CN110265087A - Chromosome abnormality detection model, its detection system and chromosome abnormality detection method - Google Patents

Abstract

The present invention provides a kind of chromosome abnormality detection system, includes image acquisition unit and non-transitory machine-readable medium.Target chromosome metaphase in cell division image of the image acquisition unit to obtain subject.Non-transitory machine-readable medium is to store program, when program is executed by processing unit to judge whether subject has chromosome abnormality.Whereby, chromosome abnormality detection system of the invention can effectively promote accuracy and the susceptibility of chromosome abnormality detection, and can shorten the assessment time whether subject has chromosome abnormality.

Description

Chromosomal abnormality detection model, detection system thereof, and chromosomal abnormality detection method

technical field

The present invention relates to a medical information analysis model, system and method, in particular to a chromosome abnormality detection model, a chromosome abnormality detection system and a chromosome abnormality detection method.

Background technique

Chromosomal abnormalities are mostly used to screen for genetic diseases, or to detect mutations in cancer cells such as blood cancer and lymphoma. Among them, genetic disease screening is mainly for pregnant women to receive relevant tests during pregnancy. Because the fetus carries the chromosomes from the father's sperm cells and the mother's egg cells through cell meiosis, so every embryo's life process occurs. , it is possible to produce chromosomal mutations in the embryo, and it is necessary to confirm the health status of the fetus by detecting whether the fetal chromosome is abnormal or not.

Chromosomal abnormalities can generally be divided into abnormal chromosome number, abnormal chromosome structure, and abnormal chromosome patchwork. Among them, the abnormal number of chromosomes means that when the germ cells undergo meiosis, if a chromosome nondisjunction occurs, it will lead to abnormal number of chromosomes in sperm or egg cells, and the number of chromosomes will be haploid or polyploid after conception. embryos and give birth to deformed fetuses. Common chromosomal abnormalities include trisomy 21 (Down syndrome), trisomy 18 (Edward's syndrome), and monosomy X (Turner's syndrome). Abnormal chromosome structure is caused by one or more defects or abnormal combination of chromosome structure. The more common chromosomal patchwork abnormalities are 46,XX/47,XX,+21 Down syndrome patchwork, 45,X/46,XX, 45,X/46,XY or 45,X/46,X, i(Xq) is a patchwork of Turner's disease. Generally, it is a patchwork of cells containing some normal chromosomes, and its symptoms are usually milder than a single pure chromosomal abnormality.

A well-known chromosomal abnormality detection method is to take pictures of chromosomes in the middle of cell division, manually arrange them into a karyotype map, and then judge whether there is haploidy or polyploidy in the chromosomes to determine whether there is abnormal chromosome number, and Whether the chromosome has loss, ring, inversion or dislocation to determine whether there is a chromosome structural abnormality, there are great differences in the evaluation results of chromosome abnormality detection among different inspectors, and the process is also cumbersome and time-consuming. Therefore, how to develop a chromosomal abnormality detection system with high accuracy and rapid detection is a technical subject with commercial value.

SUMMARY OF THE INVENTION

In view of this, an object of the present invention is to provide a chromosomal abnormality detection model, a chromosomal abnormality detection method, and a chromosomal abnormality detection system, which can objectively and accurately determine whether a subject has a chromosomal abnormality, and can thereby perform disease classification. and risk assessment.

One aspect of the present invention is to provide a chromosomal abnormality detection model, comprising the following steps of establishing: obtaining a reference database, performing image conversion, performing preliminary classification, performing feature selection, and performing training. The reference database includes a plurality of metaphase images of reference chromosomes. The image conversion step is to use an unsupervised learning method classifier to arrange 23 pairs of chromosomes in the metaphase image of the reference chromosome to obtain a plurality of karyotype images of the reference chromosome. The preliminary classification step is to classify according to the number of chromosomes in the reference chromosome karyotype image. If the number of chromosomes is 46, the number of chromosomes is classified as normal; if the number of chromosomes is greater than or less than 46, it is classified as Abnormal number of chromosomes. The feature selection step is to use the feature selection module to analyze the reference karyotype image to obtain at least one image feature value. The training step is to train the aforementioned at least one image feature value through a convolutional neural network learning classifier to achieve convergence, so as to obtain the chromosomal abnormality detection model, wherein the chromosomal abnormality detection model is used to determine the subject. Whether the test subject has chromosomal structural abnormalities or chromosomal patchwork abnormalities.

According to the aforementioned chromosome abnormality detection model, the unsupervised learning method classifier may be a Generative Adversarial Network (GAN).

According to the aforementioned chromosome abnormality detection model, the at least one image feature value may include chromosome size, chromosome position or chromosome shape.

