Gastroscope image pixel segmentation method and device, electronic equipment and storage medium

Through the improved U-Net neural network model and feature extraction technology, the problems of insufficient research on early gastric cancer segmentation and noise influence in gastroscopy are solved, and efficient and accurate gastroscopy image pixel segmentation is achieved, which is suitable for low-level medical equipment and improved Diagnostic efficiency and treatment popularity.

CN118968045APending Publication Date: 2024-11-15BEIJING HOSPITAL +1
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
CN202410872929.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-01
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

The prior art studies on early gastric cancer segmentation in gastroscopy are insufficient, and noise or blurred edges in the image lead to algorithm errors, affecting the auxiliary diagnosis results.

Method used

The improved U-Net neural network model is adopted, and through lightweight processing and feature extraction technology, combined with Focal Loss loss function, the pixel-level segmentation of gastroscopic images is performed to reduce the number of model parameters and improve the segmentation accuracy.

Benefits of technology

It realizes efficient and accurate pixel segmentation of gastroscopy images, shortens diagnosis time, is suitable for low-level medical equipment, and improves the popularity of treatment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118968045A_ABST
    Figure CN118968045A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of gastroscope image processing, in particular to a gastroscope image pixel segmentation method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining a stomach image and an image corresponding to a gastric cancer segmentation tag, and constructing an image data set; preprocessing the image data set to obtain a preprocessed image data set; carrying out lightweight processing on the U-Net neural network, and constructing a gastroscope image pixel segmentation model by utilizing the U-Net neural network subjected to the lightweight processing; training the gastroscope image pixel segmentation model by using the preprocessed image data set to obtain a trained gastroscope image pixel segmentation model; and predicting a to-be-detected gastroscope image by using the trained gastroscope image pixel segmentation model to obtain a prediction result. A medical image segmentation model U-Net is utilized, improvement is carried out on the basis of the medical image segmentation model U-Net, a deep learning model is used for carrying out pixel-level segmentation on an image, and efficient and accurate gastroscope image pixel segmentation is achieved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of gastroscopic image processing, and in particular to a gastroscopic image pixel segmentation method, device, electronic device and storage medium. Background Art

[0002] At present, gastroscopy is still the most direct, accurate and reliable diagnostic method for gastric cancer. The precancerous lesions of gastric cancer often involve a wide area of ​​the gastric mucosa. Conventional endoscopic biopsy adopts a segmented, random multi-point sampling method, and there is a high possibility of missing local cancerous tissues, especially for inexperienced endoscopists. It is more likely to miss the diagnosis; secondly, the determination of atypical hyperplasia and early gastric cancer under the microscope requires a large number of case accumulations. There are certain differences between different pathologists or the same pathologist who diagnoses the same group of gastroscopic biopsy specimens multiple times. In the segmentation process, because early cancer develops slowly in the body and the lesion size is small, it is not easy to be found. At present, the research on early gastric cancer segmentation at home and abroad is not very sufficient. Secondly, when there is noise or blurred edges in the image, the algorithm may have errors, which may affect the results of our subsequent auxiliary diagnosis;

[0003] In summary, how to design an efficient and accurate gastroscopic image pixel segmentation method is a problem that needs to be solved urgently. Summary of the invention

[0004] The present application aims to solve one of the technical problems in the related art at least to some extent.

[0005] To this end, the first purpose of this application is to propose a gastroscopic image pixel segmentation method to solve the problem that the research on early gastric cancer segmentation by existing technical means is not sufficient, and secondly, when there is noise or blurred edges in the image, the algorithm may have errors and other problems.

[0006] The second object of the present application is to provide a device.

[0007] The third objective of the present application is to provide an electronic device.

[0008] A fourth objective of the present application is to provide a computer-readable storage medium.

[0009] The fifth objective of this application is to provide a computer program product.

[0010] To achieve the above-mentioned purpose, the first embodiment of the present application proposes a gastroscopic image pixel segmentation method, comprising:

[0011] Obtain stomach images and images corresponding to gastric cancer segmentation labels to construct an image dataset;

[0012] Preprocessing the image data set to obtain a preprocessed image data set;

[0013] The U-Net neural network is lightweighted and the pixel segmentation model of gastroscope images is constructed using the lightweight U-Net neural network.

[0014] Using the preprocessed image data set to train the gastroscopic image pixel segmentation model to obtain a trained gastroscopic image pixel segmentation model;

[0015] The trained gastroscope image pixel segmentation model is used to predict the gastroscope image to be detected to obtain a prediction result.

[0016] Preferably, the preprocessing the image data set to obtain the preprocessed image data set comprises:

[0017] Performing denoising, image smoothing and image enhancement processing on the image data set to obtain a clear image data set;

[0018] Color correction is performed on the clear image data set to eliminate color cast or white balance in the image and then perform edge enhancement processing to obtain a preprocessed image data set.

[0019] Preferably, the lightweight processing of the U-Net neural network includes:

[0020] BSConv-SS is used to replace some ordinary convolutional layers in the U-Net neural network to obtain a new encoder and decoder; a feature transfer module based on a pooling layer is used to fuse deep and shallow layer features, and then BSConv-SS with a large convolution kernel is used for feature extraction.

[0021] Preferably, the using the preprocessed image data set to train the gastroscopic image pixel segmentation model to obtain the trained gastroscopic image pixel segmentation model comprises:

[0022] Based on the characteristics of early gastric cancer segmentation, a Focal Loss loss function is selected, the model parameters are updated according to the gradient information of the loss function, the error is calculated according to the Focal Loss loss function, the gastroscopic image pixel segmentation model is trained, and it is iterated repeatedly until the loss function converges.

[0023] Preferably, the Focal Loss loss function calculation formula is:

[0024]

[0025] Among them, p i is the predicted probability that the i-th pixel belongs to its true category, and γ is the adjustment factor.

[0026] Preferably, the method further includes: obtaining the harmonic mean of the model precision and recall by measuring the model precision and recall, and the calculation formula is:

[0027]

[0028] Among them, F1 is the harmonic mean, Precision is the model precision, and Recall is the model recall.

[0029] To achieve the above-mentioned purpose, the second embodiment of the present application proposes a gastroscopic image pixel segmentation device, comprising:

[0030] An image acquisition module acquires stomach images and images corresponding to gastric cancer segmentation labels to construct an image dataset;

[0031] A preprocessing module, preprocessing the image data set to obtain a preprocessed image data set;

[0032] The model building module performs lightweight processing on the U-Net neural network and uses the lightweight U-Net neural network to build a gastroscope image pixel segmentation model;

[0033] A training module, using the preprocessed image data set to train the gastroscopic image pixel segmentation model to obtain a trained gastroscopic image pixel segmentation model;

[0034] The prediction module uses the trained gastroscope image pixel segmentation model to predict the gastroscope image to be detected to obtain a prediction result.

[0035] To achieve the above-mentioned purpose, the third aspect of the present application provides an electronic device, including: a processor, and a memory communicatively connected to the processor;

[0036] The memory stores computer-executable instructions;

[0037] The processor executes the computer-executable instructions stored in the memory to implement any of the above methods.

[0038] To achieve the above-mentioned purpose, the fourth aspect embodiment of the present application proposes a computer-readable storage medium, including computer-executable instructions stored in the computer-readable storage medium, and the computer-executable instructions are used to implement any of the methods described above when executed by a processor.

[0039] In a fifth aspect, the present application discloses a computer program product, including a computer program / instructions, characterized in that when the computer program / instructions are executed by a processor, the method described in the first aspect is implemented.

[0040] The present application provides a method for pixel segmentation of gastroscopy images, which utilizes and improves the classic medical image segmentation model U-Net. The detection accuracy is basically guaranteed. The time required for pixel-level segmentation of images using a deep learning model is very short, and this time can even be ignored. In the process of auxiliary diagnosis, doctors do not need to spend extra time waiting for the system's prediction results. At the same time, compared with the original model, the improved U-Net has a reduced number of model parameters while maintaining the same accuracy. This allows our system to be installed on low-level, temporary medical equipment, which helps promote the popularization of therapeutic medical care.

[0041] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0043] Figure 1 This is a flow chart of a first specific embodiment of a gastroscope image pixel segmentation method provided by the present invention;

[0044] Figure 2 This is the flow chart of gastroscopic image pixel segmentation based on improved U-Net;

[0045] Figure 3 It is a schematic diagram of the U-Net network model structure;

[0046] Figure 4 A structural block diagram of a gastroscope image pixel segmentation device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The core of the present invention is to provide a method, device, electronic device and storage medium for gastroscopic image pixel segmentation. After lightweight processing of the U-Net segmentation model, the image is segmented at the pixel level using the lightweight deep learning model to achieve efficient and accurate gastroscopic image pixel segmentation.

[0048] In order to enable those skilled in the art to better understand the scheme of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0049] Please refer to Figure 1 , Figure 1This is a flow chart of a first specific embodiment of a gastroscope image pixel segmentation method provided by the present invention; the specific operation steps are as follows:

[0050] Step S101: Acquire a stomach image and an image corresponding to a gastric cancer segmentation label to construct an image data set;

[0051] Step S102: preprocessing the image data set to obtain a preprocessed image data set;

[0052] Performing denoising, image smoothing and image enhancement processing on the image data set to obtain a clear image data set;

[0053] Color correction is performed on the clear image data set to eliminate color cast or white balance in the image and then perform edge enhancement processing to obtain a preprocessed image data set.

[0054] Step S103: performing lightweight processing on the U-Net neural network, and using the lightweight processed U-Net neural network to construct a gastroscope image pixel segmentation model;

[0055] BSConv-SS is used to replace some ordinary convolutional layers in the U-Net neural network to obtain a new encoder and decoder; a feature transfer module based on a pooling layer is used to fuse deep and shallow layer features, and then BSConv-SS with a large convolution kernel is used for feature extraction.

[0056] Step S104: training the gastroscopic image pixel segmentation model using the preprocessed image data set to obtain a trained gastroscopic image pixel segmentation model;

[0057] Based on the characteristics of early gastric cancer segmentation, a Focal Loss loss function is selected, the model parameters are updated according to the gradient information of the loss function, the error is calculated according to the Focal Loss loss function, the gastroscopic image pixel segmentation model is trained, and it is iterated repeatedly until the loss function converges.

[0058] The Focal Loss loss function calculation formula is:

[0059]

[0060] Among them, p i is the predicted probability that the i-th pixel belongs to its true category, and γ is an adjustment factor that helps the model pay more attention to pixels that are difficult to classify. In this study, γ = 2.

[0061] By measuring the precision and recall of the model, the harmonic mean of the precision and recall of the model is obtained. The calculation formula is:

[0062]

[0063] Among them, F1 is the harmonic mean, Precision is the model precision, and Recall is the model recall.

[0064] Step S106: using the trained gastroscope image pixel segmentation model to predict the gastroscope image to be detected, to obtain a prediction result.

[0065] The improved U-Net model can automatically learn abstract features in images, from simple edges and textures to more advanced semantic and shape information, through a multi-level neural network structure. This feature learning is end-to-end, without the need to manually design feature extractors, but instead automatically learns feature representations in images through training with a large amount of data. As training progresses, the model can gradually understand the spatial relationship and semantic information between different objects and backgrounds in the image, thereby achieving pixel-level segmentation of the image. By inputting the image into a trained deep learning model, the model can predict each pixel and determine its semantic category or segmentation label, thereby achieving pixel-level segmentation tasks.

[0066] This embodiment provides a method for pixel segmentation of gastroscopy images, which utilizes and improves the classic medical image segmentation model U-Net, and the detection accuracy is basically guaranteed. By inputting the image into the trained deep learning model, the model can predict each pixel and determine its semantic category or segmentation label, thereby realizing the pixel-level segmentation task. It takes very little time to perform pixel-level segmentation of the image using the deep learning model. In the process of auxiliary diagnosis, the physician does not need to spend extra time waiting for the system's prediction results. At the same time, compared with the original model, the improved U-Net has a reduced number of model parameters while maintaining the same accuracy. This allows our system to be installed on low-level, temporary medical equipment, which helps promote the popularization of therapeutic medical care.

[0067] Based on the above embodiment, this embodiment describes a gastroscope image pixel segmentation method. Figure 2 As shown, the details are as follows:

[0068] In order to improve the quality and readability of gastroscopic images and improve the accuracy and effectiveness of our subsequent image segmentation work, the collected gastroscopic images are preprocessed as follows:

[0069] Denoising: In gastroscopy images, noise and artifacts may be affected, which will reduce image quality and readability. Therefore, denoising is required to improve image quality and clarity.

[0070] Image smoothing: Gastroscopic images may have problems such as color changes and uneven brightness. Image smoothing is needed to eliminate these problems and make the image more uniform.

[0071] Image enhancement: For some blurry gastroscopy images, image enhancement processing is required to improve their contrast and clarity.

[0072] Color correction: Perform color correction based on actual conditions to eliminate color cast or white balance problems in the image, making the image color more realistic and accurate.

[0073] Edge enhancement: In order to better locate and identify the edges and detail features in gastroscopic images, edge enhancement techniques such as the Sobel operator or the Canny edge detector can be used to highlight the edge information in the image, which is helpful for subsequent image segmentation.

[0074] The number of parameters of ordinary convolution is directly related to the number of channels of input features. In U-Net, the number of channels of input features and the number of convolution kernels increase as the network deepens, resulting in a significant increase in the number of network parameters, which becomes one of the main reasons for the increase in model size. Therefore, in order to eliminate the negative impact of ordinary convolution in the deep part of the network, this system replaces the ordinary convolution in the deep part of the network with the BSConv-SS proposed by this system, such as Figure 3 As shown in the figure, this design significantly reduces the number of parameters, but the negative impact on accuracy is inevitable. In order to make up for this part of the loss of accuracy, this system designed a PIT Network based on pooling operation. This design completed the effective fusion of features without adding additional parameters. In addition, considering that large convolution kernels can help the model better extract the features of the lesion area, this system further developed the BLC Network. Under the influence of PIT Network and BLC Network, the accuracy of the model has been significantly enhanced.

[0075] U-Net is a classic deep learning model for image semantic segmentation tasks. Its structural features are encoder-decoder structure and skip connection. The encoder of U-Net consists of a series of convolutional layers and pooling layers, which are used to gradually extract features of the input image and reduce the spatial dimension. Each convolutional layer usually includes convolution operation, activation function and batch normalization, as well as an optional Dropout layer to enhance the generalization ability of the model. The decoder of U-Net is the opposite of the encoder structure. It consists of a series of upsampling layers and convolutional layers to restore the feature map extracted by the encoder to the same size as the input image. Each upsampling layer usually includes upsampling operations (such as deconvolution or transposed convolution) and convolution operations, as well as activation functions and batch normalization. The goal of the decoder is to convert the high-level semantic features extracted by the encoder back to the spatial dimensions of the original input image and generate a segmentation result of the same size as the input image. In the encoder, through convolution and pooling operations, the size of the image is gradually reduced and the number of channels is gradually increased, thereby extracting more and more abstract feature information.

[0076] In U-Net, the feature maps of the corresponding layers in the encoder are connected to the feature maps in the decoder through jump connections, so that the decoder can obtain feature information from different scales. This multi-scale feature fusion helps the model better understand the semantic information of the image, thereby achieving accurate segmentation of each pixel. The encoder gradually captures the local and global information of the image through convolution and pooling operations and encodes it into feature vectors. These feature vectors contain the semantic information of each position in the image, and the decoder decodes this information to the original image size through upsampling operations to achieve segmentation of each pixel. Jump connections connect the low-level and high-level feature maps in the encoder with the feature maps in the decoder, allowing the decoder to utilize more contextual information, including image details and boundary information. This connection method helps the U-Net model better understand the structure of the image and achieve accurate segmentation at the pixel level.

