Image classification methods, apparatus, computer equipment and storage media
By training an image recognition model using the SVM algorithm and a sample image classification catalog management method, the problem of poor training performance of traditional methods with a small number of samples is solved, thereby improving the accuracy of image classification and simplifying the process.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- GUANGDONG BAIYUN UNIV
- Filing Date
- 2023-07-12
- Publication Date
- 2026-07-17
AI Technical Summary
Traditional neural network-based image classification methods perform poorly when training with a small number of samples, are resource-intensive, and have low accuracy.
An image recognition model based on the SVM algorithm is adopted. The model is trained by a sample image classification directory management method. The training process is simplified by combining the visual operation of the operation interface. The folder management method of the sample image classification directory is used to automatically label images and train the model.
The image classification was optimized with a small number of samples, improving the accuracy of image classification, saving manpower and resources, and simplifying the training and application process through the visual operation of the user interface.
Smart Images
Figure CN117115507B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of computer technology, and in particular to an image classification method, apparatus, computer equipment, storage medium, and computer program product. Background Technology
[0002] With the development of artificial intelligence, its error rate in some aspects of computer vision has already been lower than that of human eye recognition, showing good application prospects, such as the ability to identify disease types based on medical images.
[0003] Traditional techniques typically employ neural network-based model training, which requires manual labeling of training samples beforehand. The training effect relies on a large number of samples, resulting in high manpower and material resources, high training costs, poor training performance with a small number of samples, and poor image classification accuracy. Summary of the Invention
[0004] Therefore, it is necessary to provide an image classification method, apparatus, computer equipment, storage medium, and computer program product that can improve the accuracy of image classification when training with a small number of samples, in order to address the above-mentioned technical problems.
[0005] In a first aspect, this application provides an image classification method, the method comprising:
[0006] The interface of the image classification program is displayed; the interface includes training trigger controls and training directory information for model training, the training directory information being used to indicate at least one sample image classification directory to be called during training;
[0007] In response to the operation of the training trigger control, the model is trained using sample medical images from the at least one sample image classification directory to obtain an image recognition model; the feature category to which the sample medical image in each sample image classification directory belongs is consistent with the feature category corresponding to each sample image classification directory.
[0008] When an execution event of an image classification task is detected, the image recognition model is used to identify the target medical image in the image classification task, and the feature category classification result of the target medical image is displayed in the result display area of the operation interface.
[0009] In one embodiment, the user interface further includes a training catalog configuration control for model training. Prior to the step of training the model using sample medical images from the at least one sample image classification catalog in response to an operation of the training trigger control to obtain an image recognition model, the method further includes:
[0010] In response to an operation on the training directory configuration control, at least one sample image classification directory for training invocation is determined;
[0011] The directory location of the at least one sample image classification directory is displayed as the training directory information.
[0012] In one embodiment, prior to the step of training the model using sample medical images from the at least one sample image classification catalog in response to an operation of the training trigger control to obtain an image recognition model, the method further includes:
[0013] According to the preset feature category classification information, create sample image classification directories corresponding to different feature categories; the directory name of the sample image classification directory is used to represent the corresponding feature category;
[0014] In response to the operation of adding an image to any sample image classification directory, the sample medical images belonging to the feature category corresponding to the any sample image classification directory are stored in the any sample image classification directory.
[0015] In one embodiment, the method further includes:
[0016] During the operation of the image classification program, the medical images under each of the sample image classification directories are labeled according to the directory name of each sample image classification directory, thus obtaining the labeled medical images.
[0017] The step of training a model using sample medical images from at least one sample image classification catalog to obtain an image recognition model includes:
[0018] The labeled medical images from the at least one sample image classification directory are input into the training model based on the SVM algorithm to obtain the image recognition model.
[0019] In one embodiment, the user interface further includes a task execution control and a task directory configuration control for the classification task, wherein detecting the execution event of the image classification task includes:
[0020] In response to an operation on the task directory configuration control, the task image directory for the image classification task is determined, and the directory location of the task image directory is displayed.
[0021] The target medical image is obtained by reading the image according to its directory location in the task image directory;
[0022] In response to an operation on the task execution control, the image classification task is executed.
[0023] In one embodiment, the target medical image includes multiple images, and the feature category classification results of the target medical image are displayed in the result display area of the operation interface, including:
[0024] For any given image, the result display area of the operation interface displays the directory location of the task image directory corresponding to that image and the classification probability of that image under at least one feature category.
[0025] The user interface also includes export controls and export directory information for exporting results, and the method further includes:
[0026] Obtain object description information from the image classification task; the object description information includes description information of at least one object;
[0027] In response to the operation of the export control, the description information of the same object and the feature category classification result of the target medical image are associated and stored according to the export directory information.
