CN114566276A - Lung color Doppler ultrasound-based training method and device for child pneumonia auxiliary diagnosis model - Google Patents

Abstract

The present application provides a training method and training device for an auxiliary diagnosis model of childhood pneumonia based on lung color Doppler ultrasound. The training method includes: acquiring a diagnosis and treatment information database including color Doppler images of multiple pediatric pneumonia patients; acquiring a plurality of sample sets according to the database; Perform training processing on the training set in each sample set to obtain a training model; test the training model according to the test set of each sample set to obtain the accuracy of the training model; determine the diagnostic model based on the accuracy of multiple training models ; wherein, each sample set includes a training set and a corresponding test set. The diagnostic model obtained by the training method of the present application has high accuracy, which can help clinicians in diagnosis, reduce the rate of missed diagnosis and misdiagnosis, and can also use the diagnostic model as a tool for doctors to learn and diagnose, and can also provide medical services for the rapid growth of doctors. Huge boost.

Description

Training method and training device for auxiliary diagnosis model of children's pneumonia based on lung color Doppler ultrasound

technical field

The present application relates to the technical field of medical computers, and in particular, to a training method and a training device for a child pneumonia auxiliary diagnosis model based on lung color ultrasound.

Background technique

According to the World Health Organization (WHO), pediatric pneumonia is the leading cause of death in children. In 2015, WHO estimated that more than 920,000 children worldwide died of pneumonia, of which children under the age of 5 accounted for about 15%. Diagnosing pneumonia can sometimes be difficult because symptoms can vary depending on the child's age and the cause of the infection. In addition, some symptoms do not only point to childhood pneumonia, but can also be related clinical manifestations of other diseases. Pneumonia is a very threatening disease for infants and young children. Once found, an accurate diagnosis should be given, and then symptomatic treatment should be given. The earlier the diagnosis and the higher the accuracy, the better, which will help improve the cure rate of children. Choices should be given a strong focus. In particular, for the diagnosis of pneumonia in children, the community needs more accurate pneumonia diagnosis methods.

However, in the past, the main imaging methods for pediatric pneumonia were chest X-ray and CT examination. For the same medical image of chest X-ray and CT examination, the judgment results of different time points or different doctors are highly inconsistent, which has a high degree of inconsistency. large observer differences.

SUMMARY OF THE INVENTION

In order to solve or at least partially solve the above-mentioned technical problems, the present application provides a training method for a child pneumonia auxiliary diagnosis model based on lung color Doppler ultrasound, which includes: acquiring a diagnosis and treatment information database including color Doppler images of multiple pediatric pneumonia patients; The database obtains multiple sample sets; performs training processing on the training set in each sample set to obtain a training model; tests the training model according to the test set of each sample set to obtain the accuracy of the training model; based on multiple training models The accuracy of the diagnostic model is determined; wherein, each sample set includes a training set and a corresponding test set.

The training method of the child pneumonia auxiliary diagnosis model based on lung color ultrasound in the present application includes: acquiring a diagnosis and treatment information database including color ultrasound images of multiple children with pneumonia patients. The database can provide data support for the training of the diagnosis model and improve the accuracy of the diagnosis model as much as possible.

Next, multiple sample sets are obtained according to the database. For each sample set, there is a training set and a corresponding test set. One sample set may contain all the data of the database, or may only contain part of the data. The diagnosis and treatment information in the training set is used for model training, and the number of test sets is used to test the accuracy of the model. The training model obtained after training is then tested on the test set to obtain the accuracy of the training model.

The training process is performed on the training set in each sample set to obtain a training model, and then the training model is tested according to the test set of each sample set to obtain the accuracy of the training model. Since the number of sample sets is multiple, when each sample set is trained, multiple training models will be obtained, and after the multiple training models are tested separately, the accuracy of multiple training models will be obtained. .

