Air cavity mirror image processing device, method, equipment and medium

Abstract

The invention discloses an air endoscope image processing device and method, equipment and a medium, and relates to the field of medical image recognition, and the device comprises an image acquisition module, a preprocessing module and a tumor recognition module. The image acquisition module is used for acquiring a to-be-processed gas cavity mirror image; the preprocessing module sequentially performs enhancement and segmentation processing on the to-be-processed gas cavity mirror image; the tumor recognition module adopts a tumor recognition model to classify the preprocessed images; the tumor identification model comprises an encoder, a spatial reconstruction model, a student model and a discriminator; the encoder carries out multiple times of down-sampling on the preprocessed image; the spatial reconstruction model carries out feature space reconstruction on the coded image; the student model adopts a convolution attention mechanism to carry out multiple times of up-sampling on the spatial reconstruction image; and the discriminator compares the difference between the final reconstructed image and the preprocessed image so as to determine the category of the to-be-processed gas cavity mirror image. According to the invention, the processing precision and speed of the gas cavity mirror image are improved.

Description

Technical Field

[0001] The present application relates to the field of medical image recognition, and in particular to a laparoscope image processing device, method, equipment and medium. Background Art

[0002] Airway abnormalities can be attributed to a range of pathogenic mechanisms. Lesions located in the central airways include malignant tumors, foreign bodies, infectious diseases and other rare causes, all of which present with different symptoms. Central airway tumors require precise and safe diagnostic and treatment techniques to a large extent because they may lead to dangerous situations. In actual clinical medicine, the clinical expertise and experience of bronchoscopists are the main factors affecting the results of lesion identification. However, since the recognition of lesion images is currently too dependent on human eye observation, once there is interference from lack of experience, unclear images or other subjective and objective factors, the human eye recognition results will have omissions, errors and other problems, and manual recognition is slow and inefficient, resulting in the inability to ensure the accuracy and speed of recognition in daily recognition. Summary of the invention

[0003] The purpose of the present application is to provide a laparoscope image processing device, method, equipment and medium, which can improve the processing accuracy and speed of laparoscope images.

[0004] To achieve the above objectives, this application provides the following solutions:

[0005] In a first aspect, the present application provides a laparoscope image processing device, comprising:

[0006] An image acquisition module, used to acquire the laparoscopic image to be processed;

[0007] A preprocessing module, connected to the image acquisition module, for sequentially performing enhancement and segmentation processing on the laparoscopic image to be processed to obtain a preprocessed image;

[0008] A tumor recognition module, connected to the preprocessing module, is used to classify the preprocessed image using a tumor recognition model to determine the category of the laparoscopic image to be processed; the category is abnormal or normal;

[0009] Among them, the tumor recognition model includes an encoder, a spatial reconstruction model, a student model and a discriminator; the encoder is used to downsample the preprocessed image multiple times to obtain an encoded image; the spatial reconstruction model is used to reconstruct the feature space of the encoded image to obtain a spatial reconstructed image; the student model is used to upsample the spatial reconstructed image multiple times using a convolutional attention mechanism to obtain a final reconstructed image; the discriminator is used to compare the difference between the final reconstructed image and the preprocessed image to determine the category of the laparoscopic image to be processed.

[0010] In a second aspect, the present application provides a laparoscope image processing method, comprising:

[0011] Acquiring a laparoscope image to be processed;

[0012] Performing enhancement and segmentation processing on the to-be-processed laparoscopic image in sequence to obtain a preprocessed image;

[0013] Using a tumor recognition model to classify the pre-processed image to determine the category of the laparoscopic image to be processed; the category is abnormal or normal;

[0014] Among them, the tumor recognition model includes an encoder, a spatial reconstruction model, a student model and a discriminator; the encoder is used to downsample the preprocessed image multiple times to obtain an encoded image; the spatial reconstruction model is used to reconstruct the feature space of the encoded image to obtain a spatial reconstructed image; the student model is used to upsample the spatial reconstructed image multiple times using a convolutional attention mechanism to obtain a final reconstructed image; the discriminator is used to compare the difference between the final reconstructed image and the preprocessed image to determine the category of the laparoscopic image to be processed.

[0015] In a third aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned laparoscopic image processing method.

[0016] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which implements the above-mentioned laparoscopic image processing method when executed by a processor.

[0017] According to the specific embodiments provided in this application, this application has the following technical effects:

[0018] The present application provides a laparoscopic image processing device, method, equipment and medium, which can improve the contrast of the pixels in the lesion area in the laparoscopic image by enhancing and segmenting the laparoscopic image to be processed, so as to make the edge of the lesion area clearer. The encoder, spatial reconstruction model, student model and discriminator are further used to classify the preprocessed image, and the convolution attention mechanism is used in the student model to upsample the spatial reconstructed image, which improves the discrimination accuracy and speed of the tumor recognition model, thereby improving the processing accuracy and speed of the laparoscopic image. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 A schematic diagram of functional modules of a laparoscope image processing device provided in one embodiment of the present application;

[0021] Figure 2 This is a schematic diagram of the structure of a tumor recognition model in one embodiment of the present application;

[0022] Figure 3 A schematic flow chart of a laparoscope image processing method provided in one embodiment of the present application. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0024] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0025] In an exemplary embodiment, a laparoscope image processing apparatus is provided, which is a device having image processing capabilities, such as a computer device. Figure 1 As shown, the laparoscopic image processing device includes: an image acquisition module 101, a preprocessing module 102 and a tumor recognition module 103. The functions and principles of each module are described below.

[0026] (i) The image acquisition module 101 is used to acquire the laparoscopic image to be processed.

[0027] In a specific application example, the laparoscopic image to be processed is a CT image, a magnetic resonance image or a hyperspectral image.

[0028] (ii) The preprocessing module 102 is connected to the image acquisition module 101, and is used to sequentially enhance and segment the laparoscopic image to be processed to obtain a preprocessed image.

[0029] In a specific application example, the preprocessing module 102 includes: an edge detection submodule, a background removal submodule, an image enhancement submodule and an image segmentation submodule.

[0030] The edge detection submodule is used to scan the pixel values ​​of the cavitoscope image to be processed and compare the difference between adjacent pixel values. If the difference between adjacent pixel values ​​is less than a set threshold, the adjacent pixels are removed to obtain the cavitoscope frame.

[0031] Specifically, the laparoscopic image to be processed is first scanned horizontally and compared left and right. If the difference between adjacent pixel values ​​is less than the set threshold, they are the same part and are removed. If it is greater than or equal to the set threshold, they are different parts and are retained. Then the laparoscopic image to be processed is scanned vertically and compared up and down. If the difference between adjacent pixel values ​​is less than the set threshold, they are the same part and are removed. If it is greater than or equal to the set threshold, they are different parts and are retained. The images obtained in the above two steps are merged to obtain the laparoscopic frame.

[0032] The background removal submodule is connected to the frame detection submodule, and is used to remove pixels located outside the frame of the cavitoscope in the cavitoscope image to be processed to obtain a cavitoscope foreground image.

[0033] The image enhancement submodule is connected to the background removal submodule, and is used to convert the laparoscopic foreground image into a grayscale image, and enhance the laparoscopic foreground image based on the grayscale image to obtain an enhanced image.

[0034] The image segmentation submodule is connected to the image enhancement submodule, and is used to perform organ segmentation and blood vessel segmentation on the enhanced image in sequence to obtain a preprocessed image.

[0035] In a specific application example, the above-mentioned image enhancement submodule includes: a grayscale conversion unit, a probability parameter determination unit, a grayscale factor determination unit, a feature parameter determination unit, a variation coefficient determination unit, an adjustment factor determination unit and an enhancement unit.

[0036] The grayscale conversion unit is used to convert the cavitoscope foreground image into a grayscale image, and perform sliding sampling on the grayscale image based on a set sliding window.

[0037] The probability parameter determination unit is connected to the grayscale conversion unit, and is used to determine the probability adjustment parameter of each pixel point according to the distribution of the pixel points in each sliding window and the distribution of the pixel points in the grayscale image.

[0038] Specifically, firstly, according to the distribution of pixels in each sliding window and the distribution of pixels in the grayscale image, the probability parameter of each pixel is determined: ;in, is the probability parameter of the i-th pixel, is the discrete degree of the grayscale value of the pixel in the sliding window corresponding to the i-th pixel, is the discrete degree of grayscale values ​​of all pixels in the grayscale image, is the total number of pixels, is the discrete degree of the grayscale value of the pixel in the sliding window corresponding to the vth pixel. Then, the probability adjustment parameter of each pixel is determined according to the difference between the probability parameter of each pixel and the discrete degree of the grayscale value of the pixel in the corresponding sliding window and the discrete degree of the grayscale value of the pixel in the grayscale image: ;in, The probability adjustment parameter for the i-th pixel.

[0039] The grayscale factor determination unit is connected to the grayscale conversion unit, and is used to determine the grayscale control factor of each pixel point according to the grayscale value of each pixel point in each sliding window and the grayscale value of each pixel point in the grayscale map.

[0040] Specifically, firstly, the grayscale coefficient of each pixel is determined according to the grayscale value mean of the pixel in each sliding window and the grayscale value mean of the pixel in the grayscale image: Among them, Ga i is the grayscale coefficient of the i-th pixel, C is the length of the sliding window, is the gray value of the jth pixel, is the mean grayscale value of all pixels in the grayscale image. Then, according to the difference between the mean value of the pixels in each sliding window and the mean value of the pixels in the grayscale image and the grayscale coefficient, the grayscale control factor of each pixel is determined: ;in, is the grayscale control factor of the i-th pixel.

[0041] The characteristic parameter determination unit is connected to the probability parameter determination unit and the grayscale factor determination unit respectively, and the characteristic parameter determination unit is used to determine the characteristic parameter of each pixel point according to the probability adjustment parameter and the grayscale control factor of each pixel point: ;in, is the characteristic parameter of the i-th pixel.

[0042] The variation coefficient determination unit is connected to the characteristic parameter determination unit, and the variation coefficient determination unit is used to determine the variation coefficient of each pixel point according to the difference between the characteristic parameters of adjacent pixel points.

[0043] Specifically, the change parameter of each pixel is determined according to the difference between the characteristic parameters of each pixel and the characteristic parameters of adjacent pixels: ;in, is the change parameter of the i-th pixel, is the change parameter of the i-1th pixel, is the characteristic parameter of the j-th pixel, is the characteristic parameter of the j+1th pixel. Then, according to the difference between the variation parameter of each pixel and the mean variation parameter of all pixels, the variation coefficient of each pixel is determined: ; Among them, T i is the coefficient of change of the i-th pixel.

[0044] The adjustment factor determination unit is used to determine the adjustment factor of each pixel point according to the probability adjustment parameter, grayscale control factor and variation coefficient of each pixel point: Among them, A i is the adjustment factor of the i-th pixel.

[0045] The enhancement unit is connected to the adjustment factor determination unit, and is used to enhance the cavitoscope foreground image according to the adjustment factor of each pixel point to obtain an enhanced image.

[0046] Specifically, the gray value of each pixel in the laparoscopic foreground image is multiplied by the corresponding adjustment factor to obtain an enhanced image.

[0047] The present application sets a sliding window to adaptively adjust each pixel point, thereby avoiding the problem of too low or too high contrast caused by global pixel adjustment. By analyzing the distribution and aggregation of the grayscale values ​​of the pixels in each window and the size of the grayscale values, the grayscale values ​​of the pixels are adjusted to enhance the contrast of the grayscale values ​​of the pixels in the lesion area. The grayscale values ​​of the pixels are further adjusted according to the noise probability to reduce the contrast of the grayscale images of the normal tissue area and the noise area. The corresponding pixels of the windows are adjusted according to the changes in the distribution characteristics of the pixels in consecutive adjacent windows to improve the contrast of the pixels at the edge of the lesion area and make the edge of the lesion area clearer.

[0048] (III) The tumor recognition module 103 is connected to the pre-processing module 102, and is used to classify the pre-processed image using a tumor recognition model to determine the category of the laparoscopic image to be processed, which is abnormal or normal.

[0049] Among them, Figure 2 As shown, the tumor recognition model includes an encoder 201, a space reconstruction model 202, a student model 203 and a discriminator 204. The student model 203 is obtained by distillation training using a pre-trained teacher model 204. The functions and structures of each part are described below.

[0050] (1) The encoder 201 is used to perform multiple downsampling on the preprocessed image to obtain an encoded image.

[0051] The encoder 201 includes a first downsampling layer, a second downsampling layer, a third downsampling layer and a fourth downsampling layer connected in sequence. The first downsampling layer includes a convolution layer, a regularization layer and an activation function layer connected in sequence. The second downsampling layer, the third downsampling layer and the fourth downsampling layer each include a deformable convolution layer, a regularization layer and an activation function layer connected in sequence. The parameters of the regularization layer and the activation function layer in the encoder 201 are fixed.

[0052] Among them, the formula for deformable convolution is:

[0053] ;

[0054] Among them, x is the input feature map of the deformable convolution, y is the output feature map of the deformable convolution, and p o is the center point of the sliding window, p n is the sampling point of the convolution kernel, Δp n is the offset, Δm n is the learnable weight for each position, R is the total number of sampling points, and w is the weight matrix.

[0055] (2) The spatial reconstruction model 202 is used to reconstruct the feature space of the encoded image to obtain a spatially reconstructed image.

[0056] The spatial reconstruction model includes a memory unit and a reconstruction unit. The memory unit is used to store historical features using a memory queue. The reconstruction unit is connected to the memory unit and the fourth downsampling layer respectively, and the reconstruction unit is used to perform feature space reconstruction on the encoded image according to the historical features stored in the memory queue to obtain a spatially reconstructed image.

[0057] During the training of the spatial reconstruction model 202, the memory unit stores historical features in a queue structure, which acts as a dictionary of normal image samples. During the reconstruction period, the reconstruction unit performs reconstruction of the feature space according to the similarity of the features.

[0058] (3) The student model 203 is used to use a convolutional attention mechanism to perform multiple upsampling on the spatially reconstructed image to obtain a final reconstructed image.

[0059] The student model 203 includes a first memory layer, a first upsampling layer, a second memory layer, a second upsampling layer and a third upsampling layer which are connected in sequence.

[0060] The first memory layer and the second memory layer both include a convolutional attention mechanism and a memory matrix. The input of the first memory layer is the result of splicing the spatially reconstructed image and the output of the third downsampling layer. The input of the second memory layer is the result of splicing the output of the second downsampling layer and the output of the first upsampling layer.

[0061] The first upsampling layer, the second upsampling layer and the third upsampling layer each include a convolution layer, a regularization layer and an activation function layer connected in sequence.

[0062] (4) The discriminator 204 is used to compare the difference between the final reconstructed image and the pre-processed image to determine the category of the laparoscopic image to be processed.

[0063] Specifically, the discriminator 204 uses the formula Calculate the anomaly score; where A is the anomaly score, Φ is the Sigmoid function, μ is the mean of the anomaly scores calculated by the discriminator 204 on the training samples, δ is the standard deviation of the anomaly scores calculated by the discriminator 204 on the training samples, To preprocess the image, For encoder 201, Model 203 for students, is the discriminator 204.

[0064] (5) The teacher model 204 includes a fourth upsampling layer, a fifth upsampling layer and a sixth upsampling layer connected in sequence. The fourth upsampling layer, the fifth upsampling layer and the sixth upsampling layer all include a convolution layer, a regularization layer and an activation function layer.

[0065] The input of the fourth upsampling layer is the result of the concatenation of the output of the third downsampling layer and the output of the fourth downsampling layer. The input of the fifth upsampling layer is the result of the concatenation of the output of the second downsampling layer and the output of the fourth upsampling layer. The concatenation of the output of the activation function layer of the fourth upsampling layer and the output of the regularization layer of the sixth upsampling layer serves as the input of the activation function layer of the sixth upsampling layer.

[0066] The above activation function layers are all ReLU activation functions.

[0067] In a specific application example, the loss function involved in the tumor recognition model training process includes the following parts:

[0068] ① The result of the fourth upsampling layer in the teacher model 204 and the result of the first upsampling layer in the student model 203 are used to calculate the loss function: ;in, The first loss function value, is the result of the fourth upsampling layer, is the result of the first upsampling layer, and K is the feature level of the intermediate layer.

[0069] ② The result of the fifth upsampling layer in the teacher model 204 and the result of the second upsampling layer in the student model 203 are used to calculate the loss function: ;in, is the second loss function value, is the result of the fifth upsampling layer, is the result of the second upsampling layer.

[0070] ③ The result of the sixth upsampling layer in the teacher model 204 and the loss function calculated from the preprocessed image: ;in, is the third loss function value, Model 204 for teachers.

[0071] ④ The result of the third upsampling layer in the student model 203 and the loss function calculated by the preprocessed image: ;in, is the value of the fourth loss function.

[0072] ⑤ The loss function is calculated separately for the result of the third upsampling layer in the student model 203: ;in, is the fifth loss function value.

[0073] ⑥ The discriminator 204 calculates the loss function based on its discrimination result: ;in, is the value of the sixth loss function.

[0074] This application applies an unsupervised anomaly detection algorithm based on feature recombination and knowledge distillation to the task of laparoscopic tumor detection, uses a convolutional attention mechanism to assist the memory matrix in feature focusing, and helps the student model 203 store more meaningful normal sample features during the training process, which can improve the discrimination accuracy of the tumor recognition model. It uses deformable convolution to optimize the encoding process, increases the computational discreteness of the matrix space during the derivation of the tumor recognition model, and can preserve the tumor features inside the trachea with a more appropriate receptive field, thereby improving the accuracy and speed of laparoscopic image processing.

[0075] Based on the same inventive concept, the embodiment of the present application also provides a laparoscopic image processing method using the laparoscopic image processing device mentioned above. The implementation scheme for solving the problem provided by the method is similar to the implementation scheme recorded in the above-mentioned device, so the specific limitations in one or more laparoscopic image processing method embodiments provided below can refer to the limitations of the laparoscopic image processing device above, and will not be repeated here.

[0076] In an exemplary embodiment, Figure 3 As shown, a method for processing laparoscopic images is provided, comprising steps 301 to 303.

[0077] Step 301, obtaining a laparoscope image to be processed.

[0078] Step 302, performing enhancement and segmentation processing on the laparoscopic image to be processed in sequence to obtain a preprocessed image.

[0079] Step 303: Classify the pre-processed image using a tumor recognition model to determine the category of the laparoscopic image to be processed, which is abnormal or normal.

[0080] The tumor recognition model includes an encoder 201, a spatial reconstruction model 202, a student model 203 and a discriminator 204. The encoder 201 is used to perform multiple downsampling on the preprocessed image to obtain an encoded image. The spatial reconstruction model 202 is used to reconstruct the feature space of the encoded image to obtain a spatial reconstructed image. The student model 203 is used to perform multiple upsampling on the spatial reconstructed image using a convolutional attention mechanism to obtain a final reconstructed image. The discriminator 204 is used to compare the difference between the final reconstructed image and the preprocessed image to determine the category of the laparoscopic image to be processed.

[0081] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiment when executing the computer program.

[0082] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiment are implemented.

[0083] In an exemplary embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiment when executed by a processor.

[0084] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0085] In this application, all actions to obtain signals, information or data are carried out in compliance with the relevant data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.

[0086] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0087] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0088] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0089] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A laparoscope image processing device, characterized in that: The laparoscopic image processing device comprises: An image acquisition module, used for acquiring the laparoscopic image to be processed; A preprocessing module, connected to the image acquisition module, for sequentially enhancing and segmenting the laparoscopic image to be processed to obtain a preprocessed image; A tumor recognition module, connected to the preprocessing module, is used to classify the preprocessed image using a tumor recognition model to determine the category of the laparoscopic image to be processed; the category is abnormal or normal; Among them, the tumor recognition model includes an encoder, a spatial reconstruction model, a student model and a discriminator; the encoder is used to downsample the preprocessed image multiple times to obtain an encoded image; the spatial reconstruction model is used to reconstruct the feature space of the encoded image to obtain a spatial reconstructed image; the student model is used to upsample the spatial reconstructed image multiple times using a convolutional attention mechanism to obtain a final reconstructed image; the discriminator is used to compare the difference between the final reconstructed image and the preprocessed image to determine the category of the laparoscopic image to be processed.

2. The laparoscopic image processing device according to claim 1, characterized in that: The preprocessing module comprises: An edge detection submodule, used for scanning the pixel values ​​of the to-be-processed cavitoscope image, and comparing the difference between adjacent pixel values. If the difference between adjacent pixel values ​​is less than a set threshold, the adjacent pixels are removed to obtain a cavitoscope frame; A background removal submodule, connected to the frame detection submodule, for removing pixels located outside the frame of the cavitoscope in the cavitoscope image to be processed to obtain a cavitoscope foreground image; An image enhancement submodule, connected to the background removal submodule, for converting the laparoscopic foreground image into a grayscale image, and enhancing the laparoscopic foreground image based on the grayscale image to obtain an enhanced image; The image segmentation submodule is connected to the image enhancement submodule and is used to perform organ segmentation and blood vessel segmentation on the enhanced image in sequence to obtain a preprocessed image.

3. The laparoscopic image processing device according to claim 2, characterized in that: The image enhancement submodule comprises: A grayscale conversion unit, used for converting the cavitoscope foreground image into a grayscale image, and slidingly sampling the grayscale image based on a set sliding window; A probability parameter determination unit, connected to the grayscale conversion unit, for determining a probability adjustment parameter for each pixel according to the distribution of pixels in each sliding window and the distribution of pixels in the grayscale image; A grayscale factor determination unit, connected to the grayscale conversion unit, for determining a grayscale control factor of each pixel point according to the grayscale value of each pixel point in each sliding window and the grayscale value of each pixel point in the grayscale map; a characteristic parameter determination unit, connected to the probability parameter determination unit and the grayscale factor determination unit respectively, and configured to determine the characteristic parameter of each pixel point according to the probability adjustment parameter and the grayscale control factor of each pixel point; a variation coefficient determination unit, connected to the characteristic parameter determination unit, for determining the variation coefficient of each pixel point according to the difference between the characteristic parameters of adjacent pixel points; An adjustment factor determination unit, used to determine the adjustment factor of each pixel point according to the probability adjustment parameter, grayscale control factor and variation coefficient of each pixel point; The enhancement unit is connected to the adjustment factor determination unit and is used to enhance the cavitoscope foreground image according to the adjustment factor of each pixel point to obtain an enhanced image.

4. The laparoscopic image processing device according to claim 1, characterized in that: The encoder comprises a first downsampling layer, a second downsampling layer, a third downsampling layer and a fourth downsampling layer connected in sequence; The first downsampling layer includes a convolution layer, a regularization layer, and an activation function layer connected in sequence; The second downsampling layer, the third downsampling layer and the fourth downsampling layer each include a deformable convolution layer, a regularization layer and an activation function layer connected in sequence.

5. The laparoscopic image processing device according to claim 4, characterized in that: The spatial reconstruction model includes: A memory unit for storing historical features using a memory queue; The reconstruction unit is connected to the memory unit and the fourth downsampling layer respectively, and is used to reconstruct the feature space of the encoded image according to the historical features stored in the memory queue to obtain a spatially reconstructed image.

6. The laparoscopic image processing device according to claim 5, characterized in that: The student model comprises a first memory layer, a first upsampling layer, a second memory layer, a second upsampling layer and a third upsampling layer connected in sequence; The first memory layer and the second memory layer both include a convolutional attention mechanism and a memory matrix; the input of the first memory layer is the result of splicing the spatially reconstructed image and the output of the third downsampling layer; the input of the second memory layer is the result of splicing the output of the second downsampling layer and the output of the first upsampling layer; The first upsampling layer, the second upsampling layer and the third upsampling layer each include a convolution layer, a regularization layer and an activation function layer connected in sequence.

7. The laparoscopic image processing device according to claim 1, characterized in that: The student model is obtained by distilling and training a pre-trained teacher model.

8. A method for processing laparoscopic images, using the laparoscopic image processing device according to any one of claims 1 to 7, characterized in that: The laparoscopic image processing method comprises: Acquiring a laparoscope image to be processed; Performing enhancement and segmentation processing on the to-be-processed laparoscopic image in sequence to obtain a preprocessed image; Using a tumor recognition model to classify the pre-processed image to determine the category of the laparoscopic image to be processed; the category is abnormal or normal; Among them, the tumor recognition model includes an encoder, a spatial reconstruction model, a student model and a discriminator; the encoder is used to downsample the preprocessed image multiple times to obtain an encoded image; the spatial reconstruction model is used to reconstruct the feature space of the encoded image to obtain a spatial reconstructed image; the student model is used to upsample the spatial reconstructed image multiple times using a convolutional attention mechanism to obtain a final reconstructed image; the discriminator is used to compare the difference between the final reconstructed image and the preprocessed image to determine the category of the laparoscopic image to be processed.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the laparoscopic image processing method described in claim 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the laparoscopic image processing method described in claim 8 is implemented.

Images