According to the aforementioned chromosomal abnormality detection model, the convolutional neural network learning classifier may be an Inception-ResNet-v2 convolutional neural network or an Inception V3 convolutional neural network.

Another aspect of the present invention is to provide a chromosomal abnormality detection method, which includes the following steps. A chromosomal abnormality detection model is provided as described in the preceding paragraph. Provides a metaphase image of the subject's target chromosome. The 23 pairs of chromosomes in the metaphase image of the target chromosome are arranged by using the unsupervised learning method classifier to obtain the target chromosome karyotype image. Using the aforementioned chromosomal abnormality detection model to analyze the aforementioned target chromosomal karyotype image to determine whether the subject has a chromosomal abnormality.

According to the aforementioned chromosomal abnormality detection method, the chromosomal abnormality may include abnormal chromosome number, abnormal chromosome structure or abnormal chromosome patchwork type. Preferably, the abnormal number of chromosomes can include that the target chromosome of the subject is haploid or polyploid, and the abnormal chromosome structure can include that the target chromosome of the subject is a chromosomal deletion, a circular chromosome, a chromosomal translocation, a chromosomal inversion or a chromosome repeat.

Another aspect of the present invention is to provide a chromosomal abnormality detection system, including an image capture unit and a non-transitory machine-readable medium. The image acquisition unit is used for acquiring the target chromosome mid-division image of the subject. The non-transitory machine-readable medium is signal-connected to the image capturing unit, wherein the non-transitory machine-readable medium is used for storing a program, and when the foregoing program is executed by the processing unit, it is used to determine whether the subject has a chromosomal abnormality, and the foregoing The program includes a reference database acquisition module, a reference image conversion module, a reference preliminary classification module, a reference feature selection module, a training module, a target image conversion module, a target preliminary classification module, a target feature selection module and a comparison module. The reference database obtaining module is used for obtaining a reference database, and the aforementioned reference database is established from a plurality of reference chromosomes metaphase images. The reference image conversion module uses an unsupervised learning classifier to arrange 23 pairs of chromosomes in the metaphase image of the reference chromosome to obtain a plurality of karyotype images of the reference chromosome. Refer to the preliminary classification module to classify the reference chromosome karyotype images according to the number of reference chromosomes. If the number of reference chromosomes is 46, the number of chromosomes is classified as normal. If the number of reference chromosomes is greater than or less than 46, it is classified Abnormal number of chromosomes. The reference feature selection module is used for analyzing the reference chromosome karyotype image to obtain at least one reference image feature value. The training module is used to train at least one reference image feature value through the convolutional neural network learning classifier to achieve convergence, so as to obtain a chromosome abnormality detection model. The target image conversion module uses an unsupervised learning classifier to arrange 23 pairs of chromosomes in the target chromosome mid-cell division image to obtain the target chromosome karyotype image. The target preliminary classification module is used to classify the target chromosome karyotype image according to the target chromosome number. If the target chromosome number is 46, the number of chromosomes is normal; if the target chromosome number is greater than or less than 46, it is classified as Abnormal number of chromosomes. The target feature selection module is used for analyzing the target karyotype image to obtain at least one target image feature value. The comparison module is used to analyze the target image feature value with the chromosome abnormality detection model to obtain the target image feature value weight data, and judge whether the subject has a chromosome structural abnormality or a chromosome patch abnormality according to the target image feature value weight data. .

According to the aforementioned chromosome abnormality detection system, the unsupervised learning method classifier may be a Generative Adversarial Network (GAN).

According to the aforementioned chromosome abnormality detection system, at least one reference image feature value may include chromosome size, chromosome position or chromosome shape, and at least one target image feature value may include chromosome size, chromosome position or chromosome shape.

According to the aforementioned chromosomal abnormality detection system, the convolutional neural network learning classifier may be an Inception-ResNet-v2 convolutional neural network or an Inception V3 convolutional neural network.

According to the aforementioned chromosomal abnormality detection system, the non-transitory machine-readable medium may further comprise an evaluation module for calculating the risk value of the subject having chromosomal abnormality according to the target image feature value weight data.

Thereby, the chromosome abnormality detection model, the chromosome abnormality detection system and the chromosome abnormality detection method of the present invention automatically convert the target chromosome metaphase image into the target chromosome karyotype image, and use the target feature selection module to analyze the target chromosome karyotype image. Then, the method of obtaining at least one target image feature value can effectively reduce the error caused by the subjective consciousness of different judges in the detection of chromosome abnormality. Furthermore, the chromosomal abnormality detection model with the learning function of deep neural network can not only effectively improve the accuracy and sensitivity of chromosomal abnormality detection, but also greatly shorten the judgment time of chromosomal abnormality, so that the chromosomal abnormality detection model and chromosomal abnormality detection model of the present invention can be used. The abnormality detection system and the chromosomal abnormality detection method are more efficient in chromosomal abnormality detection.

The above summary is intended to provide a simplified summary of the disclosure to provide the reader with a basic understanding of the disclosure. This summary is not an exhaustive overview of the disclosure, and it is not intended to identify key/critical elements of embodiments of the invention or to delineate the scope of the invention.

Description of drawings

In order to make the above-mentioned and other objects, features, advantages and embodiments of the present invention more obvious and easy to understand, the descriptions are as follows in conjunction with the accompanying drawings:

FIG. 1 is a flowchart showing the steps of establishing a chromosomal abnormality detection model according to an embodiment of the present invention;

FIG. 2 is a flowchart showing the steps of a method for detecting chromosomal abnormalities according to another embodiment of the present invention;

3 is a block diagram illustrating a chromosomal abnormality detection system according to still another embodiment of the present invention;

FIG. 4 is a diagram showing the result of converting a target chromosome metaphase image into a target chromosome karyotype image;

5 is a schematic diagram illustrating the architecture of a convolutional neural network learning classifier of a chromosome abnormality detection model according to an embodiment of the present invention;

6 is a schematic diagram illustrating the architecture of a convolutional neural network learning classifier of a chromosome abnormality detection model according to another example of an embodiment of the present invention; and

FIG. 7 is a confusion matrix used by the chromosomal abnormality detection model of the present invention to determine the chromosomal abnormality of a subject.

Detailed ways

Various embodiments of the present invention are discussed in greater detail below. However, this embodiment can be an application of various inventive concepts and can be embodied in various specific scopes. The specific embodiments are for illustrative purposes only, and are not intended to limit the scope of the disclosure.

Please refer to FIG. 1 , which is a flowchart illustrating a step 100 of establishing a chromosomal abnormality detection model according to an embodiment of the present invention. The step 100 of establishing a chromosomal abnormality detection model of the present invention includes steps 110, 120, 130, 140 and 150, and the established chromosomal abnormality detection model can be used to determine whether the subject has abnormal chromosome number, abnormal chromosome structure or Chromosomal patchwork abnormalities.

Step 110 is to obtain a reference database, where the reference database includes a plurality of metaphase images of reference chromosomes. In non-dividing cells, the chromatin is mostly distributed in the nucleus in the state of 30nm to 300nm. When the cell enters the mitotic stage, the chromosomes will begin to gradually and closely arrange. In the metaphase of the cell, the nuclear envelope of the cell disappears completely, and the spindle filaments begin to become clear. The centromere on each chromosome is attached to the spindle wire (or star ray), and the centromere begins to move up and down due to the pulling force of its two poles. Finally, the pulling force of the two poles reaches equilibrium, and the centromere is arranged on the equatorial plate in the center of the cell. Chromosome clarity is at its highest point. Therefore, before obtaining the metaphase image of the reference chromosome, the cells of the reference subject are injected into the metaphase of cell division by administering hormones, and then the specific cells of the reference subject are extracted, and the reference chromosome cells are obtained by staining and microscope observation. Metaphase image.

Step 120 is an image conversion step, using an unsupervised learning classifier to arrange 23 pairs of chromosomes in the metaphase image of the reference chromosome to obtain a plurality of karyotype images of the reference chromosome. The reference chromosome karyotype image is obtained by analyzing, comparing, sorting and numbering the chromosomes according to the chromosome length, centromere position, length and short arm ratio, presence or absence of satellites and other characteristics of the above-mentioned reference chromosome cell division metaphase images. obtained image. The unsupervised learning method classifier may be a Generative Adversarial Network (GAN).

Step 130 is a preliminary classification step, which is to classify according to the number of chromosomes in the reference chromosome karyotype image. If the number of chromosomes is 46, the number of chromosomes is classified as normal; if the number of chromosomes is greater than or less than 46, then the classification is performed. Abnormal number of chromosomes.

Step 140 is a feature selection step, which uses a feature selection module to analyze the reference karyotype image to obtain at least one image feature value. The at least one image feature value may include chromosome size, chromosome position or chromosome shape.

Step 150 is a training step, which is to train the aforementioned at least one image feature value through a convolutional neural network learning classifier to achieve convergence, so as to obtain the chromosome abnormality detection model. The convolutional neural network learning classifier may be an Inception-ResNet-v2 convolutional neural network or an Inception V3 convolutional neural network.

Please refer to FIG. 2 , which shows a flow chart of steps of a method 200 for detecting chromosomal abnormalities according to another embodiment of the present invention. The chromosome abnormality detection method 200 of the present invention includes step 210 , step 220 , step 230 and step 240 .

Step 210 is to provide a chromosome abnormality detection model, and the chromosome abnormality detection model is established through the aforementioned steps 110 to 140 .

Step 220 is to provide the target chromosome metaphase image of the subject. Before obtaining the target chromosome metaphase image, the subject's cells are injected into the metaphase by applying hormones, and then specific cells of the subject are extracted. , and obtained the target chromosome cell division metaphase image by staining and microscopic observation.

Step 230 uses an unsupervised learning method classifier to arrange the 23 pairs of chromosomes in the metaphase image of the target chromosome to obtain a karyotype image of the target chromosome. The target chromosome karyotype image is to analyze, compare, sort and number the chromosomes according to the chromosome length, centromere position, length and short arm ratio, the presence or absence of satellites and other characteristics of the aforementioned target chromosome cell division metaphase image. image obtained after. The unsupervised learning method classifier may be a Generative Adversarial Network (GAN).

Step 240 is to use a chromosome abnormality detection model to analyze the target chromosome karyotype image to determine whether the subject has a chromosome abnormality. The chromosomal abnormality may include abnormal chromosome number, abnormal chromosome structure or abnormal chromosome patchwork type. Preferably, the abnormal number of chromosomes can include that the target chromosome of the subject is haploid or polyploid, and the abnormal chromosome structure can include that the target chromosome of the subject is a chromosomal deletion, a circular chromosome, a chromosomal translocation, a chromosomal inversion or a chromosome repeat.

Thereby, the chromosome abnormality detection model and the chromosome abnormality detection method of the present invention automatically convert the target chromosome metaphase image into the target chromosome karyotype image, and use the feature selection module to analyze the target chromosome karyotype image to obtain at least one image. The method of eigenvalue can effectively reduce the error caused by the subjective consciousness of different judges when detecting chromosomal abnormalities. Furthermore, the chromosomal abnormality detection model with the deep neural network learning function can not only effectively improve the accuracy and sensitivity of chromosomal abnormality detection, but also greatly shorten the judgment time of chromosomal abnormality. Anomaly detection methods are more efficient in the detection of chromosomal abnormalities.

Please refer to FIG. 3 and FIG. 4 again. FIG. 3 shows a block diagram of a chromosome abnormality detection system 300 according to still another embodiment of the present invention. FIG. 4 shows the conversion of the target chromosome metaphase image 610 into the target chromosome karyotype image. 620 result graph. The chromosome abnormality detection system 300 of the present invention includes an image capture unit 400 and a non-transitory machine-readable medium 500 . The chromosomal abnormality detection system 300 can be used to determine whether a subject has an abnormal chromosome number, a chromosome structural abnormality, or a chromosome patchwork abnormality.

The image capturing unit 400 is used for obtaining the target chromosome metaphase image 610 of the subject. The image capturing unit can be an image capturing device matched with a microscope for capturing chromosome images observed by the microscope.

The non-transitory machine-readable medium 500 is signally connected to the image capturing unit 400, wherein the non-transitory machine-readable medium is used for storing a program, and when the aforementioned program is executed by the processing unit, it is used for evaluating whether the subject has a chromosomal abnormality, The aforementioned program includes a reference database acquisition module 510 , a reference image conversion module 520 , a reference preliminary classification module 530 , a reference feature selection module 540 , a training module 550 , a target image conversion module 560 , a target preliminary classification module 570 , and a target feature selection module 580 and an alignment module 590.

The reference database obtaining module 510 is used for obtaining a reference database, and the aforementioned reference database is established from a plurality of reference chromosomes metaphase images.

The reference image conversion module 520 uses an unsupervised learning classifier to arrange 23 pairs of chromosomes in the metaphase image of the reference chromosome to obtain a plurality of karyotype images of the reference chromosome. The unsupervised learning method classifier may be a generative adversarial neural network.

The reference preliminary classification module 530 is used for classifying the reference chromosome karyotype image according to the number of reference chromosomes. If the number of reference chromosomes is 46, the number of chromosomes is classified as normal; if the number of reference chromosomes is greater than or less than 46, the number of chromosomes is classified as abnormal. Preferably, the abnormal number of chromosomes may comprise that the subject's target chromosome is haploid or polyploid.

The reference feature selection module 540 is used to obtain at least one reference image feature value after analyzing the reference chromosome karyotype image. The at least one reference image feature value may include chromosome size, chromosome position or chromosome shape.

The training module 550 is used to train at least one reference image feature value through a convolutional neural network learning classifier to achieve convergence, so as to obtain a chromosome abnormality detection model. The convolutional neural network learning classifier may be an Inception-ResNet-v2 convolutional neural network or an Inception V3 convolutional neural network.

The target image conversion module 560 uses an unsupervised learning classifier to arrange 23 pairs of chromosomes in the target chromosome metaphase image 610 to obtain the target chromosome karyotype image 620 . The unsupervised learning method classifier may be a generative adversarial neural network.

The target preliminary classification module 570 is used for classifying the target chromosome karyotype image according to the target chromosome number. If the number of target chromosomes is 46, the number of chromosomes is classified as normal; if the number of target chromosomes is greater than or less than 46, the number of chromosomes is classified as abnormal. Preferably, the abnormal number of chromosomes may comprise that the subject's target chromosome is haploid or polyploid.

The target feature selection module 580 is configured to analyze the target karyotype image to obtain at least one target image feature value. The at least one target image feature value may include chromosome size, chromosome position or chromosome shape.

The comparison module 590 is used to analyze the target image feature value with the chromosome abnormality detection model to obtain the target image feature value weight data, and determine whether the subject has a chromosome structural abnormality or a chromosome patch type according to the target image feature value weight data abnormal. Preferably, the chromosomal structural abnormality may comprise that the subject's target chromosome is a chromosomal deletion, a circular chromosome, a chromosomal translocation, a chromosomal inversion, or a chromosomal duplication.

In addition, the non-transitory machine-readable medium 500 may further include an evaluation module (not shown) for further calculating the risk value of the subject having chromosomal abnormalities according to the target image feature value weight data.

According to the above-mentioned embodiments, specific test examples are proposed below and described in detail with reference to the accompanying drawings.

<Test example>

1. Reference database

The reference database used in the present invention is the retrospectively delinked clinical content of the subjects collected by China Medical University Hospital (CMUH), which is a clinical trial approved by the Research Ethics Committee of China Medical University and Affiliated Hospitals Program, its number is: CMUH107-REC3-151. The aforementioned reference database includes 30,000 images of the reference chromosome metaphase cell division of the subjects, and the gender of the subjects to which the aforementioned reference chromosome metaphase cell division images belong is not particularly limited, nor is there any particular age range.

2. The chromosome abnormality detection model of the present invention

After the chromosome abnormality detection model of the present invention obtains the reference database, each reference chromosome metaphase cell division image will use the reference image conversion module to convert each reference chromosome metaphase cell division image to the metaphase cell division image of each reference chromosome by an unsupervised learning method classifier. The 23 pairs of chromosomes in the image are arranged to obtain multiple reference chromosome karyotype images.

In detail, the operation of the current deep neural network model requires a large amount of training data (Training Data, ie each reference chromosome metaphase image of the chromosome abnormality detection model of the present invention) to achieve stable convergence and high classification accuracy. If the number of training data is not sufficient, the deep neural network will be over-fitting, which will lead to an excessively high error value of the judgment result, resulting in a low reliability of the deep neural network model. In order to solve the aforementioned problems, the chromosome abnormality detection model of the present invention further includes an image preprocessing step, performing black and white contrast correction on each reference chromosome karyotype image, and normalizing the image values so that the image values are between 0 and 1.

A preliminary classification step is first performed to determine whether the subject has an abnormal number of chromosomes, which is classified according to the number of chromosomes in each reference chromosome karyotype image. If the number of chromosomes is 46, the number of chromosomes is classified as normal; if the number of chromosomes is greater or less than 46, the number of chromosomes is classified as abnormal.

Next, each reference chromosome karyotype image will be analyzed by a feature selection module to obtain at least one image feature value. Specifically, the feature selection module can further distinguish the image feature values of chromosome size, chromosome position or chromosome shape in each reference chromosome karyotype image.

Next, the aforementioned image feature values will be trained by the convolutional neural network learning classifier to achieve convergence, so as to obtain the chromosome abnormality detection model of the present invention. In this test case, the chromosomal abnormality detection model will be applied to determine whether the subject has abnormal chromosome number, chromosomal structure or chromosomal patchwork abnormality. The convolutional neural network learning classifier can be Inception-ResNet-v2 convolutional neural network or Inception V3 convolutional neural network.

Please refer to FIG. 5 , which is a schematic diagram illustrating the structure of the convolutional neural network learning classifier 700 of the chromosome abnormality detection model of the present invention. In the experimental example of FIG. 5 , the convolutional neural network learning classifier 700 is an Inception-ResNet-v2 convolutional neural network, which includes multiple convolutional layers (Convolution), multiple maximum pooling layers (MaxPool), Multiple average pooling layers (AvgPool) and multiple cascade layers (Concat) are used to train and analyze image feature values.

In detail, the Inception-ResNet-v2 convolutional neural network is a large-scale visual recognition convolutional neural network based on the ImageNet visual data database, and the image data in the ImageNet visual data database are all two-dimensional color images, so it is well known The GoogLeNet convolutional neural network model has RGB three-channel filters in its first convolutional layer. However, the original image files of each reference karyotype image are all three-dimensional gray-scale images, so the chromosomal abnormality detection model of the present invention further uses the GoogLeNet convolutional neural network model including RGB three-channel filters through the arithmetic mean method. And convert it into a single channel, and apply Stochastic Gradient Descent (SGD) to the pre-training model neural network of the chromosome abnormality detection model of the present invention to optimize its training process, and the number of training can be 100 ( Epochs) and the gradient descent method using 96Mini-Batch Size, and modulate by changing the initial learning rate (Learning Rates), where the learning rate is to control the weight and bias (bias) changes when training the neural network The important parameter is that the chromosome abnormality detection model of the present invention can further ensure that the loss function can achieve stable convergence by adjusting the value of the learning rate.

In the process of training the image feature value of the chromosome abnormality detection model of the present invention, the image feature value of each reference chromosome karyotype image is processed by two layers of convolution layers and one layer of max pooling layer (MaxPool) to extract the extracted After repeating the aforementioned two-layer convolutional layer and one-layer maximum pooling layer output, use multiple convolutional layers for parallel towers training to complete the primary image feature value. Training (Inception).

After completing the above-mentioned primary training, the image feature values of each reference karyotype image will be processed 10 times (10×), 20 times (20×) and 10 times (10×) at different depths, different levels and different aspects. Residual module training is used to train the image feature values of each reference karyotype image and achieve convergence. In detail, after the Inception-ResNet convolutional neural network has undergone multiple layers of weight operations, each residual module performs different operations and judgments on the image feature values of each standardized medial foot X-ray image data. , resulting in the accumulation of errors. Therefore, the training of the Inception-ResNet convolutional neural network will pull the node operation value of a specific layer back to the input end of the layer for operation again, so as to prevent the convolutional neural network from learning the classifier 700 to the aforementioned The gradient disappearance phenomenon occurs after the multi-layer weight operation training is performed on the image feature values of the image eigenvalues, and the information loss caused by the accumulation of errors can be avoided, and the training efficiency of the convolutional neural network learning classifier 700 can be effectively improved.

After the deep and repeated residual module training is completed, a convolutional layer, an average pooling layer, a global average pooling layer (Global Average Pooling 2D, GloAvePool2D) and a linear rectifier unit training layer (Rectified Linear 2D) will be replaced in sequence. Unit, ReLU) performs final training and processing on the converged image eigenvalues, so as to judge the chromosomal abnormality of the subjects. Among them, the average pooling layer can first calculate the image feature values that have completed the training of the residual module, so as to obtain the average value of each image feature value. Instead of the global average pooling layer, the convolutional neural network can learn the classifier 700 The overall network architecture is regularized to prevent the convolutional neural network learning classifier 700 from overfitting in the training mode that pursues low error, resulting in an excessively high error value of the judgment result. Finally, linear rectification is performed. The unit training layer further activates the image feature values after training, and outputs the target image feature value weight data 701 for subsequent comparison and analysis. The aforementioned linear rectification unit training layer can prevent the target image feature value weight data 701 output by the foot deformity detection model from approaching zero or approaching infinity, so as to facilitate the subsequent comparison steps, thereby improving the chromosomal abnormality detection of the present invention. The judgment accuracy of the model.

Next, the chromosomal abnormality judgment results of the aforementioned subjects will be further integrated into the reference database to optimize the chromosomal abnormality detection model of the present invention, thereby further improving the training effect and judgment accuracy of the chromosomal abnormality detection model of the present invention.

Please refer to FIG. 6 again, which shows a schematic diagram of the structure of the convolutional neural network learning classifier 800 of the chromosome abnormality detection model of the present invention. In the test example of FIG. 6 , the convolutional neural network learning classifier 800 is an Inception V3 convolutional neural network, which includes multiple convolutional layers (Convolution), multiple average pooling layers (AvgPool), multiple maximum Pooling layer (MaxPool) and multiple cascade layers (Concat), and use dropout layer (Dropout), fully connected layer (Fullyconnected) and normalization layer (Softmax) to solve the problem of over-fitting in machine learning, to solve the problem of overfitting in machine learning. Image feature values for training and analysis.

A single-layer neural network will cause overfitting in machine learning due to too many parameters. The InceptionV3 convolutional neural network is a factorization of the convolutional network based on the large filter size. The parallel parameter reduction can not only solve the problem of overfitting, but also increase the network depth by increasing the network depth. The numbers then more closely approximate the mathematical model that was originally intended to be approximated.