Finally, based on the accuracy of multiple training models, the diagnosis model is determined. The accuracy of the diagnosis model is high, which can help doctors in diagnosis, reduce the work pressure of pediatricians, improve the diagnosis rate of pneumonia in hospitals at all levels, and assist clinicians. Reducing the rate of missed diagnosis and misdiagnosis, the diagnostic model can also be used as a tool for doctors to learn diagnosis, and it can also provide a huge impetus for the rapid growth of doctors.

It is worth noting that the training method for the auxiliary diagnosis model of childhood pneumonia based on lung color ultrasound provided in this application involves acquiring data, not as a means of diagnosis.

Optionally, each diagnosis and treatment information includes an ultrasound image, a corresponding test result, and corresponding diagnostic information.

Optionally, the step of obtaining a database consisting of diagnosis and treatment information of a plurality of children with pneumonia specifically includes: obtaining diagnosis and treatment information of a plurality of children with pneumonia; performing image preprocessing on the ultrasound images in the plurality of diagnosis and treatment information to obtain a database; Among them, image preprocessing includes cropping, flipping, rotation, and scaling.

Optionally, the step of acquiring multiple sample sets according to the database specifically includes: randomly acquiring multiple sample sets from the database; wherein the ratios of the training set and the test set in any two sample sets are different.

Optionally, the multiple training models include a first training model and a second training model, and the step of determining the diagnosis model based on the accuracy of the multiple training models specifically includes: the accuracy of the first training model is greater than that of the second training model. Accuracy, determine that the first training model is a diagnostic model.

Optionally, the network structure of the diagnosis model includes an AlexNet network structure, a Resnet18 network structure or a Resnet50 network structure.

The application also provides a training device for a child pneumonia auxiliary diagnosis model based on lung color Doppler ultrasound, characterized in that it includes:

an acquisition module, used for acquiring a diagnosis and treatment information database including color Doppler ultrasound images of multiple children with pneumonia, and acquiring multiple sample sets according to the database;

The training module is used to perform training processing on the training set in each sample set to obtain the training model;

The test module tests the training model according to the test set of each sample set to obtain the accuracy of the training model;

Determine the module, determine the diagnosis model based on the accuracy of the multiple training models;

Wherein, each sample set includes a training set and a corresponding test set.

The apparatus for training an auxiliary diagnosis model for childhood pneumonia in this application includes an acquisition module, a training module, a test module and a determination module, wherein the acquisition module is used to acquire a diagnosis and treatment information database including color ultrasound images of multiple children with pneumonia patients, and the source of the database can be For the pediatric department of a hospital, the data of children with pneumonia obtained during the consultation period, the database can provide data support for the training of the diagnosis model, and improve the accuracy of the diagnosis model as much as possible.

Next, the acquisition module can also acquire multiple sample sets according to the database. For each sample set, there are training sets and corresponding test sets. A sample set can contain all the data of the database, or only part of the data. The diagnosis and treatment information in the training set is used for model training, and the number of test sets is used to test the accuracy of the model. The training model obtained after training is then tested on the test set to obtain the accuracy of the training model.

The training module can perform training processing on the training set in each sample set to obtain a training model, and then through the training module, test the training model according to the test set of each sample set to obtain the accuracy of the training model. Since the number of sample sets is multiple, when each sample set is trained, multiple training models will be obtained, and after the multiple training models are tested separately, the accuracy of multiple training models will be obtained. .

Finally, the determination module can determine the diagnosis model based on the accuracy of the multiple training models, that is to say, select the training model with higher accuracy from the multiple training models as the diagnosis model, the accuracy of the diagnosis model is higher, and can be It provides help in the diagnosis of doctors, reduces the work pressure of pediatricians, improves the diagnosis rate of pneumonia in hospitals at all levels, and assists clinicians to reduce the rate of missed diagnosis and misdiagnosis. Growth provides a huge boost.

Optionally, the obtaining module is further configured to: randomly obtain multiple sample sets from the database, wherein the proportions of the training set and the test set in any two sample sets are different.

The present application also provides a computer device, including a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that the processor is used to perform the aforementioned auxiliary diagnosis of childhood pneumonia based on lung color Doppler ultrasound The steps of the model's training method.