[0028] Secondly, this application also provides an image classification apparatus, the apparatus comprising:
[0029] The image classification interface display module is used to display the operation interface of the image classification program; the operation interface includes training trigger controls and training directory information for model training, and the training directory information is used to indicate at least one sample image classification directory called during training.
[0030] The model training module is used to train the model using sample medical images from the at least one sample image classification directory in response to the operation of the training trigger control, so as to obtain an image recognition model; the feature category to which the sample medical image in each sample image classification directory belongs is consistent with the feature category corresponding to each sample image classification directory;
[0031] The image classification task execution module is used to identify the target medical image in the image classification task using the image recognition model when the execution event of the image classification task is detected, and to display the feature category classification result of the target medical image in the result display area of the operation interface.
[0032] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:
[0033] The interface of the image classification program is displayed; the interface includes training trigger controls and training directory information for model training, the training directory information being used to indicate at least one sample image classification directory to be called during training;
[0034] In response to the operation of the training trigger control, the model is trained using sample medical images from the at least one sample image classification directory to obtain an image recognition model; the feature category to which the sample medical image in each sample image classification directory belongs is consistent with the feature category corresponding to each sample image classification directory.
[0035] When an execution event of an image classification task is detected, the image recognition model is used to identify the target medical image in the image classification task, and the feature category classification result of the target medical image is displayed in the result display area of the operation interface.
[0036] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:
[0037] The interface of the image classification program is displayed; the interface includes training trigger controls and training directory information for model training, the training directory information being used to indicate at least one sample image classification directory to be called during training;
[0038] In response to the operation of the training trigger control, the model is trained using sample medical images from the at least one sample image classification directory to obtain an image recognition model; the feature category to which the sample medical image in each sample image classification directory belongs is consistent with the feature category corresponding to each sample image classification directory.
[0039] When an execution event of an image classification task is detected, the image recognition model is used to identify the target medical image in the image classification task, and the feature category classification result of the target medical image is displayed in the result display area of the operation interface.
[0040] Fifthly, this application also provides a computer program product. The computer program product includes a computer program, which, when executed by a processor, performs the following steps:
[0041] The interface of the image classification program is displayed; the interface includes training trigger controls and training directory information for model training, the training directory information being used to indicate at least one sample image classification directory to be called during training;
[0042] In response to the operation of the training trigger control, the model is trained using sample medical images from the at least one sample image classification directory to obtain an image recognition model; the feature category to which the sample medical image in each sample image classification directory belongs is consistent with the feature category corresponding to each sample image classification directory.
[0043] When an execution event of an image classification task is detected, the image recognition model is used to identify the target medical image in the image classification task, and the feature category classification result of the target medical image is displayed in the result display area of the operation interface.
[0044] The aforementioned image classification method, apparatus, computer device, storage medium, and computer program product, by displaying an operation interface for the image classification program, includes a training trigger control and training directory information for model training. The training directory information indicates at least one sample image classification directory for training. Then, in response to the operation of the training trigger control, the model is trained using sample medical images from at least one sample image classification directory to obtain an image recognition model. The feature category of each sample medical image in each sample image classification directory is consistent with the feature category corresponding to each sample image classification directory. Furthermore, when an execution event of the image classification task is detected, the image recognition model identifies the target medical image in the image classification task, and the feature category classification result of the target medical image is displayed in the result display area of the operation interface. This achieves image classification optimization in scenarios with a small number of samples. The image recognition model can be trained using a sample image classification directory management method, and the related processing of training and application can be simplified based on the visual operation of the operation interface. It can achieve automatic image recognition and accurate classification, effectively improving the accuracy of image classification. Attached Figure Description
[0045] Figure 1 This is a flowchart illustrating an image classification method in one embodiment;
[0046] Figure 2a This is a schematic diagram of the image classification processing flow in one embodiment;
[0047] Figure 2b This is a schematic diagram of the user interface in one embodiment;
[0048] Figure 3a This is a schematic diagram of a sample image classification catalog in one embodiment;
[0049] Figure 3b This is a schematic diagram of the SVM algorithm in one embodiment;
[0050] Figure 4 This is a schematic diagram of the user interface in another embodiment;
[0051] Figure 5 This is a flowchart illustrating an image classification method in another embodiment;
[0052] Figure 6 This is a structural block diagram of an image classification device in one embodiment;
[0053] Figure 7 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation
[0054] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0055] In one embodiment, such as Figure 1 As shown, an image classification method is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, and further to a system including both a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:
[0056] Step 101: Display the user interface of the image classification program;
[0057] As an example, the user interface can be displayed in response to a trigger action that runs an image classification program, such as... Figure 2a As shown, users can double-click the executable program to run the system, which triggers the image classification program. Alternatively, other triggering conditions can be used, but no specific restrictions are imposed in this embodiment.
[0058] The user interface may include training trigger controls and training directory information for model training. The training directory information can be used to indicate at least one sample image classification directory to be called during training.
[0059] In practical applications, the user interface of an image classification program can be displayed in response to a trigger operation that runs the program; for example, such as... Figure 2b As shown, after starting the system program, the displayed interface may include training trigger controls for model training (such as...). Figure 2b (In the middle of "Start Training") and training catalog information (such as "Start Training") and training catalog information (such as Figure 2b The "Training Image Directory Location" field allows you to obtain the directory location of at least one sample image classification directory required for further training, such as the path of a folder directory, based on the training directory information.
[0060] Step 102: In response to the operation of the training trigger control, the model is trained using sample medical images from the at least one sample image classification catalog to obtain an image recognition model;
[0061] In this context, the feature category to which the sample medical image belongs in each sample image classification directory is consistent with the feature category corresponding to each sample image classification directory; for example, taking the sample medical image as a cervical medical examination image, the feature category can be the classification level of cervical lesions, and different levels can represent different degrees of lesions.
[0062] As an example, the sample image classification directory can be a pre-built folder based on different feature categories, and the name of the folder can be set to the corresponding feature category.
[0063] In the specific implementation, based on the operation interface of the image classification program, the image recognition model can be started by responding to the operation of the training trigger control. By using sample medical images from at least one sample image classification directory to train the model, the image recognition model can be obtained.
[0064] In one example, such as Figure 2b As shown, users can click the "Start Training" button (i.e., the training trigger control) to initiate the training operation, and then the system can... Figure 2b The "Training Image Directory Location" option retrieves images for training (i.e., at least one sample medical image from a sample image classification directory) to train the model, resulting in an image recognition model. The model template file can be automatically saved after training, for example, the trained model can be saved as my_model.m in the current program directory.
[0065] In another example, the image recognition model can be obtained based on the SVM (support vector machine) algorithm in machine learning. Since the SVM algorithm, with the support of the OpenCV image processing function library, has a better training effect than neural networks when there are few samples, the SVM algorithm can be used to achieve automatic and accurate identification of cervical diseases in a small number of samples.
[0066] In one alternative embodiment, an image recognition model can be obtained for image classification when applying a classification task by pre-training the model; alternatively, an image recognition model can be performed after training the model for the current classification task; alternatively, the training model can be updated intermittently. For example, the training image recognition model can be updated when a new medical image is detected in the sample image classification directory or when a medical image is replaced or changed. Alternatively, a preset time range can be set for updating the training, which can help maintain the effectiveness of the model.
[0067] Step 103: When an execution event of an image classification task is detected, the image recognition model is used to identify the target medical image in the image classification task, and the feature category classification result of the target medical image is displayed in the result display area of the operation interface.
[0068] In practical applications, taking the automatic identification of cervical diseases as an example, when the execution event of the image classification task is detected, the image recognition model can be used to identify the target medical image in the image classification task. For example, the cervical medical examination image to be classified can be identified based on the trained model, and the feature category classification result of the target medical image can be displayed in the result display area of the operation interface, such as the probability value of the category to which the cervical medical examination image to be classified belongs.
[0069] Optionally, recognition can be performed on a single cervical medical examination image, or images can be recognized in batches; during the single image recognition process, such as Figure 2b As shown, the patient's CT images (i.e., target medical images) can be stored in the set test image directory, and the patient's information (such as...) can be entered sequentially. Figure 2b This allows users to view various patient information, and then click "Start Classification" to perform image recognition. The recognition information corresponding to the patient's CT images will be displayed in the lower right corner of the interface (e.g., ...). Figure 2b (Results displayed in the area).
[0070] Compared to traditional methods that require annotation and preprocessing of a large number of training samples, the technical solution in this embodiment, by using folder management for the sample image classification directory, can efficiently train the model with a small number of samples, saving manpower and resources, and also providing users with a customizable sample data configuration function; the visual operation based on the user interface simplifies the related processing of training and application, facilitates the automation of image classification processing, and the image recognition model built based on the SVM algorithm enables the automatic image classification and recognition system to achieve automatic and accurate recognition with small samples, effectively improving the image classification accuracy.
