CN113724793A - Chromosome important feature visualization method and device based on convolutional neural network - Google Patents
Chromosome important feature visualization method and device based on convolutional neural network Download PDFInfo
- Publication number
- CN113724793A CN113724793A CN202111279350.2A CN202111279350A CN113724793A CN 113724793 A CN113724793 A CN 113724793A CN 202111279350 A CN202111279350 A CN 202111279350A CN 113724793 A CN113724793 A CN 113724793A
- Authority
- CN
- China
- Prior art keywords
- chromosome
- picture
- neural network
- convolutional neural
- importance information
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16B—BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY
- G16B45/00—ICT specially adapted for bioinformatics-related data visualisation, e.g. displaying of maps or 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16B—BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY
- G16B40/00—ICT specially adapted for biostatistics; ICT specially adapted for bioinformatics-related machine learning or data mining, e.g. knowledge discovery or pattern finding
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Biophysics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Biotechnology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Bioethics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Databases & Information Systems (AREA)
- Epidemiology (AREA)
- Mathematical Physics (AREA)
- Public Health (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Image Analysis (AREA)
- Investigating Or Analysing Biological Materials (AREA)
Abstract
The invention provides a chromosome important feature visualization method based on a convolutional neural network. The method comprises the following steps: training a preset convolutional neural network model by using a training data set; inputting the training data set into the trained convolutional neural network model to obtain a classification result, a weight result and a characteristic result of each chromosome picture; multiplying the characteristic result of each chromosome picture by the weight result of each chromosome picture to obtain first importance information of each chromosome picture; averagely mapping the first importance information of each chromosome picture to the chromosome strip corresponding to the chromosome picture to obtain longitudinal importance information; and carrying out statistical analysis on each chromosome picture according to the longitudinal importance information and the classification result of each chromosome picture to obtain and display important band features of the convolutional neural network model for identifying each chromosome, so that the visual display of the important band features of the chromosome for judging the chromosome category by the convolutional neural network can be realized.
Description
Technical Field
The invention relates to the technical field of intelligent medical treatment, in particular to a chromosome important feature visualization method and device based on a convolutional neural network.
Background
Human chromosome recognition is an important research topic of medical genetics, has wide application in the fields of medical clinical diagnosis, auxiliary teaching, scientific research and the like, and is an important basis for judging human genetic diseases. With the development of artificial intelligence, the convolutional neural network is widely and effectively applied in the field of image processing. The convolutional neural network has good effect on the aspects of chromosome recognition and the like, has higher accuracy and higher speed, and can effectively reduce the burden of doctors.
However, in the prior art, although the convolutional neural network achieves high precision in the recognition of chromosomes, the convolutional neural network cannot completely replace manual recognition, cannot enable people to understand the extracted features, and cannot realize the visualization display of important banding features of the chromosomes.
Disclosure of Invention
The invention aims to provide a chromosome important feature visualization method and a chromosome important feature visualization device based on a convolutional neural network, which can realize the visualization display of important chromosome band features for distinguishing chromosome types by the convolutional neural network, facilitate the establishment of trust of doctors on the network, assist doctors in identifying chromosomes, and facilitate researchers to discover and learn important feature differences among chromosomes of different types.
In order to achieve the above object, the present invention provides a method for visualizing important features of chromosomes based on a convolutional neural network, comprising the following steps:
step S1, providing a plurality of original chromosome pictures, straightening each original chromosome picture along a skeleton line by using a preset straightening algorithm to obtain a plurality of training pictures, and further generating a training data set comprising the plurality of training pictures;
step S2, training a preset convolution neural network model by using the training data set;
step S3, inputting a training data set into the trained convolutional neural network model to obtain a classification result, a weight result and a feature result of each chromosome picture; the classification result of the chromosome picture is the category of the chromosome in the chromosome picture output by the convolutional neural network model; the weight result of the chromosome picture is a weight value corresponding to the category of the chromosome in the chromosome picture output by the convolutional neural network model; the characteristic result of the chromosome picture is a characteristic picture value of the chromosome picture output by the convolutional neural network model;
step S4, weighting the characteristic result of each chromosome picture and the weight result thereof to obtain first importance information of each chromosome picture;
step S5, mapping the first importance information of each chromosome picture on the chromosome strip corresponding to the chromosome picture on average to obtain the longitudinal importance information of each chromosome picture;
and S6, performing statistical analysis on each chromosome picture according to the longitudinal importance information of each chromosome picture and the classification result of each chromosome picture to obtain and display important stripe characteristics of each chromosome recognized by the convolutional neural network model.
