image with the classification number to the chromosome image with clear stripes, arranging the single chromosomes according to the classification number, the polarity number and the structural variation number of the single chromosome image and the positions of the centromere to obtain a standard karyotype chart, marking the chromosome number with abnormal quantity and the single chromosome with abnormal structure on the karyotype chart, and visually displaying the karyotype chart.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111547786.5A CN114219786B (en) | 2021-12-16 | 2021-12-16 | Chromosome karyotype analysis method and system based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111547786.5A CN114219786B (en) | 2021-12-16 | 2021-12-16 | Chromosome karyotype analysis method and system based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114219786A CN114219786A (en) | 2022-03-22 |
| CN114219786B true CN114219786B (en) | 2022-07-05 |
Family
ID=80703355
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111547786.5A Active CN114219786B (en) | 2021-12-16 | 2021-12-16 | Chromosome karyotype analysis method and system based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114219786B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115082474B (en) * | 2022-08-22 | 2023-03-03 | 湖南自兴智慧医疗科技有限公司 | Chromosome segmentation method and device based on homologous same-class chromosome information |
| CN116977338B (en) * | 2023-09-25 | 2023-12-12 | 湖南自兴智慧医疗科技有限公司 | Chromosome case-level abnormality prompting system based on visual semantic association |
| CN118115996A (en) * | 2024-04-30 | 2024-05-31 | 四川大学华西第二医院 | Distributed chromosome karyotype data labeling method based on artificial intelligence algorithm assistance |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109285174A (en) * | 2017-07-19 | 2019-01-29 | 塔塔咨询服务公司 | Based on the segmentation of the chromosome of crowdsourcing and deep learning and karyotyping |
| WO2020168511A1 (en) * | 2019-02-21 | 2020-08-27 | 中国医药大学附设医院 | Chromosome abnormality detection model, chromosome abnormality detection system, and chromosome abnormality detection method |
| CN112487930A (en) * | 2020-11-26 | 2021-03-12 | 华南师范大学 | Method, system and storage medium for automated karyotype analysis |
-
2021
- 2021-12-16 CN CN202111547786.5A patent/CN114219786B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109285174A (en) * | 2017-07-19 | 2019-01-29 | 塔塔咨询服务公司 | Based on the segmentation of the chromosome of crowdsourcing and deep learning and karyotyping |
| WO2020168511A1 (en) * | 2019-02-21 | 2020-08-27 | 中国医药大学附设医院 | Chromosome abnormality detection model, chromosome abnormality detection system, and chromosome abnormality detection method |
| CN112487930A (en) * | 2020-11-26 | 2021-03-12 | 华南师范大学 | Method, system and storage medium for automated karyotype analysis |
Non-Patent Citations (3)
| Title |
|---|
| 基于粒度空间的最优聚类指标研究;唐旭清等;《系统工程理论与实践》;20180325(第03期);全文 * |
| 大兴安岭产胎生蜥蜴的染色体组型研究;朱丹等;《中国农学通报》;20110905(第22期);全文 * |
| 小麦族仲彬草属植物的研究进展;张利等;《四川农业大学学报》;20040930(第03期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114219786A (en) | 2022-03-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114219786B (en) | Chromosome karyotype analysis method and system based on deep learning | |
| CN112288706B (en) | Automatic chromosome karyotype analysis and abnormality detection method | |
| AU2018102232A4 (en) | Bone marrow cell marking method and system | |
| CN109523520B (en) | Chromosome automatic counting method based on deep learning | |
| US20190156202A1 (en) | Model construction in a neural network for object detection | |
| CN107430771B (en) | System and method for image segmentation | |
| CN105989347B (en) | Objective item intelligently reading method and system | |
| CN103518224B (en) | Method for analysing microbial growth | |
| CN109285174A (en) | Based on the segmentation of the chromosome of crowdsourcing and deep learning and karyotyping | |
| CN102651128B (en) | Image set partitioning method based on sampling | |
| CN108074243A (en) | A kind of cellular localization method and cell segmentation method | |
| CN111209864B (en) | Power equipment target identification method | |
| CN110728666B (en) | Typing method and system for chronic nasosinusitis based on digital pathological slide | |
| CN110796661B (en) | Fungal microscopic image segmentation detection method and system based on convolutional neural network | |
| CN110992384B (en) | Semi-automatic image data labeling method, electronic device and storage medium | |
| CN114241326B (en) | Progressive intelligent production method and system for ground feature elements of remote sensing images | |
| CN107977624A (en) | A kind of semantic segmentation method, apparatus and system | |
| CN110414317B (en) | Full-automatic leukocyte classification counting method based on capsule network | |
| CN110232337B (en) | Chinese character image stroke extraction method and system based on full convolution neural network | |
| CN111898677A (en) | Plankton automatic detection method based on deep learning | |
| CN112614104B (en) | Segmentation counting method and terminal for red blood cell overlapping | |
| CN114519799A (en) | Real-time detection method and system for multi-feature seat state | |
| CN114170218B (en) | Chromosome image instance label generation method and system | |
| CN110348404A (en) | A kind of road landscape visual evaluation analysis method | |
| CN116863099B (en) | Building automatic modeling method and system based on point cloud data |
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 |