Image enhancement model training method, physical examination report image processing method and equipment
By training the image enhancement model and evaluation model, the clarity of the physical examination report image is improved by using historical physical examination report data, solving the problem of poor image processing effect of physical examination report in the prior art, and providing convenience for information identification.
Patent Information
- Application Number
- CN202411600097.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-11
- Publication Date
- 2025-03-25
AI Technical Summary
In the prior art, the image processing effect of physical examination reports is poor, the clarity is low, and the historical physical examination reports cannot be effectively utilized, resulting in difficulty in identifying information and structured analysis.
By obtaining historical physical examination report data, identifying and determining true images, generating training data sets, training image enhancement models and evaluation models, alternately updating model parameters, improving the performance of image enhancement models, and enhancing the clarity of physical examination report images.
It improves the clarity of the physical examination report image, restores the image information, provides convenience for the identification of subsequent physical examination report information, and solves the problem of poor image processing effect of the physical examination report image.
Smart Images
Figure CN119693253A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of image processing technology, and in particular to an image enhancement model training method, a physical examination report image processing method and equipment. Background Art
[0002] The joint interpretation and analysis of multiple physical examination reports accumulated during regular physical examinations can reflect the health development trend of patients and promptly detect possible lesions so that preventive or therapeutic measures can be taken in advance. However, due to the long time span of multiple physical examination reports, affected by storage conditions, ink fixation, shooting environment and shooting conditions, the physical examination reports may have wrinkles, shadows, stains, perspective deformation, dynamic blur and other problems, which will have a negative impact on the subsequent information recognition, structuring and comprehensive analysis of the physical examination reports.
[0003] Existing physical examination report detection methods are usually based on the results of a single physical examination, lack the use of historical physical examination reports, and usually only perform simple preprocessing. However, physical examination reports from previous years have poor image quality due to long storage time. Simple preprocessing cannot effectively eliminate the negative impact of low-quality images on the research, and the image processing effect is poor and the clarity is low. Summary of the invention
[0004] The present application provides an image enhancement model training method, a physical examination report image processing method and a device to solve the problems of poor effect and low clarity of physical examination report image processing in the prior art.
[0005] The first aspect of the present application provides an image enhancement model training method, comprising the following steps: acquiring historical physical examination report data; identifying physical examination report images in the historical physical examination report data, and determining the true value images of the physical examination report images, and generating a training data set based on the physical examination report images and the true value images; using the training data set to train the image enhancement model, inputting the output image of the image enhancement model and the true value image into an evaluation model, evaluating the image feature differences of the model output image, and alternately updating the model parameters of the image enhancement model and the evaluation model based on the image feature differences.
[0006] Optionally, the image enhancement model includes an encoder and a decoder, wherein the encoder includes a feature extraction layer, a connection layer, a transition layer and an attention layer, the feature extraction layer is used to extract features from the physical examination report image to obtain a feature map, the connection layer is used to connect the input and output between layers, the transition layer is used to reduce the number of channels and spatial resolution of the feature map, and the attention layer is used to filter invalid information in the feature map; the decoder includes a residual layer and an attention layer, the residual layer is used to restore the features and spatial resolution of the feature map to obtain a restored feature map, and the attention layer is used to filter out invalid information in the restored feature map to obtain an output image.
[0007] Optionally, the evaluation model includes an input layer, a function layer and an output layer, wherein the input layer inputs the output image of the image enhancement model and the true value image into the function layer, the function layer calculates the image feature differences between images using a multi-scale feature loss function, and the output layer outputs the image feature differences.
[0008] Optionally, the calculation formula of the multi-scale feature loss function is:
[0009]
[0010] Among them, l mae represents the mean absolute error, x n represents the nth physical examination report image in a dataset consisting of N images, S(x n ) represents the label distribution output by the image enhancement network model, y n represents the true label distribution, f c () represents the multi-level features extracted by the evaluation model.
[0011] Optionally, model parameters of the image enhancement model and the evaluation model are updated alternately according to the image feature difference, including: updating the model parameters of the image enhancement model by back propagation according to the image feature difference, wherein the update goal of the image enhancement model is to minimize the multi-scale feature loss; updating the model parameters of the evaluation model by back propagation according to the image feature difference, wherein the update goal of the evaluation model is to maximize the multi-scale feature loss.
[0012] The second aspect of the present application provides a method for processing a physical examination report image, comprising the following steps: acquiring target physical examination report data; identifying a target physical examination report image in the target physical examination report data; inputting the target physical examination report image into an image enhancement model, and the image enhancement model outputs an enhanced image, wherein the image enhancement model is a model trained by the above embodiment.