In the computer equipment in the present application, the processor it includes is used to execute the steps of the training method of the child pneumonia auxiliary diagnosis model based on lung color Doppler in any of the above designs, and thus, the computer equipment can realize the child pneumonia based on lung color Doppler ultrasound. All the beneficial effects of the training method of the auxiliary diagnosis model will not be repeated here.

The present application also provides a computer-readable storage medium on which a computer program is stored, and is characterized in that: when the computer program is executed by the processor, the steps of the training method for the child pneumonia auxiliary diagnosis model based on lung color ultrasound as described above are realized. .

The computer-readable storage medium in the present application, when the computer program stored on it is executed by the processor, realizes the steps of the training method of the child pneumonia auxiliary diagnosis model based on lung color ultrasound in any of the above designs, so the computer can Reading the storage medium can achieve all the beneficial effects of the training method, which will not be repeated.

Description of drawings

In order to illustrate the embodiments of the present application more clearly, a brief introduction to the related drawings will be made below. It can be understood that the drawings in the following description are only used to illustrate some embodiments of the present application, and those of ordinary skill in the art can also obtain many other technical features and connection relationships not mentioned herein based on these drawings.

FIG. 1 is a schematic flowchart of a training method for a child pneumonia auxiliary diagnosis model based on lung color Doppler ultrasound according to an embodiment of the present application.

FIG. 2 is a schematic flowchart of a training method for a child pneumonia auxiliary diagnosis model based on lung color Doppler ultrasound according to another embodiment of the present application.

FIG. 3 is a schematic block diagram of a training device for a child pneumonia auxiliary diagnosis model based on lung color ultrasound according to an embodiment of the present application.

FIG. 4 is a schematic block diagram of a computer device according to an embodiment of the present application.

Detailed ways

The technical solutions in the embodiments of the present application will be described in detail below with reference to the accompanying drawings in the embodiments of the present application.

The inventors of the present application found that, in the related art, the diagnosis of pneumonia in children mainly relies on medical images. However, for the same medical image, the judgment results of different observers are highly inconsistent, and there is a possibility of missed diagnosis and misdiagnosis.

In view of this, the present application provides a training method for a child pneumonia auxiliary diagnosis model based on lung color Doppler ultrasound, so as to assist clinicians to reduce the rate of missed diagnosis and misdiagnosis, and at the same time, it can also reduce the work intensity of clinicians, and improve the accuracy and efficiency of diagnosis. .

Embodiment 1

As shown in FIG. 1 , a schematic flowchart of a training method for a child pneumonia auxiliary diagnosis model based on lung color Doppler ultrasound according to an embodiment of the present application. Among them, the training methods include:

S102, acquiring a diagnosis and treatment information database including color ultrasound images of multiple children with pneumonia;

S104, obtaining multiple sample sets according to the database;

S106, perform training processing on the training set in each sample set to obtain a training model;

S108, test the training model according to the test set of each sample set to obtain the accuracy of the training model;

S110, determining a diagnosis model based on the accuracy of the multiple training models;

Wherein, each sample set includes a training set and a corresponding test set.

The training method of the child pneumonia auxiliary diagnosis model based on lung color ultrasound in the present application includes: acquiring a diagnosis and treatment information database including color ultrasound images of multiple children with pneumonia patients. The database can provide data support for the training of the diagnosis model and improve the accuracy of the diagnosis model as much as possible.

Next, multiple sample sets are obtained according to the database. For each sample set, there is a training set and a corresponding test set. One sample set may contain all the data of the database, or may only contain part of the data. The diagnosis and treatment information in the training set is used for model training, and the number of test sets is used to test the accuracy of the model. The training model obtained after training is then tested on the test set to obtain the accuracy of the training model.

The training process is performed on the training set in each sample set to obtain a training model, and then the training model is tested according to the test set of each sample set to obtain the accuracy of the training model. Since the number of sample sets is multiple, when each sample set is trained, multiple training models will be obtained, and after the multiple training models are tested separately, the accuracy of multiple training models will be obtained. .

