CN110610757B - Chromosome image processing method based on combination of artificial intelligence and graphics - Google Patents
Chromosome image processing method based on combination of artificial intelligence and graphics Download PDFInfo
- Publication number
- CN110610757B CN110610757B CN201910903924.5A CN201910903924A CN110610757B CN 110610757 B CN110610757 B CN 110610757B CN 201910903924 A CN201910903924 A CN 201910903924A CN 110610757 B CN110610757 B CN 110610757B
- Authority
- CN
- China
- Prior art keywords
- chromosome
- image
- processing
- segmented
- segmentation
- 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
Images
Classifications
-
- 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
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H30/00—ICT specially adapted for the handling or processing of medical images
- G16H30/40—ICT specially adapted for the handling or processing of medical images for processing medical images, e.g. editing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
Abstract
The invention relates to a chromosome processing technology, and discloses a chromosome image processing method based on the combination of artificial intelligence and graphics, which can be used for rapidly and accurately processing the segmentation and classification identification of a chromosome image, improving the working efficiency, reducing the labor cost and ensuring the processing quality. The method comprises the following steps: a. training a segmentation model according to a chromosome original image data set; b. when the chromosome image is processed, the original medical image of the chromosome to be processed is segmented by utilizing a segmentation model; c. performing secondary to multi-time segmentation on the chromosome cluster image which is not completely segmented in the step b to obtain a single chromosome; d. training a classification model according to the extended single chromosome image data set processed by the graphical algorithm; e. classifying the single chromosome after segmentation and straightening by using a classification model; f. and executing a typesetting function suitable for the medical report according to the classification result, and outputting the typesetted chromosome image.
Description
Technical Field
The invention relates to a chromosome processing technology, in particular to a chromosome image processing method based on combination of artificial intelligence and graphics.
Background
The chromosome image is almost the only means for ensuring whether the genes of the fetus are normal, and various gene problems and the hidden danger of acquired diseases of the fetus can be found in time in the pregnancy. In the context of genomics-related research, this means of analyzing fetal conditions from a genetic perspective is of great instructive significance, and the difficulty of making images thereof is also conceivable. It is understood that in the period of purely manual chromosome image analysis, each image (i.e. each sample) requires more than 40 minutes of processing time, and the final result of each fetus requires counting the ratio of normal to abnormal from dozens of samples (changing the number of samples requiring additional counting according to the number of normal samples), which is not very efficient.
At present, the imaging process of the chromosome image is converted into the operation of an instrument, and the fetus cell is manually sliced and then placed under the instrument for imaging, so that the process of manually observing a microscope for each sample is reduced, and the image processing auxiliary function of the basis can be realized on a computer, so that the working efficiency is greatly improved. However, under the population pressure of the present day, such a speed has become more difficult to meet the medical requirement, because the imaged pictures still need to be manually processed, and the labor cost is high.
Under the current explosive background of artificial intelligence, if can be applied to the chromosome image processing of medical field with artificial intelligence, replace artificial process with machine operation as far as possible, then can greatly promote work efficiency.
The deep learning is widely applied to artificial intelligence, and is concerned by all fields in recent years, so that the adoption of the deep learning method for a computer to solve some intelligentization problems is a mainstream direction for the development of the artificial intelligence. The deep learning can be combined in various fields, and has good application prospect in computer graphics, so the invention uses the deep learning as a main tool for processing computer graphics problems, and the combination of the deep learning and the computer graphics is believed to have a great promoting effect on the research of related fields.
The difficult problem in the chromosome image processing process is the segmentation and identification work of the chromosome image, and although a lot of researches are made on the problem at home and abroad at present, the problem still has several defects due to practical applicability:
(1) how to improve the existing algorithm system to achieve the purpose of labor cost reduction: at present, no matter the algorithm architecture is based on traditional graphics or deep learning, a large amount of manpower is needed for participation. This is because current research work is done under a single condition, either for a single problem or for a single data set. Such research conditions determine processing systems that cannot be used for packaging in a short time;
(2) how to resolve clustering or masking in the segmentation process: in the current traditional algorithm research, the segmentation documents are few, and the difficulty is sufficient. In deep learning, the method of U-net is worth borrowing, but the achievement of the chromosome image is also very little. As for the clustering and masking cases, some corresponding processing means need to be researched certainly;
(3) how to design the evaluation standard after the automatic processing of the chromosome image is as follows: at present, the requirement of artificial intelligence processing in the market is full-automatic, automatic processing in the whole process can meet many special conditions, a small amount of manual assistance is needed, and then how to verify that a certain image needs manual assistance needs to be defined by a more innovative theory.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the chromosome image processing method based on the combination of artificial intelligence and graphics is provided, the segmentation and classification recognition of the chromosome image are rapidly and accurately processed, the working efficiency is improved, the labor cost is reduced, and the processing quality is ensured.
