CN111951266A - Artificial intelligence recognition analysis method for chromosome aberration - Google Patents
Artificial intelligence recognition analysis method for chromosome aberration Download PDFInfo
- Publication number
- CN111951266A CN111951266A CN202010906567.0A CN202010906567A CN111951266A CN 111951266 A CN111951266 A CN 111951266A CN 202010906567 A CN202010906567 A CN 202010906567A CN 111951266 A CN111951266 A CN 111951266A
- Authority
- CN
- China
- Prior art keywords
- image
- chromosome
- model
- artificial intelligence
- analysis
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 208000031404 Chromosome Aberrations Diseases 0.000 title claims abstract description 32
- 231100000005 chromosome aberration Toxicity 0.000 title claims abstract description 32
- 238000004458 analytical method Methods 0.000 title claims abstract description 30
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 23
- 238000000034 method Methods 0.000 claims abstract description 44
- 210000000349 chromosome Anatomy 0.000 claims abstract description 39
- 238000012549 training Methods 0.000 claims abstract description 11
- 230000008569 process Effects 0.000 claims abstract description 10
- 210000002230 centromere Anatomy 0.000 claims description 33
- 238000001514 detection method Methods 0.000 claims description 26
- 238000010606 normalization Methods 0.000 claims description 17
- 238000001914 filtration Methods 0.000 claims description 15
- 238000012545 processing Methods 0.000 claims description 11
- 239000012634 fragment Substances 0.000 claims description 10
- 230000002159 abnormal effect Effects 0.000 claims description 8
- 230000005945 translocation Effects 0.000 claims description 8
- 238000009499 grossing Methods 0.000 claims description 3
- 239000012535 impurity Substances 0.000 claims description 3
- 239000002245 particle Substances 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 210000002500 microbody Anatomy 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 5
- 238000002372 labelling Methods 0.000 description 5
- 210000004027 cell Anatomy 0.000 description 4
- 230000002285 radioactive effect Effects 0.000 description 3
- 208000037088 Chromosome Breakage Diseases 0.000 description 2
- 230000004075 alteration Effects 0.000 description 2
- 230000032823 cell division Effects 0.000 description 2
- 230000002759 chromosomal effect Effects 0.000 description 2
- 230000005886 chromosome breakage Effects 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 238000003745 diagnosis Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000002068 genetic effect Effects 0.000 description 2
- 230000021121 meiosis Effects 0.000 description 2
- 230000031864 metaphase Effects 0.000 description 2
- 230000011278 mitosis Effects 0.000 description 2
- 230000005855 radiation Effects 0.000 description 2
- 208000028571 Occupational disease Diseases 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 210000003793 centrosome Anatomy 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000003889 chemical engineering Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 210000005104 human peripheral blood lymphocyte Anatomy 0.000 description 1
- 238000000338 in vitro Methods 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 210000003470 mitochondria Anatomy 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000000007 visual effect Effects 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
- 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
- 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
- G06T5/00—Image enhancement or restoration
- G06T5/73—Deblurring; Sharpening
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10056—Microscopic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20032—Median filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30072—Microarray; Biochip, DNA array; Well plate
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention provides an artificial intelligence recognition analysis method for chromosome aberration, which comprises the following steps: inputting original historical chromosome distortion form image data, carrying out data marking according to geometric figures and making type marks so as to complete learning training, and storing marked data into a database; inputting a chromosome distortion form target image to be identified into a backbone network backbone model to realize the position positioning of the candidate object; continuously classifying and identifying the objects by positioning the objects, inputting the processed images into a local pre-selection frame model, and classifying and identifying actual conditions of different sizes of the target objects; and outputting a chromosome aberration form automatic analysis result according to the classified and identified data. The method can solve the problem that the accuracy rate of automatically identifying the chromosome distortion is low in the prior art, so that the identification and classification process is full-automatic, fast and high in accuracy. The accuracy rate of the method is not greatly different from that of manual identification.
Description
Technical Field
The invention relates to the field of chromosome aberration identification, in particular to an artificial intelligence identification and analysis method for chromosome aberration.