Finally, based on the accuracy of multiple training models, the diagnosis model is determined. The accuracy of the diagnosis model is high, which can help doctors in diagnosis, reduce the work pressure of pediatricians, improve the diagnosis rate of pneumonia in hospitals at all levels, and assist clinicians. Reducing the rate of missed diagnosis and misdiagnosis, the diagnostic model can also be used as a tool for doctors to learn diagnosis, and it can also provide a huge impetus for the rapid growth of doctors.

Optionally, each diagnosis and treatment information includes an ultrasound image, a corresponding test result, and corresponding diagnostic information.

In this embodiment, the database contains a plurality of diagnosis and treatment information, each diagnosis and treatment information corresponds to a patient, and each diagnosis and treatment information includes ultrasound images, test results and diagnosis information corresponding to each patient.

Wherein, the ultrasound image is obtained by using ultrasound imaging technology. Specifically, ultrasound imaging uses ultrasound to scan the human body, and obtains images of internal organs by receiving and processing reflected signals. There are many kinds of commonly used ultrasonic instruments: type A (amplitude modulation type) indicates the strength of the reflected signal by the amplitude of the wave, and it shows an "echo graph". Ultrasound images are non-invasive and non-radioactive, which can effectively improve the radiation damage of patients, and is feasible. Ultrasound examination has the advantages of simplicity, speed, multi-directionality and less restriction.

Among them, the test results include blood test results during the patient's hospitalization.

The diagnostic information includes the doctor making a diagnosis conclusion on the patient's condition according to the ultrasound image and the test result. Specifically, the diagnostic information includes pneumothorax, interstitial pulmonary syndrome, lung consolidation, acute lung injury, respiratory distress syndrome, neonatal transient tachypnea, pneumonia, and the like.

Specifically, healthy human lung tissue contains a large amount of gas and a small amount of water. The interface between the pleura and lung tissue constitutes the interface between soft tissue and gas. When sound waves pass through this interface, a strong reflection occurs. The main pathological pathology of pediatric pneumonia is The change is that the lung tissue is exudative, and the normal alveolar air is filled with exudate, inflammatory cells, etc., which is the basis for the diagnosis of pneumonia by ultrasound.

When there is inflammation in the lungs, the air content in the lungs decreases, the inflammatory exudation increases, and the acoustic impedance between air and liquid increases, and the ultrasound produces strong reverberation at the gas-liquid junction, forming a comet tail sign, and extending to the far field is the B line. . The number of B-lines depends on the degree of lung ventilation loss, and the intensity of its echoes increases with inspiratory motion. Among them, multiple B lines with a distance of ≤3mm are called B3 lines, which are related to ground glass shadows displayed on chest CT, indicating the possibility of alveolar pulmonary edema, while multiple B lines with a distance of >7mm are called B7 lines. This suggests the possibility of interlobular septal thickening. If the child's disease progresses to the point where the gas in the alveoli disappears and is filled with a large amount of fibrin, red blood cells, white blood cells and other exudates, it becomes the pathological basis for the diagnosis of lung consolidation by ultrasound. When the consolidated lung is in direct contact with the pleura or through water to form an acoustic window, it provides conditions for ultrasonography.

Embodiment 2

As shown in FIG. 2 , according to another embodiment of the present application, a schematic flowchart of a training method for a child pneumonia auxiliary diagnosis model based on lung color ultrasound. Among them, the training methods include:

S202, obtaining diagnosis and treatment information of multiple children with pneumonia;

S204, performing image preprocessing on the ultrasound images in the plurality of diagnosis and treatment information to obtain a database;

S206, obtaining multiple sample sets according to the database;

S208, performing training processing on the training set in each sample set to obtain a training model;

S210, test the training model according to the test set of each sample set to obtain the accuracy of the training model;

S212, determining a diagnosis model based on the accuracy of the multiple training models.

In this embodiment, the step of obtaining a database consisting of diagnosis and treatment information of a plurality of children with pneumonia specifically includes: obtaining diagnosis and treatment information of a plurality of children with pneumonia, where the diagnosis and treatment information of each child pneumonia patient includes ultrasound images, corresponding Test results and corresponding diagnostic information. In the subsequent training process, a sample set composed of multiple diagnosis and treatment information is used as the training basis. Therefore, the accuracy of diagnosis and treatment information is particularly important.

Among them, the test results and diagnosis information in the diagnosis and treatment information are all composed of data text, and the accuracy is relatively high, which can be directly processed by training. There is room for improvement in the accuracy of ultrasound images.

In this embodiment, image preprocessing is performed on the ultrasound image in each diagnosis and treatment information, so as to reduce noise in the ultrasound image, make up for data defects of the ultrasound image, improve the accuracy of the ultrasound image, and further improve the efficiency of training processing.

It is worth noting that, in the ultrasound image, the main noise comes from speckle, which is caused by the interference effect caused by the scattering of the sound beam in the uneven fine tissue, which appears as granular in the image, It does not reflect the actual organizational structure, but affects the detail resolution ability of the image. This is not conducive to the quantitative analysis of the image, so it is necessary to suppress the speckle noise in the image.

Optionally, the image preprocessing includes cropping, flipping, rotation, and scaling.

Among them, before the training process, in order to reduce the workload of data calculation and improve the training speed, the original image can be cropped for the feature part of interest, so that each ultrasound image not only retains the feature set to be extracted, but also reduces the size of the original image. The overall size can effectively shorten the time spent training the model and improve the training efficiency.

Optionally, the step of acquiring multiple sample sets according to the database specifically includes: randomly acquiring multiple sample sets from the database; wherein the ratios of the training set and the test set in any two sample sets are different.

In this embodiment, the step of obtaining multiple sample sets according to the database specifically includes: randomly obtaining multiple sample sets from the database, so as to ensure as much as possible the unique properties of the training set in each sample set during training processing , to obtain different training models as much as possible to provide more choices for obtaining the final diagnostic model.

Optionally, the proportions of the training set and the test set in any two sample sets are different, that is, the composition of any two sample sets is different, at this time, for these two sample sets, the obtained training model also has heterogeneity.

For example, the first sample set, the second sample set and the third sample set are obtained in the database, wherein the ratio of the training set and the test set in the first sample set is 5:5, and the training set and the test set in the second sample set are The ratio of the test set is 8:2, and the ratio of the training set and the test set in the third sample set is 9:1. Then, the first training model can be obtained by training on the test set of the first sample set. The second training model can be obtained by training on the test set of the second sample set, the third training model can be obtained by training on the test set of the third sample set, and then the accuracy of the obtained training model can be checked through the respective test sets The test is performed to obtain the respective corresponding accuracies, and the diagnostic model is determined according to the accuracies of the multiple training models.

When the diagnostic model is applied to the actual ultrasound examination process, the patient's lung ultrasound images are input into the diagnostic model, which can assist doctors to reduce the rate of missed diagnosis and misdiagnosis, and achieve the purpose of early diagnosis and early treatment.

Optionally, the multiple training models include a first training model and a second training model, and the step of determining the diagnosis model based on the accuracy of the multiple training models specifically includes: the accuracy of the first training model is greater than that of the second training model. Accuracy, determine that the first training model is a diagnostic model.

In this embodiment, the multiple training models include a first training model and a second training model, and the step of determining the diagnosis model based on the accuracy of the multiple training models specifically includes: the accuracy of the first training model is greater than that of the second training model The accuracy of the model determines that the first training model is a diagnosis model, and the diagnosis model is a training model with higher accuracy.

Among them, the higher accuracy of the training model indicates that its performance is excellent, and the diagnostic information can be obtained according to the ultrasound images, which can effectively assist clinicians and reduce work intensity.

It is worth noting that the number of multiple training models can be a positive integer without an upper limit. The number of sample sets can be selected according to the actual situation, so as to obtain the corresponding number of training models, and then select the best among the multiple training models. high as the final diagnostic model.

Optionally, the network structure of the diagnosis model includes an AlexNet network structure, a Resnet18 network structure or a Resnet50 network structure.

Embodiment 3

FIG. 3 shows a schematic block diagram of a training device for a child pneumonia auxiliary diagnosis model based on lung color ultrasound according to an embodiment of the present application. As shown in FIG. 3, the training device 300 includes:

The acquisition module 302 is used to acquire a diagnosis and treatment information database including color Doppler ultrasound images of multiple children with pneumonia, and acquire multiple sample sets according to the database;

A training module 304, configured to perform training processing on the training set in each sample set to obtain a training model;

The testing module 306 checks the training model according to the test set of each sample set to obtain the accuracy of the training model;

determining module 308, determining a diagnosis model based on the accuracy of the multiple training models;

Wherein, each sample set includes a training set and a corresponding test set.

The training device 300 of the child pneumonia auxiliary diagnosis model based on lung color Doppler in the present application includes an acquisition module 302, a training module 304, a test module 306 and a determination module 308, wherein the acquisition module 302 is used to obtain the color Doppler ultrasound including multiple children with pneumonia patients The image diagnosis and treatment information database, the source of the database can be the pediatric department of a hospital, the medical treatment information of children with pneumonia obtained during the consultation period, the database can provide data support for the training of the diagnosis model, and improve the accuracy of the diagnosis model as much as possible.

Next, the acquisition module 302 can also acquire multiple sample sets according to the database, and each sample set has a training set and a corresponding test set, and a sample set may contain all the data of the database, or only a part of the data. The diagnosis and treatment information in the training set is used for model training, and the number of test sets is used to test the accuracy of the model. The training model obtained after training is then tested on the test set to obtain the accuracy of the training model.

The training module 304 can perform training processing on the training set in each sample set to obtain a training model, and then pass the testing module 306 to test the training model according to the test set of each sample set to obtain the accuracy of the training model. Since the number of sample sets is multiple, when each sample set is trained, multiple training models will be obtained, and after the multiple training models are tested separately, the accuracy of multiple training models will be obtained. .

Finally, the determination module 308 can determine the diagnosis model based on the accuracy of the multiple training models. The diagnosis model has a high accuracy, which can provide help for the doctor's diagnosis, reduce the work pressure of the pediatrician, and improve the diagnosis rate of pneumonia in hospitals at all levels. , assisting clinicians to reduce the rate of missed diagnosis and misdiagnosis, and can also use the diagnostic model as a tool for doctors to learn diagnosis, and can also provide a huge impetus for the rapid growth of doctors.

Optionally, the obtaining module 302 is further configured to: randomly obtain multiple sample sets from the database, wherein the ratios of the training set and the test set in any two sample sets are different.

Optionally, each diagnosis and treatment information includes an ultrasound image, a corresponding test result, and corresponding diagnostic information.

Optionally, the obtaining module 302 is further configured to: obtain the diagnosis and treatment information of multiple children with pneumonia; perform image preprocessing on the ultrasound images in the multiple diagnosis and treatment information to obtain a database; wherein, the image preprocessing includes cropping, flipping, rotating, Zoom processing.

Optionally, the multiple training models include a first training model and a second training model, and the determining module 308 is further configured to: the accuracy of the first training model is greater than the accuracy of the second training model, and determining that the first training model is a diagnostic model .

Optionally, the network structure of the diagnosis model includes an AlexNet network structure, a Resnet18 network structure or a Resnet50 network structure.

Embodiment 4

FIG. 4 shows a schematic block diagram of a computer device 400 in an embodiment of the present application. The computer device 400 includes a memory 402, a processor 404, and a computer program stored in the memory 402 and running on the processor 404, characterized in that: the processor 404 is used to perform the aforementioned lung color ultrasound-based child pneumonia assistance Steps in the training method of the diagnostic model.

In the computer device 400 in the present application, the processor 404 included in it is configured to execute the steps of the training method of the child pneumonia auxiliary diagnosis model based on lung color ultrasound in any of the above designs. Therefore, the computer device 400 can realize the lung-based diagnosis All the beneficial effects of the training method of the color Doppler ultrasound auxiliary diagnosis model of childhood pneumonia will not be repeated here.