The technical scheme adopted by the invention for solving the technical problems is as follows:
the chromosome image processing method based on the combination of artificial intelligence and graphics comprises the following steps:
a. training a segmentation model according to a chromosome original image data set;
b. when the chromosome image is processed, the original medical image of the chromosome to be processed is segmented by utilizing a segmentation model;
c. performing secondary to multi-time segmentation on the chromosome cluster image which is not completely segmented in the step b to obtain a single chromosome;
d. training a classification model according to the image data set of the straightened single chromosome processed by the graphical algorithm;
e. classifying the single chromosome after segmentation and straightening by using a classification model;
f. and executing the typesetting function suitable for the medical report according to the classification result, and outputting the typesetted chromosome image.
As a further optimization, in step a, before training the segmentation model, firstly, a chromosome original image dataset is created, specifically including: marking the obtained original chromosome image by a data marking tool, marking a bounding box for each visible chromosome on the image, further generating a binary mask, outputting JSON format data which is conveniently processed by a convolution network, and then performing rotation, displacement and scaling transformation on each image including the mask image.
As a further optimization, the number of chromosomes in the acquired original images of chromosomes is not limited.
In the step a, a U-net model framework based on deep learning is adopted for the segmentation model, a manufactured chromosome original image data set is used for dividing a training set and a verification set, the training set is adopted for training the model, the verification set is adopted for performing verification test on the trained model, and finally the segmentation model meeting the requirements is obtained.
As a further optimization, in the step b, after the original medical image of the chromosome to be processed is segmented by using the segmentation model, a binary mask of the segmented chromosome image is obtained, and the segmented chromosome image is extracted on the original medical image of the chromosome to be processed according to the binary mask; the segmented chromosome images comprise most of single chromosomes and a few of incompletely segmented chromosome cluster images.
As a further optimization, in step c, if the chromosome cluster image which is not completely segmented in step b is still unsuccessfully segmented for a certain number of times, a manual assistance interface is called to give a prompt to be subjected to manual assistance processing, and an operator finishes the segmentation task through checking conditions or directly drawing lines and points to operate an assistance program.
As a further optimization, the chromosome cluster images which are not completely segmented are preprocessed in a gray scale equalization mode before the artificial auxiliary interface is called.
As a further optimization, in step d, the graphical algorithm processing includes:
d1. the pretreatment process comprises the following steps: carrying out gray level equalization processing on the image of the single chromosome;
d2. midline extraction and head-to-tail cutting: in the acquisition of the middle axis segment, a Delaunay triangle method is adopted to take out the middle axis of the chromosome for pixel correlation processing;
d3. direction matching and pixel association: for each pixel point on the middle axis, taking a normal of the curve of the associated middle axis, wherein each effective pixel point passing along the normal is a pixel associated with the pixel point of the middle axis; in direction matching, approximate normals adapting to 16 different directions are determined through pixel distribution of 3 × 3 neighborhoods of the centering axis pixels; in addition, secondary point supplementing operation is carried out on the pixel points at the bending position of the middle shaft, namely, one point is selected from the front and the back of the bending trend of the point, and the pixels are related according to the original slope and then are sequentially added into the queue to be straightened;
d4. angle correction and head-to-tail attachment: firstly, the average length of a queue to be straightened needs to be measured so as to judge whether a certain queue needs to be corrected; if correction is needed, the position of the central axis in the original image is retrieved aiming at the queue, the current angle information of the central axis is discarded, and the pixels are re-associated by using 16 defined angles until the queue with the most suitable length is found; if the angle information of the new queue needs to be supplemented with points, point supplementing processing is carried out, and then all newly generated queues are uniformly replaced into the extended image;
after the angle correction process is completed, the head and tail cutting parts need to be integrated, and a boundary expansion method is used for processing: providing a cross-shaped 1-pixel expansion template for each effective pixel point from a cutting line, attaching the obtained pixels of the cutting part to the head end and the tail end of the associated part of the straightened middle shaft after expansion, using the newly obtained pixels as the starting points of expansion in the next round, and repeating the steps until no new pixel point is obtained;
and (d) processing the steps d1-d4 to obtain a single chromosome image after straightening.
