CN113327238A - Gastroscope image classification model construction method and gastroscope image classification method - Google Patents
Gastroscope image classification model construction method and gastroscope image classification method Download PDFInfo
- Publication number
- CN113327238A CN113327238A CN202110649737.6A CN202110649737A CN113327238A CN 113327238 A CN113327238 A CN 113327238A CN 202110649737 A CN202110649737 A CN 202110649737A CN 113327238 A CN113327238 A CN 113327238A
- Authority
- CN
- China
- Prior art keywords
- classification model
- image
- gastroscope
- training
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30092—Stomach; Gastric
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Medical Informatics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Image Analysis (AREA)
Abstract
The application provides a gastroscope image classification model construction method and a gastroscope image classification method, wherein the gastroscope image classification model construction method comprises the following steps: acquiring a diagnosis result and a gastroscope image set of a patient; wherein the gastroscopic image set comprises a plurality of gastroscopic images; inputting the diagnosis result and the gastroscope image set into a preset original classification model to pre-train the original classification model to obtain a pre-training classification model; and constructing a target classification model according to the model parameters of the pre-training classification model. According to the gastroscope image classification model building method provided by the technical scheme, the target classification model is built according to the model parameters of the classification model pre-trained by the multiple pictures, the subsequent training time of the target classification model is saved, the building efficiency of the target classification model is improved, and the multiple gastroscope images corresponding to the diagnosed patient are collected by taking the patient as a sample statistical unit, so that the labor labeling cost is saved.
Description
Technical Field
The application relates to the technical field of image recognition, in particular to a gastroscope image classification model construction method and a gastroscope image classification method.
Background
At present, gastroscopy is an important means for diagnosing stomach diseases, and in order to improve the detection efficiency of gastroscopy images, a machine learning model is generally adopted to automatically detect and classify the gastroscopy images. Therefore, how to improve the reliability of the machine learning model through model training has become a content of hot research.
In the prior art, a model training method generally needs to collect massive labeled samples and then sequentially perform model training by taking a single labeled sample as a main body.
However, the labeling process of the sample consumes a large amount of human resources, and the subsequent model training process also consumes a large amount of time and cost, so that the model construction efficiency is reduced.
Disclosure of Invention
The application provides a gastroscope image classification model construction method and a gastroscope image classification method, which aim to overcome the defects of low model construction efficiency and the like in the prior art.
The first aspect of the present application provides a gastroscope image classification model construction method, including:
acquiring a diagnosis result and a gastroscope image set of a patient; wherein the gastroscopic image set comprises a plurality of gastroscopic images;
inputting the diagnosis result and the gastroscope image set into a preset original classification model so as to pre-train the original classification model to obtain a pre-training classification model;
and constructing a target classification model according to the model parameters of the pre-training classification model.
Optionally, the method further includes:
acquiring a plurality of gastroscopic image samples; wherein the gastroscope image sample comprises a gastroscope image and a sample label corresponding to the gastroscope image;
and inputting the gastroscope image sample into the target classification model, and training the target classification model to optimize the target classification model.
Optionally, the pre-training the original classification model includes:
extracting image characteristics of each gastroscope image in the gastroscope image set;
constructing a feature matrix of the diagnosis result according to the image features;
determining the classification weight of each image characteristic according to the image quality of each gastroscope image;
performing dimension reduction processing on the feature matrix according to the classification weight to obtain a target feature vector corresponding to the feature matrix;
and pre-training the original classification model by using the target feature vector.
Optionally, the constructing a target classification model according to the model parameters of the pre-training classification model includes:
obtaining an initial target classification model;
migrating the parameter weights of the pre-trained classification model to the initial target classification model;
the pre-training classification model comprises an input layer, a machine learning layer and an output layer, and the model parameters at least comprise model parameters corresponding to the machine learning layer.
Optionally, before constructing the target classification model according to the model parameters of the pre-trained classification model, the method further includes:
obtaining an evaluation index of a current pre-training classification model;
judging whether the current pre-training model is in a stable state or not according to the evaluation index;
returning to the step of obtaining a diagnosis of the patient and a set of gastroscopic images when the current pre-trained model is in an unstable state.
A second aspect of the present application provides a gastroscopic image classification method comprising:
acquiring a gastroscope image to be classified;
inputting the gastroscope image to be classified into the target classification model constructed by the gastroscope image classification model construction method according to the first aspect and various possible designs of the first aspect, so as to obtain the classification result of the gastroscope image to be classified.
The third aspect of the present application provides a gastroscope image classification model construction device, including:
a first acquisition module for acquiring a diagnosis result of a patient and a gastroscope image set; wherein the gastroscopic image set comprises a plurality of gastroscopic images;
the pre-training module is used for inputting the diagnosis result and the gastroscope image set into a preset original classification model so as to pre-train the original classification model to obtain a pre-training classification model;
and the construction module is used for constructing a target classification model according to the model parameters of the pre-training classification model.
Optionally, the building module is further configured to:
acquiring a plurality of gastroscopic image samples; wherein the gastroscope image sample comprises a gastroscope image and a sample label corresponding to the gastroscope image;
and inputting the gastroscope image sample into the target classification model, and training the target classification model to optimize the target classification model.
Optionally, the pre-training module is specifically configured to:
extracting image characteristics of each gastroscope image in the gastroscope image set;
constructing a feature matrix of the diagnosis result according to the image features;
determining the classification weight of each image characteristic according to the image quality of each gastroscope image;
performing dimension reduction processing on the feature matrix according to the classification weight to obtain a target feature vector corresponding to the feature matrix;
and pre-training the original classification model by using the target feature vector.
Optionally, the building module is specifically configured to:
obtaining an initial target classification model;
migrating the parameter weights of the pre-trained classification model to the initial target classification model;
the pre-training classification model comprises an input layer, a machine learning layer and an output layer, and the model parameters at least comprise model parameters corresponding to the machine learning layer.
Optionally, the building module is further configured to:
obtaining an evaluation index of a current pre-training classification model;
judging whether the current pre-training model is in a stable state or not according to the evaluation index;
returning to the step of obtaining a diagnosis of the patient and a set of gastroscopic images when the current pre-trained model is in an unstable state.
A fourth aspect of the present application provides a gastroscopic image classification device comprising:
the second acquisition module is used for acquiring gastroscope images to be classified;
and the classification module is used for inputting the gastroscope image to be classified into the target classification model constructed by the gastroscope image classification model construction device according to the third aspect and various possible designs of the third aspect so as to obtain the classification result of the gastroscope image to be classified.
A fifth aspect of the present application provides an electronic device, comprising: at least one processor and memory;
the memory stores computer-executable instructions;
the at least one processor executing the computer-executable instructions stored by the memory causes the at least one processor to perform the method as set forth in the first aspect and various possible designs of the first aspect, or as set forth in the second aspect and various possible designs of the second aspect.
A sixth aspect of the present application provides a computer-readable storage medium having stored thereon computer-executable instructions that, when executed by a processor, perform a method as set forth in the first aspect and various possible designs of the first aspect, or as set forth in the second aspect and various possible designs of the second aspect.
This application technical scheme has following advantage:
according to the gastroscope image classification model construction method and the gastroscope image classification method, the diagnosis result of a patient and a gastroscope image set are obtained; wherein the gastroscopic image set comprises a plurality of gastroscopic images; inputting the diagnosis result and the gastroscope image set into a preset original classification model to pre-train the original classification model to obtain a pre-training classification model; and constructing a target classification model according to the model parameters of the pre-training classification model. According to the gastroscope image classification model building method provided by the technical scheme, the target classification model is built according to the model parameters of the classification model pre-trained by the multiple pictures, the subsequent training time of the target classification model is saved, the building efficiency of the target classification model is improved, and the multiple gastroscope images corresponding to the diagnosed patient are collected by taking the patient as a sample statistical unit, so that the labor labeling cost is saved.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art according to these drawings.
FIG. 1 is a schematic structural diagram of a gastroscope image classification model construction system based on an embodiment of the application;
FIG. 2 is a schematic structural diagram of a gastroscopic image classification system based on an embodiment of the present application;
FIG. 3 is a schematic flow chart of a gastroscope image classification model construction method provided in the embodiments of the present application;
FIG. 4 is a schematic structural diagram of an exemplary pre-trained classification model provided in an embodiment of the present application;
FIG. 5 is a schematic structural diagram of an exemplary deep learning feature extraction module provided in an embodiment of the present application;
FIG. 6 is a schematic flow chart illustrating a gastroscopic image classification method according to an embodiment of the present application;
FIG. 7 is a schematic structural diagram of a gastroscope image classification model construction device provided by an embodiment of the application;
FIG. 8 is a schematic structural diagram of a gastroscopic image classification device provided by an embodiment of the present application;
fig. 9 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
With the above figures, there are shown specific embodiments of the present application, which will be described in more detail below. These drawings and written description are not intended to limit the scope of the disclosed concepts in any way, but rather to illustrate the concepts of the disclosure to those skilled in the art by reference to specific embodiments.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some embodiments of the present application, but not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
Furthermore, the terms "first", "second", etc. are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. In the description of the following examples, "plurality" means two or more unless specifically limited otherwise.
In the prior art, a model training method generally needs to collect massive labeled samples and then sequentially perform model training by taking a single labeled sample as a main body. However, the labeling process of the sample consumes a large amount of human resources, and the subsequent model training process also consumes a large amount of time and cost, so that the model construction efficiency is reduced.
In order to solve the above problems, the gastroscope image classification model construction method and the gastroscope image classification method provided by the embodiment of the application acquire a diagnosis result of a patient and a gastroscope image set; wherein the gastroscopic image set comprises a plurality of gastroscopic images; inputting the diagnosis result and the gastroscope image set into a preset original classification model to pre-train the original classification model to obtain a pre-training classification model; and constructing a target classification model according to the model parameters of the pre-training classification model. According to the gastroscope image classification model building method provided by the technical scheme, the target classification model is built according to the model parameters of the classification model pre-trained by the multiple pictures, the subsequent training time of the target classification model is saved, the building efficiency of the target classification model is improved, and the multiple gastroscope images corresponding to the diagnosed patient are collected by taking the patient as a sample statistical unit, so that the labor labeling cost is saved.
The following several specific embodiments may be combined with each other, and details of the same or similar concepts or processes may not be repeated in some embodiments. Embodiments of the present invention will be described below with reference to the accompanying drawings.
First, the structure of a gastroscope image classification model construction system based on the present application will be explained:
the gastroscope image classification model building method provided by the embodiment of the application is suitable for building a target classification model capable of classifying gastroscope images. As shown in fig. 1, a schematic structural diagram of a gastroscope image classification model construction system according to an embodiment of the present application mainly includes a sample collection device and a gastroscope image classification model construction device. Specifically, a sample collecting device is used for collecting a diagnosis result of a patient with confirmed diagnosis and a corresponding gastroscope image, and the collected data are sent to a gastroscope image classification model building device, so that the gastroscope image classification model can use the diagnosis result and the corresponding gastroscope image to carry out model training, and further build a target classification model.
Next, a description will be given of a configuration of a gastroscopic image classification system based on the present application:
the gastroscope image classification method provided by the embodiment of the application is suitable for classifying gastroscope images. Fig. 2 is a schematic structural diagram of a gastroscopic image classification system based on an embodiment of the present application, and mainly includes an image acquisition device and a gastroscopic image classification device. Specifically, an image acquisition device is used for acquiring gastroscope images to be classified, the acquired gastroscope images are sent to a gastroscope image classification device, the gastroscope images to be classified are classified by the gastroscope image classification device, and corresponding classification results are obtained.
The embodiment of the application provides a gastroscope image classification model construction method which is used for constructing a target classification model capable of classifying gastroscope images. The implementation subject of the embodiment of the present application is an electronic device, such as a server, a desktop computer, a notebook computer, a tablet computer, and other electronic devices that can be used to construct a gastroscopic image classification model.
As shown in fig. 3, a schematic flowchart of a method for constructing a gastroscope image classification model according to an embodiment of the present application is shown, where the method includes:
Wherein the gastroscopic image set comprises a plurality of gastroscopic images.
It should be noted that the diagnosis result is at least divided into gastric cancer, gastric ulcer and normal, and the gastroscopic image set includes a plurality of gastroscopic images with different angles or different resolutions.
Specifically, the diagnosis result and the corresponding gastroscope image set are directly input into the original classification model, and identification does not need to be added to the training samples before the diagnosis result and the original classification model provided by the embodiment of the application can automatically and uniformly mark the gastroscope images in the gastroscope image set according to the diagnosis result so as to complete the subsequent multi-image pre-training process.
It should be noted that the pre-training classification model adopts a multi-picture pre-training network, and the target classification model adopts a single-picture adaptive network.
Specifically, the weights obtained by the multi-picture pre-training network training can be migrated to the single-picture adaptive network, so as to obtain the target classification model.
For the target classification model provided by the embodiment of the application, the model parameters of the pre-training classification model subjected to multi-picture pre-training are obtained before single-picture training, so that higher single-picture classification precision can be achieved only by inputting a small amount of single-picture sample training, the demand of subsequent training samples is reduced, and the cost of sample labeling is reduced.
Specifically, in one embodiment, to further optimize the target classification model, multiple gastroscopic image samples may be acquired; the gastroscope image sample comprises a gastroscope image and a sample label corresponding to the gastroscope image; and inputting the gastroscope image sample into the target classification model, and training the target classification model to optimize the target classification model.
It should be noted that, since the input of the pre-trained classification model provided in the above embodiment is multi-image, that is, each time a plurality of gastroscope images are required to be input, the whole classification result can be obtained, but in practical application, a single gastroscope image is generally classified, and therefore, in order to further improve the generalization ability of the target classification model, even if the target classification model already has the image classification ability of the pre-trained classification, further optimization training is required. The specific optimization training process may refer to the prior art, and the embodiments of the present application are not limited.
On the basis of the foregoing embodiment, in order to ensure the reliability of the pre-trained classification model, as an implementable manner, in an embodiment, the pre-training of the original classification model includes:
step 3021, extracting image features of each gastroscopic image in the gastroscopic image set;
step 3022, constructing a feature matrix of the diagnosis result according to the image features;
step 3023, determining a classification weight of each image feature according to the image quality of each gastroscopic image;
step 3024, performing dimension reduction processing on the feature matrix according to the classification weight to obtain a target feature vector corresponding to the feature matrix;
and step 3025, pre-training the original classification model by using the target feature vector.
It should be noted that the gastroscope images in the gastroscope image set are gastroscope images corresponding to different acquisition angles or different resolutions, some gastroscope images have lesion features, some gastroscope images may not have lesion features, and each gastroscope image obtained cannot be guaranteed to be clear due to the influence of the performance of the gastroscope image acquisition equipment.
Therefore, in order to avoid the influence of the low-quality images on the reliability of the pre-trained classification model and adapt to the characteristic of single-image input of the subsequent target classification model, in the pre-training process, the classification weight of the image features corresponding to each gastroscope image can be determined according to factors such as the image quality of each gastroscope image, and the like, so that a reliable target feature vector is obtained, and the reliability of the pre-trained classification model is improved.
As shown in fig. 4, a schematic structural diagram of an exemplary pre-training classification model provided in the embodiment of the present application is shown, where the pre-training classification model is composed of a deep learning feature extraction module + attention, where the deep learning feature extraction module has multiple implementation manners, such as ResNet, EfficientNet, and the like.
Illustratively, the embodiment of the present application may use ResNet-50-V2-BiT to implement a deep learning feature extraction module, which mainly consists of ResNet-50 base, a total of 50 weighted layers, and convolution kernels 1 × 1, 3 × 3 and 1 × 1. Fig. 5 is a schematic structural diagram of an exemplary deep learning feature extraction module provided in an embodiment of the present application, where ResNet-50 is mainly divided into 6 parts, which are respectively composed of conv1, conv2_ x, conv3_ x, conv4_ x, conv5_ x, and a final global average pooling layer and a full connection layer part. The global average pooling layer maps the two-dimensional characteristic map into vectors and inputs the vectors into a full-connection layer, the full-connection layer comprises two layers, the first layer of the full-connection layer comprises 1000 neurons, the second layer of the full-connection layer comprises 100 neurons, and the classification result P corresponding to the input group of images is obtained after the two layers of the full-connection layer are put into an attention layer. Wherein P is [ P1, P2], P1 is the probability that the gastroscope image is gastric cancer, P2 is the probability that the gastroscope image is gastric ulcer, P1+ P2 is 1, when P1> is 0.5, P2 ═ 0.5 represents that the gastroscope image is judged to be gastric cancer image by the pre-training classification model, when P1<0.5, P2>0.5 represents that the gastroscope image is judged to be gastric ulcer image by the pre-training classification model, and the prediction result r' and the labeling result r are fitted by Cross entopy Loss function.
The training loss function formula adopted in the embodiment of the application is as follows:
where CE is the Cross Encopy loss function, ytrue(xi) Is picture xiOutput probability of ypred(xi) Is picture xiThe corresponding sample true label.
On the basis of the foregoing embodiment, as an implementable manner, in an embodiment, constructing a target classification model according to model parameters of a pre-trained classification model includes:
step 3031, obtaining an initial target classification model;
step 3032, transferring the parameter weight of the pre-training classification model to the initial target classification model;
the pre-training classification model comprises an input layer, a machine learning layer and an output layer, and the model parameters at least comprise model parameters corresponding to the machine learning layer.
It should be noted that the machine learning layer corresponds to the deep learning feature extraction module.
Illustratively, the construction process of the target classification model is also a migration process from a multi-picture pre-training network to a single-picture adaptation network, and the single-picture adaptation network is composed of the same deep learning feature extraction module and an output layer as those in the multi-picture pre-training network, where the output layer may have multiple implementation manners, such as a linear network and a softmax layer, or an attention layer in the multi-picture pre-training network is also used. If the attribute layer is selected to be used as well in the embodiment of the application, the structure of the model does not need to be modified, and only when the corpus is put into the model, the model is modified to be put into a single picture for training at a time, and finally, the classification result P of one image is correspondingly input.
On the basis of the foregoing embodiment, in order to further ensure the reliability of the constructed target classification model, as an implementable manner, in an embodiment, before constructing the target classification model according to the model parameters of the pre-trained classification model, the method further includes:
step 401, obtaining an evaluation index of a current pre-training classification model;
step 402, judging whether the current pre-training model is in a stable state or not according to the evaluation index;
in step 403, when the current pre-trained model is in an unstable state, the method returns to the step of obtaining the diagnosis result and the gastroscope image set of the patient.
Conversely, when the current pre-trained model is in a stable state, the step of constructing the target classification model according to the model parameters of the pre-trained classification model is performed.
It should be noted that the evaluation index may be an F1 score of the model, where F1 ═ 2 × (precision rate ×/(precision rate + recall rate), and the evaluation index may also be another index that can characterize the current performance of the model, and this embodiment of the present application is not limited specifically.
Specifically, if the evaluation index changes to a small extent in the evaluation period, that is, the evaluation index has reached equilibrium and has reached the expected standard, it may be determined that the current pre-trained model is already in a stable state, and at this time, further weight migration work (constructing the target classification model) may be performed. Otherwise, training is continued.
Specifically, in one embodiment, in order to ensure the image quality of the training sample, before inputting the training sample (gastroscope image set or gastroscope image sample) into the model to be trained (original classification model or target classification model), the training sample is first subjected to image preprocessing, specifically, the picture may be compressed to 256 × 3 size, then cut to 224 × 3 size, cut out four redundant black edges of the gastroscope image, then subjected to random horizontal flipping to correct the gastroscope image, and finally subjected to normalization.
For example, the classification model provided in this embodiment of the present application may be set to have an initial learning rate of 0.00001, beta (0.9,0.999), weight _ decade (0.01), and an Adam optimizer is used to optimize parameters of the network, and a label-level Loss function is used as a Loss function of the model, and a Cross Entropy Loss function (Cross Entropy Loss) is used as the label-level Loss function. The batch sizes were set to 16 (multiple pictures) and 1 (single picture). To prevent overfitting, Dropout was used to randomly ignore some of the neurons in the fully connected layer. During training, the multi-picture gastroscope classification network is trained on the basis of a multi-picture training sample (gastroscope image set), the network is enabled to learn a multi-picture gastric cancer/gastric ulcer detection task, and then an optimizer is used for minimizing a loss function of a characteristic level and a label level of image classification, so that the network is finally converged. And then migrating to a single-picture task, training by using a single-picture few samples, and minimizing the loss function of the feature level and the label level of image classification by using an optimizer to finally converge the network.
According to the gastroscope image classification model construction method provided by the embodiment of the application, a diagnosis result of a patient and a gastroscope image set are obtained; wherein the gastroscopic image set comprises a plurality of gastroscopic images; inputting the diagnosis result and the gastroscope image set into a preset original classification model to pre-train the original classification model to obtain a pre-training classification model; and constructing a target classification model according to the model parameters of the pre-training classification model. According to the gastroscope image classification model building method provided by the technical scheme, the target classification model is built according to the model parameters of the classification model pre-trained by the multiple pictures, the subsequent training time of the target classification model is saved, the building efficiency of the target classification model is improved, and the multiple gastroscope images corresponding to the diagnosed patient are collected by taking the patient as a sample statistical unit, so that the labor labeling cost is saved. Moreover, the reliability of the pre-training classification model is prevented from being influenced by the low-quality images, and the reliability of the pre-training classification model is improved.
The embodiment of the application provides a gastroscope image classification method which is used for automatically classifying gastroscope images. The execution subject of the embodiment of the application is an electronic device, such as a server, a desktop computer, a notebook computer, a tablet computer and other electronic devices which can be used for gastroscopic image classification.
As shown in fig. 6, a schematic flowchart of a gastroscopic image classification method provided in an embodiment of the present application is shown, and the method includes:
601, acquiring gastroscope images to be classified;
The gastroscope image to be classified is the gastroscope image of the current subject, and the gastroscope image to be classified is classified by utilizing the target classification model so as to determine the diagnosis result corresponding to the gastroscope image.
According to the gastroscope image classification method provided by the embodiment of the application, the gastroscope image to be classified is obtained; and inputting the gastroscope image to be classified into the target classification model constructed by the gastroscope image classification model construction method of the embodiment to obtain a classification result of the gastroscope image to be classified. The gastroscope image classification method provided by the scheme is realized based on the target classification model obtained by the embodiment, the reliability of the target classification model is high, the gastroscope image classification precision is improved, and the accuracy of the classification result is further improved.
The embodiment of the application provides a gastroscope image classification model construction device which is used for executing the gastroscope image classification model construction method provided by the embodiment.
Fig. 7 is a schematic structural diagram of a gastroscope image classification model construction device provided in the embodiment of the present application. The gastroscope image classification model construction device 70 comprises a first acquisition module 701, a pre-training module 702 and a construction module 703.
The first acquisition module is used for acquiring a diagnosis result and a gastroscope image set of a patient; wherein the gastroscopic image set comprises a plurality of gastroscopic images; the pre-training module is used for inputting the diagnosis result and the gastroscope image set into a preset original classification model so as to pre-train the original classification model to obtain a pre-training classification model; and the construction module is used for constructing a target classification model according to the model parameters of the pre-training classification model.
With regard to the gastroscopic image classification model construction apparatus in the present embodiment, the specific manner in which the respective modules perform operations has been described in detail in the embodiment related to the method, and will not be explained in detail here.
The gastroscope image classification model construction device provided by the embodiment of the application is used for executing the gastroscope image classification model construction method provided by the embodiment, the implementation mode and the principle are the same, and the details are not repeated.
The embodiment of the application provides a gastroscope image classification device which is used for executing the gastroscope image classification method provided by the embodiment.
Fig. 8 is a schematic structural diagram of a gastroscopic image classification device according to an embodiment of the present application. The gastroscopic image classification device 80 includes a second acquisition module 801 and a classification module 802.
The second acquisition module is used for acquiring gastroscope images to be classified; and the classification module is used for inputting the gastroscope image to be classified into the target classification model constructed by the gastroscope image classification model construction device provided by the embodiment so as to obtain the classification result of the gastroscope image to be classified.
With regard to the gastroscopic image classification apparatus in the present embodiment, the specific manner in which the respective modules perform the operations has been described in detail in the embodiment related to the method, and will not be explained in detail here.
The gastroscope image classification device provided by the embodiment of the application is used for executing the gastroscope image classification method provided by the embodiment, the implementation mode and the principle are the same, and the repeated description is omitted.
The embodiment of the application provides electronic equipment for executing the gastroscope image classification model construction method or the gastroscope image classification method provided by the embodiment.
Fig. 9 is a schematic structural diagram of an electronic device according to an embodiment of the present application. The electronic device 90 includes: at least one processor 91 and memory 92;
the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored by the memory to cause the at least one processor to perform a gastroscopic image classification model construction method or a gastroscopic image classification method as provided by the above embodiments.
The electronic device provided by the embodiment of the application is used for executing the gastroscope image classification model construction method or the gastroscope image classification method provided by the embodiment, the implementation mode and the principle are the same, and the details are not repeated.
The embodiment of the application provides a computer-readable storage medium, wherein a computer executing instruction is stored in the computer-readable storage medium, and when a processor executes the computer executing instruction, the gastroscope image classification model building method or the gastroscope image classification method provided by any one of the above embodiments is realized.
The storage medium containing the computer-executable instructions of the embodiment of the present application may be used to store the computer-executable instructions of the gastroscope image classification model construction method or the gastroscope image classification method provided in the foregoing embodiments, and the implementation manner and the principle thereof are the same and are not described again.
It is obvious to those skilled in the art that, for convenience and simplicity of description, the foregoing division of the functional modules is merely used as an example, and in practical applications, the above function distribution may be performed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to perform all or part of the above described functions. For the specific working process of the device described above, reference may be made to the corresponding process in the foregoing method embodiment, which is not described herein again.
Finally, it should be noted that: the above embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present application.
Claims (10)
1. A gastroscope image classification model construction method is characterized by comprising the following steps:
acquiring a diagnosis result and a gastroscope image set of a patient; wherein the gastroscopic image set comprises a plurality of gastroscopic images;
inputting the diagnosis result and the gastroscope image set into a preset original classification model so as to pre-train the original classification model to obtain a pre-training classification model;
and constructing a target classification model according to the model parameters of the pre-training classification model.
2. The method of claim 1, further comprising:
acquiring a plurality of gastroscopic image samples; wherein the gastroscope image sample comprises a gastroscope image and a sample label corresponding to the gastroscope image;
and inputting the gastroscope image sample into the target classification model, and training the target classification model to optimize the target classification model.
3. The method of claim 1, wherein pre-training the raw classification model comprises:
extracting image characteristics of each gastroscope image in the gastroscope image set;
constructing a feature matrix of the diagnosis result according to the image features;
determining the classification weight of each image characteristic according to the image quality of each gastroscope image;
performing dimension reduction processing on the feature matrix according to the classification weight to obtain a target feature vector corresponding to the feature matrix;
and pre-training the original classification model by using the target feature vector.
4. The method of claim 1, wherein constructing a target classification model from the model parameters of the pre-trained classification model comprises:
obtaining an initial target classification model;
migrating the parameter weights of the pre-trained classification model to the initial target classification model;
the pre-training classification model comprises an input layer, a machine learning layer and an output layer, and the model parameters at least comprise model parameters corresponding to the machine learning layer.
5. The method of claim 1, wherein prior to constructing a target classification model from the model parameters of the pre-trained classification model, the method further comprises:
obtaining an evaluation index of a current pre-training classification model;
judging whether the current pre-training model is in a stable state or not according to the evaluation index;
returning to the step of obtaining a diagnosis of the patient and a set of gastroscopic images when the current pre-trained model is in an unstable state.
6. A gastroscopic image classification method comprising:
acquiring a gastroscope image to be classified;
inputting the gastroscope image to be classified into the target classification model constructed by the gastroscope image classification model construction method according to any one of claims 1 to 5 to obtain the classification result of the gastroscope image to be classified.
7. A gastroscope image classification model construction device is characterized by comprising:
a first acquisition module for acquiring a diagnosis result of a patient and a gastroscope image set; wherein the gastroscopic image set comprises a plurality of gastroscopic images;
the pre-training module is used for inputting the diagnosis result and the gastroscope image set into a preset original classification model so as to pre-train the original classification model to obtain a pre-training classification model;
and the construction module is used for constructing a target classification model according to the model parameters of the pre-training classification model.
8. A gastroscopic image classification device comprising:
the second acquisition module is used for acquiring gastroscope images to be classified;
a classification module, configured to input the gastroscope image to be classified into the target classification model constructed by the gastroscope image classification model construction device according to claim 7, so as to obtain a classification result of the gastroscope image to be classified.
9. An electronic device, comprising: at least one processor and memory;
the memory stores computer-executable instructions;
the at least one processor executing the computer-executable instructions stored by the memory causes the at least one processor to perform the method of any one of claims 1 to 5 or the method of claim 6.
10. A computer-readable storage medium having computer-executable instructions stored thereon which, when executed by a processor, implement the method of any one of claims 1 to 5 or the method of claim 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110649737.6A CN113327238A (en) | 2021-06-10 | 2021-06-10 | Gastroscope image classification model construction method and gastroscope image classification method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110649737.6A CN113327238A (en) | 2021-06-10 | 2021-06-10 | Gastroscope image classification model construction method and gastroscope image classification method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113327238A true CN113327238A (en) | 2021-08-31 |
Family
ID=77420512
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110649737.6A Pending CN113327238A (en) | 2021-06-10 | 2021-06-10 | Gastroscope image classification model construction method and gastroscope image classification method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113327238A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113689430A (en) * | 2021-10-26 | 2021-11-23 | 紫东信息科技(苏州)有限公司 | Image processing method and device for enteroscopy state monitoring |
| CN114795258A (en) * | 2022-04-18 | 2022-07-29 | 浙江大学 | Child hip joint dysplasia diagnosis system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200104984A1 (en) * | 2018-09-29 | 2020-04-02 | Shanghai United Imaging Intelligence Co., Ltd. | Methods and devices for reducing dimension of eigenvectors |
| WO2020215593A1 (en) * | 2019-04-22 | 2020-10-29 | 武汉楚精灵医疗科技有限公司 | Artificial intelligence-based automatic evaluation method and system on quality check of gastrointestinal endoscopy |
| CN112364926A (en) * | 2020-11-17 | 2021-02-12 | 苏州大学 | Gastroscope picture classification method and device based on ResNet-50 time compression and storage medium |
| CN112784801A (en) * | 2021-02-03 | 2021-05-11 | 紫东信息科技(苏州)有限公司 | Text and picture-based bimodal gastric disease classification method and device |
| CN112786160A (en) * | 2021-02-03 | 2021-05-11 | 紫东信息科技(苏州)有限公司 | Multi-image input multi-label gastroscope image classification method based on graph neural network |
-
2021
- 2021-06-10 CN CN202110649737.6A patent/CN113327238A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200104984A1 (en) * | 2018-09-29 | 2020-04-02 | Shanghai United Imaging Intelligence Co., Ltd. | Methods and devices for reducing dimension of eigenvectors |
| WO2020215593A1 (en) * | 2019-04-22 | 2020-10-29 | 武汉楚精灵医疗科技有限公司 | Artificial intelligence-based automatic evaluation method and system on quality check of gastrointestinal endoscopy |
| CN112364926A (en) * | 2020-11-17 | 2021-02-12 | 苏州大学 | Gastroscope picture classification method and device based on ResNet-50 time compression and storage medium |
| CN112784801A (en) * | 2021-02-03 | 2021-05-11 | 紫东信息科技(苏州)有限公司 | Text and picture-based bimodal gastric disease classification method and device |
| CN112786160A (en) * | 2021-02-03 | 2021-05-11 | 紫东信息科技(苏州)有限公司 | Multi-image input multi-label gastroscope image classification method based on graph neural network |
Non-Patent Citations (2)
| Title |
|---|
| 张菁，钟绿，杜岗等: "基于迁移学习的胃镜图像识别模型的构建及其在胃癌诊断中的应用", 第二军医大学学报, vol. 40, no. 05, pages 1 - 9 * |
| 蒋强;沈林;张伟;何旭;: "基于深度学习的故障诊断方法研究", 计算机仿真, no. 07 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113689430A (en) * | 2021-10-26 | 2021-11-23 | 紫东信息科技(苏州)有限公司 | Image processing method and device for enteroscopy state monitoring |
| CN114795258A (en) * | 2022-04-18 | 2022-07-29 | 浙江大学 | Child hip joint dysplasia diagnosis system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112580439B (en) | Large-format remote sensing image ship target detection method and system under small sample condition | |
| CN109685765B (en) | X-ray film pneumonia result prediction device based on convolutional neural network | |
| CN112633382B (en) | Method and system for classifying few sample images based on mutual neighbor | |
| CN110738235B (en) | Pulmonary tuberculosis judging method, device, computer equipment and storage medium | |
| CN113327238A (en) | Gastroscope image classification model construction method and gastroscope image classification method | |
| CN110765882B (en) | Video tag determination method, device, server and storage medium | |
| CN110647802A (en) | Remote sensing image ship target detection method based on deep learning | |
| CN110838125A (en) | Target detection method, device, equipment and storage medium of medical image | |
| CN111291825A (en) | Focus classification model training method and device, computer equipment and storage medium | |
| CN111507399A (en) | Cloud recognition and model training method, device, terminal and medium based on deep learning | |
| CN111612051A (en) | Weak supervision target detection method based on graph convolution neural network | |
| CN111027576A (en) | Cooperative significance detection method based on cooperative significance generation type countermeasure network | |
| CN113435269A (en) | Improved water surface floating object detection and identification method and system based on YOLOv3 | |
| CN110807362A (en) | Image detection method and device and computer readable storage medium | |
| CN111274972A (en) | Dish identification method and device based on metric learning | |
| CN111192678A (en) | Pathological microscopic image diagnosis and model training method, device, equipment and medium | |
| CN114494195A (en) | Small sample attention mechanism parallel twinning method for fundus image classification | |
| CN116168240A (en) | Arbitrary-direction dense ship target detection method based on attention enhancement | |
| CN112419326A (en) | Image segmentation data processing method, device, equipment and storage medium | |
| CN116452810A (en) | Multi-level semantic segmentation method and device, electronic equipment and storage medium | |
| CN113487610B (en) | Herpes image recognition method and device, computer equipment and storage medium | |
| CN112991281B (en) | Visual detection method, system, electronic equipment and medium | |
| CN114742750A (en) | Abnormal cell detection method, abnormal cell detection device, terminal device and readable storage medium | |
| WO2021155661A1 (en) | Image processing method and related device | |
| CN116091946A (en) | Yolov 5-based unmanned aerial vehicle aerial image target detection method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |