1. A method for constructing a chromosome abnormality prediction model based on fetal ultrasound image characteristic omics comprises the following steps:

respectively obtaining three ROI target area images in an ultrasonic image of a fetus with normal chromosomes and three ROI target area images in an ultrasonic image of a fetus with abnormal chromosomes, wherein the three ROI target area images comprise a neck back area, a face area and a midbrain area of the fetus;

respectively extracting the features of the three ROI target area images;

fusing the extracted features of the three ROI target area images;

inputting the result into a classifier to obtain a classification result that the fetus in the ultrasonic image is normal or abnormal;

comparing the obtained classification result with the manual classification result of the doctor, and optimizing a classifier according to the comparison result;

obtaining a chromosome abnormality prediction model.

2. The method for constructing the chromosome abnormality prediction model according to claim 1, characterized in that three ROI target region images in an ultrasound image of a fetus are obtained by using an ROI target region detection model;

the construction method of the ROI target region detection model comprises the following steps:

acquiring an ultrasonic image of a fetus;

extracting a ROI area from the ultrasonic image;

comparing the obtained ROI with a fetal retrocervical region, a facial region and a midbrain region manually defined by a doctor respectively to generate a loss value, performing back propagation, and performing ROI region extraction optimization;

and obtaining an ROI target region detection model.

3. The method for constructing the chromosome abnormality prediction model according to claim 1, wherein the extraction of the features of the three images of the ROI target region can be realized by one or more of the following methods: VGG, inclusion, Xception, MobileNet, AlexNet, LeNet, ZF _ Net, ResNet, ResNeXt, ResNeSt; the classifier may employ one or more of the following classification models: KNN, decision tree, random forest, SVM, logistic regression, Ensemble-Boosting, Ensemble-Bagging.

4. The method for constructing a chromosome abnormality prediction model according to claim 1, wherein the fusing the features of the extracted three ROI target region images further comprises fusing the features of the extracted three ROI target region images and clinical information features.

5. The method for constructing the chromosome abnormality prediction model according to claim 4, wherein the fused fusion function is F = W [ (E) ]_n(i_n), E_h(i_h), E_f(i_f), E_c(i_c) B) where i) is present_n、i_h、i_f、i_cRespectively representing a retrocervical region, a facial region, a midbrain region and clinical information, E representing a feature extraction module, b representing a bias term, and W representing a weight.

6. A chromosome abnormality risk diagnosis device based on fetal ultrasound image characterimics, the device comprising: a memory and a processor;

the memory is to store program instructions;

the processor is configured to invoke program instructions that, when executed, are configured to:

acquiring an ultrasonic image of a fetus of a sample to be detected or acquiring an ultrasonic image and clinical information of the fetus of the sample to be detected;

inputting the data into a chromosome abnormality prediction model, wherein the chromosome abnormality prediction model is constructed by adopting the construction method of the chromosome abnormality prediction model according to any one of claims 1 to 5;

and obtaining the classification result that the sample to be detected is normal chromosome or abnormal chromosome.

7. The chromosome abnormality risk diagnosis device according to claim 6, wherein an ultrasound image of a fetus of a sample to be measured is acquired and input to the fetus ultrasound image standard judgment model to obtain a classification result of whether the ultrasound image is standard or not;

when the classification result is nonstandard, stopping the prediction of the chromosome abnormality risk, and when the classification result is standard, inputting the ultrasound image of the fetus of the sample to be detected or the ultrasound image and the clinical information of the fetus of the sample to be detected into a chromosome abnormality prediction model to obtain the classification result that the sample to be detected is normal chromosome or abnormal chromosome;

and when the sample to be detected is abnormal chromosome, stopping the prediction of the abnormal chromosome risk, and when the result is abnormal chromosome, inputting the ultrasonic image of the fetus, which is obtained from the sample to be detected, into the ultrasonic image of the fetus, namely the abnormal chromosome multi-classification model, so as to obtain the classification result of abnormal chromosome diseases.

8. A chromosome abnormality risk diagnosis system based on fetal ultrasound image characteristic omics comprises:

the acquisition unit is used for acquiring an ultrasonic image of a fetus of a sample to be detected or an ultrasonic image and clinical information of the fetus;

a processing unit, configured to input the ultrasound image of the fetus of the sample to be tested or the ultrasound image of the fetus and clinical information into a chromosome abnormality prediction model, where the chromosome abnormality prediction model is constructed by using the construction method of the chromosome abnormality prediction model according to any one of claims 1 to 5;

and the display unit is used for outputting the classification result that the sample to be detected is normal chromosome or abnormal chromosome.

9. The system for diagnosing risk of chromosomal abnormality according to claim 8, further comprising a preprocessing unit for obtaining a classification result whether the ultrasound image is standard or not; the chromosome abnormality risk diagnosis system further comprises a classification unit for outputting a classification result of the chromosome abnormality diseases.

10. A computer-readable storage medium, on which a computer program is stored, wherein the computer program, when executed by a processor, implements the method for constructing a chromosome abnormality prediction model based on fetal ultrasound image characterics according to any one of claims 1 to 5.

Background

Chromosomal abnormalities include chromosome aberrations such as trisomy 21, trisomy 18, trisomy 13, and sex chromosome abnormalities, and particularly Down Syndrome (DS), also known as trisomy 21, is the most common chromosomal aneuploidy abnormality, and the risk of developing the disease is increased in pregnant women of advanced age. The fetal chromosomal abnormality disease not only brings heavy burden to families of children patients, but also brings burden to society, and no effective treatment method for the fetal chromosomal abnormality disease exists at present, so that the prenatal screening and diagnosis level is improved, and the screening strength is increased, thereby having great significance for preventing birth defects. In recent years, the number of elderly women in childbirth has increased dramatically, and prenatal screening for fetal chromosomal abnormalities has presented new challenges. When the chromosome abnormality of the fetus is found by screening in the middle of pregnancy, the optimal intervention time is missed, the early screening of the fetus with the chromosome abnormality is an important link for realizing the early intervention, the screening time of the birth defect can be shifted forward, and an early informed selection opportunity is provided for the pregnant woman.

The ultrasonic image characteristic omics technology can extract massive quantitative image information from an ultrasonic image in a high-flux manner, and the massive image data information is processed, predicted and analyzed by constructing a big data model, so that doctors are assisted to make more accurate judgment on diseases. In recent years, the imaging technology is widely applied to the ultrasound field, such as identifying ovarian tumor from ultrasound image, classifying the local and diffuse liver disease, and the like. However, an intelligent diagnosis system constructed for clinical fetal ultrasound images of chromosome abnormalities does not exist at present.

Disclosure of Invention

The research develops a chromosome abnormality prediction model based on fetal ultrasound image characteristic omics for the first time, and the model can be used for predicting the risk of chromosome abnormality and assisting doctors in making more accurate judgment and treatment decisions on diseases.

A method for constructing a chromosome abnormality prediction model based on fetal ultrasound image characteristic omics comprises the following steps:

respectively obtaining three ROI target area images in an ultrasonic image of a fetus with normal chromosomes and three ROI target area images in an ultrasonic image of a fetus with abnormal chromosomes, wherein the three ROI target area images comprise a neck back area, a face area and a midbrain area of the fetus;

respectively extracting the features of the three ROI target area images;

fusing the extracted features of the three ROI target area images;

inputting the result into a classifier to obtain a classification result that the fetus in the ultrasonic image is normal or abnormal;

comparing the obtained classification result with the manual classification result of the doctor, and optimizing a classifier according to the comparison result;

obtaining a chromosome abnormality prediction model.

Chromosomal abnormalities are also known as chromosomal disorders and refer to a large group of diseases caused by various chromosomal abnormalities. According to the nature of the chromosomal abnormality, there are two major categories, chromosome number abnormality and chromosome structure abnormality.

Chromosomal numerical abnormalities include trisomy 21, trisomy 18, trisomy 13, trisomy 8, trisomy 9, klinfelter, tunner, trisomy X, XYY, sexual development disorders, and the like.

Chromosomal structural abnormalities include chromosomal deletions, chromosomal duplications, chromosomal translocations, chromosomal inversions, circular chromosomes, isochromosomes, chromosomal insertions, small extra-marker chromosomes, and the like.

Preferably, the fetal cervical region comprises structures such as a transparent cervical collar layer, a skin behind the neck and the like of the fetus; the face area of the fetus needs to comprise structures of forehead, nasal bone, maxilla, mandible and the like of the fetus; the midbrain region includes structures of the fetal mesencephalon, brainstem, fourth ventricle, medulla oblongata, etc.

Further, three ROI target area images in the ultrasonic image of the fetus are obtained by adopting the ROI target area detection model.

Further, the construction method of the ROI target region detection model comprises the following steps:

acquiring an ultrasonic image of a fetus;

extracting a ROI area from the ultrasonic image;

comparing the obtained ROI with a fetal retrocervical region, a facial region and a midbrain region manually defined by a doctor respectively to generate a loss value, performing back propagation, and performing ROI region extraction optimization;

and obtaining an ROI target region detection model.

Further, the ROI area is preliminarily extracted from the ultrasonic image through a fast-RCNN model and the like.

Further, the retrocervical region, the facial region and the midbrain region of the fetus manually defined by the doctor are defined by the doctor in the image processing software, and optionally, the point selected for the first time is set as the center in the definition.

The image processing software may be OPENCV in which the ROI region is delineated by a data structure Rect.

Further, the extraction of the features of the three ROI target region images may employ one or more of the following methods: VGG, Inception V3, Xception, MobileNet, AlexNet, LeNet, ZF _ Net, ResNet.

Further, the classifier may employ one or more of the following classification models: KNN, Bayesian, Decision Tree (Decision Tree), Random Forest (Random Forest), SVM, Logistic regression, Ensemble-Boosting, Ensemble-Bagging.

Decision trees (Decision trees) are generally the cornerstones of other algorithms, different types of data can be processed simultaneously, and the model is easier to understand the influence degree of different attributes on results.

Random Forest (Random Forest) is a Random integration of decision trees, and the model is suitable for the time when the data dimension is not too high and the accuracy is high.

The Ensemble-Boosting searches for a classifier capable of solving the current error each time, and finally performs weight summation, and the model has self-contained feature selection, finds effective features and is convenient for understanding high-dimensional data.

The Ensemble-Bagging trains a plurality of weak classifiers to vote and solve, and a training set is randomly selected, so that overfitting is avoided.

The fusion comprises different fusion modes, one is the fusion carried out at the feature level, and is called front-end fusion or feature fusion; another is fusion at the outcome level, called back-end fusion or outcome fusion.

The fusion comprises the steps of carrying out front-end fusion or rear-end fusion on the image features of the ROI target region, wherein the front-end fusion is carried out on the image features of the ROI target region, and then the image features are sent into a classifier; and the rear-end fusion is to respectively send the three ROI target region image characteristics into respective classifiers, and further fuse the results of the three ROI target region image classifiers.

Further, the fusion function of fusing the features of the extracted three ROI target region images is F = W (concat (E)_n(i_n), E_h(i_h), E_f(i_f) B) where i) is present_n、i_h、i_fRespectively representing a retrocervical region, a face region and a midbrain region, E representing a feature extraction module, b representing a bias term and W representing a weight.

Further, the fusion of the extracted features of the three ROI target area images also comprises the fusion of the extracted features of the three ROI target area images and clinical information features; optionalThe fusion function is F = W (concat (E)_n(i_n), E_h(i_h), E_f(i_f), E_c(i_c) B) where i) is present_n、i_h、i_f、i_cRespectively representing a retrocervical region, a facial region, a midbrain region and clinical information, E representing a feature extraction module, b representing a bias term, and W representing a weight.

Another object of the present invention is to provide a chromosome abnormality risk diagnosis apparatus based on fetal ultrasound image characterics, the apparatus comprising: a memory and a processor;

the memory is to store program instructions;

the processor is configured to invoke program instructions that, when executed, are configured to:

acquiring an ultrasonic image of a fetus of a sample to be detected or acquiring an ultrasonic image and clinical information of the fetus of the sample to be detected;

inputting the chromosome abnormality prediction model into a chromosome abnormality prediction model, wherein the chromosome abnormality prediction model is constructed by adopting the construction method of the chromosome abnormality prediction model;

and obtaining the classification result that the sample to be detected is normal chromosome or abnormal chromosome.

Further, the processor is configured to invoke program instructions that, when executed, are configured to:

acquiring an ultrasonic image of a fetus of a sample to be detected or an ultrasonic image and clinical information of the fetus;

inputting the ultrasonic image of the fetus of the sample to be detected into a fetus ultrasonic image standard judgment model to obtain a classification result of whether the ultrasonic image is standard or not; when the classification result is nonstandard, stopping the prediction of the chromosome abnormality risk, and when the classification result is standard, inputting the ultrasonic image of the fetus of the sample to be detected or the ultrasonic image and the clinical information of the fetus of the sample to be detected into a chromosome abnormality prediction model;

and obtaining the classification result that the sample to be detected is normal chromosome or abnormal chromosome.