As a further optimization, in step d, the classification model adopts a single-input 24-class-output convolution network structure to learn each single chromosome image in the extended single chromosome image data set.
As a further optimization, in step f, the outputted typeset chromosome image is a single chromosome image which is segmented from the chromosome original medical image and is not straightened.
The invention has the beneficial effects that:
(1) The invention makes automatic improvement on the graphical algorithm, so that the preprocessing processes such as thresholding processing, gray level equalization processing and the like can be completely separated from manual operation. The method avoids errors caused by personal characteristics, and can bring improvement meeting human vision to the image. Meanwhile, the invention also supports manual readjustment to meet the requirements of manual error correction and specificity, and the final processing quality of the method is not influenced.
(2) The invention introduces an artificial intelligence algorithm which is effective for image segmentation and classification, namely deep learning, can completely take the banding characteristics and the pixel information of the chromosome into consideration, has the accuracy rate of more than 95 percent, and tests show that even if an original chromosome image is not processed, the result obtained by a network is still very accurate when human eyes can not distinguish at all;
(3) The invention adopts a non-end-to-end deep learning network model for image processing for the first time, namely, a graphical algorithm chromosome straightening operation is added between segmentation and classification work, aiming at increasing the accuracy of a classification network. The operation eliminates the influence of chromosome morphology on classification work, and the operation proves that the operation can achieve good effect.
(4) The invention has high integral automation degree, does not need excessive manual participation, and can generate a final result by inputting the original image, thereby greatly improving the working efficiency and reducing the labor cost.
Drawings
Fig. 1 is a flow chart of a chromosome image processing method based on artificial intelligence and graphics combined in the invention.
Detailed Description
The invention aims to provide a chromosome image processing method based on combination of artificial intelligence and graphics, which can be used for rapidly and accurately processing the segmentation and classification identification of chromosome images, improving the working efficiency, reducing the labor cost and ensuring the processing quality.
Aiming at the defects of the research results of the segmentation and identification work of the chromosome image at home and abroad, the invention solves the problems one by one through corresponding means:
(1) aiming at the problems of how to improve an algorithm system and reducing labor cost: the invention adopts clinical original images to train a deep learning model, or the original images are preprocessed by a graphics method and then are sent to a network for learning. Such operations make the trained model applicable to most chromosome images, since the quality of the original images varies widely and a training set with a certain characteristic is not carefully chosen. In addition, the solution of a large amount of manpower problems is mainly embodied on data labeling, the method adopted by the scheme is to transform an original chromosome image, and the image is changeable through operations such as rotation, displacement, scaling and the like, so that the capacity of data is greatly expanded, and a network structure which can achieve high accuracy by only needing a small amount of data in other fields is used for reference, and the manpower cost can be reduced on the premise of ensuring the quality.
(2) To solve the problem of clustering or masking in the segmentation process: since clustering and masking phenomena are quite common in practice, which is highly related to the level of a slide operator, a small amount of the phenomenon does not affect the diagnosis of a doctor, and an overlapped part is directly selected in a sleeving manner in actual operation. Therefore, the phenomenon needs to be processed independently, if the part which cannot be identified appears in the result after the primary segmentation, the part is segmented twice or even for many times, the output result is ensured to be a single chromosome, and meanwhile, the overlapping selection of the overlapping area is met by utilizing the segmentation characteristic of the U-net network.
(3) Aiming at the problem of how to design the judgment standard after the automatic processing of the chromosome image: in image segmentation, if clusters or masking phenomena which cannot be successfully separated all the time exist, manual assistance is required. In the last classification task, the scheme utilizes the confidence coefficient output by deep learning, if the confidence coefficient is within the set threshold range, manual intervention is not needed, the threshold can be adjusted in the system, and the system can support reprocessing (automatic or manual) of each image. In addition, the system will output indicative log information (e.g., one less chromosome is detected, etc.) for the operator to view at any time to determine whether human intervention is required, depending on the situation during the treatment process.