Background
One of the deep-stained structures that chromosomes can see in dividing cells, usually within a spheroid, is observed only during cell division (mitosis or meiosis). The number of chromosomes in the cell is constant and the morphological characteristics are stable. Chromosomal aberrations are structural aberrations due to chromosome breakage, and reconnection of broken ends, producing new structures of abnormal chromosome length, chromosomal location and genetic makeup in an illegal way, including: centromere, secondary constriction, long arm, short arm, and satellite. In view of the chromosome characteristics, the chromosome aberration morphology induced by radiation is: non-centromere fragments, minute bodies, non-centromere rings, double centromere bodies (three centromere bodies), inverted positions, and mutual translocation. The method is characterized in that a chromosome sample is prepared in vitro by using human peripheral blood lymphocytes for chromosome aberration analysis, and is a preferred detection method for diagnosing radiation damage; the method is widely applied to daily work of nuclear power, chemical engineering, airports in hospitals, railway stations and the like, and has occupational disease protection evaluation and diagnosis indexes of radioactive ray contact personnel. The chromosome aberration examination is characterized by high time consumption and visual analysis, so that a professional skill is required, an experienced doctor can visually identify and analyze metaphase terms (analysis cells) under a microscope, the time consumption is long, the efficiency is low, and the diagnosis is easy to miss; one sample was analyzed for 100 metaphase terms (cells analyzed) per day.
The existing automatic analysis method can be used for identifying part of chromosomes with clear forms, the correct detection rate is not more than 50% at most, and 50% or more of chromosomes are lost; meanwhile, the resolved morphology is mainly double. Other classes, such as fragments, rings, micro-bodies, etc., are less desirable or unrecognizable.
The prior art has the problem that the accuracy of automatic identification of chromosome aberrations is lower than the above case.
Disclosure of Invention
In order to solve the problem of low accuracy of automatic identification of chromosome distortion in the prior art, the artificial intelligent identification and analysis method for chromosome distortion provided by the invention can solve the problem of low accuracy of automatic identification of chromosome distortion, so that the identification and classification process is full-automatic, high in speed and high in accuracy.
In a first aspect, the present application provides a method for artificial intelligence recognition and analysis of chromosome aberration, comprising:
s100: inputting original historical chromosome distortion form image data, carrying out data marking according to geometric figures and making type marks so as to complete learning training, and storing marked data into a database;
s200: inputting a chromosome distortion form target image to be identified into a backbone network backbone model to realize the position positioning of the candidate object;
s300: continuously classifying and identifying the objects by positioning the objects through S200, inputting the images processed through S200 into a local preselection frame model, and classifying and identifying actual conditions of different sizes of the target objects;
s400: and outputting a chromosome aberration form automatic analysis result according to the classified and identified data.
Further, before step S100, normalization processing of the image is performed.
Further, the normalization processing of the image includes the following methods:
removing background noise of the image: removing background noise by Gaussian filtering; then, watermarking noise and impurity noise by low-frequency filtering, and finally, sharpening the image and matching with filtering, selecting one of median filtering or Gaussian filtering to process unclear silk particles;
processing of background noise of image: the image value square is analyzed, corresponding pixels of the value square of the nearly white background are removed, the background can be removed, and if oversaturation and undersaturation exposure exist, the image is considered to be abandoned;
image value histogram normalization: stretching and moving pixel values of the value square map, and normalizing the value square map to be in the same x-axis area range;
image entropy normalization: and calculating the entropy of the image, removing the image with too large or too small entropy, sharpening the image with too small entropy, smoothing the image with too large entropy, and performing normalization processing.
Further, the data marking in S100 marks chromosome aberration morphology by marking required content, and marks double centromeres, triple centromeres, non-centromeres fragments, minute bodies, non-centromeres rings, inversions, and mutual translocations;
marking the concerned object according to the geometric figure, and making a type mark;
the data quantity of the marks can meet the accuracy requirement set by the user, and the quantity and the accuracy are in positive correlation.
Further, the backbone network backbone model comprises Darknet, CSPDarkNet, EfficientNet, ResNet and ResNet Xt models; the local pre-selection frame model comprises an FPN model, a BiFPN model, an ASFF model, an RFB model and an SPP model.
Further, the target detection artificial intelligence method of the model combination of S200-S300 is adopted to perform positioning and classification recognition of the target object, namely predicting the target.
Further, in S300, 3 feature maps with different scales are output according to S200, 3 prediction frames with different sizes are set for the feature maps with each scale, and the prediction frames with 3 sizes from large to small are clustered to obtain 9 prediction frames with different sizes.
Furthermore, a backbone network backbone model and a local pre-selection frame model form a complete target detection model, the images after the database standardization are trained to obtain training parameters, then the concerned objects in the test images are predicted and judged through the training parameters, and positioning and classification marking are carried out.
Further, in S400, each image of the prediction frame is identified, whether the image is a chromosome or not and whether the image is normal or not are determined, which type of abnormal chromosome and double centromeres, triple centromeres, non-centromeres fragments, minute bodies, non-centromeres rings, inversions, and reciprocal translocation are determined, and information of the abnormal chromosome, including position and type information, is output.
Compared with the prior art, the artificial intelligent identification and analysis method for chromosome distortion provided by the invention solves the problem of low accuracy of automatic identification of chromosome distortion by utilizing a backbone network backbone model, a local preselected frame model and a complete target detection model, and carries out automatic identification and counting of computer images aiming at specific images scanned by a chromosome microscope, so that the identification and classification process is full-automatic, high in speed and high in accuracy. Compared with the existing manual method, the detection speed can be improved by tens of times, and in the full-automatic identification process, manual intervention is not needed in the working period. And comparing the automatic classification result with the manual judgment: statistical accuracy, false positive rate, false negative rate, no significant difference. The identification of chromosome distortion is not dependent on manual operation any more, can be completely identified by computer images, and can quickly and accurately detect personnel in a large area and a large range after a large-scale radioactive accident occurs, so that the problem that professional medical personnel are urgently needed to detect the personnel is solved. Artificial intelligence identification of chromosome aberrations is also an emerging technology in the future and has a significant contribution.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a schematic flow chart of an embodiment of a method for artificial intelligence recognition and analysis of chromosome aberrations provided by the present invention;
FIG. 2 is an image before image normalization;
FIG. 3 is an image after image normalization;
FIG. 4 is an image of a mid-split phase;
FIG. 5 is a schematic illustration of a tagged image making type tags;
FIGS. 6A-6B are images of double-sized mitochondria;
FIGS. 7A-7B are images of three centrosomes;
FIG. 8 is an image of a non-centromeric segment;
FIG. 9 is an image of a minute body;
FIG. 10 is a schematic diagram of YOLOv 3;
fig. 11 is a schematic diagram of 3 prediction blocks of 13 × 13 feature maps;
fig. 12 is a schematic diagram of 3 prediction blocks of 26-by-26 feature maps;
fig. 13 is a schematic diagram of 3 prediction blocks of the 52-by-52 feature map;
FIG. 14 is a schematic illustration of a partial pre-selected frame model image input to output;
FIG. 15 is a diagram showing the comparison between the results of the present method and those of other methods.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Furthermore, the technical features designed in the different embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
Example one
The invention provides an artificial intelligence identification and analysis method of chromosome aberration, as shown in figure 1, comprising the following steps:
s100: inputting original historical chromosome distortion form image data, carrying out data marking according to geometric figures and making type marks so as to complete learning training, and storing marked data into a database;
s200: inputting a chromosome distortion form target image to be identified into a backbone network backbone model to realize the position positioning of the candidate object;
s300: continuously classifying and identifying the objects by positioning the objects through S200, inputting the images processed through S200 into a local preselection frame model, and classifying and identifying actual conditions of different sizes of the target objects;
s400: and outputting a chromosome aberration form automatic analysis result according to the classified and identified data.
The collected images in the analysis method are divided into two specifications of color and black and white.
Acquisition devices include, but are not limited to, an automatic scanning microscope, a digital slide scanner, a CCD optical microscope; the acquired chromosome image may be in color or black and white.
The analysis method can analyze color and black-white images and obtain an automatic analysis result.
In a specific implementation, before step S100, image normalization processing is performed. As shown in fig. 2-3.
Specifically, the normalization processing of the image includes the following methods:
removing background noise of the image: removing background noise by Gaussian filtering; then, watermarking noise and impurity noise by low-frequency filtering, and finally, sharpening the image and matching with filtering, selecting one of median filtering or Gaussian filtering to process unclear silk particles;
processing of background noise of image: the image value square is analyzed, corresponding pixels of the value square of the nearly white background are removed, the background can be removed, and if oversaturation and undersaturation exposure exist, the image is considered to be abandoned;
image value histogram normalization: stretching and moving pixel values of the value square map, and normalizing the value square map to be in the same x-axis area range;
image entropy normalization: and calculating the entropy of the image, removing the image with too large or too small entropy, sharpening the image with too small entropy, smoothing the image with too large entropy, and performing normalization processing.