[0077] Using BSConv-SS to replace some standard convolutions that process deep features will inevitably lead to accuracy loss, and the fundamental reason is that the deep features are not fully extracted. Under the condition of adding as few additional parameters as possible, this method proposes PIT Network to restore this part of the loss as much as possible. PIT Network consists of a large number of Maxpooling. In this network, the shallow features obtained by the U-Net encoder are first Maxpooled to make the dimensions of the processed features consistent with the deep features, and then they are concat-ed in the channel dimension. This operation enables this system to not only transfer shallow feature information to deep features, but also Maxpooling does not add additional parameters. Finally, BSConv-SS is used to reduce the dimension of the deep features after concat, so that the shallow feature information and the deep feature information are further integrated. The basic design dimension of the convolutional network—Kernel Size—is crucial to the performance of the model, especially for downstream tasks such as segmentation and detection. Large convolution kernels can cover a larger receptive field, which helps to capture the global features of the input data, which means that the model can better understand the overall structure and contextual information; at the same time, large convolution kernels enhance the model's ability to learn translation invariance to a certain extent, and even if the input position changes by a certain amount, the model can maintain good recognition ability. However, the increase in the number of parameters brought by large convolution kernels is criticized. This system proposes BLC Network (BSConv-S based large convolution network) to solve this pain point. BSConv-S has excellent lightweight capabilities, and even with large convolution kernels, the parameter increase is acceptable. BLC Network consists of multiple BSConv-S with large convolution kernels to further extract the input features. In the network of this system, this system sets the convolution kernel size to 21. At the same time, this system uses residual connections to make the network have better generalization ability. After passing through the BLC Network, the decoder will fuse the features and finally output the prediction results of the model. This process is consistent with U-Net.

[0078] Model training is a crucial step in deep learning. It refers to adjusting the model's parameters by learning the model with labeled data so that it can accurately predict unknown data. According to the characteristics of early gastric cancer segmentation, a suitable loss function is selected to measure the difference between the model's prediction results and the true label. A suitable optimization algorithm is selected to update the model parameters according to the gradient information of the loss function. Common optimization algorithms include stochastic gradient descent (SGD), Adam, RMSprop, etc. Through training, the model can learn appropriate parameters based on the training data, so that the model can accurately predict new unknown data. During the training process, the model is continuously verified by the validation data set, which can effectively improve the generalization ability of the model and make the model have good adaptability and generalization for new data. Through the model training process, the model can be tuned and hyperparameters can be adjusted to improve the performance and stability of the model. During the training process, the model continuously performs error back propagation according to the true label, so that the model gradually converges to the optimal solution and finally obtains a better prediction effect.

[0079] Model validation is a very important step in deep learning. It is used to evaluate the performance of the model on unseen data. The validation process can help us understand the generalization ability and stability of the model, so as to determine whether the model can be well applied to data in real scenarios. In this system, we use the following indicators and loss functions:

[0080] Accuracy is the most intuitive performance indicator, indicating the proportion of instances predicted correctly by the model, including both positive and negative classes.

[0081]

[0082] Precision measures the proportion of instances that are actually positive out of all the instances that the model predicts to be positive. A high precision means fewer false positives.

[0083]

[0084] Recall measures the proportion of instances that the model correctly predicts to be positive out of all instances that are actually positive. A high recall indicates fewer false negatives.

[0085]

[0086] Among them, TP is the number of positive samples correctly identified as positive samples by the model, TN is the number of negative samples that the model correctly did not identify as positive samples, FP is the number of negative samples that the model incorrectly identified as positive samples, and FN is the number of positive samples that the model incorrectly identified as negative samples;

[0087] The F1 score is the harmonic average of precision and recall, and it attempts to provide a single metric that balances the performance of precision and recall. The F1 score is particularly useful for scenarios where precision and recall are equally important. In some scenarios, improving precision may reduce recall, and vice versa, and the F1 score helps measure the overall effectiveness of the model.

[0088]

[0089] In the gastroscopy images of early gastric cancer, the ratio of foreground pixels to background pixels in the lesion area is very different, and the foreground pixels account for a small number. Therefore, we use Focal Loss as the loss function, and its calculation formula is:

[0090]

[0091] Among them, p i is the predicted probability that the i-th pixel belongs to its true category, and γ is the adjustment factor.

[0092] The improved U-Net model can automatically learn abstract features in images, from simple edges and textures to more advanced semantic and shape information, through a multi-level neural network structure. This feature learning is end-to-end, without the need to manually design feature extractors, but instead automatically learns feature representations in images through training with a large amount of data. As training progresses, the model can gradually understand the spatial relationship and semantic information between different objects and backgrounds in the image, thereby achieving pixel-level segmentation of the image. By inputting the image into a trained deep learning model, the model can predict each pixel and determine its semantic category or segmentation label, thereby achieving pixel-level segmentation tasks.

[0093] A gastroscopic image pixel segmentation method provided in an embodiment of the present invention utilizes and improves the classic medical image segmentation model U-Net, thereby ensuring the most basic accuracy of detection. The time required for pixel-level segmentation of images using a deep learning model is very short, and even negligible. In the process of auxiliary diagnosis, doctors do not need to spend extra time waiting for the system's prediction results. As training progresses, the model can gradually understand the spatial relationship and semantic information between different objects and backgrounds in the image, thereby achieving pixel-level segmentation of the image.

[0094] Please refer to Figure 4 , Figure 4 A structural block diagram of a gastroscope image pixel segmentation device provided by an embodiment of the present invention; the specific device may include:

[0095] An image acquisition module 100 acquires a stomach image and an image corresponding to a gastric cancer segmentation label to construct an image data set;

[0096] A preprocessing module 200 preprocesses the image data set to obtain a preprocessed image data set;

[0097] The model building module 300 performs lightweight processing on the U-Net neural network and uses the lightweight processed U-Net neural network to build a gastroscope image pixel segmentation model;

[0098] A training module 400 is used to train the gastroscopic image pixel segmentation model using the preprocessed image data set to obtain a trained gastroscopic image pixel segmentation model;

[0099] The prediction module 500 uses the trained gastroscope image pixel segmentation model to predict the gastroscope image to be detected to obtain a prediction result.

[0100] A gastroscopic image pixel segmentation device of the present embodiment is used to implement the aforementioned gastroscopic image pixel segmentation method. Therefore, the specific implementation method of a gastroscopic image pixel segmentation device can be seen in the embodiment part of a gastroscopic image pixel segmentation method in the previous text. For example, the image acquisition module 100, the preprocessing module 200, the model building module 300, the training module 400, and the prediction module 500 are respectively used to implement steps S101, S102, S103, S104, and S105 in the aforementioned gastroscopic image pixel segmentation method. Therefore, its specific implementation method can refer to the description of the corresponding various parts of the embodiment, which will not be repeated here.

[0101] In order to implement the above embodiments, the present application also proposes an electronic device, comprising: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method provided by the above embodiments.

[0102] In order to implement the above embodiments, the present application also proposes a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the methods provided by the above embodiments.

[0103] In order to implement the above embodiments, the present application also proposes a computer program product, including a computer program, which implements the methods provided by the above embodiments when executed by a processor.

[0104] The collection, storage, use, processing, transmission, provision and disclosure of user personal information involved in this application are in compliance with relevant laws and regulations and do not violate public order and good morals.

[0105] It should be noted that personal information from users should be collected for legitimate and reasonable purposes and should not be shared or sold outside of these legitimate uses. In addition, such collection / sharing should be carried out after receiving the user's informed consent, including but not limited to notifying the user to read the user agreement / user notice and sign the agreement / authorization including authorization of relevant user information before the user uses the function. In addition, any necessary steps should be taken to protect and safeguard access to such personal information data and ensure that others who have access to personal information data comply with its privacy policy and procedures.

[0106] The present application is expected to provide an implementation scheme for users to selectively block the use or access of personal information data. That is, the present disclosure is expected to provide hardware and / or software to prevent or block access to such personal information data. Once the personal information data is no longer needed, the risk can be minimized by limiting data collection and deleting the data. In addition, when applicable, such personal information is de-identified to protect the privacy of the user.

[0107] In the description of the aforementioned embodiments, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0108] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0109] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0110] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute the instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0111] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0112] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0113] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0114] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A gastroscopic image pixel segmentation method, It is characterized in that include: Obtain stomach images and images corresponding to gastric cancer segmentation labels to construct an image dataset; Preprocessing the image data set to obtain a preprocessed image data set; The U-Net neural network is lightweighted and the pixel segmentation model of gastroscope images is constructed using the lightweight U-Net neural network. Using the preprocessed image data set to train the gastroscopic image pixel segmentation model to obtain a trained gastroscopic image pixel segmentation model; The trained gastroscope image pixel segmentation model is used to predict the gastroscope image to be detected to obtain a prediction result.

2. The gastroscope image pixel segmentation method according to claim 1, It is characterized in that The preprocessing of the image data set to obtain the preprocessed image data set comprises: Performing denoising, image smoothing and image enhancement processing on the image data set to obtain a clear image data set; Color correction is performed on the clear image data set to eliminate color cast or white balance in the image and then perform edge enhancement processing to obtain a preprocessed image data set.

3. The gastroscope image pixel segmentation method according to claim 1, It is characterized in that The lightweight processing of the U-Net neural network includes: BSConv-SS is used to replace some ordinary convolutional layers in the U-Net neural network to obtain a new encoder and decoder; a feature transfer module based on a pooling layer is used to fuse deep and shallow layer features, and then BSConv-SS with a large convolution kernel is used for feature extraction.

4. The gastroscope image pixel segmentation method according to claim 1, It is characterized in that The method of training the gastroscope image pixel segmentation model by using the preprocessed image data set to obtain the trained gastroscope image pixel segmentation model comprises: Based on the characteristics of early gastric cancer segmentation, the FocalLoss loss function is selected, the model parameters are updated according to the gradient information of the loss function, the error is calculated according to the FocalLoss loss function, the gastroscopic image pixel segmentation model is trained, and it is iterated repeatedly until the loss function converges.

5. The gastroscope image pixel segmentation method according to claim 4, It is characterized in that The Focal Loss loss function calculation formula is: Among them, p i is the predicted probability that the i-th pixel belongs to its true category, and γ is the adjustment factor.

6. The gastroscope image pixel segmentation method according to claim 4, It is characterized in that Also includes: By measuring the precision and recall of the model, the harmonic mean of the precision and recall of the model is obtained. The calculation formula is: Among them, F1 is the harmonic mean, Precision is the model precision, and Recall is the model recall.

7. A gastroscope image pixel segmentation device, It is characterized in that include: An image acquisition module acquires stomach images and images corresponding to gastric cancer segmentation labels to construct an image dataset; A preprocessing module, preprocessing the image data set to obtain a preprocessed image data set; The model building module performs lightweight processing on the U-Net neural network and uses the lightweight U-Net neural network to build a gastroscope image pixel segmentation model; A training module, using the preprocessed image data set to train the gastroscopic image pixel segmentation model to obtain a trained gastroscopic image pixel segmentation model; The prediction module uses the trained gastroscope image pixel segmentation model to predict the gastroscope image to be detected to obtain a prediction result.

8. An electronic device, It is characterized in that include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 6 when executed by a processor.

10. A computer program product comprising a computer program / instructions, It is characterized in that When the computer program / instructions are executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Citation Information

Patent Citations

  • Image detection segmentation method and system, storage medium, computer program and terminal

    CN111476778A

  • Gastric cancer focus recognition method, device and equipment based on image processing and storage medium

    CN112967287A

  • Method, device and system for generating medical image segmentation model

    CN113159040A

  • Lightweight image segmentation method and system, medium, terminal and application

    CN114612477A

  • Endoscope image segmentation method based on single-image deep learning

    CN116309651A