In a specific implementation, as shown in fig. 1, the method for processing a chromosome image based on artificial intelligence and graphics in the present invention includes the following implementation steps:
1. training a segmentation model according to a chromosome original image data set;
in this step, before training the segmentation model, the chromosome original image dataset needs to be prepared: the obtained chromosome original image is labeled through a data labeling tool (such as label me), the number of chromosomes in one image is not required to be limited, the randomness of the sample is ensured, a surrounding frame is labeled for each visible chromosome on the image, a binary mask is further generated, and JSON format data which is convenient to process through a network is output. Each image, including its mask image, is then transformed by rotation, displacement, scaling, etc., which exponentially increases its data size.
The training segmentation model mainly adopts a U-net model framework based on deep learning to scan a chromosome image, analyzes each pixel of the image through a full-convolution network structure, solves the pixel positioning problem at a shallower level, and is used for pixel classification (whether the pixel belongs to a mask part or not). The final network will output the segmentation of the image and thus generate the required mask.
2. When the chromosome image is processed, the original medical image of the chromosome to be processed is segmented by utilizing a segmentation model;
in the step, after a chromosome original medical image to be processed is segmented by a segmentation model, a binary mask of the segmented chromosome image is obtained, and the segmented chromosome image is extracted from the chromosome original medical image to be processed according to the binary mask; the segmented chromosome images comprise most of single chromosomes and a few of incompletely segmented chromosome cluster images.
3. Performing secondary to multi-time segmentation on the chromosome cluster image which is not completely segmented in the step 2 to obtain a single chromosome;
in the step, a single chromosome image can be obtained by secondary segmentation or tertiary segmentation aiming at a small part of incompletely segmented chromosome cluster images, manual intervention is required if the segmentation frequency reaches a preset threshold value, a prompt for manual auxiliary treatment can be given by calling a manual auxiliary interface, and an operator finishes a segmentation task by checking conditions or directly marking and dotting operation auxiliary programs. However, in order to facilitate human observation, the chromosome cluster image which is not completely segmented is preprocessed by gray level equalization before the artificial auxiliary interface is called.
4. Training a classification model according to the extended single chromosome image data set processed by the graphical algorithm;
in this step, the graphics algorithm processing mainly includes:
(1) The pretreatment process comprises the following steps: gray level equalization processing is carried out on the image of the single chromosome so as to highlight the characteristics of the pixels and the correlation among the pixels, so that classification processing is conveniently carried out in the step 5, and the visualization of final result output is also met;
(2) Midline extraction and head-to-tail cutting: and in the acquisition of the middle axis segment, taking out the middle axis of the chromosome by adopting a Delaunay triangle method for pixel association processing. Because the obtained middle axis segment can not be associated with all pixel points of the chromosome, and the middle axis is hopefully avoided from being expanded, the chromosome parts matched with the length of the middle axis are cut, namely, the parts are not processed at first, and the parts are jointed after the middle axis part is straightened.
(3) Direction matching and pixel association: for each pixel point on the medial axis, the normal of the pixel point about the medial axis curve is taken, and each effective pixel point passing along the normal is the pixel associated with the medial axis pixel point. Regarding the matching of directions, a similar inverse process is designed by taking the theory of screen drawing lines in computer graphics as reference, and approximate normals adaptive to 16 different directions are determined by pixel distribution in the neighborhood of the central axis pixel 3 x 3. In addition, pixel loss due to the fact that the pixel density at the outer curve of the curve is less than the pixel density on the central axis needs to be considered. Due to the relationship between pixels and lines in the image, it can be determined that a relatively large pixel loss is caused only when the normal slope at a certain point of the central axis curvature is 1 or-1. Therefore, secondary point supplementing operation is needed for the middle axis pixel point, namely, one point is selected from the front and the back of the middle axis pixel point in the bending trend of the point, the pixels are related according to the original slope and then are added into the queue to be straightened in sequence, and the number of lost pixels can be ensured to be less.
(4) Angle correction and head-to-tail attachment: the average length of the queue to be straightened needs to be measured first and is used as a decision whether correction is needed for a certain queue. If processing is required, the central axis position in the original image is retrieved for the queue, and the current angle information is discarded, and the pixels are re-associated with the previously defined 16 angles until a queue with the most suitable length is found (if all queues fail to satisfy the condition, the point is discarded as an invalid central axis point). And if the angle information of the new queue needs point filling, performing point filling processing, and then uniformly replacing all newly generated queues into the extended image. After the angle correction process is completed, the head and tail cutting parts need to be integrated, and the boundary expansion method is used for processing. And (3) providing a cross-shaped 1-pixel expansion template for each effective pixel point from the cutting line, attaching the obtained pixels of the cutting part to the head and tail ends of the associated part of the straightened middle shaft after expansion, using the newly obtained pixels as the starting points of expansion in the next round, and repeating the steps until new pixel points are not obtained.
By applying the graphical algorithm, a single chromosome image after straightening can be obtained. Of course the original single image still needs to be retained, this is the visual image that is the final output report.
When the classification model is trained, a network structure of single input-24 types of output is designed, and each single chromosome image in the data set is learned. And the data set is made as an intermediate result of the preorder process and can be directly applied.
5. Classifying the single chromosome after segmentation and straightening by using a classification model;
in the step, after straightening treatment is carried out on the single chromosome which is segmented out, the single chromosome can be classified by applying a trained classification model.
6. And executing the typesetting function suitable for the medical report according to the classification result, and outputting the typesetted chromosome image.
In this step, the outputted typeset chromosome image is a single chromosome image which is segmented from the chromosome original medical image and is not straightened.
Claims (8)
1. A chromosome image processing method based on the combination of artificial intelligence and graphics is characterized in that,
the method comprises the following steps:
a. training a segmentation model according to a chromosome original image data set;
b. when the chromosome image is processed, the original medical image of the chromosome to be processed is segmented by utilizing a segmentation model;
c. b, performing secondary to multiple segmentation on the chromosome cluster image which is not completely segmented in the step b to obtain a single chromosome;
d. training a classification model according to the extended single chromosome image data set processed by the graphical algorithm;
e. classifying the single chromosome after segmentation and straightening by using a classification model;
f. executing a typesetting function suitable for the medical report according to the classification result, and outputting the typesetted chromosome image;
in step d, the graphical algorithm processing includes:
d1. the pretreatment process comprises the following steps: carrying out gray level equalization processing on the image of the single chromosome;
d2. midline extraction and head-to-tail cutting: on the acquisition of the middle axis segment, taking out the middle axis of the chromosome by adopting a Delaunay triangle method for pixel correlation processing; cutting the chromosome part matched to the outside of the length of the middle shaft;
d3. direction matching and pixel association: for each pixel point on the middle axis, taking a normal line of a curve of the associated middle axis, wherein each effective pixel point passing along the normal line is a pixel associated with the pixel point of the middle axis; in direction matching, approximate normals adapting to 16 different directions are determined through pixel distribution of 3 × 3 neighborhoods of the centering axis pixels; in addition, secondary point supplementing operation is carried out on the pixel points at the bending position of the middle shaft, namely, one point is selected from the front and the back of the bending trend of the point, and the pixels are related according to the original slope and then are sequentially added into the queue to be straightened;
d4. angle correction and head-to-tail attachment: firstly, the average length of a queue to be straightened needs to be measured so as to judge whether a certain queue needs to be corrected; if correction is needed, the central axis position in the original image is retrieved aiming at the queue, the current angle information is abandoned, and the pixels are re-associated by using 16 defined angles until a queue with the most proper length is found; if the angle information of the new queue needs to be supplemented with points, point supplementing processing is carried out, and then all newly generated queues are uniformly replaced into the extended image;
after the angle correction process is completed, the head and tail cutting parts need to be integrated, and a boundary expansion method is used for processing: providing a cross-shaped 1-pixel expansion template for each effective pixel point from a cutting line, attaching the obtained pixels of the cutting part to the head end and the tail end of the associated part of the straightened middle shaft after expansion, using the newly obtained pixels as the starting points of expansion in the next round, and repeating the steps until no new pixel point is obtained;
and d1-d4, obtaining a straightened single chromosome image.
2. The method for processing chromosome image based on artificial intelligence and graphics in combination as claimed in claim 1,
the method is characterized in that in the step a, before training a segmentation model, firstly, a chromosome original image data set is manufactured, and the method specifically comprises the following steps: marking the obtained original chromosome image by a data marking tool, marking a surrounding frame for each visible chromosome on the image, further generating a binary mask, outputting JSON format data which is convenient to process by a rolling network, and then carrying out rotation, displacement and scaling transformation on each image including the mask image.
3. The method for processing chromosome image based on artificial intelligence and graphics combination as claimed in claim 2,
the method is characterized in that in the step a, a U-net model framework based on deep learning is adopted for the segmentation model, a training set and a verification set are divided by utilizing a manufactured chromosome original image data set, the model is trained by adopting the training set, the trained model is subjected to verification test by adopting the verification set, and finally the segmentation model meeting the requirements is obtained.