[0071] In the above image classification method, the operation interface of the image classification program is displayed. Then, in response to the operation of the training trigger control, the model is trained using sample medical images from at least one sample image classification directory to obtain an image recognition model. Then, when the execution event of the image classification task is detected, the image recognition model is used to identify the target medical image in the image classification task. The feature category classification result of the target medical image is displayed in the result display area of the operation interface. This method realizes image classification optimization in scenarios with a small number of samples. The image recognition model can be trained by managing the sample image classification directory. The related processing of training and application can be simplified by the visual operation of the operation interface. It can achieve automatic recognition and accurate image classification, effectively improving the accuracy of image classification.
[0072] In one embodiment, the user interface may further include a training catalog configuration control for model training. Before the step of training the model using sample medical images from the at least one sample image classification catalog in response to an operation of the training trigger control to obtain an image recognition model, the following steps may also be included:
[0073] In response to an operation on the training directory configuration control, at least one sample image classification directory for training is determined; the directory location of the at least one sample image classification directory is displayed as the training directory information.
[0074] In practical applications, such as Figure 2b As shown, users can click the "Set Training Image Directory" button (i.e., the training directory configuration control) and specify the location of the training images (i.e., the directory location of at least one sample image classification directory) in the pop-up menu. Figure 2b Location of the training image directory.
[0075] In one alternative embodiment, the image format of the images used for training (i.e., sample medical images) can be adjusted before model training, such as unifying image parameters like image size and resolution, to facilitate subsequent model training.
[0076] In this embodiment, by responding to the operation of the training directory configuration control, at least one sample image classification directory for training is determined, and then the directory location of at least one sample image classification directory is displayed as training directory information, which can provide users with flexible configuration functions for customizing sample data.
[0077] In one embodiment, prior to the step of training the model using sample medical images from the at least one sample image classification catalog in response to an operation of the training trigger control to obtain an image recognition model, the following steps may be included:
[0078] According to the preset feature category classification information, create sample image classification directories corresponding to different feature categories; the directory name of the sample image classification directory is used to represent the corresponding feature category; in response to the operation of adding an image under any sample image classification directory, store the sample medical image of the feature category corresponding to the any sample image classification directory into the any sample image classification directory.
[0079] As an example, the sample medical image can be a medical imaging image, such as a CT scan; or it can be a medically related examination image obtained through other means.
[0080] In one example, such as Figure 2aAs shown, a sample image classification directory can be pre-configured during the preparation work before operation. Taking the recognition of cervical medical examination images as an example, preset feature category classification information can be obtained, and file directories can be set according to multiple categories of cervical diseases (i.e., creating sample image classification directories corresponding to different feature categories), which can be used to store the images required for training.
[0081] For example, cervical lesions can be classified into different grades, such as five categories: mild, CIN1, CIN2, CIN3, and CA, corresponding to mild cervical lesions, cervical intraepithelial neoplasia (CIN I), cervical intraepithelial neoplasia (CIN II), cervical intraepithelial neoplasia (CIN III), and cervical cancer, respectively. Figure 3a As shown, folders (i.e., sample image classification directories) corresponding to light, CIN1, CIN2, CIN3, and CA can be created. The folders can be named light, CIN1, CIN2, CIN3, and CA. Cervical medical examination images (i.e. sample medical images) that have been diagnosed by doctors as belonging to the above five categories can be stored in the corresponding directories according to their categories. That is, sample medical images of the feature category corresponding to any sample image classification directory can be stored in any sample image classification directory.
[0082] In this embodiment, by creating sample image classification directories corresponding to different feature categories according to preset feature category classification information, and then responding to the operation of adding an image under any sample image classification directory, the sample medical images of the feature category corresponding to any sample image classification directory are stored in any sample image classification directory, providing data support for subsequent system operation and processing.
[0083] In one embodiment, the following steps may also be included:
[0084] During the operation of the image classification program, the medical images under each of the sample image classification directories are labeled according to the directory name of each sample image classification directory, thus obtaining the labeled medical images.
[0085] In one example, when the system (i.e., the image classification program) is running, by reading the images in the relevant directories, the feature categories of the stored images (i.e., the sample medical images) can be labeled according to the directory names (i.e., the directory names of the sample image classification directories) for further use in model training.
[0086] The step of training a model using sample medical images from at least one sample image classification catalog to obtain an image recognition model includes:
[0087] The labeled medical images from the at least one sample image classification directory are input into the training model based on the SVM algorithm to obtain the image recognition model.
[0088] In practical implementation, images can be automatically labeled using a folder management approach for the sample image classification directory, such as... Figure 2a As shown, at least one labeled medical image from a sample image classification directory can be obtained for model training to obtain an image recognition model based on the SVM algorithm.