Further, the processor is configured to invoke program instructions that, when executed, are configured to:

acquiring an ultrasonic image of a fetus of a sample to be detected or an ultrasonic image and clinical information of the fetus;

inputting an ultrasonic image of a fetus, which is obtained from a sample to be detected, into a fetus image standard judgment model to obtain a classification result of whether the ultrasonic image is standard or not;

when the classification result is nonstandard, stopping the prediction of the chromosome abnormality risk, and when the classification result is standard, inputting the ultrasound image of the fetus of the sample to be detected or the ultrasound image and the clinical information of the fetus of the sample to be detected into a chromosome abnormality prediction model to obtain the classification result that the sample to be detected is normal chromosome or abnormal chromosome;

and when the sample to be detected is abnormal chromosome, stopping the prediction of the abnormal chromosome risk, and when the result is abnormal chromosome, inputting the ultrasonic image of the fetus, which is obtained from the sample to be detected, into the ultrasonic image of the fetus, namely the abnormal chromosome multi-classification model, so as to obtain the classification result of abnormal chromosome diseases.

Optionally, the chromosome abnormality disease of the sample as the classification result is one of the following diseases: 21-trisomy syndrome, 18-trisomy syndrome, 13-trisomy syndrome, 8-trisomy syndrome, 9-trisomy syndrome, klinfelter syndrome, tunner syndrome, trisomy X syndrome, XYY syndrome, sexual development disease, and the like;

optionally, the chromosome abnormality disease of the sample as the classification result is one of the following diseases: chromosomal deletions, chromosomal duplications, chromosomal translocations, chromosomal inversions, circular chromosomes, isochromosomes, chromosomal insertions, small extra-marker chromosomes, and the like.

Optionally, the chromosome abnormality disease of the sample as the classification result is one of the following diseases: 21-trisomy syndrome, 18-trisomy syndrome, 13-trisomy syndrome, 8-trisomy mosaic syndrome, 9-trisomy mosaic syndrome, klinfelter syndrome, tunner syndrome, trisomy X syndrome, XYY syndrome, sexual development disease, chromosome deletion, chromosome duplication, chromosome translocation, chromosome inversion, circular chromosomes, isoarm chromosomes, chromosome insertions, small extra-marker chromosomes, and the like.

Another object of the present invention is to provide a system for diagnosing a risk of a chromosomal abnormality based on fetal ultrasound image characterics, comprising:

the acquisition unit is used for acquiring an ultrasonic image of a fetus of a sample to be detected or an ultrasonic image and clinical information of the fetus;

the processing unit is used for inputting the ultrasonic image of the fetus of the sample to be detected or the ultrasonic image and the clinical information of the fetus into a chromosome abnormality prediction model, and the chromosome abnormality prediction model is constructed by adopting the construction method;

and the display unit is used for outputting the classification result that the sample to be detected is normal chromosome or abnormal chromosome.

Further, a chromosome abnormality risk diagnosis system based on fetal ultrasound image characteristic omics includes:

the acquisition unit is used for acquiring an ultrasonic image of a fetus of a sample to be detected or an ultrasonic image and clinical information of the fetus;

the preprocessing unit is used for obtaining a classification result whether the ultrasonic image is standard or not;

the processing unit is used for inputting the ultrasonic image of the fetus of the sample to be detected or the ultrasonic image and the clinical information of the fetus into a chromosome abnormality prediction model, and the chromosome abnormality prediction model is constructed by adopting the construction method;

and the display unit is used for outputting the classification result that the sample to be detected is normal chromosome or abnormal chromosome.

Further, a chromosome abnormality risk diagnosis system based on fetal ultrasound image characteristic omics includes:

the acquisition unit is used for acquiring an ultrasonic image of a fetus of a sample to be detected or an ultrasonic image and clinical information of the fetus;

the preprocessing unit is used for obtaining a classification result whether the ultrasonic image is standard or not;

the processing unit is used for inputting the ultrasonic image of the fetus of the sample to be detected or the ultrasonic image and the clinical information of the fetus into a chromosome abnormality prediction model, and the chromosome abnormality prediction model is constructed by adopting the construction method;

the display unit is used for outputting the classification result that the sample to be detected is normal chromosome or abnormal chromosome;

and the classification unit is used for outputting the classification result of the abnormal chromosome diseases.

A computer-readable storage medium, on which a computer program is stored, wherein the computer program, when being executed by a processor, implements the above method for constructing a chromosome abnormality prediction model based on fetal ultrasound image characterics.

The application has the advantages that:

1. according to the characteristics of the chromosome abnormality ultrasonic image, the specific ROI target region (the fetal neck back region, the facial region and the midbrain region) image is processed by combining the disease ultrasonic image key part, so that on one hand, the detection accuracy of the model is improved, unnecessary ultrasonic image information is filtered, and on the other hand, the calculation power of the model is reduced;

2. in order to obtain a more accurate prediction result, the image data feature is processed, and meanwhile, clinical information is combined, the three ROI target area image features and the clinical information features are fused, so that an accurate classification result is obtained;

3. in order to solve the problems of misdiagnosis and the like caused by the existing clinical fetus ultrasonic image acquisition nonstandard, the invention judges whether the image is standard or not by a fetus image standard judgment model before risk diagnosis, thereby avoiding misdiagnosis;

4. in order to provide a more accurate prediction result for a clinician, the invention also provides a fetus ultrasonic image chromosome abnormality multi-classification model, when the output result of the chromosome abnormality prediction model is chromosome abnormality, the fetus ultrasonic image of the sample to be detected is input into the fetus ultrasonic image chromosome abnormality multi-classification model, and the chromosome abnormality disease classification result is obtained.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

Fig. 1 is a schematic flow chart for constructing a chromosome abnormality prediction model based on fetal ultrasound image characterics provided by an embodiment of the present invention;

FIG. 2 is a schematic flowchart of the construction of a constant ROI target region detection model provided by the embodiment of the invention;

fig. 3 is a schematic flowchart of chromosome abnormality risk diagnosis based on fetal ultrasound image characterics provided by an embodiment of the present invention;

fig. 4 is a schematic diagram of a chromosome abnormality risk diagnosis system based on fetal ultrasound image characterics provided by an embodiment of the present invention;

fig. 5 is a schematic diagram of a chromosome abnormality risk diagnosis device based on fetal ultrasound image characterics provided by an embodiment of the present invention;

FIG. 6 is an image of three ROI target regions of a fetus with chromosome abnormality defined by a doctor in image processing software, wherein A is 21-the facial region of a trisomy fetus, B: 21-midbrain region of trisomy fetus, C: 21-retrocervical region of trisomy;

FIG. 7 is an image of three ROI target areas of a normal chromosome fetus defined by a doctor in image processing software, wherein A is a facial area of the normal chromosome fetus, B: chromosomal normal fetal midbrain region, C: chromosome normal fetal retrocervical region.

Detailed Description

In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention.

In some of the flows described in the present specification and claims and in the above-described figures, a number of operations are included that occur in a particular order, but it should be clearly understood that these operations may be performed out of order or in parallel as they occur herein, with the order of the operations, e.g., S101, S102, etc., merely being used to distinguish between various operations, and the order of the operations itself does not represent any order of performance. Additionally, the flows may include more or fewer operations, and the operations may be performed sequentially or in parallel. It should be noted that, the descriptions of "first", "second", etc. in this document are used for distinguishing different messages, devices, modules, etc., and do not represent a sequential order, nor limit the types of "first" and "second" to be different.

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, 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 invention.

Fig. 1 is a schematic flow chart for constructing a chromosome abnormality prediction model based on imaging omics according to an embodiment of the present invention, specifically, the method includes the following steps:

s101: respectively acquiring three ROI target area images in an ultrasonic image of a fetus with normal chromosomes and three ROI target area images in an ultrasonic image of a fetus with abnormal chromosomes, wherein the three ROI target area images comprise a retrocervical region, a facial region and a midbrain region of the fetus;

in one embodiment, the fetal cervical region includes structures such as the transparent cervical collar layer, the skin behind the neck, etc. of the fetus; the face area of the fetus needs to comprise structures of forehead, nasal bone, maxilla, mandible and the like of the fetus; the midbrain region includes structures of the fetal mesencephalon, brainstem, fourth ventricle, medulla oblongata, etc.

In one embodiment, three ROI target area images in an ultrasound image of a fetus are acquired using an ROI target area detection model.

In one embodiment, the method for constructing the ROI target region detection model includes:

acquiring an ultrasonic image of a normal fetus;

extracting a ROI area from the ultrasonic image;

comparing the obtained ROI with a fetal retrocervical region, a facial region and a midbrain region manually defined by a doctor respectively to generate a loss value, performing back propagation, and performing ROI region extraction optimization;

and obtaining a chromosome abnormality ROI detection model.