4. The method for processing chromosome image based on artificial intelligence and graphics in combination as claimed in claim 1,
the method is characterized in that in the step b, after a chromosome original medical image to be processed is segmented by a segmentation model, a binary mask of the segmented chromosome image is obtained, and the segmented chromosome image is extracted from the chromosome original medical image to be processed according to the binary mask; the segmented chromosome images comprise most single chromosomes and a few incompletely segmented chromosome cluster images.
5. The method for processing chromosome image based on artificial intelligence and graphics combination as claimed in claim 1,
the method is characterized in that in the step c, if the chromosome cluster images which are not completely segmented in the step b are still not successfully segmented for a certain number of times, a manual auxiliary interface is called to give a prompt to be subjected to manual auxiliary processing, and an operator finishes the segmentation task through a checking condition or a direct scribing and dotting operation auxiliary program.
6. The method for processing chromosome image based on artificial intelligence and graphics combination as claimed in claim 5,
the method is characterized in that the gray level equalization pretreatment is carried out on the incompletely segmented chromosome cluster image before the manual auxiliary interface is called.
7. The method for processing chromosome image based on artificial intelligence and graphics combination as claimed in claim 1,
the method is characterized in that in the step d, the classification model adopts a convolution network structure with single input and 24 types of output to learn each single chromosome image in the straightening single chromosome image data set.
8. The method for processing chromosome images based on combination of artificial intelligence and graphics according to any one of claims 1 to 7, wherein in step f, the outputted typeset chromosome images are single chromosome images which are segmented from chromosome original medical images and are not straightened.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910903924.5A CN110610757B (en) | 2019-09-24 | 2019-09-24 | Chromosome image processing method based on combination of artificial intelligence and graphics |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910903924.5A CN110610757B (en) | 2019-09-24 | 2019-09-24 | Chromosome image processing method based on combination of artificial intelligence and graphics |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110610757A CN110610757A (en) | 2019-12-24 |
| CN110610757B true CN110610757B (en) | 2023-02-03 |
Family
ID=68892010
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910903924.5A Active CN110610757B (en) | 2019-09-24 | 2019-09-24 | Chromosome image processing method based on combination of artificial intelligence and graphics |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110610757B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111223084A (en) * | 2020-01-07 | 2020-06-02 | 华南师范大学 | Chromosome cutting data processing method, system and storage medium |
| CN111489336B (en) * | 2020-04-07 | 2023-07-25 | 内蒙古工业大学 | Method and device for detecting length of carding cashmere based on pixel calculation |
| CN111986183B (en) * | 2020-08-25 | 2023-07-04 | 中国科学院长春光学精密机械与物理研究所 | Chromosome scattered image automatic segmentation and identification system and device |
| CN112052813B (en) * | 2020-09-15 | 2023-12-19 | 中国人民解放军军事科学院军事医学研究院 | Method and device for identifying translocation between chromosomes, electronic equipment and readable storage medium |
| CN112487930A (en) * | 2020-11-26 | 2021-03-12 | 华南师范大学 | Method, system and storage medium for automated karyotype analysis |
| CN114580487A (en) * | 2020-11-30 | 2022-06-03 | 深圳市瑞图生物技术有限公司 | Chromosome recognition method, device, equipment and storage medium based on deep learning |
| CN113158950B (en) * | 2021-04-30 | 2022-04-05 | 天津深析智能科技发展有限公司 | Automatic segmentation method for overlapped chromosomes |
| CN114973241A (en) * | 2022-04-25 | 2022-08-30 | 上海科莫生医疗科技有限公司 | Chromosome image screening method based on artificial intelligence |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2005215751A (en) * | 2004-01-27 | 2005-08-11 | Jgs:Kk | Method for identifying non-linear data by genetic algorithm, non-linear data recording method, non-linear data recording device, non-linear data reproducing device and program |
| WO2017087018A1 (en) * | 2015-11-18 | 2017-05-26 | University Of Louisville Research Foundation, Inc. | Automated methods for the objective quantification of retinal characteristics by retinal region and diagnosis of retinal pathology |