[0089] In one example, the SVM (Support Vector Machine) algorithm is a binary classification model whose goal is to find a hyperplane to segment the samples. The principle of segmentation is to maximize the margin, which is then transformed into a convex quadratic programming problem to be solved, such as... Figure 3b The support vectors and maximum gap r are shown.
[0090] According to the definition of the SVM algorithm, the following can be solved by computer, such as the maximum gap r;
[0091]
[0092] Transformed into:
[0093]
[0094] This can then be transformed into:
[0095]
[0096]
[0097] Simultaneously satisfying the KKT conditions:
[0098]
[0099] In another example, the kernel function involved in SVM may include the polynomial kernel function (PF), the radial basis function (RBF), the sigmoid kernel function (SF), and the linear kernel function (LF). Since the polynomial kernel function is more effective than the other three kernel functions, this embodiment may use the polynomial kernel function.
[0100] In this embodiment, by labeling the sample medical images under each sample image classification directory according to the directory name of each sample image classification directory during the operation of the image classification program, the labeled sample medical images are obtained. The images can be automatically labeled based on the folder management method, which improves the processing efficiency.
[0101] In one embodiment, the user interface may further include a task execution control and a task directory configuration control for the classification task. Detecting the execution event of the image classification task may include the following steps:
[0102] In response to an operation on the task directory configuration control, the task image directory for the image classification task is determined, and the directory location of the task image directory is displayed; the image is read according to the directory location of the task image directory to obtain the target medical image; in response to an operation on the task execution control, the image classification task is executed.
[0103] In practical applications, such as Figure 2b As shown, users can click the "Set Test Image Directory" button (i.e., the task directory configuration control) to configure the location of the test image directory where the patient CT images to be processed (i.e., the target medical images) are stored. Then, the patient CT images can be read according to the corresponding directory location. Then, the image recognition can be performed by clicking "Start Classification" (i.e., the task execution control). In other words, the image classification task is executed in response to the operation of the task execution control.
[0104] In this embodiment, by responding to the operation of the task directory configuration control, the task image directory of the image classification task is determined and the directory position of the task image directory is displayed. Then, the image is read according to the directory position of the task image directory to obtain the target medical image. Then, in response to the operation of the task execution control, the image classification task is executed, which can flexibly configure the patient CT images to be processed.
[0105] In one embodiment, the target medical image may include multiple images, and displaying the feature category classification result of the target medical image in the result display area of the operation interface may include the following steps:
[0106] For any given image, the result display area of the operation interface displays the directory location of the task image directory corresponding to that image and the classification probability of that image under at least one feature category.
[0107] In one example, for batch image recognition, such as Figure 4 As shown, the results of batch image recognition can be displayed in the lower right corner of the operation interface (i.e., the result display area), and the recognized results can be saved as files to the file directory corresponding to the configured result save path.
[0108] For example, such as Figure 4The batch image recognition results shown, taking image 1 in the batch processing as an example, can display the directory location and name of image 1 (i.e., the directory location of the task image directory corresponding to any image), and the probability 1, probability 2, and probability 3 corresponding to image 1 (i.e., the classification probability of any image under at least one feature category).
[0109] The user interface may also include export controls and export directory information for exporting results. The method further includes:
[0110] Obtain object description information from the image classification task; the object description information includes description information of at least one object; in response to the operation of the export control, associate and store the description information of the same object with the feature category classification result of the target medical image according to the export directory information.
[0111] In the specific implementation, by obtaining the input patient information (i.e., object description information in the image classification task), the patient's relevant information can be saved as a file to the file directory corresponding to the configured result saving path; for example... Figure 4 As shown, if archiving is required, the patient information can be archived by clicking the Export Results button (i.e., the Export Control) based on the identified feature category classification results and the patient's relevant information.
[0112] For example, for batch image recognition, if archiving is required, the patient information can be archived based on the basic patient information input for batch image recognition and the feature category classification results obtained by clicking the export results button.
[0113] To enable those skilled in the art to better understand the above steps, the following is combined with... Figure 2a An example is provided to illustrate an embodiment of this application, but it should be understood that the embodiments of this application are not limited thereto.
[0114] Let's take recognizing a cervical medical examination image as an example:
[0115] 1. Preparations before system operation: Relevant folders (i.e., sample image classification directories) can be created according to the classification of cervical diseases. Then, the cervical disease images (i.e. sample medical images) identified by hospital experts in the early stage can be stored in the corresponding file directories according to their categories.
[0116] 2. Running the system: The system (image classification program) can be run by double-clicking the executable program, and the location of the training images (i.e., training directory information) can be specified in the pop-up menu by clicking the "Set Training Image Directory" button.
[0117] 3. Training Images: By clicking the Start Training button (i.e., the training trigger control), the system will start training the image recognition model and can save the trained model to the current directory with the file name my_model.m;
[0118] 4. Automatic cervical disease identification: It can identify single images or batch images;
[0119] For single-image recognition, the patient's CT image (i.e., the target medical image) can be stored in a set test image directory. The user can then input the patient's information (i.e., object description information) and click "Start Classification" (i.e., task execution control) to perform image recognition. The recognition information corresponding to the patient's CT image can be displayed in the lower right corner of the interface (i.e., the results display area). The patient's relevant information can be saved as a file to the file directory corresponding to the configured results save path. If archiving is required, the user can click the "Export Results" button (i.e., export control) to complete the archiving operation of the patient information based on the recognized feature category classification results and the patient's relevant information.
[0120] For batch image recognition: The results of batch image recognition can be displayed in the lower right corner of the operation interface, and the results can be saved as files to the file directory corresponding to the configured result save path; if archiving is required, the patient information can be archived based on the basic information of multiple patients entered in the batch image recognition and the feature category classification results obtained.
[0121] In this embodiment, for any image, the directory location of the task image directory corresponding to any image and the classification probability of any image under at least one feature category are displayed in the result display area of the operation interface, so as to clearly show the identified feature category classification results.
[0122] In one embodiment, such as Figure 5 The diagram illustrates another image classification method. In this embodiment, the method includes the following steps:
[0123] In step 501, sample image classification directories corresponding to different feature categories are created according to preset feature category classification information. In step 502, in response to the operation of adding an image to any sample image classification directory, the sample medical images of the feature category corresponding to any sample image classification directory are stored in that sample image classification directory. In step 503, the operation interface of the image classification program is displayed. In response to the operation of the training directory configuration control, at least one sample image classification directory to be called for training is determined, and the directory position of at least one sample image classification directory is displayed as training directory information. In step 504, in response to the operation of the training trigger control, the labeled sample medical images under at least one sample image classification directory are input into the training model based on the SVM algorithm for training to obtain an image recognition model. In step 505, in response to the operation of the task directory configuration control, the task image directory of the image classification task is determined, and the directory position of the task image directory is displayed. Images are read according to the directory position of the task image directory to obtain the target medical image. In step 506, in response to the operation of the task execution control, the image classification task is executed, and the image recognition model is used to identify the target medical image in the image classification task. In step 507, the target medical image includes multiple images. For any given image, the result display area of the operation interface displays the directory location of the corresponding task image directory and the classification probability of any given image under at least one feature category. It should be noted that the specific limitations of the above steps can be found in the specific limitations of an image classification method described above, and will not be repeated here.
[0124] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.
[0125] Based on the same inventive concept, this application also provides an image classification apparatus for implementing the image classification method described above. The solution provided by this apparatus is similar to the implementation described in the above method; therefore, the specific limitations in one or more image classification apparatus embodiments provided below can be found in the limitations of the image classification method described above, and will not be repeated here.
[0126] In one embodiment, such as Figure 6 As shown, an image classification device is provided, comprising:
[0127] The image classification interface display module 601 is used to display the operation interface of the image classification program; the operation interface includes training trigger controls and training directory information for model training, and the training directory information is used to indicate at least one sample image classification directory called during training.
[0128] Model training module 602 is used to train a model using sample medical images from the at least one sample image classification directory in response to an operation of the training trigger control, thereby obtaining an image recognition model; the feature category to which the sample medical image in each sample image classification directory belongs is consistent with the feature category corresponding to each sample image classification directory.
[0129] The image classification task execution module 603 is used to identify the target medical image in the image classification task using the image recognition model when an image classification task execution event is detected, and to display the feature category classification result of the target medical image in the result display area of the operation interface.
[0130] In one embodiment, the user interface further includes a training directory configuration control for model training, and the device further includes:
[0131] The training call directory determination module is used to determine at least one sample image classification directory for training calls in response to an operation on the training directory configuration control.
[0132] The training directory location display module is used to display the directory location of the at least one sample image classification directory as the training directory information.
[0133] In one embodiment, the apparatus further includes:
[0134] The sample image classification directory creation module is used to create sample image classification directories corresponding to different feature categories according to preset feature category classification information; the directory name of the sample image classification directory is used to represent the corresponding feature category;
[0135] The image addition module in the directory is used to store sample medical images of the feature category corresponding to any sample image classification directory in response to the operation of adding an image under any sample image classification directory.
[0136] In one embodiment, the apparatus further includes:
[0137] An automatic annotation module is used to annotate the sample medical images under each sample image classification directory according to the directory name of each sample image classification directory during the operation of the image classification program, so as to obtain the annotated sample medical images.
[0138] The model training module 602 includes:
[0139] The image recognition model obtains a sub-module, which is used to input the labeled sample medical images from the at least one sample image classification directory into the training model based on the SVM algorithm for training, so as to obtain the image recognition model.
[0140] In one embodiment, the user interface further includes task execution controls and task directory configuration controls for classification tasks, and the image classification task execution module 603 includes:
[0141] The task directory determination submodule is used to determine the task image directory of the image classification task in response to the operation of the task directory configuration control, and to display the directory location of the task image directory.
[0142] The target medical image acquisition submodule is used to read images according to the directory position of the task image directory to obtain the target medical image;
[0143] The task execution triggering submodule is used to execute the image classification task in response to the operation of the task execution control.
[0144] In one embodiment, the target medical image includes multiple images, and the image classification task execution module 603 includes:
[0145] The results display submodule is used to display, in the results display area of the operation interface, the directory location of the task image directory corresponding to any image and the classification probability of any image under at least one feature category for any image.
[0146] The user interface also includes export controls and export directory information for exporting results, and the device further includes:
[0147] An object description information acquisition module is used to acquire object description information in the image classification task; the object description information includes description information of at least one object.
[0148] The associated storage module is used to, in response to the operation of the export control, associate and store the description information of the same object and the feature category classification result of the target medical image according to the export directory information.
[0149] Each module in the aforementioned image classification device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.
[0150] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 7 As shown, the computer device includes a processor, memory, input / output interfaces, a communication interface, a display unit, and an input device. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When executed by the processor, the computer program implements an image classification method. The display unit is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.
[0151] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0152] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:
[0153] The interface of the image classification program is displayed; the interface includes training trigger controls and training directory information for model training, the training directory information being used to indicate at least one sample image classification directory to be called during training;
[0154] In response to the operation of the training trigger control, the model is trained using sample medical images from the at least one sample image classification directory to obtain an image recognition model; the feature category to which the sample medical image in each sample image classification directory belongs is consistent with the feature category corresponding to each sample image classification directory.
[0155] When an execution event of an image classification task is detected, the image recognition model is used to identify the target medical image in the image classification task, and the feature category classification result of the target medical image is displayed in the result display area of the operation interface.
[0156] In one embodiment, the processor, when executing the computer program, also implements the steps of the image classification method described in the other embodiments above.
[0157] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:
[0158] The interface of the image classification program is displayed; the interface includes training trigger controls and training directory information for model training, the training directory information being used to indicate at least one sample image classification directory to be called during training;
[0159] In response to the operation of the training trigger control, the model is trained using sample medical images from the at least one sample image classification directory to obtain an image recognition model; the feature category to which the sample medical image in each sample image classification directory belongs is consistent with the feature category corresponding to each sample image classification directory.
[0160] When an execution event of an image classification task is detected, the image recognition model is used to identify the target medical image in the image classification task, and the feature category classification result of the target medical image is displayed in the result display area of the operation interface.
[0161] In one embodiment, when the computer program is executed by a processor, it also implements the steps of the image classification method described in the other embodiments above.
[0162] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:
[0163] The interface of the image classification program is displayed; the interface includes training trigger controls and training directory information for model training, the training directory information being used to indicate at least one sample image classification directory to be called during training;
[0164] In response to the operation of the training trigger control, the model is trained using sample medical images from the at least one sample image classification directory to obtain an image recognition model; the feature category to which the sample medical image in each sample image classification directory belongs is consistent with the feature category corresponding to each sample image classification directory.
[0165] When an execution event of an image classification task is detected, the image recognition model is used to identify the target medical image in the image classification task, and the feature category classification result of the target medical image is displayed in the result display area of the operation interface.
[0166] In one embodiment, when the computer program is executed by a processor, it also implements the steps of the image classification method described in the other embodiments above.
[0167] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data shall comply with the relevant laws, regulations and standards of the relevant countries and regions.
[0168] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.
[0169] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0170] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. An image classification method, characterized in that, The method includes: The interface of the image classification program is displayed. The interface includes a training trigger control, a training directory configuration control, and training directory information. The training directory information indicates at least one sample image classification directory to be used during training. This at least one sample image classification directory is created according to the feature category classification information of cervical medical examination images. The feature category of each sample medical image in each sample image classification directory is consistent with the feature category corresponding to each sample image classification directory. The feature category classification information includes mild cervical lesions, cervical intraepithelial neoplasia (CIN I), cervical intraepithelial neoplasia (CIN II), cervical intraepithelial neoplasia (CIN III), and cervical cancer. In response to an operation on the training directory configuration control, at least one sample image classification directory for the training call is determined; the directory location of the at least one sample image classification directory is displayed as the training directory information; In response to the operation of the training trigger control, the model is trained using sample medical images from the at least one sample image classification catalog to obtain an image recognition model; When an execution event of an image classification task is detected, the image recognition model is used to identify the target medical image in the image classification task, and the feature category classification result of the target medical image is displayed in the result display area of the operation interface. The image classification task includes a single image recognition task and a batch image recognition task. The result display area of the operation interface is used to display the feature category classification result of the target medical image corresponding to the single image recognition task, or the feature category classification result of the target medical image corresponding to the batch image recognition task, and the object description information in the image classification task is saved to the file directory corresponding to the configured result save path. The feature category classification result includes the directory location of the target medical image and the classification probability of the feature category to which the target medical image belongs.
2. The method according to claim 1, characterized in that, Before the step of training the model using sample medical images from the at least one sample image classification catalog in response to an operation of the training trigger control to obtain an image recognition model, the method further includes: According to the preset feature category classification information, create sample image classification directories corresponding to different feature categories; the directory name of the sample image classification directory is used to represent the corresponding feature category; In response to the operation of adding an image to any sample image classification directory, the sample medical images belonging to the feature category corresponding to the any sample image classification directory are stored in the any sample image classification directory.
3. The method according to claim 2, characterized in that, The method further includes: During the operation of the image classification program, the medical images under each of the sample image classification directories are labeled according to the directory name of each sample image classification directory, thus obtaining the labeled medical images. The step of training a model using sample medical images from at least one sample image classification catalog to obtain an image recognition model includes: The labeled medical images from the at least one sample image classification directory are input into the training model based on the SVM algorithm to obtain the image recognition model.
4. The method according to claim 1, characterized in that, The user interface also includes task execution controls and task directory configuration controls for classification tasks. The detected execution event of the image classification task includes: In response to an operation on the task directory configuration control, the task image directory for the image classification task is determined, and the directory location of the task image directory is displayed. The target medical image is obtained by reading the image according to its directory location in the task image directory; In response to an operation on the task execution control, the image classification task is executed.
5. The method according to any one of claims 1 to 4, characterized in that, The target medical image includes multiple images, and the feature category classification results of the target medical image are displayed in the result display area of the operation interface, including: For any given image, the result display area of the operation interface displays the directory location of the task image directory corresponding to that image and the classification probability of that image under at least one feature category. The user interface also includes export controls and export directory information for exporting results, and the method further includes: Obtain object description information from the image classification task; the object description information includes description information of at least one object; In response to the operation of the export control, the description information of the same object and the feature category classification result of the target medical image are associated and stored according to the export directory information.
6. An image classification device, characterized in that, The device includes: The image classification interface display module is used to display the operation interface of the image classification program. The operation interface includes training trigger controls, training directory configuration controls, and training directory information for model training. The training directory information is used to indicate at least one sample image classification directory to be called during training. The at least one sample image classification directory is created according to the feature category classification information of cervical medical examination images. The feature category to which the sample medical images in each sample image classification directory belong is consistent with the feature category corresponding to each sample image classification directory. The feature category classification information includes mild cervical lesions, cervical intraepithelial neoplasia (CIN I), cervical intraepithelial neoplasia (CIN II), cervical intraepithelial neoplasia (CIN III), and cervical cancer. A training call directory determination module is used to determine at least one sample image classification directory of the training call in response to an operation on the training directory configuration control; The training directory location display module is used to display the directory location of the at least one sample image classification directory as the training directory information; The model training module is used to train the model using sample medical images from the at least one sample image classification directory in response to the operation of the training trigger control, so as to obtain an image recognition model. The image classification task execution module is used to identify the target medical image in the image classification task using the image recognition model when an image classification task execution event is detected, and to display the feature category classification result of the target medical image in the result display area of the operation interface; the image classification task includes a single image recognition task and a batch image recognition task, and the result display area of the operation interface is used to display the feature category classification result of the target medical image corresponding to the single image recognition task or the target medical image corresponding to the batch image recognition task, and to save the object description information in the image classification task to the file directory corresponding to the configured result save path; the feature category classification result includes the directory location of the target medical image and the classification probability of the feature category to which the target medical image belongs.
7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.
8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.
9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.