[0013] The third aspect of the present application provides a method for processing a physical examination report image, comprising the following steps: obtaining target physical examination report data; identifying a target physical examination report image in the target physical examination report data; inputting the target physical examination report image into an image enhancement model, and the image enhancement model outputs an enhanced image, wherein the image enhancement model comprises an encoder and a decoder, wherein the encoder comprises a feature extraction layer, a connection layer, a transition layer and an attention layer, the feature extraction layer is used to extract features from the physical examination report image to obtain a feature map, the connection layer is used to connect the input and output between layers, the transition layer is used to reduce the number of channels and the spatial resolution of the feature map, and the attention layer is used to filter out invalid information in the feature map; the decoder comprises a residual layer and an attention layer, the residual layer is used to restore the features and spatial resolution of the feature map to obtain a restored feature map, and the attention layer is used to filter out invalid information in the restored feature map to obtain an output image.
[0014] The fourth aspect of the present application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to execute an image enhancement model training method or a physical examination report image processing method as described in the above-mentioned embodiments.
[0015] The fifth aspect of the present application provides a computer-readable storage medium on which a computer program or instruction is stored. The computer program or instruction is executed by a processor to perform the image enhancement model training method or the physical examination report image processing method as described in the above-mentioned embodiment.
[0016] The sixth aspect of the present application provides a computer program product, including a computer program or instructions. When the computer program or instructions are executed, they can implement the image enhancement model training method or physical examination report image processing method as described in the above embodiments.
[0017] Therefore, this application has at least the following beneficial effects:
[0018] The embodiment of the present application can determine the corresponding true value image based on the physical examination report image in the historical physical examination report data, train the image enhancement model based on the training data set generated by the physical examination report image and the true value image, and alternately update the model parameters of the image enhancement model and the evaluation model according to the image feature difference between the output image of the image enhancement model and the true value image output by the evaluation model, thereby improving the performance of the image enhancement model and the evaluation model respectively, and subsequently enhance the physical examination report image using the image enhancement model to improve the clarity of the physical examination report image, restore the image information, and facilitate the subsequent recognition of the physical examination report information. Thus, the technical problems of poor effect and low clarity of physical examination report image processing in the prior art are solved.
[0019] 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
[0020] 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:
[0021] Figure 1 A flowchart of an image enhancement model training method provided according to an embodiment of the present application;
[0022] Figure 2 A structural diagram of an image enhancement model provided according to an embodiment of the present application;
[0023] Figure 3 A structural diagram of an encoder provided according to an embodiment of the present application;
[0024] Figure 4 A structural diagram of a decoder provided according to an embodiment of the present application;
[0025] Figure 5 A flowchart of a method for processing a physical examination report image according to an embodiment of the present application;
[0026] Figure 6 A flowchart of a method for processing a physical examination report image according to another embodiment of the present application;
[0027] Figure 7 A specific flow chart of a method for processing a physical examination report image according to an embodiment of the present application;
[0028] Figure 8 It is a schematic diagram of the structure of an electronic device provided according to an embodiment of the present application. DETAILED DESCRIPTION
[0029] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.
[0030] The following describes the image enhancement model training method, physical examination report image processing method and device of the embodiment of the present application with reference to the accompanying drawings. The existing physical examination report detection method mentioned in the above background technology is usually based on the research of a physical examination result, lacks the use of historical physical examination reports, and usually only performs simple preprocessing. However, the physical examination reports of previous years have poor image quality due to long storage time. Simple preprocessing cannot effectively eliminate the negative impact of low-quality images on the research, and the image processing effect is poor and the clarity is low.
[0031] In order to solve the problem of the above, the present application provides an image enhancement model training method, in which the physical examination report image in the historical physical examination report data can be used to determine the corresponding true value image according to the physical examination report image, and the image enhancement model and the evaluation model can be trained based on the training data set generated by the physical examination report image and the true value image to complete the training of the image enhancement model and the evaluation model, and the image enhancement model is subsequently used to enhance the physical examination report image, improve the clarity of the physical examination report image, restore the image information, and facilitate the subsequent physical examination report information recognition. In this way, the problems of poor effect and low clarity of physical examination report image processing in the prior art are solved.
[0032] Specifically, Figure 1 A flowchart of an image enhancement model training method provided in an embodiment of the present application.
[0033] like Figure 1 As shown, the image enhancement model training method includes the following steps:
[0034] In step S101, historical physical examination report data is obtained.
[0035] The historical physical examination data may be obtained from a database, and the historical physical examination data includes physical examination report images.
[0036] In step S102, the physical examination report images in the historical physical examination report data are identified, and the true value images of the physical examination report images are determined, and a training data set is generated according to the physical examination report images and the true value images.
[0037] The true value image is an image with true label distribution corresponding to the physical examination report image in the historical physical examination report data.
[0038] It can be understood that the embodiments of the present application can identify physical examination report images in historical physical examination report data and determine their corresponding true value images, and generate a training data set based on the physical examination report images and the true value images to facilitate the subsequent training of the image enhancement model.
[0039] In step S103, the image enhancement model is trained using the training data set, the output image of the image enhancement model and the true value image are input into the evaluation model, the image feature difference of the output image of the evaluation model is evaluated, and the model parameters of the image enhancement model and the evaluation model are updated alternately according to the image feature difference.
[0040] It can be understood that the embodiments of the present application can use the training data set to train the image enhancement model, and input the output image of the image enhancement model and the true value image into the evaluation model, evaluate the image feature difference of the output image of the model, and then alternately update the model parameters of the image enhancement model and the evaluation model according to the image feature difference, so as to improve the enhancement performance of the image enhancement model and the evaluation performance of the evaluation model, so as to subsequently use the image enhancement model to enhance the physical examination report image, improve the clarity of the physical examination report image, and restore the image information.
[0041] It should be noted that before using the training data set to train the image enhancement model, the image is preprocessed, such as being unified into 1920pixel*1440pixel images. In addition, the training data set can also be divided into a training set, a validation set, and a test set for model training.
[0042] In an embodiment of the present application, the image enhancement model includes an encoder and a decoder, wherein the encoder includes a feature extraction layer, a connection layer, a transition layer and an attention layer, the feature extraction layer is used to extract features from the physical examination report image to obtain a feature map, the connection layer is used to connect the input and output between layers, the transition layer is used to reduce the number of channels and spatial resolution of the feature map, and the attention layer is used to filter out invalid information in the feature map; the decoder includes a residual layer and an attention layer, the residual layer is used to restore the features and spatial resolution of the feature map to obtain a restored feature map, and the attention layer is used to filter out invalid information in the restored feature map to obtain an output image.
[0043] It can be understood that the image enhancement model in the embodiment of the present application includes an encoder and a decoder, and the specific structure is as follows: Figure 2 As shown, the encoder includes a feature extraction layer, a connection layer, a transition layer and an attention layer. The feature extraction layer is used to extract features from the physical examination report image to obtain a feature map. The connection layer is used to connect the input and output between layers. The transition layer is used to reduce the number of channels and spatial resolution of the feature map. The attention layer is used to filter out invalid information in the feature map. The decoder includes a residual layer and an attention layer. The residual layer is used to restore the features and spatial resolution of the feature map to obtain a restored feature map. The attention layer is used to filter out invalid information in the restored feature map to obtain an output image.
[0044] Specifically, the generator part of the image enhancement model is an image enhancement network, which consists of an encoder and a decoder. The encoder structure is as follows: Figure 2 As shown in the figure. The encoder refers to the dense connection module of DenseNet and adds an attention module after each transition layer. Each dense connection module contains 6 dense connection layers, and each dense connection layer contains a 3×3 convolution layer. The transition layer mainly contains a 1×1 convolution layer and a maximum pooling layer with a stride of 2, which is used to reduce the number of channels of the feature map to prevent the network model from being too complicated and complete the downsampling of the feature map. The decoder structure is shown in the figure. Figure 3 As shown in the figure. The decoder uses a residual structure. Each layer contains a residual block. Each residual block contains a cross-layer data link. While using deconvolution to complete upsampling, it retains more image details and realizes the rapid propagation of input information. The attention modules of the encoder and decoder both use CBAM modules, focusing on valid information from both channel and spatial aspects and filtering invalid information.
[0045] In an embodiment of the present application, the evaluation model includes an input layer, a function layer and an output layer, wherein the input layer inputs the output image of the image enhancement model and the true value image into the function layer, the function layer calculates the multi-scale feature loss between images using a multi-scale feature loss function, and the output layer outputs the multi-scale feature loss.
[0046] It can be understood that the evaluation model of the embodiment of the present application is used to calculate the image feature difference between the output image of the image enhancement network and the true labeled image. The input layer inputs the output image of the image enhancement model and the true value image into the function layer. The function layer calculates the image feature difference between the images using a multi-scale feature loss function, and the output layer outputs the image feature difference.
[0047] Among them, the calculation formula of the multi-scale feature loss function is:
[0048]
[0049] l mae represents the mean absolute error, x n represents the nth physical examination report image in a dataset consisting of N images, S(x n ) represents the label distribution output by the image enhancement network model, y n represents the true label distribution, f c () represents the multi-level features extracted by the evaluation model.
[0050] l mae It can be expressed as:
[0051]
[0052] Among them, L represents the number of network layers of the evaluation model, Represents the feature distribution extracted from the i-th layer.
[0053] It should be noted that the evaluation model uses three convolutional layers with a step size of 2 to implement 3 times downsampling to obtain the feature difference between the predicted distribution of the image enhancement network model and the actual distribution. The specific implementation is to input the physical examination report image and the true value image output by the image enhancement network into the evaluation model respectively, and evaluate the performance of the image enhancement network model by calculating the difference in the output feature image.
[0054] In an embodiment of the present application, model parameters of an image enhancement model and an evaluation model are updated alternately according to image feature differences, including: updating the model parameters of the image enhancement model by back propagation according to image feature differences, wherein the update goal of the image enhancement model is to minimize a multi-scale feature loss function; updating the model parameters of the evaluation model by back propagation according to image feature differences, wherein the update goal of the evaluation model is to maximize the multi-scale feature loss function.
[0055] It can be understood that the embodiments of the present application can update the model parameters of the image enhancement model according to the back propagation of image feature differences to improve the enhancement performance of the image enhancement model, and because the evaluation network may also have inaccurate evaluations, the model parameters of the evaluation model can also be updated according to the back propagation of image feature differences, wherein the update goal of the image enhancement model is to minimize the multi-scale feature loss function, and the update goal of the evaluation model is to maximize the multi-scale feature loss function.
[0056] It should be noted that the evaluation model is not updated during the update process of the image enhancement model, and the image enhancement model and the evaluation model are trained alternately in adversarial fashion.
[0057] Specifically, when training the evaluation model, the preprocessed physical examination report image is input into the image enhancement network, and then the output image and the true labeled image are respectively input into the evaluation network to calculate the multi-scale feature loss, and the gradient of the evaluation model is updated through back propagation. The image enhancement network model does not participate in parameter update during the evaluation model training process. When training the image enhancement network model, the preprocessed physical examination report image is input into the image enhancement network, and the output image of the image enhancement network and the true labeled image are respectively input into the evaluation network to calculate the multi-scale feature loss, and the gradient of the image enhancement network model is updated through back propagation. In this process, the evaluation model does not participate in parameter update. The process of training the image enhancement network model and the evaluation model is actually the process of minimizing and maximizing the loss function. The image enhancement network hopes that the extracted predicted distribution is closer to the original label, so it needs to minimize the loss function; while the evaluation model hopes to better distinguish the gap between the predicted distribution and the true labeled distribution, so it needs to maximize multiple loss functions.
[0058] Through multiple rounds of alternating training, that is, multiple rounds of games between the image enhancement network model and the evaluation model, both the image enhancement network and the evaluation model can learn good performance. The image enhancement network model gradually improves the ability to convert low-quality images into high-quality images, and ultimately obtains smoother, clearer, and more accurate physical examination report images, providing data support for subsequent physical examination report information recognition and structural and comprehensive analysis.
[0059] According to the image enhancement model training method proposed in the embodiment of the present application, the corresponding true value image can be determined based on the physical examination report image in the historical physical examination report data, and the image enhancement model can be trained based on the training data set generated by the physical examination report image and the true value image. The model parameters of the image enhancement model and the evaluation model are updated alternately based on the image feature differences between the output image of the image enhancement model and the true value image output by the evaluation model to improve the performance of the image enhancement model and the evaluation model respectively. The image enhancement model is then used to enhance the physical examination report image to improve the clarity of the physical examination report image and restore the image information, thereby facilitating the subsequent recognition of the physical examination report information.
[0060] The above embodiments focus on describing the training process of the image enhancement model, and the following embodiments focus on describing from the application perspective of the image enhancement model. The incomplete parts between the embodiments can be referenced to each other.
[0061] An embodiment of the present application also provides a method for processing a physical examination report image.
[0062] Figure 5 A flowchart of a method for processing a physical examination report image is also provided according to one embodiment of the application.
[0063] like Figure 5 As shown, the physical examination report image processing method includes the following steps:
[0064] In step S201, target physical examination report data is obtained.
[0065] Among them, the target physical examination report data is the physical examination report data that needs to be processed.
[0066] In step S202, a target physical examination report image in the target physical examination report data is identified.
[0067] In step S203, the target physical examination report image is input into the image enhancement model, and the image enhancement model outputs an enhanced image, wherein the image enhancement model is a model trained in the above embodiment.
[0068] It can be understood that the embodiment of the present application can input the target physical examination report image into the image enhancement model, and the image enhancement model outputs the enhanced physical examination report image to quickly improve the clarity of the physical examination report image, restore the image information, and facilitate the subsequent physical examination report information recognition.
[0069] According to the physical examination report image processing method proposed in the embodiment of the present application, the target physical examination report image in the target physical examination report data can be identified, and the target physical examination report image can be processed using the image enhancement model to output an enhanced physical examination report image, quickly improve the clarity of the physical examination report image, restore the image information, and facilitate subsequent physical examination report information recognition.
[0070] Another embodiment of the present application also provides a method for processing physical examination report images.
[0071] Figure 6 A flowchart of a method for processing a physical examination report image is also provided according to another embodiment of the application.
[0072] like Figure 6 As shown, the physical examination report image processing method includes the following steps:
[0073] In step S301, target physical examination report data is obtained.
[0074] In step S302, a target physical examination report image in the target physical examination report data is identified.
[0075] In step S303, the target physical examination report image is input into the image enhancement model, and the image enhancement model outputs an enhanced image, wherein the image enhancement model includes an encoder and a decoder, wherein the encoder includes a feature extraction layer, a connection layer, a transition layer and an attention layer, the feature extraction layer is used to extract features from the physical examination report image to obtain a feature map, the connection layer is used to connect the input and output between layers, the transition layer is used to reduce the number of channels and spatial resolution of the feature map, and the attention layer is used to filter out invalid information in the feature map; the decoder includes a residual layer and an attention layer, the residual layer is used to restore the features and spatial resolution of the feature map to obtain a restored feature map, and the attention layer is used to filter out invalid information in the restored feature map to obtain an output image.
[0076] It can be understood that the embodiment of the present application can enhance the physical examination report image through the image enhancement model, and the encoder and decoder parts in the image enhancement model both use attention layers to focus on valid information from both channel and spatial aspects, and filter out invalid information to improve the accuracy of image enhancement.
[0077] According to the physical examination report image processing method proposed in the embodiment of the present application, the target physical examination report image in the target physical examination report data can be identified, and the target physical examination report image can be processed using the image enhancement model to output an enhanced physical examination report image, quickly improve the clarity of the physical examination report image, restore the image information, and facilitate subsequent physical examination report information recognition.
[0078] The following is a specific example to describe the medical examination report image processing method of the present application. Figure 7 As shown, including:
[0079] S1: Input the physical examination report image.
[0080] S2: Image preprocessing.
[0081] Unify the image into an image of 1920pixel*1440pixel.
[0082] S3: Image enhancement model based on generative adversarial network
[0083] The generator part of the model is an image enhancement network consisting of an encoder and a decoder, which is used to improve the clarity of the physical examination report image and restore the image information.
[0084] S4: Output clear physical examination report image through image enhancement model.
[0085] Figure 8 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device may include:
[0086] A memory 801 , a processor 802 , and a computer program stored in the memory 801 and executable on the processor 802 .
[0087] When the processor 802 executes the program, the image enhancement model training method or the physical examination report image processing method provided in the above embodiments is implemented.
[0088] Furthermore, the electronic device further comprises:
[0089] The communication interface 803 is used for communication between the memory 801 and the processor 802 .
[0090] The memory 801 is used to store computer programs that can be executed on the processor 802 .
[0091] The memory 801 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.
[0092] If the memory 801, the processor 802 and the communication interface 803 are implemented independently, the communication interface 803, the memory 801 and the processor 802 can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 8 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.
[0093] Optionally, in a specific implementation, if the memory 801, the processor 802 and the communication interface 803 are integrated on a chip, the memory 801, the processor 802 and the communication interface 803 can communicate with each other through an internal interface.
[0094] The processor 802 may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.
[0095] An embodiment of the present application also provides a computer-readable storage medium on which a computer program or instruction is stored. When the computer program or instruction is executed by a processor, the above-mentioned image enhancement model training method or physical examination report image processing method is implemented.
[0096] An embodiment of the present application also provides a computer program product, including a computer program or instructions, which, when executed, implements the above-mentioned image enhancement model training method or physical examination report image processing method.
[0097] In the description of this specification, 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 N 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.
[0098] 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 technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise clearly and specifically defined.
[0099] 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 N 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 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.
[0100] 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 embodiment, the N 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 or a combination of multiple of the following technologies known in the art: 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.
[0101] 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.
Claims
1. A method for training an image enhancement model, characterized in that: The following steps are involved: Obtain historical physical examination report data; Identify the physical examination report image in the historical physical examination report data, determine the true value image of the physical examination report image, and generate a training data set according to the physical examination report image and the true value image; The image enhancement model is trained using the training data set, the output image of the image enhancement model and the true value image are input into an evaluation model, the image feature difference of the evaluation model output image is obtained, and the model parameters of the image enhancement model and the evaluation model are alternately updated according to the image feature difference.
2. The image enhancement model training method according to claim 1, characterized in that: The image enhancement model includes an encoder and a decoder, wherein: The encoder includes a feature extraction layer, a connection layer, a transition layer and an attention layer, wherein the feature extraction layer is used to extract features from the physical examination report image to obtain a feature map, the connection layer is used to connect the input and output between layers, the transition layer is used to reduce the number of channels and spatial resolution of the feature map, and the attention layer is used to filter invalid information in the feature map; The decoder includes a residual layer and an attention layer. The residual layer is used to restore the features and spatial resolution of the feature map to obtain a restored feature map, and the attention layer is used to filter out invalid information in the restored feature map to obtain an output image.
3. The image enhancement model training method according to claim 1, characterized in that: The evaluation model includes an input layer, a function layer and an output layer, wherein: The input layer inputs the output image of the image enhancement model and the true value image to the function layer, the function layer calculates the image feature difference between images using a multi-scale feature loss function, and the output layer outputs the image feature difference.
4. The image enhancement model training method according to claim 3, characterized in that: The calculation formula of the multi-scale feature loss function is: Among them, l mae represents the mean absolute error, x n represents the nth physical examination report image in a dataset consisting of N images, S(x n ) represents the label distribution output by the image enhancement network model, y n represents the true label distribution, f C () represents the multi-level features extracted by the evaluation model.
5. The image enhancement model training method according to claim 4, characterized in that: The alternately updating the model parameters of the image enhancement model and the evaluation model according to the image feature difference comprises: Back-propagating the image feature difference to update the model parameters of the image enhancement model, wherein the updating target of the image enhancement model is to minimize the multi-scale feature loss function; The model parameters of the evaluation model are updated according to the back propagation of the image feature difference, wherein the updating goal of the evaluation model is to maximize the multi-scale feature loss function.
6. A method for processing a physical examination report image, characterized in that: The following steps are involved: Obtain target physical examination report data; Identifying a target physical examination report image in the target physical examination report data; The target physical examination report image is input into an image enhancement model, and the image enhancement model outputs an enhanced image, wherein the image enhancement model is a model trained by the image enhancement model training method according to any one of claims 1 to 5.
7. A method for processing a physical examination report image, characterized in that: The following steps are involved: Obtain target physical examination report data; Identifying a target physical examination report image in the target physical examination report data; The target physical examination report image is input into the image enhancement model, and the image enhancement model outputs an enhanced image, wherein the image enhancement model includes an encoder and a decoder, wherein the encoder includes a feature extraction layer, a connection layer, a transition layer and an attention layer, the feature extraction layer is used to extract features from the physical examination report image to obtain a feature map, the connection layer is used to connect the input and output between layers, the transition layer is used to reduce the number of channels and spatial resolution of the feature map, and the attention layer is used to filter out invalid information in the feature map; the decoder includes a residual layer and an attention layer, the residual layer is used to restore the features and spatial resolution of the feature map to obtain a restored feature map, and the attention layer is used to filter out invalid information in the restored feature map to obtain an output image.
8. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the image enhancement model training method as described in any one of claims 1 to 5, or the physical examination report image processing method as described in any one of claims 6 to 7.
9. A computer-readable storage medium having a computer program or instruction stored thereon, characterized in that: The computer program or instruction is executed by a processor to implement the image enhancement model training method as described in any one of claims 1-5, or the physical examination report image processing method as described in any one of claims 6-7.
10. A computer program product comprising a computer program or instructions, characterized in that When the computer program or instruction is executed, it implements the image enhancement model training method as described in any one of claims 1-5, or the physical examination report image processing method as described in any one of claims 6-7.