Optionally, the convolutional neural network model preset in step S2 includes: the system comprises a preprocessing module, a first channel convolution layer, a first maximum pooling layer, a second channel convolution layer, a second maximum pooling layer, a third channel convolution layer, a fourth channel convolution layer, a third maximum pooling layer, a fifth channel convolution layer, a sixth channel convolution layer, a fourth maximum pooling layer, a seventh channel convolution layer, an eighth pass convolution layer, a ninth pass convolution layer, a global average pooling layer and a full-connection layer which are connected in sequence;
the classification result of the chromosome picture is the type of the chromosome in the chromosome picture output by the full-connection layer; the weighting result of the chromosome picture is the weighting value corresponding to the category of the chromosome in the chromosome picture output by the global average pooling layer; and the characteristic result of the chromosome picture is the characteristic map value of the chromosome picture output by the ninth channel convolution layer.
Optionally, the preprocessing module is configured to change the size of the training picture to a preset size, normalize the pixel point values of the training picture, and perform operations of randomly turning the horizontal direction and increasing the horizontal offset on the training picture.
Optionally, the first channel convolutional layer is a 64-channel convolutional layer, the second channel convolutional layer is a 128-channel convolutional layer, the third and fourth channel convolutional layers are 256-channel convolutional layers, the fifth, sixth, seventh and eighth channel convolutional layers are 512-channel convolutional layers, and the ninth channel convolutional layer is a 1024-channel convolutional layer;
the sizes of convolution kernels of the first to ninth channel convolution layers are all 3 multiplied by 3, and the activation functions are all linear rectification functions;
the pooling areas of the first to fourth largest pooling layers are all 2 × 2;
the activation function of the full connection layer is a normalized exponential function;
and L2 regularization terms with the parameter of 0.001 are added into the first to ninth channel convolution layers, the first to fourth maximum pooling layers, the global average pooling layer and the full-connection layer to prevent overfitting.
Optionally, in step S2, a convolutional neural network model preset by using a cross entropy loss function and a stochastic gradient descent method is used.
Optionally, the first importance information comprises a plurality of dot values distributed in an array; in step S5, the point location values of each row in the first importance information are compressed in a horizontal average manner, so as to obtain the vertical importance information of each chromosome picture.
Optionally, the performing statistical analysis on each chromosome picture according to the longitudinal importance information of each chromosome picture and the classification result of each chromosome picture in step S6 specifically includes:
normalizing the longitudinal importance information of each chromosome picture;
and carrying out average calculation on the normalized longitudinal importance information of the chromosome pictures with the same classification result to obtain the average longitudinal importance information of each type of chromosome picture.
Optionally, the performing statistical analysis on each chromosome picture according to the longitudinal importance information of each chromosome picture and the classification result of each chromosome picture in step S6 specifically includes:
marking the longitudinal importance information of each chromosome picture, adding a first mark to n point positions with the largest value in the longitudinal importance information, and adding a second mark to the other point positions;
and (3) carrying out mark statistics on the chromosome pictures with the same classification result, and calculating the mark proportion of each point position, wherein the calculation formula is as follows:
Pij=Aij/Biwherein P is the mark proportion of the jth point of the longitudinal importance information of the chromosome picture with all classification results being i, AijThe number of the first mark added to the jth point of the longitudinal importance information of all chromosome images with classification result i, BiThe total number of chromosome images with the classification result i.
Optionally, the step S6 of displaying that the identification of the important strip features of each type of chromosome by the convolutional neural network model specifically includes:
and comparing the statistical result obtained by the statistical analysis in the step S6 with the pattern graph corresponding to the training data set or the average graph of the training data set in a histogram mode.
The invention also provides a chromosome important characteristic visualization device based on the convolutional neural network, and the method is adopted for realizing the chromosome important characteristic visualization device.
The invention has the beneficial effects that: the invention provides a chromosome important feature visualization method based on a convolutional neural network, which comprises the following steps: step S1, providing a plurality of original chromosome pictures, straightening each original chromosome picture along a skeleton line by using a preset straightening algorithm to obtain a plurality of training pictures, and further generating a training data set comprising the plurality of training pictures; step S2, training a preset convolution neural network model by using the training data set; step S3, inputting a training data set into the trained convolutional neural network model to obtain a classification result, a weight result and a feature result of each chromosome picture; step S4, weighting the characteristic result of each chromosome picture and the weight result thereof to obtain first importance information of each chromosome picture; step S5, mapping the first importance information of each chromosome picture on the chromosome strip corresponding to the chromosome picture on average to obtain the longitudinal importance information of each chromosome picture; and step S6, performing statistical analysis on each chromosome picture according to the longitudinal importance information of each chromosome picture and the classification result of each chromosome picture to obtain and display the important strip characteristics of each chromosome recognized by the convolutional neural network model, so that the visual display of the important strip characteristics of the chromosomes for judging the chromosome types by the convolutional neural network can be realized, the trust of doctors on the network can be conveniently established, the doctors are assisted in recognizing the chromosomes, and meanwhile, researchers can conveniently find and learn the important characteristic differences among the chromosomes of different types.
Drawings
For a better understanding of the nature and technical aspects of the present invention, reference should be made to the following detailed description of the invention, taken in conjunction with the accompanying drawings, which are provided for purposes of illustration and description and are not intended to limit the invention.
In the drawings, there is shown in the drawings,
FIG. 1 is a flow chart of a method for visualizing important features of a chromosome based on a convolutional neural network according to the present invention;
FIG. 2 is a schematic diagram of the architecture of a convolutional neural network in the method for visualizing important features of chromosomes based on the convolutional neural network according to the present invention;
FIG. 3 is a result display diagram of an embodiment of the method for visualizing important features of chromosomes based on a convolutional neural network according to the present invention
Fig. 4 is a display effect diagram of another embodiment of the method for visualizing important features of chromosomes based on the convolutional neural network of the present invention.
Detailed Description
To further illustrate the technical means and effects of the present invention, the following detailed description is given with reference to the preferred embodiments of the present invention and the accompanying drawings.
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. 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 application.
In the description of the present application, it is to be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", and the like indicate orientations or positional relationships based on those shown in the drawings, and are used merely for convenience of description and for simplicity of description, and do not indicate or imply that the referenced device or element must have a particular orientation, be constructed in a particular orientation, and be operated, and thus should not be considered as limiting the present application. Furthermore, the terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more features. In the description of the present application, "a plurality" means two or more unless specifically limited otherwise.
In this application, the word "exemplary" is used to mean "serving as an example, instance, or illustration. Any embodiment described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments. The following description is presented to enable any person skilled in the art to make and use the application. In the following description, details are set forth for the purpose of explanation. It will be apparent to one of ordinary skill in the art that the present application may be practiced without these specific details. In other instances, well-known structures and processes are not set forth in detail in order to avoid obscuring the description of the present application with unnecessary detail. Thus, the present application is not intended to be limited to the embodiments shown, but is to be accorded the widest scope consistent with the principles and features disclosed herein.
Referring to fig. 1, the present invention first provides a method for visualizing important features of a chromosome based on a convolutional neural network, which includes the following steps:
and step S1, providing a plurality of original chromosome pictures, straightening each original chromosome picture along a skeleton line by using a preset straightening algorithm to obtain a plurality of training pictures, and further generating a training data set comprising the plurality of training pictures.
Specifically, the step S1 further includes a step of removing the training pictures, which are processed by the straightening algorithm and have the effect not meeting the preset processing standard, from the training data set, and further, the step of removing may be manually completed by a human.
Preferably, in some embodiments of the present invention, the step S1 provides 612 cases, 1224 original chromosome pictures, and the ratio of the number of X, Y sex chromosomes in the 1224 original chromosome pictures is about 3: 1.
And step S2, training a preset convolutional neural network model by using the training data set.
Specifically, in some embodiments of the present invention, the preset convolutional neural network model includes: the device comprises a preprocessing module, a plurality of channel convolution layers, a plurality of maximum pooling layers, a global average pooling layer and a full-connection layer which are connected in sequence.
The preprocessing module is used for changing the size of the training picture to a preset size, normalizing pixel point values of the training picture, and performing random horizontal turning and horizontal offset increasing operations on the training picture.
Further, as shown in fig. 2, in some embodiments of the present invention, the convolutional neural network model preset in step S2 includes: the device comprises a preprocessing module, a first channel convolution layer 10, a first maximum pooling layer 20, a second channel convolution layer 30, a second maximum pooling layer 40, a third channel convolution layer 50, a fourth channel convolution layer 60, a third maximum pooling layer 70, a fifth channel convolution layer 80, a sixth channel convolution layer 90, a fourth maximum pooling layer 100, a seventh channel convolution layer 110, an eighth pass convolution layer 120, a ninth pass convolution layer 130, a global average pooling layer 140 and a full-connection layer 150 which are connected in sequence;
preferably, as shown in fig. 2, in some embodiments of the present invention, the first channel convolutional layer 10 is a 64-channel convolutional layer, the second channel convolutional layer 30 is a 128-channel convolutional layer, the third channel convolutional layer 50 and the fourth channel convolutional layer 60 are 256-channel convolutional layers, the fifth channel convolutional layer 80, the sixth channel convolutional layer 90, the seventh channel convolutional layer 110 and the eighth channel convolutional layer 120 are 512-channel convolutional layers, and the ninth channel convolutional layer 130 is a 1024-channel convolutional layer;
preferably, the convolution kernels of the first to ninth channel convolution layers are all 3 × 3 in size, and the activation functions are all Linear rectification functions (ReLU); the pooling areas of the first to fourth largest pooling layers are all 2 × 2; the activation function of the fully-connected layer is a normalized exponential function (Softmax); and L2 regularization terms with the parameter of 0.001 are added into the first to ninth channel convolution layers, the first to fourth maximum pooling layers, the global average pooling layer and the full-connection layer to prevent overfitting.
Further, in step S2, a convolutional neural network model preset by using a cross entropy loss function and a stochastic gradient descent method is used.
In detail, in some embodiments of the invention, the pre-processing module will change the size of the trained picture to (240, 80), normalize the pixel point values to between 0-1, randomly flip horizontally, add horizontal offset, vertically for stripe correspondence and prevent stripe loss from doing nothing.
Step S3, inputting a training data set into the trained convolutional neural network model to obtain a classification result, a weight result and a feature result of each chromosome picture; the classification result of the chromosome picture is the category of the chromosome in the chromosome picture output by the convolutional neural network model; the weight result of the chromosome picture is a weight value corresponding to the category of the chromosome in the chromosome picture output by the convolutional neural network model; and the characteristic result of the chromosome picture is the characteristic picture value of the chromosome picture output by the convolutional neural network model.
Specifically, the classification result of the chromosome picture is the category of the chromosome in the chromosome picture output by the full-link layer; the weighting result of the chromosome picture is the weighting value corresponding to the category of the chromosome in the chromosome picture output by the global average pooling layer; and the characteristic result of the chromosome picture is the characteristic map value of the chromosome picture output by the ninth channel convolution layer.
And step S4, weighting the feature result of each chromosome picture and the weight result thereof to obtain first importance information of each chromosome picture.
For example, a chromosome picture is input into a trained convolutional neural network model to obtain 1024-dimensional 15 × 5 feature maps and weights of corresponding categories, and the two are weighted to obtain 15 × 5 overall feature maps shown in table 1, point values on the overall feature maps can reflect importance of different spatial positions of the chromosome picture to chromosome classification and identification, wherein the 1024-dimensional 15 × 5 feature maps are feature results of the chromosome picture, the weights of the corresponding categories are weight results of the chromosome picture, and the 15 × 5 overall feature map is first importance information of the chromosome picture.
TABLE 1 first importance information of a chromosome map
And step S5, averagely mapping the first importance information of each chromosome picture to the chromosome strip corresponding to the chromosome picture to obtain the longitudinal importance information of each chromosome picture.
Specifically, the first importance information includes a plurality of dot values distributed in an array; in step S5, the point location values of each row in the first importance information are compressed in a horizontal average manner, so as to obtain the vertical importance information of each chromosome picture.
Taking the embodiment shown in table 1 as an example, the first importance information of the chromosome image includes 75 point values distributed in 15 rows and 5 columns, and in step S5, the point values of each row in the first importance information are transversely and equally compressed to obtain the longitudinal importance information of the chromosome image, as shown in table 2, the longitudinal importance information of the chromosome image includes 15 point values distributed in 15 rows and 1 columns.
TABLE 2 longitudinal importance information of a chromosome map
And S6, performing statistical analysis on each chromosome picture according to the longitudinal importance information of each chromosome picture and the classification result of each chromosome picture to obtain and display important stripe characteristics of each chromosome recognized by the convolutional neural network model.
Specifically, the step S6 of performing statistical analysis on each chromosome picture according to the longitudinal importance information of each chromosome picture and the classification result of each chromosome picture specifically includes:
normalizing the longitudinal importance information of each chromosome picture;
and carrying out average calculation on the normalized longitudinal importance information of the chromosome pictures with the same classification result to obtain the average longitudinal importance information of each type of chromosome picture.
Taking table 2 as an example, the specific method for normalizing the longitudinal importance information of each chromosome image is as follows: dividing each dot value in the longitudinal importance information of the chromosome picture by the maximum dot value in the longitudinal importance information to obtain the normalized longitudinal importance information of the chromosome picture shown in table 3.
Table 3, normalized longitudinal importance information of one chromosome picture;
the normalized longitudinal importance information of the chromosome pictures with the same classification result is averaged to obtain the average longitudinal importance information of each type of chromosome picture, for example, if the chromosome picture with the type of B number is 100, the chromosome picture with the type of B number corresponds to 100 pieces of normalized longitudinal importance information, and the average value of the 100 pieces of normalized longitudinal importance information is calculated, the average longitudinal importance information of the chromosome pictures with the type of B number shown in table 4 can be obtained, where B is a positive integer.
TABLE 4 average longitudinal importance information of chromosome images of type B
Specifically, the step S6, which shows that the identification of the important strip features of each type of chromosome by the convolutional neural network model, specifically includes: and comparing the statistical result obtained by the statistical analysis in the step S6 with the pattern graph corresponding to the training data set or the average graph of the training data set in a histogram mode.
Taking table 4 as an example, the statistical result obtained by the statistical analysis in step S6 is displayed by comparing the histogram with the pattern diagram corresponding to the training data set or the average diagram of the training data set, specifically, the statistical result is displayed by comparing the histogram with the pattern diagram corresponding to the training data set or the average diagram of the training data set, the display result is shown in fig. 3, and according to fig. 3, the researcher can find the important band feature of the chromosome of the identification category B number.
Further, in other embodiments of the present invention, the performing, in step S6, statistical analysis on each chromosome picture according to the longitudinal importance information of each chromosome picture and the classification result of each chromosome picture specifically includes:
marking the longitudinal importance information of each chromosome picture, adding a first mark to n points with the largest value in the longitudinal importance information, and adding a second mark to the rest positions;
and (3) carrying out mark statistics on the chromosome pictures with the same classification result, and calculating the mark proportion of each point position, wherein the calculation formula is as follows:
Pij=Aij/Biwherein P is the mark proportion of the jth point of the longitudinal importance information of the chromosome picture with all classification results being i, AijThe number of the first mark added to the jth point of the longitudinal importance information of all chromosome images with classification result i, BiThe total number of chromosome images with the classification result i.
Taking table 2 as an example, a first marker is added to the 4 largest-valued point positions in the longitudinal importance information, the first marker is "1", the second marker is "0", and the longitudinal importance information of one chromosome picture after marking is shown in table 5.
Table 5, longitudinal importance information of one marked chromosome picture;
taking table 5 as an example, the chromosome pictures with the same classification result are subjected to marker statistics, and the marker proportion of each point location is calculated, for example, if the chromosome picture with the type B number is 100, the chromosome picture with the type B number corresponds to the longitudinal importance information after 100 markers, and the ratio of each point location added with the marker "1" is counted, so as to obtain table 6.
Table 6, marker ratio of chromosome images of category B;
taking table 4 as an example, the statistical result obtained by the statistical analysis in step S6 is displayed by comparing the histogram with the pattern diagram corresponding to the training data set or the average diagram of the training data set, specifically, the statistical result is displayed by comparing the histogram with the pattern diagram corresponding to the training data set or the average diagram of the training data set in table 6, the display result is shown in fig. 4, and according to fig. 4, the researcher can find the important band feature of the chromosome of the identification category B number.
Based on the above inventive idea, the invention further provides a chromosome important feature visualization device based on the convolutional neural network, which is realized by adopting the method.
In summary, the invention provides a chromosome important feature visualization method based on a convolutional neural network, which comprises the following steps: step S1, providing a plurality of original chromosome pictures, straightening each original chromosome picture along a skeleton line by using a preset straightening algorithm to obtain a plurality of training pictures, and further generating a training data set comprising the plurality of training pictures; step S2, training a preset convolution neural network model by using the training data set; step S3, inputting a training data set into the trained convolutional neural network model to obtain a classification result, a weight result and a feature result of each chromosome picture; step S4, weighting the characteristic result of each chromosome picture and the weight result thereof to obtain first importance information of each chromosome picture; step S5, mapping the first importance information of each chromosome picture on the chromosome strip corresponding to the chromosome picture on average to obtain the longitudinal importance information of each chromosome picture; and step S6, performing statistical analysis on each chromosome picture according to the longitudinal importance information of each chromosome picture and the classification result of each chromosome picture to obtain and display the important strip characteristics of each chromosome recognized by the convolutional neural network model, so that the visual display of the important strip characteristics of the chromosomes for judging the chromosome types by the convolutional neural network can be realized, the trust of doctors on the network can be conveniently established, the doctors are assisted in recognizing the chromosomes, and meanwhile, researchers can conveniently find and learn the important characteristic differences among the chromosomes of different types.
As described above, it will be apparent to those skilled in the art that other various changes and modifications may be made based on the technical solution and concept of the present invention, and all such changes and modifications are intended to fall within the scope of the appended claims.
Claims (10)
1. A chromosome important feature visualization method based on a convolutional neural network is characterized by comprising the following steps:
step S1, providing a plurality of original chromosome pictures, straightening each original chromosome picture along a skeleton line by using a preset straightening algorithm to obtain a plurality of training pictures, and further generating a training data set comprising the plurality of training pictures;
step S2, training a preset convolution neural network model by using the training data set;
step S3, inputting a training data set into the trained convolutional neural network model to obtain a classification result, a weight result and a feature result of each chromosome picture; the classification result of the chromosome picture is the category of the chromosome in the chromosome picture output by the convolutional neural network model; the weight result of the chromosome picture is a weight value corresponding to the category of the chromosome in the chromosome picture output by the convolutional neural network model; the characteristic result of the chromosome picture is a characteristic picture value of the chromosome picture output by the convolutional neural network model;
step S4, weighting the characteristic result of each chromosome picture and the weight result thereof to obtain first importance information of each chromosome picture;
step S5, mapping the first importance information of each chromosome picture on the chromosome strip corresponding to the chromosome picture on average to obtain the longitudinal importance information of each chromosome picture;
and S6, performing statistical analysis on each chromosome picture according to the longitudinal importance information of each chromosome picture and the classification result of each chromosome picture to obtain and display important stripe characteristics of each chromosome recognized by the convolutional neural network model.
2. The method for visualizing important features of chromosomes based on convolutional neural network as claimed in claim 1, wherein the convolutional neural network model preset in step S2 comprises: the system comprises a preprocessing module, a first channel convolution layer, a first maximum pooling layer, a second channel convolution layer, a second maximum pooling layer, a third channel convolution layer, a fourth channel convolution layer, a third maximum pooling layer, a fifth channel convolution layer, a sixth channel convolution layer, a fourth maximum pooling layer, a seventh channel convolution layer, an eighth pass convolution layer, a ninth pass convolution layer, a global average pooling layer and a full-connection layer which are connected in sequence;
the classification result of the chromosome picture is the type of the chromosome in the chromosome picture output by the full-connection layer; the weighting result of the chromosome picture is the weighting value corresponding to the category of the chromosome in the chromosome picture output by the global average pooling layer; and the characteristic result of the chromosome picture is the characteristic map value of the chromosome picture output by the ninth channel convolution layer.
3. The method for visualizing important features of chromosomes based on convolutional neural network as claimed in claim 2, wherein said preprocessing module is used for changing the size of the training picture to a preset size, normalizing the pixel point values of the training picture, and performing operations of random horizontal flipping and horizontal offset adding on the training picture.
4. The convolutional neural network-based chromosome significant feature visualization method according to claim 2, wherein the first channel convolutional layer is a 64-channel convolutional layer, the second channel convolutional layer is a 128-channel convolutional layer, the third and fourth channel convolutional layers are 256-channel convolutional layers, the fifth, sixth, seventh and eighth channel convolutional layers are 512-channel convolutional layers, and the ninth channel convolutional layer is a 1024-channel convolutional layer;
the sizes of convolution kernels of the first to ninth channel convolution layers are all 3 multiplied by 3, and the activation functions are all linear rectification functions;
the pooling areas of the first to fourth largest pooling layers are all 2 × 2;
the activation function of the full connection layer is a normalized exponential function;
and L2 regularization terms with the parameter of 0.001 are added into the first to ninth channel convolution layers, the first to fourth maximum pooling layers, the global average pooling layer and the full-connection layer to prevent overfitting.
5. The method for visualizing important features of chromosomes based on convolutional neural network as claimed in claim 1, wherein the step S2 adopts a convolutional neural network model preset by cross entropy loss function and stochastic gradient descent method.
6. The method for visualizing important features of chromosomes based on convolutional neural network as claimed in claim 1, wherein the first importance information comprises a plurality of locus values distributed in an array; in step S5, the point location values of each row in the first importance information are compressed in a horizontal average manner, so as to obtain the vertical importance information of each chromosome picture.
7. The visualization method of chromosome important features based on convolutional neural network as claimed in claim 1, wherein the step S6 of performing statistical analysis on each chromosome picture according to the longitudinal importance information of each chromosome picture and the classification result of each chromosome picture specifically comprises:
normalizing the longitudinal importance information of each chromosome picture;
and carrying out average calculation on the normalized longitudinal importance information of the chromosome pictures with the same classification result to obtain the average longitudinal importance information of each type of chromosome picture.
8. The chromosome importance feature visualization method based on the convolutional neural network as claimed in claim 1, wherein the step S6 of performing statistical analysis on each chromosome picture according to the longitudinal importance information of each chromosome picture and the classification result of each chromosome picture specifically comprises:
marking the longitudinal importance information of each chromosome picture, adding a first mark to n point positions with the largest value in the longitudinal importance information, and adding a second mark to the other point positions;
and (3) carrying out mark statistics on the chromosome pictures with the same classification result, and calculating the mark proportion of each point position, wherein the calculation formula is as follows:
Pij=Aij/Biwherein P is the mark proportion of the jth point of the longitudinal importance information of the chromosome picture with all classification results being i, AijThe number of the first mark added to the jth point of the longitudinal importance information of all chromosome images with classification result i, BiThe total number of chromosome images with the classification result i.
9. The chromosome important feature visualization method based on the convolutional neural network as claimed in claim 1, wherein the step S6 of displaying that the convolutional neural network model identifies important strip features of various chromosomes specifically includes:
and comparing the statistical result obtained by the statistical analysis in the step S6 with the pattern graph corresponding to the training data set or the average graph of the training data set in a histogram mode.
10. A chromosome significance signature visualization device based on a convolutional neural network, which is realized by the method of any one of claims 1 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111279350.2A CN113724793B (en) | 2021-11-01 | 2021-11-01 | Chromosome important stripe feature visualization method and device based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111279350.2A CN113724793B (en) | 2021-11-01 | 2021-11-01 | Chromosome important stripe feature visualization method and device based on convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113724793A true CN113724793A (en) | 2021-11-30 |
| CN113724793B CN113724793B (en) | 2022-01-28 |
Family
ID=78686215
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111279350.2A Active CN113724793B (en) | 2021-11-01 | 2021-11-01 | Chromosome important stripe feature visualization method and device based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113724793B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114820596A (en) * | 2022-06-23 | 2022-07-29 | 西湖大学 | Curved chromosome image straightening method based on combined model |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106170558A (en) * | 2013-10-21 | 2016-11-30 | 金圣千 | Utilize the analysis method and device of the biomolecule of oligonucleotide |
| CN109886966A (en) * | 2019-05-09 | 2019-06-14 | 湖南自兴智慧医疗科技有限公司 | A kind of image processing method extracting target signature from ultrasonography |
| CN110689036A (en) * | 2018-07-06 | 2020-01-14 | 塔塔咨询服务有限公司 | Method and system for automatic chromosome classification |
| US20200303038A1 (en) * | 2019-03-19 | 2020-09-24 | The University Of Hong Kong | Variant calling in single molecule sequencing using a convolutional neural network |
| WO2021026242A1 (en) * | 2019-08-05 | 2021-02-11 | doc.ai, Inc. | Accelerated processing of genomic data and streamlined visualization of genomic insights |
| WO2021081257A1 (en) * | 2019-10-22 | 2021-04-29 | Novateur Research Solutions LLC | Artificial intelligence for personalized oncology |
-
2021
- 2021-11-01 CN CN202111279350.2A patent/CN113724793B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106170558A (en) * | 2013-10-21 | 2016-11-30 | 金圣千 | Utilize the analysis method and device of the biomolecule of oligonucleotide |
| CN110689036A (en) * | 2018-07-06 | 2020-01-14 | 塔塔咨询服务有限公司 | Method and system for automatic chromosome classification |
| US20200303038A1 (en) * | 2019-03-19 | 2020-09-24 | The University Of Hong Kong | Variant calling in single molecule sequencing using a convolutional neural network |
| CN109886966A (en) * | 2019-05-09 | 2019-06-14 | 湖南自兴智慧医疗科技有限公司 | A kind of image processing method extracting target signature from ultrasonography |
| WO2021026242A1 (en) * | 2019-08-05 | 2021-02-11 | doc.ai, Inc. | Accelerated processing of genomic data and streamlined visualization of genomic insights |
| WO2021081257A1 (en) * | 2019-10-22 | 2021-04-29 | Novateur Research Solutions LLC | Artificial intelligence for personalized oncology |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114820596A (en) * | 2022-06-23 | 2022-07-29 | 西湖大学 | Curved chromosome image straightening method based on combined model |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113724793B (en) | 2022-01-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP6788264B2 (en) | Facial expression recognition method, facial expression recognition device, computer program and advertisement management system | |
| US20200193232A1 (en) | License plate recognition method, device thereof, and user equipment | |
| EP4075324A1 (en) | Face recognition method and face recognition device | |
| WO2020103417A1 (en) | Bmi evaluation method and device, and computer readable storage medium | |
| CN110222611A (en) | Human skeleton Activity recognition method, system, device based on figure convolutional network | |
| CN109902546A (en) | Face identification method, device and computer-readable medium | |
| CN108288035A (en) | The human motion recognition method of multichannel image Fusion Features based on deep learning | |
| CN109558902A (en) | A kind of fast target detection method | |
| CN107408211A (en) | Method for distinguishing is known again for object | |
| CN111754368A (en) | College teaching evaluation method and college teaching evaluation system based on edge intelligence | |
| CN112464865A (en) | Facial expression recognition method based on pixel and geometric mixed features | |
| CN106295521A (en) | A kind of gender identification method based on multi output convolutional neural networks, device and the equipment of calculating | |
| JP2018165948A (en) | Image recognition device, image recognition method, computer program, and product monitoring system | |
| CN111368672A (en) | Construction method and device for genetic disease facial recognition model | |
| CN113221694B (en) | Action recognition method | |
| CN109377441A (en) | Tongue with privacy protection function is as acquisition method and system | |
| CN113724793B (en) | Chromosome important stripe feature visualization method and device based on convolutional neural network | |
| CN109614990A (en) | A kind of object detecting device | |
| CN114663715B (en) | Medical image quality control and classification model training method and device and computer equipment | |
| CN110399820A (en) | A kind of margin of roads scenery visual identity analysis method | |
| CN107767416A (en) | The recognition methods of pedestrian's direction in a kind of low-resolution image | |
| CN110363072A (en) | Tongue image recognition method, apparatus, computer equipment and computer readable storage medium | |
| CN113610046A (en) | Behavior identification method based on depth video linkage characteristics | |
| CN111709326A (en) | Deep learning-based cascaded granular crop quality analysis method and system | |
| CN111178187A (en) | Face recognition method and device based on convolutional neural network |
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 |