Bladder cancer early detection system based on image recognition

Abstract

The invention provides a bladder cancer early detection system based on image recognition, and the system comprises an enhancement module, a first denoising module, a second denoising module, a segmentation module and a detection module. The image contrast can be effectively improved while detail information in the image can be fully reserved and highlighted, then the image is denoised twice through the first denoising module and the second denoising module, noise and artifacts in the image can be effectively reduced, edge texture information of the image can be effectively captured, the visual quality and the image structure are improved, and the image quality is improved. And then the image is segmented through the segmentation module, so that the segmentation quality can be improved, the sensitivity to noise can be reduced, and the accuracy and precision of subsequent model identification and detection can be improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of image recognition, and particularly relates to an early detection system for bladder cancer based on image recognition. Background Art

[0002] Bladder cancer refers to the occurrence of malignant tumors in the bladder, which is one of the common malignant tumors in the urinary system. The causes of bladder cancer are relatively complex, which may be caused by genetics, or may be caused by long-term smoking, exposure to aromatic amine chemicals, environmental pollution, etc. Patients with bladder cancer will first show symptoms of hematuria. As the disease worsens, patients will gradually show manifestations such as frequent urination, urgency of urination, difficulty in urination, dysuria, and lower limb edema.

[0003] Among them, the early symptoms of bladder cancer are manifested as hematuria, difficulty in urination, or abdominal pain, mass and other symptoms. However, in the actual early identification process of bladder cancer, medical imaging is usually used as the basis for identification, such as CT, MRI, etc. In the prior art, the early identification of bladder cancer is usually based on the experience of medical practitioners. In order to improve the identification efficiency and accuracy, artificial intelligence is also used to assist in the early identification of bladder cancer by identifying bladder images. However, due to the presence of noise in the images, as well as low contrast, unclear texture details, and blurred lesion areas, the accuracy of identification is affected. Summary of the Invention

[0004] In order to solve the above technical problems, the present invention provides an early detection system for bladder cancer based on image recognition to solve the technical problems in the prior art.

[0005] On the one hand, the present invention provides the following technical solution. An early detection system for bladder cancer based on image recognition, characterized in that the system includes: An enhancement module, configured to obtain an abdominal medical image of a target object, and perform equalization enhancement processing on the abdominal medical image to obtain an enhanced image; A first denoising module, configured to construct a target denoising function based on the enhanced image, perform variable transformation on the target denoising function to obtain a transformation function, and perform iterative solution on the transformation function to obtain a first denoised image; A second denoising module, configured to perform image reconstruction denoising on the first denoised image to obtain a second denoised image; A segmentation module, configured to perform image clustering segmentation on the second denoised image to obtain a target region image; A detection module, configured to obtain preset training data, input the preset training data into a preset detection model for training, and input the target region image into the trained preset detection model for detection to output a detection result.

[0006] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention includes an enhancement module for obtaining an abdominal medical image of a target object and performing equalization enhancement processing on the abdominal medical image to obtain an enhanced image; a first denoising module for constructing a target denoising function based on the enhanced image, performing variable transformation on the target denoising function to obtain a transformation function, and performing iterative solution on the transformation function to obtain a first denoised image; a second denoising module for performing image reconstruction denoising on the first denoised image to obtain a second denoised image; a segmentation module for performing image clustering segmentation on the second denoised image to obtain a target region image; and a detection module for obtaining preset training data, inputting the preset training data into a preset detection model for training, and inputting the target region image into the trained preset detection model for detection to output a detection result. First, the present invention performs enhancement processing on the image through the enhancement module, which can effectively improve the image contrast and also fully retain and highlight the detailed information in the image. Then, the image is denoised twice through the first denoising module and the second denoising module, which can effectively reduce the noise and artifacts in the image and effectively capture the edge texture information of the image, improving the visual quality and image structure and avoiding image distortion. Then, the image is segmented through the segmentation module, which can improve the segmentation quality and reduce the sensitivity to noise, thereby improving the accuracy and precision of subsequent model recognition and detection.

[0007] Preferably, the enhancement module includes: An equalization sub-module for performing equalization processing on the abdominal medical image using a preset equalization algorithm to obtain an equalized image; A transformation sub-module for performing discrete wavelet transform on the equalized image to obtain a first low-frequency component, a first horizontal component, a first vertical component, and a first diagonal component, and performing discrete wavelet transform on the abdominal medical image to obtain a second low-frequency component, a second horizontal component, a second vertical component, and a second diagonal component; A decomposition sub-module for performing singular value decomposition on the first low-frequency component to obtain a first eigenvector , a first singular value vector , a second eigenvector , performing singular value decomposition on the second low-frequency component to obtain a third eigenvector , a second singular value vector , a fourth eigenvector ; A difference sub-module for based on the first eigenvector , the first singular value vector , the second eigenvector , the third eigenvector , the second singular value vector , the fourth eigenvector Calculate the image difference component : ; In the formula, represents the adjustable weight; The first sub-component module is used to calculate the first image component based on the image difference component : ; The second sub-component module is used to calculate the second image component, the third image component, and the fourth image component respectively based on a horizontal component, a first vertical component, a first diagonal component, a second horizontal component, a second vertical component, and a second diagonal component ; The fusion sub-module is used to fuse the first image component, the second image component, the third image component, and the fourth image component to obtain a fused image, and perform inverse discrete wavelet transform on the fused image to obtain an enhanced image Preferably, the second sub-component module includes: The second component unit is used to calculate the second image component based on the first horizontal component and the second horizontal component :

[0008] ; ; ; In the formula, respectively represent the first horizontal frequency and the second horizontal frequency, represents the image size, respectively represent the gray values of the image corresponding to the first horizontal component at ; respectively represent the gray values of the image corresponding to the second horizontal component at ; The third component unit is used to calculate the third image component based on the first vertical component and the second vertical component : ; ; ; ​​​​​​​​​​; In the formula, respectively represent the first vertical frequency and the second vertical frequency, respectively represent the gray value of the image corresponding to the first vertical component at ; respectively represent the gray value of the image corresponding to the second vertical component at ; The fourth component unit is used to calculate the fourth image component , the second diagonal component : : ; ; ; In the formula, respectively represent the first diagonal frequency and the second diagonal frequency.

[0009] Preferably, the first denoising module includes: The function conversion sub-module is used to construct the target denoising function and perform variable conversion on the target denoising function to obtain the conversion function; The iterative function sub-module is used to determine the iterative objective function based on the conversion function : ; In the formula, is the penalty parameter, is the Lagrange parameter, is the inner product calculation, is the Gaussian weight matrix, represents the estimated image, is the regularization factor, is the discrete gradient operator, represents the number of pixel points in the enhanced image, is the decision variable; The weight iteration sub-module is used to perform several iterations on in the and determine the weights of the convolutional neural network generator , the decision variable and the Lagrange parameter ; The output solution sub-module is used to input the weights of the convolutional neural network generator , the decision variable , the Lagrange parameter after the last iteration into the iterative objective function to output the first denoised image .

[0010] Preferably, the function conversion sub-module includes: A matrix unit for calculating the Gaussian weight matrix of the enhanced image : ; In the formula, represents an adjustment factor, represents a Gaussian convolution function, , respectively represent the discrete gradient operators with respect to , , represents the enhanced image; An objective function construction unit for constructing an objective denoising function based on the enhanced image and the Gaussian weight matrix , where the objective denoising function is: ; In the formula, represents the function target image, represents the estimated image, represents the weights of the convolutional neural network generator, represents the updated weights, is a regularization factor, is a discrete gradient operator; A conversion unit for performing variable conversion on the objective denoising function to obtain a conversion function: ; In the formula, represents the number of pixel points in the enhanced image, is a decision variable.

[0011] Preferably, the weight iteration sub-module includes: A first parameter iteration unit for performing several iterations on in the iterative objective function until the iteration stop condition is satisfied, stopping the iteration and determining the weights of the convolutional neural network generator of the last iteration: ; ; ; In the formula, respectively represent the first and second attenuation factors, represents the discrete gradient operator with respect to , represents the gradient loss calculation, denotes a convex function, denotes a training sample, denotes the estimated value of the training sample, respectively denote the first moment and second moment after the respectively denote the first moment and second moment after the respectively denote the weights of the convolutional neural network generator after the is a protection constant, is the learning rate; The second parameter iteration unit is used to determine the decision variable of the last iteration : ; In the formula, denotes the Lagrange parameter after the iteration, denotes the estimated image after the iteration; The third parameter iteration unit is used to determine the Lagrange parameter of the last iteration : .

[0012] Preferably, the second denoising module includes: A reconstruction sub-module for reconstructing the first denoised image to obtain a reconstructed image, and determining the similarity and the first similarity weight based on the reconstructed image and the first similarity weight ; A similarity weight sub-module for calculating the second similarity weight based on the first similarity weight and the similarity : : ; In the formula, denotes the target pixel point in the image block and the target image block the similarity between them, denotes the Gaussian standard deviation; A denoising sub-module for calculating the target denoised image block based on the second similarity weight : ; In the formula, denotes the pixel value of the pixel point in the image block ; A combinatorial sub-module for combining a number of target denoised image patches to obtain a second denoised image.

[0013] Preferably, the reconstruction sub-module includes:[[]] A reconstruction unit for reconstructing the first denoised image using a preset reconstruction algorithm to obtain a reconstructed image, decomposing the reconstructed image into a number of image patches of the same size, and estimating the noise level of each image patch to obtain the image patch noise level; A similarity unit for calculating the similarity between any two image patches :[[]] ; ; ; In the formula, respectively represent the membership difference and the hesitation difference, represents the membership of the pixel at in the image patch belonging to the image patch , represents the membership of the pixel at in the image patch belonging to the image patch , represents the hesitation of the pixel at in the image patch belonging to the image patch , represents the hesitation of the pixel at in the image patch belonging to the image patch ; A similarity weight unit for selecting any pixel in the first denoised image as a target pixel, selecting an image patch as a target image patch at the target pixel, and calculating a first similarity weight based on the target image patch :[[]] ; ; ; In the formula, represents the image floating mean, represents the image patch representation, represents the image patch noise level of the target image patch , respectively represent the first, second, and third preset values, and , represents the target image patch in The pixel value at

[0014] Preferably, the segmentation module includes: A constraint sub-module for determining local features based on the second denoised image and constraint weights ; A segmentation function sub-module for constructing a segmentation objective function based on the constraint weights and the local features : : ; In the formula, is the element in the th row and th column of the membership matrix, is the adjustment exponent, respectively represent the local features of the th pixel point and the th pixel point corresponding sub-images, is the local feature of the corresponding sub-image, is the Gaussian standard deviation, is the regularization factor, is the th cluster center, is the th pixel point and the constraint weight between them, is the element in the th row and th column of the membership matrix; A clustering iteration sub-module for performing several iterations on the in the segmentation objective function until the iteration stop condition is satisfied, stopping the iteration and determining the element in the th row and th column of the membership matrix in the last iteration and the th cluster center : ; ; In the formula, respectively represent the elements in the th row and , th column of the local fuzzy matrix, represents the th th cluster center after the iteration; An image output sub-module, configured to determine the target area image according to the principle of maximum membership degree and Determine the target area image.

[0015] Preferably, the constraint sub-module includes: An image segmentation unit, configured to segment the second denoised image into a plurality of sub-images, select any one of the sub-images as a target sub-image, and calculate a first feature factor of the target sub-image And a second feature factor : ; ; ; ; In the formula, Is the Th Pixel point within the neighborhood window of the target sub-image, Is the central pixel point within the neighborhood window Of the target sub-image, Respectively represent a first sign function and a second sign function, Is a judgment threshold; A local feature unit, configured to calculate a local feature Based on the first feature factor And the second feature factor : ; In the formula, Represents a screening threshold; A constraint weight unit, configured to calculate a constraint weight : ; ; In the formula, Represents the average gray value of the neighborhood window , Represents The corresponding gray value, Represents Euclidean distance calculation.

[0016] In a second aspect, the present invention provides the following technical solution. A computer includes a memory, a processor, and a computer program stored on the memory and executable on the processor. The processor is equipped with the above-mentioned early bladder cancer detection system based on image recognition.

[0017] In a third aspect, the present invention provides the following technical solution: a storage medium on which a computer program is stored, and the storage medium is equipped with the early bladder cancer detection system based on image recognition as described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0019] Figure 1 It is a structural block diagram of the early bladder cancer detection system based on image recognition provided in Embodiment 1 of the present invention; Figure 2 It is a schematic diagram of the hardware structure of a computer provided in another embodiment of the present invention.

[0020] The following will further illustrate the embodiments of the present invention with reference to the drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] The following details the embodiments of the present invention. The examples of the embodiments are shown in the drawings, where the same or similar reference numerals represent the same or similar elements or elements with the same or similar functions throughout. The embodiments described below with reference to the drawings are exemplary and are intended to explain the embodiments of the present invention, and should not be construed as a limitation of the present invention.

[0022] In the description of the embodiments of the present invention, it should be understood that the orientation or positional relationship indicated by terms such as "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the embodiments of the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention.

[0023] In addition, the terms "first" and "second" are only used for descriptive purposes and should not be construed as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the embodiments of the present invention, "a plurality" means two or more, unless otherwise specifically defined.

[0024] In the embodiments of the present invention, unless otherwise clearly defined and limited, terms such as "installation", "connection", "connection", "fixation", etc. shall be understood in a broad sense. For example, it may be a fixed connection, a detachable connection, or integrated; it may be a mechanical connection or an electrical connection; it may be directly connected or indirectly connected through an intermediate medium, and it may be the communication inside two components or the interaction relationship between two components. For those of ordinary skill in the art, the specific meanings of the above terms in the embodiments of the present invention can be understood according to specific circumstances.

[0025] Embodiment 1 In Embodiment 1 of the present invention, as Figure 1 shown, a bladder cancer early detection system based on image recognition includes: Enhancement module 1, configured to obtain the abdominal medical image of the target object, and perform equalization enhancement processing on the abdominal medical image to obtain an enhanced image.

[0026] Among them, the enhancement module 1 includes: Equalization sub-module 11, configured to perform equalization processing on the abdominal medical image by using a preset equalization algorithm to obtain an equalized image; Among them, the preset equalization algorithm here is the CLAHE algorithm in the prior art, which will not be elaborated here again, and the abdominal medical image here is specifically an abdominal CT image.

[0027] Transformation sub-module 12, configured to perform discrete wavelet transform on the equalized image to obtain a first low-frequency component, a first horizontal component, a first vertical component, and a first diagonal component, and perform discrete wavelet transform on the abdominal medical image to obtain a second low-frequency component, a second horizontal component, a second vertical component, and a second diagonal component; Specifically, after performing discrete wavelet transform on the equalized image and the abdominal medical image respectively, four components will be obtained, namely the low-frequency component, the horizontal component, the vertical component, and the diagonal component, and these four components are all image components, and their manifestation forms are specifically components with the same size as the original image.

[0028] Decomposition sub-module 13, configured to perform singular value decomposition on the first low-frequency component to obtain a first eigenvector , a first singular value vector , a second eigenvector , perform singular value decomposition on the second low-frequency component to obtain a third eigenvector , a second singular value vector , a fourth eigenvector ; Specifically, the singular value decomposition here is a common method in the prior art, which can decompose the low-frequency vector into an m×m eigenvector, an m×n singular value vector, and an n×n eigenvector.

[0029] The difference sub-module 14 is used to calculate the image difference component based on the first eigenvector , the first singular value vector , the second eigenvector , the third eigenvector , the second singular value vector , the fourth eigenvector : : ; In the formula, represents an adjustable weight; Specifically, the value range of the adjustable weight here is 0.05 - 0.95, and in this application, it is specifically 0.65.

[0030] The first sub-component module 15 is used to calculate the first image component based on the image difference component : : .

[0031] The second sub-component module 16 is used to calculate the second image component, the third image component , and the fourth image component respectively based on a horizontal component, a first vertical component, a first diagonal component, a second horizontal component, a second vertical component, and a second diagonal component; ; Among them, the second sub-component module 16 includes: The second component unit 161 is used to calculate the second image component based on the first horizontal component , the second horizontal component : : ; ; ; In the formula, respectively represent the first horizontal frequency and the second horizontal frequency, represents the image size, respectively represent the gray values of the image corresponding to the first horizontal component at , respectively represent the gray values of the image corresponding to the second horizontal component at ; Specifically, the first horizontal frequency and the second horizontal frequency respectively correspond to the spatial frequencies of the first horizontal component and the second horizontal component.

[0032] A third component unit 162, configured to calculate a third image component based on the first vertical component , the second vertical component : : ; ; ; wherein, respectively represent the first vertical frequency and the second vertical frequency, respectively represent the gray values of the image corresponding to the first vertical component at ; respectively represent the gray values of the image corresponding to the second vertical component at ; Specifically, the first vertical frequency and the second vertical frequency respectively correspond to the spatial frequencies of the first vertical component and the second vertical component.

[0033] A fourth component unit 163, configured to calculate a fourth image component based on the first diagonal component , the second diagonal component : : ; ; ; wherein, respectively represent the first diagonal frequency and the second diagonal frequency; Specifically, the first diagonal frequency and the second diagonal frequency respectively correspond to the spatial frequencies of the first diagonal component and the second diagonal component.

[0034] A fusion sub-module 17, configured to fuse the first image component , the second image component , the third image component , the fourth image component to obtain a fused image, and perform an inverse discrete wavelet transform on the fused image to obtain an enhanced image.

[0035] A first denoising module 2, configured to construct a target denoising function based on the enhanced image, perform variable conversion on the target denoising function to obtain a conversion function, and perform iterative solution on the conversion function to obtain a first denoised image; wherein, the first denoising module 2 includes: The function conversion sub-module 21 is used to construct a target denoising function and perform variable conversion on the target denoising function to obtain a conversion function; Among them, the function conversion sub-module 21 includes: The matrix unit 211 is used to calculate the Gaussian weight matrix of the enhanced image : ; In the formula, represents the adjustment factor, represents the Gaussian convolution function, , respectively represent the discrete gradient operators with respect to , , represents the enhanced image; Specifically, the Gaussian weight matrix here can change its weight value by changing the direction of the discrete gradient operator, so that the subsequent denoising function can diffuse along the tangent direction of the local features of the image, thereby maintaining the local features of the image, and the adjustment factor here is greater than 0.

[0036] The target function construction unit 212 is used to construct a target denoising function based on the enhanced image and the Gaussian weight matrix , where the target denoising function is: ; In the formula, represents the function target image, represents the estimated image, represents the weights of the convolutional neural network generator, represents the updated weights, is the regularization factor, is the discrete gradient operator.

[0037] The conversion unit 213 is used to perform variable conversion on the target denoising function to obtain a conversion function: ; In the formula, represents the number of pixel points in the enhanced image, is the decision variable; Specifically, since the target denoising function also includes non-smooth terms, new decision variables are introduced to assist in the iterative solution process.

[0038] The iterative function sub-module 22 is used to determine the iterative objective function based on the conversion function: ; In the formula, is the penalty parameter, is the Lagrange parameter, is the inner product calculation, is the Gaussian weight matrix, represents the estimated image, is the regularization factor, is the discrete gradient operator, represents the number of pixel points in the enhanced image, is the decision variable; Specifically, since the above conversion function is an optimization problem, the iterative objective function can be changed to minimize the parameters and maximize , where the penalty parameter is greater than 0.

[0039] The weight iteration sub-module 23 is used to perform several iterations on the iterative objective function in and determine the convolutional neural network generator weights , decision variable and Lagrange parameter ; Among them, the weight iteration sub-module 23 includes: The first parameter iteration unit 231 is used to perform several iterations on the iterative objective function in until the iteration stop condition is satisfied, stop the iteration and determine the convolutional neural network generator weights of the last iteration: ; ; ; In the formula, respectively represent the first and second decay factors, represents the discrete gradient operator with respect to , represents the gradient loss calculation, represents a convex function, represents the training sample, represents the estimated value of the training sample, respectively represent the first moment and second moment after the th iteration, respectively represent the first moment and second moment after the th iteration, respectively represent the convolutional neural network generator weights after the th iteration, is the protection constant, is the learning rate; Among them, the iteration stop condition is that the iteration satisfies a certain number of times or the difference between two adjacent iterations is less than a preset difference. The first and second attenuation factors are respectively within the range of 0 to 1, and they are 0.9 and 0.999 respectively, and the learning rate is 0.01.

[0040] The second parameter iteration unit 232 is used to determine the decision variable of the last iteration : ; In the formula, represents the Lagrangian parameter after the th iteration, represents the estimated image after the th iteration.

[0041] The third parameter iteration unit 233 is used to determine the Lagrangian parameter of the last iteration : .

[0042] The output solution sub-module 24 is used to input the weights of the convolutional neural network generator, the decision variable, and the Lagrangian parameter after the last iteration into the iterative objective function to output the first denoised image , the decision variable , the Lagrangian parameter and input them into the iterative objective function to output the first denoised image .

[0043] The second denoising module 3 is used to perform image reconstruction denoising on the first denoised image to obtain a second denoised image; Among them, the second denoising module 3 includes: A reconstruction sub-module is used to reconstruct the first denoised image to obtain a reconstructed image, and determine the similarity and the first similarity weight based on the reconstructed image and the first similarity weight ; Among them, the reconstruction sub-module includes: A reconstruction unit is used to reconstruct the first denoised image by using a preset reconstruction algorithm to obtain a reconstructed image, decompose the reconstructed image into a plurality of image blocks of the same size, and estimate the noise level of each image block to obtain the image block noise level; Specifically, the preset reconstruction algorithm here is the FBP algorithm in the prior art, so it will not be elaborated here.

[0044] A similarity unit is used to calculate the similarity between any two image blocks : ; ; ; In the formula, respectively represent the membership difference and the hesitation difference, represents the membership of the pixel point at in the image block to the image block in the pixel point at to the image block in the pixel point at to the image block in the pixel point at to the

[0045] The similarity weight unit is used to select any pixel point in the first denoised image as the target pixel point, and select an image block as the target image block at the target pixel point, and calculate the first similarity weight based on the target image block : ; ; ; In the formula, represents the floating mean of the image, represents the image block representation, represents the image block noise level of the target image block respectively represent the first, second, and third preset values, and , represents the pixel value at in the target image block Specifically, in this application, the first, second, and third preset values are 2.6, 2.2, and 1.8 respectively.

[0046] The similarity weight sub-module is used to calculate the second similarity weight based on the first similarity weight and the similarity : ; In the formula, represents the similarity between the image block where the target pixel point is located and the target image block Represents the Gaussian standard deviation.

[0047] A denoising sub-module for calculating a target denoised image block based on the second similarity weight : ; In the formula, represents the pixel value of the pixel point in the image block ;

[0048] A combining sub-module for combining a plurality of target denoised image blocks to obtain a second denoised image.

[0049] A segmentation module 4 for performing image clustering segmentation on the second denoised image to obtain a target region image; wherein, the segmentation module includes: A constraint sub-module for determining local features and constraint weights ; wherein, the constraint sub-module includes: An image segmentation unit for segmenting the second denoised image into a plurality of sub-images, selecting any one sub-image as a target sub-image, and calculating the first feature factor of the target sub-image and the second feature factor : ; ; ; ; In the formula, is the th pixel point in the neighborhood window of the target sub-image, is the central pixel point in the neighborhood window of the target sub-image, respectively represent the first sign function and the second sign function, is the judgment threshold.

[0050] A local feature unit for calculating local features based on the first feature factor and the second feature factor : ; In the formula, represents the screening threshold; Specifically, the specific meaning of is to sum the pixel points that satisfy

[0051] Constraint weight unit for calculating constraint weights : ; ; In the formula, represents the average grayscale value of the neighborhood window , represents the corresponding grayscale value, represents Euclidean distance calculation.

[0052] Segmentation function sub-module for constructing a segmentation objective function based on the constraint weight and the local feature : : ; In the formula, is the element in the th row and th column of the membership matrix, is the adjustment exponent, respectively represent the local features of the sub-images corresponding to the th pixel point and the th pixel point, is the local feature of the corresponding sub-image, is the Gaussian standard deviation, is the regularization factor, is the th cluster center, is the th pixel point and the constraint weight between them, is the element in the th row and th column of the membership matrix.

[0053] Cluster iteration sub-module for performing several iterations on the in the segmentation objective function until the iteration stop condition is met, stopping the iteration and determining the element in the th row and th column of the membership matrix of the last iteration and the th cluster center : : ; ; In the formula, respectively represent the element at the -th row and the , -th column in the local fuzzy matrix; represents the -th clustering center after the -th iteration; wherein, the iteration stopping condition is to meet a certain number of iterations or the segmentation objective function gradually converges, and , , wherein, represents the -th clustering center after the -th iteration, represents the element at the -th row, the -th column, the -th column and the -th column in the membership degree matrix after the

[0054] image output sub-module, which is used to determine the target area image according to the maximum membership degree principle and ; Specifically, after determining the element in the final membership degree matrix, the image can be segmented into regions according to the maximum membership degree principle, and the image of the bladder region can be selected to obtain the target area image.

[0055] Detection module 5, which is used to obtain preset training data, input the preset training data into a preset detection model for training, and input the target area image into the trained preset detection model for detection to output a detection result; Specifically, the preset detection model therein is the KNN model in the prior art. The model is trained with the preset training data, and then the target area image is input into the trained model to obtain the corresponding detection result.

[0056] The early bladder cancer detection system based on image recognition provided in the first embodiment of the present invention includes an enhancement module for obtaining the abdominal medical image of a target object and performing equalization enhancement processing on the abdominal medical image to obtain an enhanced image; a first denoising module for constructing a target denoising function based on the enhanced image, performing variable conversion on the target denoising function to obtain a conversion function, and performing iterative solution on the conversion function to obtain a first denoised image; a second denoising module for performing image reconstruction denoising on the first denoised image to obtain a second denoised image; a segmentation module for performing image clustering segmentation on the second denoised image to obtain a target region image; and a detection module for obtaining preset training data, inputting the preset training data into a preset detection model for training, and inputting the target region image into the trained preset detection model for detection to output a detection result. First, the present invention enhances the image through the enhancement module, which can effectively improve the image contrast and fully retain and highlight the detailed information in the image. Then, the first denoising module and the second denoising module perform two denoising processes on the image, which can effectively reduce the noise and artifacts in the image and effectively capture the edge texture information of the image, improving the visual quality and image structure and avoiding image distortion. Then, the segmentation module segments the image, which can improve the segmentation quality and reduce the sensitivity to noise, thereby improving the accuracy and precision of subsequent model recognition and detection.

[0057] In some other embodiments of the present invention, the present invention provides the following technical solution. A computer includes a memory 102, a processor 101, and a computer program stored on the memory 102 and executable on the processor 101. The processor 101 is equipped with the early bladder cancer detection system based on image recognition as described above.

[0058] Specifically, the above-mentioned processor 101 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured as one or more integrated circuits for implementing the embodiments of the present invention.

[0059] Among them, the memory 102 may include a mass storage for data or instructions. By way of example and not limitation, the memory 102 may include a hard disk drive (HDD), a floppy disk drive, a solid state drive (SSD), a flash memory, an optical disc, a magneto-optical disc, a magnetic tape, or a universal serial bus (USB) drive, or a combination of two or more of these. In appropriate cases, the memory 102 may include removable or non-removable (or fixed) media. In appropriate cases, the memory 102 may be internal or external to the data processing device. In a particular embodiment, the memory 102 is non-volatile memory. In a particular embodiment, the memory 102 includes a read-only memory (ROM) and a random access memory (RAM). In appropriate cases, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically alterable ROM (EAROM), or a flash memory, or a combination of two or more of these. In appropriate cases, the RAM may be a static random access memory (SRAM) or a dynamic random access memory (DRAM), where the DRAM may be a fast page mode dynamic random access memory (FPMDRAM), an extended data out dynamic random access memory (EDODRAM), a synchronous dynamic random access memory (SDRAM), etc.

[0060] The memory 102 can be used to store or cache various data files required for processing and / or communication, as well as possible computer program instructions executed by the processor 101.

[0061] The processor 101 reads and executes the computer program instructions stored in the memory 102 to execute the early bladder cancer detection system based on image recognition.

[0062] In some of these embodiments, the computer may further include a communication interface 103 and a bus 100. Among them, as Figure 2 shown, the processor 101, the memory 102, and the communication interface 103 are connected through the bus 100 to complete communication with each other.

[0063] The communication interface 103 is used to implement communication between the various modules, devices, units, and / or devices in the embodiments of the present invention. The communication interface 103 can also implement data communication with other components, such as external devices, image / data acquisition devices, databases, external storage, and image / data processing workstations, etc.

[0064] The bus 100 includes hardware, software, or both, and couples components of a computer device to each other. The bus 100 includes, but is not limited to, at least one of the following: Data Bus, Address Bus, Control Bus, Expansion Bus, Local Bus. By way of example and not limitation, the bus 100 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses or a combination of two or more of these. In suitable cases, the bus 100 may include one or more buses. Although embodiments of the present invention describe and illustrate specific buses, the present invention contemplates any suitable bus or interconnect.

[0065] The computer can execute the early bladder cancer detection system based on image recognition of the present invention, thereby realizing the early bladder cancer detection based on image recognition.

[0066] In still some other embodiments of the present invention, in combination with the above-mentioned early bladder cancer detection system based on image recognition, embodiments of the present invention provide the following technical solution: a storage medium, on which a computer program is stored, and the above-mentioned early bladder cancer detection system based on image recognition is carried on the storage medium.

[0067] Those skilled in the art will understand that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a definite sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus or device (such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus or device), or used in combination with these instruction execution systems, apparatus or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate or transport a program for use by or in connection with an instruction execution system, apparatus or device.

[0068] More specific examples (non-exhaustive list) of the readable medium include the following: an electrical connection portion with one or more wirings (electronic device), a portable computer disk cartridge (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpretation or other suitable processing as necessary, and then stored in a computer memory.

[0069] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above 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, any one or a combination of the following techniques well-known in the art can be used: discrete logic circuits with logic gate circuits for implementing logical functions on data signals, application-specific integrated circuits with suitable combinational logic gate circuits, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0070] The technical features of the above-described embodiments can be combined arbitrarily. For the sake of brevity of description, 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, it should be considered as within the scope described in this specification.

[0071] The above-described embodiments merely represent several implementation manners of the present invention. The description thereof is relatively specific and detailed, but it should not be construed as a limitation to the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all fall within the protection scope of the present invention. Therefore, the protection scope of the present invention patent shall be subject to the appended claims.

Claims

1. An early detection system for bladder cancer based on image recognition, characterized in that, The system includes: An enhancement module, configured to obtain an abdominal medical image of a target object, and perform equalization enhancement processing on the abdominal medical image to obtain an enhanced image; A first denoising module, configured to construct a target denoising function based on the enhanced image, perform variable transformation on the target denoising function to obtain a transformed function, and perform iterative solution on the transformed function to obtain a first denoised image; A second denoising module, configured to perform image reconstruction denoising on the first denoised image to obtain a second denoised image; A segmentation module, configured to perform image clustering segmentation on the second denoised image to obtain a target region image; A detection module, configured to obtain preset training data, input the preset training data into a preset detection model for training, and input the target region image into the trained preset detection model for detection to output a detection result.

2. The early bladder cancer detection system based on image recognition according to claim 1, wherein The enhancement module includes: An equalization sub-module, configured to perform equalization processing on the abdominal medical image by using a preset equalization algorithm to obtain an equalized image; A transformation sub-module, configured to perform discrete wavelet transformation on the equalized image to obtain a first low-frequency component, a first horizontal component, a first vertical component, and a first diagonal component, and perform discrete wavelet transformation on the abdominal medical image to obtain a second low-frequency component, a second horizontal component, a second vertical component, and a second diagonal component; A decomposition sub-module, configured to perform singular value decomposition on the first low-frequency component to obtain a first eigenvector , a first singular value vector , a second eigenvector , and perform singular value decomposition on the second low-frequency component to obtain a third eigenvector , a second singular value vector , and a fourth eigenvector ; A difference sub-module, for calculating an image difference component based on a first eigenvector , a first singular value vector , a second eigenvector , a third eigenvector , a second singular value vector , a fourth eigenvector : ​ ； In the formula, represents an adjustable weight; The first sub-component module is used to calculate a first image component based on the image difference component :​ ； A second sub-component module, configured to calculate a second image component respectively based on a horizontal component, a first vertical component, a first diagonal component, a second horizontal component, a second vertical component, and a second diagonal component , a third image component , and a fourth image component ; A fusion sub-module for fusing the first image component , the second image component , the third image component , and the fourth image component to obtain a fused image, and performing inverse discrete wavelet transform on the fused image to obtain an enhanced image.

3. The early bladder cancer detection system based on image recognition according to claim 2, characterized in that, The second component sub-module includes: A second component unit, configured to calculate a second image component based on a first horizontal component , a second horizontal component : ​ ； ； ； wherein, respectively represent the first horizontal frequency and the second horizontal frequency, represents the image size, respectively represent the gray values of the image corresponding to the first horizontal component at ; respectively represent the gray values of the image corresponding to the second horizontal component at ; A third component unit, configured to calculate a third image component based on a first vertical component and a second vertical component : ； ； ； In the formula, respectively represent the first vertical frequency and the second vertical frequency, respectively represent the gray values of the image corresponding to the first vertical component at , respectively represent the gray values of the image corresponding to the second vertical component at ; The fourth component unit is configured to calculate a fourth image component based on a first diagonal component , a second diagonal component : ​ ； ； ； In the formula, respectively represent the first diagonal frequency and the second diagonal frequency.

4. The early bladder cancer detection system based on image recognition according to claim 1, wherein The first denoising module includes: A function transformation sub-module, configured to construct a target denoising function and perform variable transformation on the target denoising function to obtain a transformed function; Iterative function sub-module, configured to determine an iterative objective function based on the conversion function : ； wherein, is the penalty parameter, is the Lagrange parameter, is the inner product calculation, is the Gaussian weight matrix, represents the estimated image, is the regularization factor, is the discrete gradient operator, represents the number of pixel points in the enhanced image, is the decision variable; A weight iteration sub-module for the iterative objective function in perform a number of iterations and determine the weights of the convolutional neural network generator, the decision variables , and the Lagrange parameters ;​ An output solving sub-module, which is used to input the weights of the convolutional neural network generator after the last iteration , decision variables , Lagrange parameters into the iterative objective function to output a first denoised image .

5. The early bladder cancer detection system based on image recognition according to claim 4, characterized in that, The function transformation sub-module includes: A matrix unit for calculating a Gaussian weight matrix of the enhanced image : ； In the formula, represents an adjustment factor, represents a Gaussian convolution function, , respectively represent the , discrete gradient operators, represents the enhanced image; The objective function construction unit is used to construct an objective denoising function based on the enhanced image and the Gaussian weight matrix where the objective denoising function is as follows: ； In the formula, represents the function target image, represents the estimated image, represents the convolutional neural network generator weights, represents the updated weights, is the regularization factor, is the discrete gradient operator; A transformation unit, configured to perform variable transformation on the target denoising function to obtain a transformed function: ； In the formula, represents the number of pixel points in the enhanced image, is a decision variable.

6. The early bladder cancer detection system based on image recognition according to claim 4, characterized in that The weight iteration sub-module includes: A first parameter iteration unit for iterating the iterative objective function in for a number of iterations until an iteration stop condition is satisfied, at which point the iteration is stopped and the convolutional neural network generator weights of the last iteration are determined : ； ； ； Wherein, respectively represent the first and second attenuation factors, represents the discrete gradient operator with respect to ; represents the gradient loss calculation, represents a convex function, represents the training sample, represents the estimated value of the training sample, respectively represent the first moment and the second moment after the -th iteration, respectively represent the first moment and the second moment after the -th iteration, respectively represent the weights of the convolutional neural network generator after the -th iteration, is a protection constant, is the learning rate; The second parameter iteration unit is used to determine the decision variable of the last iteration : ； In the formula, represents the Lagrange parameter after the -th iteration, represents the estimated image after the -th iteration; The third parameter iteration unit is used to determine the Lagrange parameter of the last iteration : 。 7. The early bladder cancer detection system based on image recognition according to claim 1, characterized in that, The second denoising module includes: A reconstruction sub-module, configured to reconstruct the first denoised image to obtain a reconstructed image, and determine a similarity based on the reconstructed image and a first similarity weight ; A similarity weight sub-module, for calculating a second similarity weight based on the first similarity weight and the similarity :​ ； In the formula, represents the target pixel point in the image block and the similarity between the target image block is represents the Gaussian standard deviation; A denoising sub-module for calculating a target denoised image block based on the second similarity weight :​ ； In the formula, represents the pixel value of a pixel point in the image block; Combination sub-module, used to combine a number of target denoised image patches to obtain a second denoised image.

8. The early bladder cancer detection system based on image recognition according to claim 7, characterized in that, The reconstruction sub-module includes: A reconstruction unit, configured to reconstruct the first denoised image by using a preset reconstruction algorithm to obtain a reconstructed image, decompose the reconstructed image into a plurality of image blocks of the same size, and estimate the noise level of each image block to obtain the image block noise level; A similarity unit for calculating the similarity between any two image patches : ； ； ； In the formula, respectively represent the membership degree difference and the hesitation degree difference, represents the image block in the membership degree of the pixel point at belonging to the image block represents the image block in the membership degree of the pixel point at belonging to the image block represents the image block in the hesitation degree of the pixel point at belonging to the image block represents the image block in the hesitation degree of the pixel point at belonging to the image block; A similarity weight unit is configured to select any pixel point in the first denoised image as a target pixel point, select an image patch at the target pixel point as a target image patch, and calculate a first similarity weight based on the target image patch : ； ； ； In the formula, represents the floating mean of the image, represents the image block representation, represents the target image block of the image block noise level, respectively represent the first, second, and third preset values, and , represents the target image block in the pixel value at the location.

9. The early bladder cancer detection system based on image recognition according to claim 1, characterized in that, The segmentation module includes: A constraint sub-module, configured to determine local features based on the second denoised image and constraint weights ; The segmentation function sub-module is used to build a segmentation objective function based on the constraint weights and the local features : ​ ； Wherein, is the element at the -th row and -th column in the membership degree matrix, is the adjustment index, respectively represent the local features of the sub-images corresponding to the -th pixel point and the -th pixel point, is the local feature of the corresponding sub-image, is the Gaussian standard deviation, is the regularization factor, is the -th clustering center, is the -th pixel point and the constraint weight therebetween, is the element at the -th row and -th column in the membership degree matrix; The clustering iteration sub-module is used to perform several iterations on the segmentation objective function in until the iteration stop condition is met, stop the iteration and determine the element at the th row and the th column in the membership matrix of the last iteration and the th cluster center : ； ； Wherein, respectively represent the row and the , column elements in the local fuzzy matrix, represents the th cluster center after the An image output sub-module, which is used to determine the target region image according to the principle of maximum membership degree and determine the target region image.

10. The early bladder cancer detection system based on image recognition according to claim 9, characterized in that, The constraint sub-module includes: An image segmentation unit, configured to segment the second denoised image into a plurality of sub-images, select any one of the sub-images as a target sub-image, and calculate a first feature factor of the target sub-image and a second feature factor : ； ； ； ； In the formula, is the th pixel point within the neighborhood window of the target sub-image, and is the central pixel point within the neighborhood window of the target sub-image, where respectively represent the first sign function and the second sign function, is the judgment threshold; Local feature unit, for calculating local features based on a first feature factor and a second feature factor :​ ； In the formula, represents the screening threshold; Constraint weight unit, for calculating constraint weights : ； ； In the formula, represents the average gray value of the neighborhood window , represents the corresponding gray value, represents the Euclidean distance calculation.

Images