In one embodiment, the retrocervical region, the facial region and the midbrain region of the fetus in the ultrasound images of the chromosome normal fetus and chromosome abnormal fetus manually delineated by the physician are delineated by the physician in the image processing software, see fig. 6 and 7.

In one embodiment, ROI regions are extracted by the physician using the data structure Rect in OPENCV, the ROI regions of each ultrasound image including the fetal retrocervical region, the facial region, and the midbrain region, respectively.

In one embodiment, the physician defines the ROI area by centering on the first selected point.

In one embodiment, the fetal ultrasound image is normalized and then the retrocervical region, facial region and midbrain region of the image are manually delineated by a skilled physician.

S102: respectively extracting the features of the three ROI target area images;

in one embodiment, the extraction of the features of the three ROI target region images can be modeled by one or more of the following methods: VGG, Inception V3, Xception, MobileNet, AlexNet, LeNet, ZF _ Net, ResNet.

S103: fusing the extracted features of the three ROI target area images;

in one embodiment, fusing the extracted features of the three ROI target region images further includes fusing the extracted features of the three ROI target region images and the clinical information features; optionally, the fusion function is F = W (concat (E)_n(i_n), E_h(i_h), E_f(i_f), E_c(i_c) B) where i) is present_n、i_h、i_f、i_cRespectively representing a retrocervical region, a facial region, a midbrain region and clinical information, E representing a feature extraction module, b representing a bias term, and W representing a weight.

S104: inputting the result into a classifier to obtain a classification result that the fetus in the ultrasonic image is normal or abnormal;

in one embodiment, the classifier may employ one or more of the following classification models: KNN, Bayesian, decision tree, random forest, SVM, logistic regression, Ensemble-Boosting, Ensemble-Bagging.

S105: comparing the obtained classification result with the manual classification result of the doctor, and optimizing the model according to the comparison result;

s106: obtaining a chromosome abnormality prediction model.

In one embodiment, the obtained classification result is compared with the manual classification result of the doctor to generate a loss value, and the model is optimized through back propagation.

In one embodiment, the data is from Beijing women's obstetrical Hospital affiliated with the university of capital medicine of Beijing, China, and pregnancies 11-13 were selected from our database⁺⁶Two-dimensional (2D) image of the median sagittal plane of the face of a fetus in a monday pregnancy. The selection criteria were as follows:

(a) fetal karyotype is determined by amniocentesis;

(b) the nasal bone, frontal bone, skin above the nasal bone, and the anterior edges of the maxilla and mandible of the fetus are clearly shown on the image.

302 trisomy 21 cases, confirmed by inclusion in amniocentesis, were matched with 322 euploid cases randomly selected using the same criteria as the trisomy 21 cases. All cases were divided into a training set (257 euploid cases +241 trisomy 21 cases) and a validation set (65 euploid cases +61 trisomy 21 cases) at a ratio of about 8: 2. The retrocervical, facial and midbrain regions of the fetus were manually delineated for training and validation set images by a skilled radiologist.

Fig. 2 is a schematic flowchart of the construction of a ROI target region detection model according to an embodiment of the present invention, and specifically, the method includes the following steps:

s201: acquiring an ultrasonic image of a fetus;

in one embodiment, the ultrasound image of the fetus may be a set of ultrasound images of a chromosome normal fetus and a chromosome abnormality fetus, or a set of ultrasound images of a chromosome normal fetus.

S202: extracting a ROI area from the ultrasonic image;

s203: comparing the obtained ROI with a fetal retrocervical region, a facial region and a midbrain region manually defined by a doctor respectively to generate a loss value, performing back propagation, and performing ROI region extraction optimization;

s204: and obtaining an ROI target region detection model.

Fig. 3 is a schematic flow chart of chromosome abnormality risk diagnosis based on imaging omics, specifically, the method comprises the following steps:

s301: acquiring an ultrasonic image of a fetus of a sample to be detected;

s302: inputting the chromosome abnormality prediction model into a chromosome abnormality prediction model, wherein the chromosome abnormality prediction model is constructed by adopting the construction method;

s303: obtaining the classification result that the sample to be detected is normal chromosome or abnormal chromosome;

in one embodiment, ultrasound images and clinical information of a fetus of a sample to be tested are acquired.

In one embodiment, the clinical information is the age of the pregnant woman (the line of demarcation between the ages of the pregnant women less than or greater than 35 years old).