In the process of training the image feature value by the chromosome abnormality detection model of the present invention, the image feature value of each reference chromosome karyotype image is respectively subjected to one layer of average pooling layer and one layer of convolution layer; five layers of convolution layer; three layers Layer convolution layer; after the operation of one layer of convolution layer, the feature matrix values of each group of operations are superimposed in cascade layers. After that, it is repeated twice to perform one layer of average pooling layer and one layer of convolution layer; five layers of convolution layer; three layers of convolution layer; Cascading layer stacking. Then perform one layer of maximum pooling layer; three layers of convolution layer; after one layer of convolution layer operation, the feature matrix values of each group of operations are superimposed in cascade layers. After that, repeat the operation for one layer of average pooling layer and one layer of convolution layer, five layers of convolution layer, three layers of convolution layer and one layer of convolution layer, respectively. Cascading layer stacking. Then perform one layer of average pooling layer, two layers of convolution layer, one layer of fully connected layer and one layer of normalization layer operations, and the feature matrix value of the operation is repeated twice for one layer of average pooling layer and one layer of volume respectively. Stacking layer; three-layer convolutional layer and one-layer concatenated layer; two-layer convolutional layer and one-layer concatenated layer; after one-layer convolutional layer operation, the feature matrix values of each group of operations are superimposed in a concatenated layer. Finally, after performing one layer of average pooling layer, one layer of discarding layer, one layer of fully connected layer and one layer of normalization layer, output target image feature value weight data 801 to obtain a trained chromosome abnormality detection model.

Next, the chromosomal abnormality judgment results of the aforementioned subjects will be further integrated into the reference database to optimize the chromosomal abnormality detection model of the present invention, thereby further improving the training effect and judgment accuracy of the chromosomal abnormality detection model of the present invention.

Please refer to FIG. 7 again, which is a confusion matrix used by the chromosomal abnormality detection model of the present invention to determine the chromosomal abnormality of a subject. In the test example of FIG. 7 , the convolutional neural network learning classifier for establishing the chromosome abnormality detection model is the convolutional neural network learning classifier 800 shown in FIG. Divided into normal and abnormal. The horizontal axis is the predicted label, and the vertical axis is the actual label. The confusion matrix can be divided into True Positive (TP), True Negative (TN), False Positive (FP) and False Negative (False Negative, FN), and calculate the accuracy, sensitivity, specificity, positive predictive value and negative predictive value of the chromosomal abnormality detection model of the present invention according to the data of TP, TN, FP and FN. The correct rate is calculated as (TP+TN)/(TP+FP+TN+FN), the sensitivity is calculated as TP/(TP+FN), and the specificity is calculated as TN/(TN+FP), The positive predictive value was calculated as TP/(TP+FP), and the negative predictive value was calculated as TN/(FN+TN).

As shown in Figure 7, the number of subjects in the TP block is 206, the number of subjects in the TN block is 201, the number of subjects in the FP block is 3, and the number of subjects in the FN block is 3. The number is 0 people. After calculation, the prediction results of the chromosomal abnormality detection model of the present invention for judging the chromosomal abnormality of the subject are shown in Table 1.

It can be seen from the above results that the chromosomal abnormality detection model of the present invention can be used to accurately determine whether a subject has a chromosomal abnormality, and the chromosomal abnormality can include abnormal chromosome number, abnormal chromosome structure, and abnormal chromosome patchwork type.

Thereby, the chromosomal abnormality detection system of the present invention can effectively improve the accuracy and sensitivity of chromosomal abnormality detection, and can shorten the evaluation time of whether a subject has chromosomal abnormality. It can be completed in 1 second, making it more widely used.

Although the present invention has been disclosed as above in embodiments, it is not intended to limit the present invention. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, this The protection scope of the invention shall be determined by the claims.

Claims (13)

1. a kind of chromosome abnormality detection model, which is characterized in that include following set-up step:

It obtains referring to database, wherein described include multiple reference chromosome cell division mid-term images referring to database；

Video conversion step is carried out, is by the multiple using non-supervisory formula learning method classifier referring to chromosome cell division 23 pairs of chromosomes arrange in mid-term image, multiple referring to karyotype image to obtain；

Preliminary classification step is carried out, is divided according to the multiple chromosome item number referring in karyotype image It is normal to be classified as chromosome number if the chromosome item number is 46 for class, if the chromosome item number is more than or less than 46 Item is then classified as numerical abnormalities of chromosomes；

Feature Selection step is carried out, is after analyzing the multiple reference karyotype image using characteristic selecting module to obtain To at least one image feature value；And

It is trained step, is to instruct at least one described image feature value by convolutional Neural network Study strategies and methods Practice and reach convergence, to obtain the chromosome abnormality detection model, wherein the chromosome abnormality detection model is to sentence Whether disconnected subject there is chromosomal structural abnormality or chromosome to piece together type exception.

2. chromosome abnormality detection model as described in claim 1, which is characterized in that the non-supervisory formula learning method classifier Make a living into confrontation neural network.

3. chromosome abnormality detection model as described in claim 1, which is characterized in that at least one described image feature value packet Size containing chromosome, chromosome location or chromosome form.

4. chromosome abnormality detection model as described in claim 1, which is characterized in that convolutional Neural network learning classification Device is Inception-ResNet-v2 convolutional Neural network or Inception V3 convolutional Neural network.

5. a kind of chromosome abnormality detection method, characterized by comprising:

Chromosome abnormality detection model as described in claim 1 is provided；

The target chromosome metaphase in cell division image of subject is provided；

Using the non-supervisory formula learning method classifier by 23 pairs of chromosomes in the target chromosome metaphase in cell division image It is arranged, to obtain target chromosome caryogram image；And

The target chromosome caryogram image is analyzed using the chromosome abnormality detection model, whether to judge the subject With chromosome abnormality.

6. chromosome abnormality detection method as claimed in claim 5, which is characterized in that the chromosome abnormality includes chromosome Numerical abnormality, chromosomal structural abnormality or chromosome piece together type exception.

7. chromosome abnormality detection method as claimed in claim 6, which is characterized in that the numerical abnormalities of chromosomes includes institute The target chromosome for stating subject is monoploid or polyploid.

8. chromosome abnormality detection method as claimed in claim 6, which is characterized in that the chromosomal structural abnormality includes institute The target chromosome for stating subject is chromosome deficiency, circular chromosome, chromosome translocation, chromosome reverses or dyeing weight It is multiple.

9. a kind of chromosome abnormality detection system, characterized by comprising:

Image acquisition unit, to obtain the target chromosome metaphase in cell division image of subject；And

Non-transitory machine-readable medium, signal connects the image acquisition unit, wherein the machine readable matchmaker of the non-transitory Body is to judge whether the subject has chromosome different when described program is executed by processing unit to store program Often, and described program includes:

Module is obtained referring to database, to obtain referring to database, and described referring to database is by multiple reference chromosomes Metaphase in cell division image is established；

It is by the multiple using non-supervisory formula learning method classifier referring to chromosome cell division referring to video conversion module 23 pairs of chromosomes arrange in mid-term image, multiple referring to karyotype image to obtain；

Referring to preliminary classification module, to be divided the multiple referring to karyotype image foundation referring to chromosome item number Class referring to chromosome item number is 46 if described, and it is normal to be classified as chromosome number, if it is described referring to chromosome item number be greater than Or less than 46, then it is classified as numerical abnormalities of chromosomes；

Reference feature chooses module, to analyze after the multiple image referring to karyotype to obtain at least one reference shadow As characteristic value；

Training module, to by it is described at least one referring to image feature value by convolutional Neural network Study strategies and methods training reach To convergence, to obtain chromosome abnormality detection model；

Target image conversion module is to utilize the non-supervisory formula learning method classifier by the target coloration somatic cell division 23 pairs of chromosomes arrange in mid-term image, to obtain target chromosome caryogram image；

Target preliminary classification module, the target chromosome caryogram image to be classified according to target chromosome item number, If the target chromosome item number is 46, it is normal to be classified as chromosome number, if the target chromosome item number be greater than or Less than 46, then numerical abnormalities of chromosomes is classified as；

Target signature chooses module, be to analyze after the target chromosome caryogram image at least one target image Characteristic value；And

Comparison module is being analyzed the target image characteristic value with the chromosome abnormality detection model to obtain Target image characteristic value weighted data, and judge whether the subject has according to the target image characteristic value weighted data Chromosomal structural abnormality or chromosome piece together type exception.

10. chromosome abnormality detection system as claimed in claim 9, which is characterized in that the non-supervisory formula learning method classification Device makes a living into confrontation neural network.

11. chromosome abnormality detection system as claimed in claim 9, which is characterized in that at least one described reference image is special Value indicative includes chromosome size, chromosome location or chromosome form, at least one described target image characteristic value includes dyeing Body size, chromosome location or chromosome form.

12. chromosome abnormality detection system as claimed in claim 9, which is characterized in that the convolutional Neural network study point Class device is Inception-ResNet-v2 convolutional Neural network or Inception V3 convolutional Neural network.

13. chromosome abnormality detection system as claimed in claim 9, which is characterized in that the machine readable matchmaker of non-transitory Body also includes:

Evaluation module has the chromosome different to calculate the subject according to the target image characteristic value weighted data Normal value-at-risk.