Embodiment 5

The present application also provides a computer-readable storage medium on which a computer program is stored, and is characterized in that: when the computer program is executed by the processor, the steps of the training method for the child pneumonia auxiliary diagnosis model based on lung color ultrasound as described above are realized. .

The computer-readable storage medium in the present application, when the computer program stored on it is executed by the processor, realizes the steps of the training method of the child pneumonia auxiliary diagnosis model based on lung color ultrasound in any of the above designs, so the computer can Reading the storage medium can achieve all the beneficial effects of the training method, which will not be repeated.

It will be apparent to those skilled in the art that the present application is not limited to the details of the above-described exemplary embodiments, but that the present application may be implemented in other specific forms without departing from the spirit or essential characteristics of the present application. Accordingly, the embodiments are to be regarded in all respects as illustrative and not restrictive, and the scope of the application is to be defined by the appended claims rather than the foregoing description, which is therefore intended to fall within the scope of the claims. All changes that come within the meaning and scope of equivalents to are included in this application. Any reference signs in the claims shall not be construed as limiting the involved claim.

Claims (10)

1. A training method of a lung color Doppler ultrasound-based children pneumonia auxiliary diagnosis model is characterized by comprising the following steps:

acquiring a diagnosis and treatment information database comprising color Doppler ultrasound images of a plurality of children pneumonia patients;

acquiring a plurality of sample sets according to the database;

training the training set in each sample set to obtain a training model;

testing the training model according to the test set of each sample set to obtain the accuracy of the training model;

determining the diagnostic model based on the accuracy of a plurality of the training models;

each sample set comprises a training set and a corresponding test set.

2. Training method according to claim 1,

each diagnosis and treatment information comprises an ultrasonic image, a corresponding test result and corresponding diagnosis information.

3. A training method according to claim 1, wherein the step of obtaining a database of diagnosis and treatment information of a plurality of children pneumonia patients specifically comprises:

acquiring diagnosis and treatment information of a plurality of children pneumonia patients;

carrying out image preprocessing on the ultrasonic images in the diagnosis and treatment information to obtain a database;

the image preprocessing comprises cutting, turning, rotating and zooming.

4. The training method according to claim 1, wherein the step of obtaining a plurality of sample sets from the database specifically comprises:

randomly acquiring a plurality of sample sets in the database;

and the proportion of the training set to the test set in any two sample sets is different.

5. The training method of claim 1, wherein the plurality of training models comprises a first training model and a second training model, and wherein the step of determining the diagnostic model based on the accuracy of the plurality of training models comprises:

and the accuracy of the first training model is greater than that of the second training model, and the first training model is determined to be the diagnosis model.

6. Training method according to claim 1,

the network structure of the diagnostic model comprises an AlexNet network structure, a Resnet18 network structure or a Resnet50 network structure.

7. A children pneumonia aided diagnosis model training device based on lung color Doppler ultrasound is characterized by comprising:

the system comprises an acquisition module, a data acquisition module and a data processing module, wherein the acquisition module is used for acquiring a diagnosis and treatment information database comprising color Doppler ultrasound images of a plurality of children pneumonia patients and acquiring a plurality of sample sets according to the database;

the training module is used for training the training set in each sample set to obtain a training model;

the testing module is used for testing the training model according to the testing set of each sample set to obtain the accuracy of the training model;

a determination module that determines the diagnostic model based on an accuracy of a plurality of the training models;

each sample set comprises a training set and a corresponding test set.

8. The training device of claim 7,

the acquisition module is further configured to: randomly acquiring a plurality of sample sets in the database, wherein the proportion of a training set to a testing set in any two sample sets is different.

9. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein: the processor is used for executing the steps of the training method of the lung color ultrasound-based children pneumonia auxiliary diagnosis model according to any one of claims 1 to 6.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that: the computer program when executed by a processor realizes the steps of the training method of the lung color ultrasound-based children pneumonia auxiliary diagnosis model according to any one of claims 1 to 6.

Images