In one embodiment, an ultrasonic image of a fetus of a sample to be detected is obtained and input to a fetus image standard judgment model, and a classification result of whether the ultrasonic image is standard or not is obtained; and when the classification result is nonstandard, terminating the prediction of the chromosome abnormality risk, and when the classification result is standard, inputting the ultrasonic image of the fetus of the sample to be detected or the ultrasonic image and the clinical information of the fetus of the sample to be detected into the chromosome abnormality prediction model.

In one embodiment, the building process of the fetal ultrasound image standard judgment model comprises the following steps:

and acquiring standard and non-standard ultrasonic images of the fetus, comparing the obtained classification result with an actual result by adopting a machine/deep learning model capable of being supervised, generating a loss value, performing back propagation, performing model optimization, and acquiring a standard judgment model of the ultrasonic image of the fetus.

In one embodiment, the construction process of the chromosome abnormality multi-classification model of the fetal ultrasound image comprises the following steps:

obtaining ultrasonic images of different types of fetal chromosome abnormality diseases, grouping according to the types of the diseases, adopting a machine/deep learning model capable of being supervised, comparing the obtained classification result with an actual result, generating a loss value, performing back propagation, performing model optimization, and obtaining a fetal ultrasonic image chromosome abnormality multi-classification model.

In one embodiment, ultrasound images of different kinds of fetal chromosomal abnormalities are obtained, the different kinds being 2 or more of the following: 21-trisomy syndrome, 18-trisomy syndrome, 13-trisomy syndrome, 8-trisomy mosaic syndrome, 9-trisomy mosaic syndrome, klinfelter syndrome, tunner syndrome, X trisomy syndrome, XYY syndrome, sexual development diseases, chromosome deletion, chromosome duplication, chromosome translocation, chromosome inversion, circular chromosomes, isoarm chromosomes, chromosome insertion, small extra marker chromosomes and the like, grouping according to disease types, adopting a supervised machine/deep learning model, comparing obtained classification results with actual results to generate loss values, carrying out reverse propagation and carrying out model optimization to obtain a fetal ultrasound image chromosome abnormality multi-classification model.

In one embodiment, the data is from Beijing gynecologic Hospital/Beijing women's child care institute affiliated with the university of capital medical, Beijing, China, and pregnancies 11-13 were selected from our database⁺⁶Two-dimensional (2D) images of the median sagittal plane of the face of a normal fetus in a single gestational fetus of a week are 120 images of the standard and nonstandard images respectively, and a senior doctor manually evaluates whether the ultrasonic image of the fetus is a standard plane. All images were divided into a training set (96 standard +96 non-standard images) and a validation set (24 standard +24 non-standard images) at a ratio of about 8:2,and comparing the obtained classification result with an actual result by adopting a machine/deep learning model capable of being supervised to generate a loss value, performing back propagation, and performing model optimization to obtain a fetus ultrasonic image standard judgment model. And performing cross validation (5-10 times) on the obtained fetus ultrasonic image standard judgment model to obtain the accuracy of model prediction, and stopping model optimization when the accuracy of the model is more than 0.8 or more than 0.85.

Fig. 4 is a schematic diagram of a chromosome abnormality risk diagnosis system based on fetal ultrasound image characterics provided by an embodiment of the present invention, where the risk diagnosis system includes:

an obtaining unit 401, configured to obtain an ultrasound image of a fetus of a sample to be detected or an ultrasound image and clinical information of the fetus;

a processing unit 402, configured to input the ultrasound image of the fetus of the sample to be detected or the ultrasound image of the fetus and clinical information into a chromosome abnormality prediction model, where the chromosome abnormality prediction model is constructed by using the above-mentioned construction method;

and a display unit 403, configured to output a classification result that the sample to be tested is normal or abnormal.

Based on the foetus ultrasonic image characteristic omics chromosome abnormality risk diagnosis system schematic diagram, the risk diagnosis system includes:

the acquisition unit is used for acquiring an ultrasonic image of a fetus of a sample to be detected or an ultrasonic image and clinical information of the fetus;

a pretreatment unit: and obtaining a classification result whether the ultrasonic image is standard or not.

The processing unit is used for inputting the ultrasonic image of the fetus of the sample to be detected or the ultrasonic image and the clinical information of the fetus into a chromosome abnormality prediction model, and the chromosome abnormality prediction model is constructed by adopting the construction method of the chromosome abnormality prediction model;

and the display unit is used for outputting the classification result that the sample to be detected is normal chromosome or abnormal chromosome.