In particular, the study chromosomes can only be observed during cell division (mitosis or meiosis), as shown in FIG. 4.
In particular, chromosomal aberrations are structural aberrations due to chromosome breakage, and reconnection of broken ends, which in an illegal way produce new structures of abnormal chromosome length, chromosomal location and genetic makeup.
In specific implementation, the data labeling in S100 is performed by labeling chromosome aberration morphology according to the content required for labeling as shown in fig. 5, and labeling double-centromere (as shown in fig. 6A-6B), triple-centromere (as shown in fig. 7A-7B), non-centromere fragment (as shown in fig. 8), minute body (as shown in fig. 9), non-centromere ring, inversion, and reciprocal translocation;
marking the concerned object according to the geometric figure, and making a type mark;
the data quantity of the marks can meet the accuracy requirement set by the user, and the quantity and the accuracy are in positive correlation.
In specific implementation, an artificial intelligent model for target detection is established, the RCNN model is applicable, and the target is to realize target detection or instance segmentation;
the target detection model includes: MASK-RCNN, VGG, YOLO (v2, v3..), SSD, etc., are very numerous, and distorted images have very high consistency compared to live-action photographs, reducing the complex reading requirements for models.
If the MASK-RCNN model, the SSD model and the Faster-RCNN model are selected to select the background, the RPN and the loss function, the custom or common collocation selection is available;
if you choose YOLO (v3), or higher, you choose their background, FPN, and loss functions, all of which have a common or common collocation.
These inherent configurations or customizations can be altered to accommodate the actual image characteristics. For example, the backbone can be changed, the emphasis of image feature recognition can be adjusted, and better targeted adaptation can be achieved.
Through the target detection artificial intelligence model, after training, the target detection artificial intelligence model identifies the concerned information such as double centromere bodies, non-centromere fragments and the like.
Specifically, the structure of YOLO v3, as shown in FIG. 10;
YOLO3 adopts a Darknet-53 network structure (containing 53 convolutional layers), and shortcut links (shortcut connections) are set between some layers;
compared with a passive high structure adopted by YOLO2 to detect fine-grained features, 3 feature maps with different scales are further adopted by YOLO3 to detect objects.
When predicting the object type, softmax is not used, and the output of logistic is used for prediction instead. This enables multi-tagged objects to be supported.
In specific implementation, the backbone network backbone model comprises Darknet, CSPDarkNet, EfficientNet, ResNet Xt and other models; the local pre-selection frame model comprises FPN, BiFPN, ASFF, RFB, SPP and other models.
In specific implementation, the target detection artificial intelligence method of the model combination of S200-S300 is adopted to carry out positioning and classification recognition on the target object, namely, the target is predicted.
The method can predict object types, support multi-label objects, and can distinguish multiple classifications of double centromere, triple centromere, non-centromere fragment, microminia, non-centromere ring, inversion and mutual translocation.
In specific implementation, in S300, 3 feature maps with different scales are output according to S200, 3 prediction frames with different sizes are set for the feature maps with each scale, and the prediction frames with 3 sizes from large to small are clustered to obtain 9 prediction frames with different sizes.
Specifically, as shown in fig. 11-14, the algorithm of the prediction box:
a prediction box of 9 scales;
as the number and scale of the output feature maps change, the size of the prediction box also needs to be adjusted accordingly. YOLO2 has begun to use K-means clustering to derive the prediction box sizes, and YOLO3 continues this method, setting 3 prediction boxes for each downsampling scale, clustering out 9 sizes of prediction boxes in total. In the COCO dataset these 9 prediction boxes are: (10x13), (16x30), (33x23), (30x61), (62x45), (59x119), (116x90), (156x198), (373x 326).
In assignment, larger prediction boxes (116x90), (156x198), (373x326) are applied on the smallest 13 x13 signature (with the largest receptive field), suitable for detecting larger objects. Medium 26 × 26 feature maps (medium receptive fields) were fitted with medium prediction boxes (30x61), (62x45), (59x119) suitable for detection of medium sized subjects. Smaller prediction boxes (10x13), (16x30), (33x23) were applied on the larger 52 x 52 signature (smaller receptive field), suitable for detecting smaller subjects.
Specifically, YOLO3 employs a one stage approach. By using the residual error network structure for reference, a deeper network layer is formed, multi-scale detection is realized, and mAP and small object detection effects are improved. When the accuracy of the YOLO3 is compared with other object detection models at the same time, the velocity of the YOLO v3 is 3 and 4 times higher than that of other models at the same time.
In specific implementation, a backbone network backbone model and a local pre-selection frame model form a complete target detection model, images after database standardization are trained to obtain training parameters, then objects of interest in the test images are predicted and judged through the training parameters, and positioning and classification labeling are carried out.
Specifically, the complete target detection model having a sense prediction (one stage) type model includes: SSD, YOLO, RetinaNet; or spark prediction (two stage) class of models include: fast-RCNN, R-FCN, Mask-RCNN.
In specific implementation, in S400, each image of the prediction frame is identified, whether the image is a chromosome or not and whether the image is normal or not is determined, which type of abnormal chromosome and double centromeres, triple centromeres, non-centromeres fragments, minute bodies, non-centromeres rings, inversions and mutual translocations are determined, and information of the abnormal chromosome, including position and type information, is output.
In specific implementation, as shown in fig. 15, in comparison between the method and other methods, if the detection rate of the manual method is 100%, the detection rate of the method is 81%, and the detection rate of other comparison methods is only 42%, it can be seen that the accuracy of the method is very high and far exceeds the detection rate of the method, and thus the method has a significant effect in practical application.
Compared with the prior art, the artificial intelligent identification and analysis method for chromosome distortion provided by the invention solves the problem of low accuracy of automatic identification of chromosome distortion by utilizing a backbone network backbone model, a local preselected frame model and a complete target detection model, and carries out automatic identification and counting of computer images aiming at specific images scanned by a chromosome microscope, so that the identification and classification process is full-automatic, high in speed and high in accuracy. Compared with the existing manual method, the detection speed can be improved by tens of times, and in the full-automatic identification process, manual intervention is not needed in the working period. And comparing the automatic classification result with the manual judgment: statistical accuracy, false positive rate, false negative rate, no significant difference. The identification of chromosome distortion is not dependent on manual operation any more, can be completely identified by computer images, and can quickly and accurately detect personnel in a large area and a large range after a large-scale radioactive accident occurs, so that the problem that professional medical personnel are urgently needed to detect the personnel is solved. Artificial intelligence identification of chromosome aberrations is also an emerging technology in the future and has a significant contribution.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.
Claims (9)
1. An artificial intelligence identification and analysis method for chromosome aberration is characterized by comprising the following steps:
s100: inputting original historical chromosome distortion form image data, carrying out data marking according to geometric figures and making type marks so as to complete learning training, and storing marked data into a database;
s200: inputting a chromosome distortion form target image to be identified into a backbone network backbone model to realize the position positioning of the candidate object;
s300: continuously classifying and identifying the objects by positioning the objects through S200, inputting the images processed through S200 into a local preselection frame model, and classifying and identifying actual conditions of different sizes of the target objects;
s400: and outputting a chromosome aberration form automatic analysis result according to the classified and identified data.
2. The method for artificial intelligence recognition and analysis of chromosome aberrations of claim 1, wherein: before step S100, image normalization processing is performed.
3. The method for artificial intelligence recognition and analysis of chromosome aberrations according to claim 2, wherein the normalization process of the image comprises the following methods:
removing background noise of the image: removing background noise by Gaussian filtering; then, watermarking noise and impurity noise by low-frequency filtering, and finally, sharpening the image and matching with filtering, selecting one of median filtering or Gaussian filtering to process unclear silk particles;
processing of background noise of image: the image value square is analyzed, corresponding pixels of the value square of the nearly white background are removed, the background can be removed, and if oversaturation and undersaturation exposure exist, the image is considered to be abandoned;
image value histogram normalization: stretching and moving pixel values of the value square map, and normalizing the value square map to be in the same x-axis area range;
image entropy normalization: and calculating the entropy of the image, removing the image with too large or too small entropy, sharpening the image with too small entropy, smoothing the image with too large entropy, and performing normalization processing.
4. The method for artificial intelligence recognition and analysis of chromosome aberrations of claim 1, wherein: the data marking in the S100 marks chromosome aberration forms through marking required contents, and marks double centromeres, three centromeres, non-centromeres fragments, microminians, non-centromeres rings, inversions and mutual translocation;
marking the concerned object according to the geometric figure, and making a type mark;
the data quantity of the marks can meet the accuracy requirement set by the user, and the quantity and the accuracy are in positive correlation.
5. The method for artificial intelligence recognition and analysis of chromosome aberrations of claim 1, wherein: the backbone network backbone model comprises Darknet, CSPDarkNet, EfficientNet, ResNet and ResNet Xt models; the local pre-selection frame model comprises an FPN model, a BiFPN model, an ASFF model, an RFB model and an SPP model.
6. The method for artificial intelligence recognition and analysis of chromosome aberrations of claim 1, wherein:
and positioning and classifying identification of the object, namely predicting the object, are carried out by adopting an artificial intelligence method for target detection of the model combination of S200-S300.
7. The method for artificial intelligence recognition and analysis of chromosome aberrations of claim 1, wherein: in the step S300, 3 feature maps with different scales are output according to the step S200, prediction frames with 3 different sizes are set for the feature maps with each scale, the prediction frames with 3 sizes are clustered from large to small, and the prediction frames with 9 sizes are clustered.
8. The method for artificial intelligence recognition and analysis of chromosome aberrations of claim 1, wherein: the backbone network backbone model and the local pre-selection frame model form a complete target detection model, the images standardized by the database are trained to obtain training parameters, then the objects concerned in the test images are predicted and judged through the training parameters, and positioning and classification marking are carried out.
9. The method for artificial intelligence recognition and analysis of chromosome aberrations of claim 1, wherein: in S400, each image of the prediction frame is identified, whether the image is a chromosome or not is determined, whether the image is normal or not is determined, which type of abnormal chromosome and double centromeres, triple centromeres, non-centromeres fragments, micro-bodies, non-centromeres rings, inversions and mutual translocations are determined, and information of the abnormal chromosome, including position and type information, is output.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010906567.0A CN111951266A (en) | 2020-09-01 | 2020-09-01 | Artificial intelligence recognition analysis method for chromosome aberration |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010906567.0A CN111951266A (en) | 2020-09-01 | 2020-09-01 | Artificial intelligence recognition analysis method for chromosome aberration |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111951266A true CN111951266A (en) | 2020-11-17 |
Family
ID=73367725
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010906567.0A Pending CN111951266A (en) | 2020-09-01 | 2020-09-01 | Artificial intelligence recognition analysis method for chromosome aberration |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111951266A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112381806A (en) * | 2020-11-18 | 2021-02-19 | 上海北昂医药科技股份有限公司 | Double centromere aberration chromosome analysis and prediction method based on multi-scale fusion method |
| CN112464652A (en) * | 2020-11-24 | 2021-03-09 | 昆明理工大学 | Method for extracting illustration and title thereof from literature |
| CN113052834A (en) * | 2021-04-20 | 2021-06-29 | 河南大学 | Pipeline defect detection method based on convolution neural network multi-scale features |
| CN113450338A (en) * | 2021-07-09 | 2021-09-28 | 湖南晟瞳科技有限公司 | Method and system for automatically detecting white band sample based on deep learning and storage medium |
| CN114973244A (en) * | 2022-06-12 | 2022-08-30 | 桂林电子科技大学 | System and method for automatically identifying mitosis of H & E staining pathological image of breast cancer |
| CN115063411A (en) * | 2022-08-04 | 2022-09-16 | 湖南自兴智慧医疗科技有限公司 | Chromosome abnormal region segmentation detection method and system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090048785A1 (en) * | 2006-01-10 | 2009-02-19 | Applied Spectral Imaging Ltd. | Methods And Systems For Analyzing Biological Samples |
| CN108467884A (en) * | 2015-06-23 | 2018-08-31 | 兹托视觉有限公司 | The method for detecting chromosome aberration |
| CN110390312A (en) * | 2019-07-29 | 2019-10-29 | 北京航空航天大学 | Chromosome automatic classification method and classifier based on convolutional neural networks |
| CN110533672A (en) * | 2019-08-22 | 2019-12-03 | 杭州德适生物科技有限公司 | A kind of chromosome sort method based on band identification |
-
2020
- 2020-09-01 CN CN202010906567.0A patent/CN111951266A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090048785A1 (en) * | 2006-01-10 | 2009-02-19 | Applied Spectral Imaging Ltd. | Methods And Systems For Analyzing Biological Samples |
| CN108467884A (en) * | 2015-06-23 | 2018-08-31 | 兹托视觉有限公司 | The method for detecting chromosome aberration |
| CN110390312A (en) * | 2019-07-29 | 2019-10-29 | 北京航空航天大学 | Chromosome automatic classification method and classifier based on convolutional neural networks |
| CN110533672A (en) * | 2019-08-22 | 2019-12-03 | 杭州德适生物科技有限公司 | A kind of chromosome sort method based on band identification |
Non-Patent Citations (1)
| Title |
|---|
| 李朝文 等: "人外周血淋巴细胞微核、染色体畸变图像自动分析系统的研制", 《中国毒理学会第九次全国毒理学大会论文集》, 17 September 2019 (2019-09-17), pages 1 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112381806A (en) * | 2020-11-18 | 2021-02-19 | 上海北昂医药科技股份有限公司 | Double centromere aberration chromosome analysis and prediction method based on multi-scale fusion method |
| CN112464652A (en) * | 2020-11-24 | 2021-03-09 | 昆明理工大学 | Method for extracting illustration and title thereof from literature |
| CN113052834A (en) * | 2021-04-20 | 2021-06-29 | 河南大学 | Pipeline defect detection method based on convolution neural network multi-scale features |
| CN113052834B (en) * | 2021-04-20 | 2023-04-18 | 河南大学 | Pipeline defect detection method based on convolution neural network multi-scale features |
| CN113450338A (en) * | 2021-07-09 | 2021-09-28 | 湖南晟瞳科技有限公司 | Method and system for automatically detecting white band sample based on deep learning and storage medium |
| CN114973244A (en) * | 2022-06-12 | 2022-08-30 | 桂林电子科技大学 | System and method for automatically identifying mitosis of H & E staining pathological image of breast cancer |
| CN114973244B (en) * | 2022-06-12 | 2023-04-11 | 桂林电子科技大学 | System and method for automatically identifying mitosis of H & E staining pathological image of breast cancer |
| CN115063411A (en) * | 2022-08-04 | 2022-09-16 | 湖南自兴智慧医疗科技有限公司 | Chromosome abnormal region segmentation detection method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111951266A (en) | Artificial intelligence recognition analysis method for chromosome aberration | |
| CN107274386B (en) | artificial intelligent auxiliary cervical cell fluid-based smear reading system | |
| DK2973397T3 (en) | Tissue-object-based machine learning system for automated assessment of digital whole-slide glass | |
| CN112257704A (en) | Cervical fluid-based cell digital image classification method based on deep learning detection model | |
| CN110765855B (en) | Pathological image processing method and system | |
| CN111986150B (en) | The method comprises the following steps of: digital number pathological image Interactive annotation refining method | |
| CN107256558A (en) | The cervical cell image automatic segmentation method and system of a kind of unsupervised formula | |
| CN112184657A (en) | Pulmonary nodule automatic detection method, device and computer system | |
| Pan et al. | Mitosis detection techniques in H&E stained breast cancer pathological images: A comprehensive review | |
| CN112215807A (en) | Cell image automatic classification method and system based on deep learning | |
| US10769432B2 (en) | Automated parameterization image pattern recognition method | |
| CN113658174B (en) | Microkernel histology image detection method based on deep learning and image processing algorithm | |
| CN111539330A (en) | Transformer substation digital display instrument identification method based on double-SVM multi-classifier | |
| Kovalev et al. | Deep learning in big image data: Histology image classification for breast cancer diagnosis | |
| Khan et al. | Comparitive study of tree counting algorithms in dense and sparse vegetative regions | |
| CN116978543A (en) | Artificial intelligent auxiliary marrow tumor pathological diagnosis device | |
| CN114782948A (en) | Global interpretation method and system for cervical liquid-based cytology smear | |
| JP4609322B2 (en) | Chromosome state evaluation method and evaluation system | |
| Riana et al. | Comparison of nucleus and inflammatory cell detection methods on Pap smear images | |
| Imran et al. | Image-Based Automatic Energy Meter Reading Using Deep Learning | |
| Merone et al. | On using active contour to segment HEp-2 cells | |
| CN113838008A (en) | Abnormal cell detection method based on attention-drawing mechanism | |
| Amitha et al. | Developement of computer aided system for detection and classification of mitosis using SVM | |
| JP6329651B1 (en) | Image processing apparatus and image processing method | |
| CN116958710B (en) | Embryo development stage prediction method and system based on annular feature statistics |
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 |