| CN114331895A (en) * | 2021-12-30 | 2022-04-12 | 电子科技大学 | Monte Carlo rendering image denoising method based on generation countermeasure network |
| CN115272055A (en) * | 2022-05-12 | 2022-11-01 | 上海科莫生医疗科技有限公司 | Chromosome image analysis method based on knowledge representation |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101520890B (en) * | 2008-12-31 | 2011-04-20 | 广东威创视讯科技股份有限公司 | Grey scale characteristic graph-based automatic separation method for conglutinated chromosomes |
| CN101625755B (en) * | 2009-08-06 | 2012-09-05 | 西安电子科技大学 | Image division method based on watershed-quantum evolution clustering algorithm |
| WO2013192355A1 (en) * | 2012-06-19 | 2013-12-27 | Health Discovery Corporation | Computer-assisted karyotyping |
| EP3432198B1 (en) * | 2017-07-19 | 2024-04-17 | Tata Consultancy Services Limited | Crowdsourcing and deep learning based segmenting and karyotyping of chromosomes |
| CN109344874B (en) * | 2018-08-31 | 2020-09-15 | 华侨大学 | Chromosome automatic analysis method and system based on deep learning |
| CN109934828B (en) * | 2019-02-28 | 2023-03-24 | 西交利物浦大学 | Double-chromosome image cutting method based on Compact SegUnet self-learning model |
-
2019
- 2019-09-24 CN CN201910903924.5A patent/CN110610757B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2005215751A (en) * | 2004-01-27 | 2005-08-11 | Jgs:Kk | Method for identifying non-linear data by genetic algorithm, non-linear data recording method, non-linear data recording device, non-linear data reproducing device and program |
| WO2017087018A1 (en) * | 2015-11-18 | 2017-05-26 | University Of Louisville Research Foundation, Inc. | Automated methods for the objective quantification of retinal characteristics by retinal region and diagnosis of retinal pathology |
| CN114331895A (en) * | 2021-12-30 | 2022-04-12 | 电子科技大学 | Monte Carlo rendering image denoising method based on generation countermeasure network |
| CN115272055A (en) * | 2022-05-12 | 2022-11-01 | 上海科莫生医疗科技有限公司 | Chromosome image analysis method based on knowledge representation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110610757A (en) | 2019-12-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110610757B (en) | Chromosome image processing method based on combination of artificial intelligence and graphics | |
| CN110415230B (en) | CT slice image semantic segmentation system and method based on deep learning | |
| CN109636808B (en) | Lung lobe segmentation method based on full convolution neural network | |
| CN110458831B (en) | Scoliosis image processing method based on deep learning | |
| CN109378052B (en) | The preprocess method and system of image labeling | |
| CN110555836A (en) | Automatic identification method and system for standard fetal section in ultrasonic image | |
| CN112215800B (en) | Machine learning-based overlapped chromosome recognition and segmentation method | |
| CN110648310B (en) | Weak supervision casting defect identification method based on attention mechanism | |
| CN115115841B (en) | Shadow spot image processing and analyzing method and system | |
| CN114663383A (en) | Blood cell segmentation and identification method and device, electronic equipment and storage medium | |
| CN109002831A (en) | A kind of breast density classification method, system and device based on convolutional neural networks | |
| CN115641323A (en) | Method and device for automatically labeling medical images | |
| CN114240883B (en) | Chromosome image processing method and system | |
| CN116468655A (en) | Brain development atlas and image processing system based on fetal magnetic resonance imaging | |
| CN113793385A (en) | Method and device for positioning fish head and fish tail | |
| CN113393456A (en) | Automatic quality control method of early pregnancy fetus standard section based on multiple tasks | |
| CN115641335B (en) | Embryo abnormity multi-cascade intelligent comprehensive analysis system based on time difference incubator | |
| Rabatel et al. | Grape berry calibration by computer vision using elliptical model fitting | |
| CN111260619A (en) | Tongue body automatic segmentation method based on U-net model | |
| CN116229236A (en) | Bacillus tuberculosis detection method based on improved YOLO v5 model | |
| CN115272055A (en) | Chromosome image analysis method based on knowledge representation | |
| CN115188413A (en) | Chromosome karyotype analysis module | |
| CN114693693A (en) | Fetal corpus callosum ultrasonic image segmentation method based on convolutional neural network | |
| CN109509189B (en) | Abdominal muscle labeling method and labeling device based on multiple sub-region templates | |
| CN113838008A (en) | Abnormal cell detection method based on attention-drawing mechanism |
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 |