Based on the foetus ultrasonic image characteristic omics chromosome abnormality risk diagnosis system schematic diagram, the risk diagnosis system includes:

the acquisition unit is used for acquiring an ultrasonic image of a fetus of a sample to be detected or an ultrasonic image and clinical information of the fetus;

a pretreatment unit: and obtaining a classification result whether the ultrasonic image is standard or not.

The processing unit is used for inputting the ultrasonic image of the fetus of the sample to be detected or the ultrasonic image and the clinical information of the fetus into a chromosome abnormality prediction model, and the chromosome abnormality prediction model is constructed by adopting the construction method of the chromosome abnormality prediction model;

the display unit is used for outputting the classification result that the sample to be detected is normal chromosome or abnormal chromosome;

and the classification unit is used for outputting the classification result of the abnormal chromosome diseases.

Fig. 5 is a schematic diagram of a chromosome abnormality risk diagnosis device based on fetal ultrasound imaging omics according to an embodiment of the present invention, where the device includes: a memory and a processor;

the memory is to store program instructions;

the processor is configured to invoke program instructions that, when executed, are configured to:

acquiring an ultrasonic image of a fetus of a sample to be detected or acquiring an ultrasonic image and clinical information of the fetus of the sample to be detected;

inputting the chromosome abnormality prediction model into a chromosome abnormality prediction model, wherein the chromosome abnormality prediction model is constructed by adopting the construction method of the chromosome abnormality prediction model;

and obtaining the classification result that the sample to be detected is normal chromosome or abnormal chromosome.

In one embodiment, the invention provides a chromosome abnormality risk diagnosis device based on fetal ultrasound image characterics, comprising: a memory and a processor;

the memory is to store program instructions;

the processor is configured to invoke program instructions that, when executed, are configured to:

acquiring an ultrasonic image of a fetus of a sample to be detected or acquiring an ultrasonic image and clinical information of the fetus of the sample to be detected;

acquiring an ultrasonic image of a fetus of a sample to be detected, and inputting the ultrasonic image into a fetus ultrasonic image standard judgment model to obtain a classification result of whether the ultrasonic image is standard or not; when the classification result is nonstandard, stopping the prediction of the chromosome abnormality risk, and when the classification result is standard, inputting the ultrasonic image of the fetus of the sample to be detected or the ultrasonic image and the clinical information of the fetus of the sample to be detected into a chromosome abnormality prediction model;

and obtaining the classification result that the sample to be detected is normal chromosome or abnormal chromosome.

In one embodiment, the invention provides a chromosome abnormality risk diagnosis device based on fetal ultrasound image characterics, comprising: a memory and a processor;

the memory is to store program instructions;

the processor is configured to invoke program instructions that, when executed, are configured to:

acquiring an ultrasonic image of a fetus of a sample to be detected or acquiring an ultrasonic image and clinical information of the fetus of the sample to be detected;

acquiring an ultrasonic image of a fetus of a sample to be detected, and inputting the ultrasonic image into a fetus image standard judgment model to obtain a classification result of whether the ultrasonic image is standard or not; when the classification result is nonstandard, stopping the prediction of the chromosome abnormality risk, and when the classification result is standard, inputting the ultrasound image of the fetus of the sample to be detected or the ultrasound image and the clinical information of the fetus of the sample to be detected into a chromosome abnormality prediction model to obtain the classification result that the sample to be detected is normal chromosome or abnormal chromosome;

and when the sample to be detected is abnormal chromosome, stopping the prediction of the abnormal chromosome risk, and when the result is abnormal chromosome, inputting the ultrasonic image of the fetus, which is obtained from the sample to be detected, into the ultrasonic image of the fetus, namely the abnormal chromosome multi-classification model, so as to obtain the classification result of abnormal chromosome diseases.

In one embodiment, the clinical information is the age of the pregnant woman.

It is an object of the present invention to provide a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, enables the above-mentioned risk prediction of chromosomal abnormalities.

The validation results of this validation example show that assigning an intrinsic weight to an indication can moderately improve the performance of the method relative to the default settings.

It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.

In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other manners. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.

Those skilled in the art will appreciate that all or part of the steps in the methods of the above embodiments may be implemented by associated hardware instructed by a program, which may be stored in a computer-readable storage medium, and the storage medium may include: read Only Memory (ROM), Random Access Memory (RAM), magnetic or optical disks, and the like.

It will be understood by those skilled in the art that all or part of the steps in the method for implementing the above embodiments may be implemented by hardware that is instructed to implement by a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.

While the invention has been described in detail with reference to specific embodiments thereof, it will be apparent to one skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention.