Lung cancer image recognition method and system

Abstract

The invention provides a lung cancer image recognition method and system, and the method comprises the steps: obtaining a lung image of a patient, and carrying out the image correction processing of the lung image, so as to obtain a corrected lung image; performing iterative denoising processing on the corrected lung image to obtain a denoised lung image; performing feature image segmentation on the denoised lung image to obtain a lung feature image; the method comprises the following steps: acquiring a historical template lung cancer image set, inputting the historical template lung cancer image set into a preset recognition model for training, inputting a lung feature image into the trained preset recognition model for recognition, and outputting a recognition result. Noise and artifacts existing in the image are effectively removed, interference information existing in the image can be eliminated at the same time, and then the recognition accuracy is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of image recognition, and in particular relates to a lung cancer image recognition method and system. Background Art

[0002] For lung cancer patients, they generally scan the patient's lung CT image and then determine whether the patient has lung cancer based on the specific CT image, and the above process is usually completed manually. There are also some methods in the prior art that input the patient's lung image into a model to identify whether the patient has lung cancer, but most of them simply pre-process the image, which results in low image quality in the input model and cannot effectively remove noise, artifacts, etc. in the image, thereby affecting the accuracy of model recognition. Summary of the invention

[0003] In order to solve the above technical problems, the present invention provides a lung cancer image recognition method and system for solving the technical problems in the prior art.

[0004] On the one hand, the present invention provides the following technical solution, a lung cancer image recognition method, comprising: Acquiring a lung image of a patient, and performing image correction processing on the lung image to obtain a corrected lung image; Performing iterative denoising processing on the corrected lung image to obtain a denoised lung image; Performing feature image segmentation on the denoised lung image to obtain a lung feature image; A historical template lung cancer image set is obtained, the historical template lung cancer image set is input into a preset recognition model for training, and the lung feature image is input into the trained preset recognition model for recognition to output a recognition result.

[0005] Compared with the prior art, the beneficial effects of the present invention are as follows: the present application first obtains a lung image of the patient, performs image correction processing on the lung image to obtain a corrected lung image; then performs iterative denoising processing on the corrected lung image to obtain a denoised lung image; then performs feature image segmentation on the denoised lung image to obtain a lung feature image; finally, obtains a historical template lung cancer image set, inputs the historical template lung cancer image set into a preset recognition model for training, and inputs the lung feature image into the trained preset recognition model for recognition to output a recognition result. The present invention first performs correction processing on the image to avoid the situation where the image appears darker and the features are not obvious, and then performs iterative denoising processing on the image. Combined with the correction process, the noise and artifacts in the image can be effectively removed. At the same time, the image is feature segmented to eliminate the interference information in the image, thereby improving the recognition accuracy.

[0006] Preferably, the step of performing image correction processing on the lung image to obtain a corrected lung image includes: The lung image is divided into equal regions to obtain a plurality of sub-images, a linear transformation is performed on the plurality of sub-images to determine a transformation equation, and a first transformation coefficient in the transformation equation is extracted. and the second transform coefficient , based on the first transform coefficient With the second transform coefficient Determine the cost function : ; In the formula, , Represents the output and input Zhang sub-image, represents the number of sub-images, is the first regularization parameter; For the cost function Perform minimization and output the optimal first transform coefficient and the optimal second transform coefficient : ; ; In the formula, , Respectively , The mean value in the corresponding window, express The number of pixels in the neighborhood window of express Variance within the corresponding window; Based on the optimal first transform coefficient and the optimal second transform coefficient Image separation and correction are performed to obtain a corrected lung image.

[0007] Preferably, the optimal first transform coefficient and the optimal second transform coefficient The steps of performing image separation and correction to obtain a corrected lung image include: Based on the optimal first transform coefficient and the optimal second transform coefficient Determine the brightness feature image Image with background reflection : ; ; For the brightness feature image Correction processing is performed to obtain a brightness-corrected image : ; ; In the formula, is the grayscale value level of the brightness feature image, is the grayscale mean of the brightness feature image, Respectively represent the length and width of the brightness feature image; The brightness correction image With the background reflection image The fusion is performed to obtain the corrected lung image.

[0008] Preferably, the step of performing iterative denoising on the corrected lung image to obtain a denoised lung image comprises: The lung image is used as an initial iterative image, and the initial iterative image is iteratively regenerated to obtain an iteratively generated image: ; ; In the formula, , Respectively represent sequence The iterative generation image after iterations is is the relaxation factor, Indicates The set of all projection ray indices corresponding to the projection angles, Indicates The projection coefficient matrix corresponding to the projection rays is: express The corresponding projection ray vector, is the transpose symbol; Based on sequence The iterative generation image after iterations calculates the first iteration difference : ; Determine the first iterative difference Is it less than the first iteration threshold? If the first iteration difference is less than the first iteration threshold, then The iterative image generated after the iteration is used as the final iterative image. If the first iterative difference If the value is not less than the first iteration threshold, the iteration process is repeated until the final iteration image is output; Iterative denoising is performed on the final iterative image to obtain a denoised lung image.

[0009] Preferably, the step of iteratively denoising the final iterative image to obtain a denoised lung image comprises: The final iteration image Convert to image matrix , the image matrix As the initial value and iteratively denoising the initial value to obtain an iterative denoising value: ; ; ; In the formula, , Respectively represent sequence Iterative denoising value after iterations, is the second regularization parameter, Indicates sequence The first variable after iterations, For the first step, represents the initial first variable, , Indicates sequence The second variable after iterations, represents the second step length, , Indicates sequence The image pixel value after iterations, represents the adjustment factor; Based on sequence The iterative denoising value after the iteration calculates the second iteration difference : ; Determine the two-iteration difference Is it less than the second iteration threshold? If the two iteration difference is less than the second iteration threshold, then The iterative denoising value after the iteration is used as the final image matrix and the final image matrix is ​​restored to an image to obtain a denoised lung image. If the two-iteration difference If it is not less than the second iteration threshold, the iteration process is repeated until the final image matrix is ​​output.

[0010] Preferably, the step of performing feature image segmentation on the denoised lung image to obtain a lung feature image comprises: Performing convolution filtering on the denoised lung image to obtain a filtered image; performing region selection, first diffusion processing and contour fitting on the filtered image in sequence to obtain an updated feature image; The updated feature image is subjected to a second diffusion process to obtain a lung feature image.

[0011] Preferably, the step of performing convolution filtering on the denoised lung image to obtain a filtered image comprises: The denoised lung image Perform Gaussian convolution to obtain the scale matrix :

[0012] ; ; ; ; ; In the formula, , , , are the first, second, third, and fourth scale parameters, respectively. represents the Gaussian function, Respectively represent the horizontal and vertical coordinates of the pixels in the denoised lung image, represents the Gaussian parameter, Represents the convolution operation; Perform eigendecomposition on the scale matrix to obtain the first eigenvalue With the second eigenvalue , based on the first eigenvalue With the second eigenvalue Determine the tuning parameters : ; In the formula, is the sensitivity coefficient, is the smoothing coefficient; The adjustment parameters The Gaussian parameters in the Gaussian filter are replaced and the denoised lung image is filtered to obtain a filtered image.

[0013] Preferably, the step of sequentially performing region selection, diffusion processing and contour fitting on the filtered image to obtain an updated feature image comprises: Determine a histogram of the filtered image, extract connected areas in the histogram whose grayscale values ​​are greater than a preset grayscale value, calculate the circularity of each of the connected areas, and use the image area corresponding to the connected area with the largest circularity as the first feature area; Identify the maximum circumscribed rectangular area of ​​the first feature area, expand the edge of the maximum circumscribed rectangular area outward by a preset number of pixels to obtain a second feature area, map the second feature area to the filtered image to obtain a basic feature image, and binarize the basic feature image to obtain a binary foreground image; Determine a rectangular window with the center position of the binary foreground image as the midpoint, select two connected areas with the largest area in the rectangular window as reference areas, and determine whether the area difference between the two reference areas is less than a first preset value and whether the centroid coordinates of the two reference areas are less than a second preset value; If the area difference between the two reference areas is not less than a first preset value and / or the centroid coordinates of the two reference areas are not less than a second preset value, the rectangular window is enlarged and the reference area is reselected; if the area difference between the two reference areas is less than the first preset value and the centroid coordinates of the two reference areas are less than the second preset value, any number of pixel points are selected in the two reference areas as reference points respectively, and diffused outward with the reference points as the center until they diffuse to the edge position of the foreground in the binary foreground image, so as to obtain a diffused image; Identifying edge points of the diffuse image and storing the edge points in a diffuse edge point set, and removing pixel points in the diffuse edge point set that are not in the second feature area to obtain an edge pixel point set; Identify the number of valid pixels of each pixel in the edge pixel point set within an 8-neighborhood range, and take the pixel point whose number of valid pixels in the 8-neighborhood range is 1 as an endpoint; Taking the endpoint as the starting point, respectively calculating the Euclidean distance between the endpoint and each pixel point in the edge pixel point set, taking the pixel point with the smallest Euclidean distance as the new starting point, and iteratively selecting the starting point and calculating the Euclidean distance to obtain a number of starting points and store the starting points in the starting point set; Minimum multiplication curve fitting is performed on the pixel points in the starting point set to obtain a fitting point set, contour updating and internal filling are performed on the diffusion image based on the fitting point set to obtain a contour update image, and the contour update image is mapped to the basic feature image to obtain an updated feature image.

[0014] Preferably, the step of performing a second diffusion process on the updated feature image to obtain a lung feature image comprises: Extracting the highlight area in the updated feature image and performing morphological processing on the highlight area to obtain the original area image, extracting the boundary points of the original area image, taking the boundary points whose grayscale values ​​are greater than the boundary grayscale threshold as diffusion points, storing the diffusion points in a diffusion point set, and storing all the pixel points in the updated feature image in a feature point set; Diffusion is performed outwards with the diffusion point in the diffusion point set as the center, and the pixel points obtained by diffusion are stored in the diffused pixel point set, and the average grayscale of each pixel point in the diffused pixel point set is calculated; Randomly select a pixel point from the feature point set as a diffusion cutoff point, scan the eight neighboring pixel points of the diffusion cutoff point, and determine whether the absolute value of the difference between the eight neighboring pixel points and the average grayscale of each pixel point in the diffused pixel point set is less than a diffusion threshold; If the absolute value of the difference between the average grayscale of each pixel in the 8-neighborhood pixel point and the diffused pixel point set is less than the diffusion threshold, the diffusion cutoff point is stored in the diffused pixel point set. If the absolute value of the difference between the average grayscale of each pixel in the 8-neighborhood pixel point and the diffused pixel point set is not less than the diffusion threshold, the diffusion cutoff point is used as a background point and removed from the feature point set. The diffusion and judgment process is iteratively repeated until the feature point set is an empty set, and the lung feature image is output.

[0015] In a second aspect, the present invention provides the following technical solution, a lung cancer image recognition system, the system comprising: A correction module, used for acquiring a lung image of a patient and performing image correction processing on the lung image to obtain a corrected lung image; A denoising module, used for iteratively denoising the corrected lung image to obtain a denoised lung image; A segmentation module, used for performing feature image segmentation on the denoised lung image to obtain a lung feature image; The recognition module is used to obtain a historical template lung cancer image set, input the historical template lung cancer image set into a preset recognition model for training, input the lung feature image into the trained preset recognition model for recognition, and output a recognition result. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 A flowchart of a lung cancer image recognition method provided in Embodiment 1 of the present invention; Figure 2 A structural block diagram of a lung cancer image recognition system provided in Embodiment 2 of the present invention; Figure 3 A schematic diagram of the hardware structure of a computer provided in another embodiment of the present invention.

[0018] The embodiments of the present invention will be further described below with reference to the accompanying drawings. DETAILED DESCRIPTION

[0019] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the embodiments of the present invention, and should not be construed as limiting the present invention.

[0020] Embodiment 1 In the first embodiment of the present invention, Figure 1 As shown, a lung cancer image recognition method comprises: S1. Acquire a lung image of a patient, and perform image correction processing on the lung image to obtain a corrected lung image; Specifically, the lung image here is a CT image of the patient's lungs, and CT scanning uses X-rays to interact with the scanned object to generate an image, so the CT image can express the patient's lung characteristics more clearly.

[0021] Wherein, the step S1 comprises: S11, dividing the lung image into equal regions to obtain a plurality of sub-images, performing linear transformation on the plurality of sub-images to determine a transformation equation, and extracting a first transformation coefficient in the transformation equation and the second transform coefficient , based on the first transform coefficient With the second transform coefficient Determine the cost function : ; In the formula, , Represents the output and input Zhang sub-image, represents the number of sub-images, is the first regularization parameter; Specifically, the equal area division here is to divide the lung image into several small rectangular areas according to the size of the lung image, and then use the image corresponding to each small rectangular area as a sub-image. As for the transformation equation, in this step, the first transformation coefficient and the second transform coefficient The output image and the output image can be approximately expressed in the form of a linear equation.

[0022] S12, the cost function Perform minimization and output the optimal first transform coefficient and the optimal second transform coefficient : ; ; In the formula, , Respectively , The mean value in the corresponding window, express The number of pixels in the neighborhood window of express Variance within the corresponding window; Specifically, the cost function is an estimation function of the optimal solution, so the first transformation coefficients can be obtained by minimizing the cost function and performing linear regression analysis. and the second transform coefficient The optimal solution of and the optimal second transform coefficient .

[0023] S13, based on the optimal first transform coefficient and the optimal second transform coefficient Performing image separation and correction to obtain a corrected lung image; Wherein, the step S13 comprises: S131, based on the optimal first transform coefficient and the optimal second transform coefficient Determine the brightness feature image Image with background reflection : ; ; Specifically, since the optimal first transform coefficient obtained above and the optimal second transform coefficient is the transformation coefficient of each sub-image. At the same time, since the pixel point at the same position may be included in multiple repeated windows, the transformation coefficients of different windows are different. Therefore, it is necessary to take the average value as the final transformation coefficient. Finally, the corresponding transformation coefficient and the corresponding sub-image can be used to obtain the corrected sub-image. After that, the corrected image is fused to obtain the complete brightness feature image. Since the brightness feature image is separated from the lung image, the background reflection image can be obtained in the same way. .

[0024] S132: the brightness feature image Correction processing is performed to obtain a brightness-corrected image : ; ; In the formula, is the grayscale value level of the brightness feature image, is the grayscale mean of the brightness feature image, Respectively represent the length and width of the brightness feature image; Specifically, by correcting the brightness feature image according to the above formula, the detail information of the image can be fully retained, and at the same time, the effect of suppressing artifacts can also be achieved.

[0025] S133: Correct the brightness of the image With the background reflection image The fusion is performed to obtain the corrected lung image.

[0026] S2, performing iterative denoising processing on the corrected lung image to obtain a denoised lung image; Wherein, the step S2 comprises: S21, taking the lung image as an initial iterative image, and iteratively regenerating the initial iterative image to obtain an iteratively generated image: ; ; In the formula, , Respectively represent sequence The iterative generation image after iterations is is the relaxation factor, Indicates The set of all projection ray indices corresponding to the projection angles, Indicates The projection coefficient matrix corresponding to the projection rays is: express The corresponding projection ray vector, is the transpose symbol; Specifically, in the above iterative process, each iteration only uses the projection data of all rays corresponding to the same projection angle to update the pixel value, which can effectively reduce image noise and has a fast iterative convergence speed.

[0027] S22, based on sequence The iterative generation image after iterations calculates the first iteration difference : ; S23, determining the first iterative difference Is it less than the first iteration threshold? If the first iteration difference is less than the first iteration threshold, then The iterative image generated after the iteration is used as the final iterative image. If the first iterative difference If the value is not less than the first iteration threshold, the iteration process is repeated until the final iteration image is output; Specifically, after a certain number of iterations, the first iteration difference is calculated to determine whether it is necessary to continue the iteration process. If the first iteration difference is less than the first iteration threshold, the iteration condition has been met at this time, and the corresponding iteratively generated image can be output as the final iterative image. If the first iteration difference is not less than the first iteration threshold, the iteration process needs to continue until it is less than the first iteration threshold. If the condition is not met after a certain number of times, it is determined whether the number of iterations exceeds the preset number. If it exceeds, the iteration is stopped and the image of the last iteration is output. If it does not exceed, the iteration process continues. In the present application, the preset number is 200.

[0028] S24, performing iterative denoising on the final iterative image to obtain a denoised lung image; Wherein, the step S24 comprises: S241, the final iterative image Convert to image matrix , the image matrix As the initial value and iteratively denoising the initial value to obtain an iterative denoising value: ; ; ; In the formula, , Respectively represent sequence Iterative denoising value after iterations, is the second regularization parameter, Indicates sequence The first variable after iterations, For the first step, represents the initial first variable, , Indicates sequence The second variable after iterations, represents the second step length, , Indicates sequence The image pixel value after iterations, represents the adjustment factor; Specifically, step S241 is specifically a primal-dual iterative process, in which the first variable is a dual variable and the second variable is an extrapolated variable.

[0029] S242, based on sequence The iterative denoising value after the iteration calculates the second iteration difference : .

[0030] S243, determining the two-iteration difference Is it less than the second iteration threshold? If the two iteration difference is less than the second iteration threshold, then The iterative denoising value after the iteration is used as the final image matrix and the final image matrix is ​​restored to an image to obtain a denoised lung image. If the two-iteration difference If the value is not less than the second iteration threshold, the iteration process is repeated until the final image matrix is ​​output; Specifically, this step is similar to step S23. After a certain number of iterations, the second iteration difference is calculated to determine whether the iteration process needs to continue. If the conditions are met, the corresponding iterative denoising value is output and restored to an image to obtain a denoised lung image. If not, the iteration process continues.

[0031] S3, performing feature image segmentation on the denoised lung image to obtain a lung feature image; Wherein, the step S3 comprises: S31. Perform convolution filtering on the denoised lung image to obtain a filtered image.

[0032] Wherein, the step S31 includes: S311, the denoised lung image Perform Gaussian convolution to obtain the scale matrix :

[0033] ; ; ; ; ; In the formula, , , , are the first, second, third, and fourth scale parameters, respectively. represents the Gaussian function, Respectively represent the horizontal and vertical coordinates of the pixels in the denoised lung image, represents the Gaussian parameter, Represents the convolution operation; Specifically, the Gaussian parameters can be adjusted. Selecting reasonable Gaussian parameters can improve the filtering effect. The selection of Gaussian parameters is shown in step S313.

[0034] S312, performing eigendecomposition on the scale matrix to obtain a first eigenvalue With the second eigenvalue , based on the first eigenvalue With the second eigenvalue Determine the tuning parameters : ; In the formula, is the sensitivity coefficient, is the smoothing coefficient.

[0035] S313, adjusting the parameters The Gaussian parameters in the Gaussian filter are replaced and the denoised lung image is filtered to obtain a filtered image.

[0036] S32, performing region selection, first diffusion processing and contour fitting on the filtered image in sequence to obtain an updated feature image; Wherein, the step S32 includes: S321, determining a histogram of the filtered image, extracting connected areas in the histogram whose grayscale values ​​are greater than a preset grayscale value, calculating the circularity of each of the connected areas, and taking the image area corresponding to the connected area with the largest circularity as the first feature area.

[0037] S323, identifying the maximum circumscribed rectangular area of ​​the first feature area, expanding the edge of the maximum circumscribed rectangular area outward by a preset number of pixels to obtain a second feature area, mapping the second feature area to the filtered image to obtain a basic feature image, and binarizing the basic feature image to obtain a binary foreground image; Specifically, after selecting the maximum circumscribed rectangular area, the situation of missing feature information cannot be avoided. Therefore, the maximum circumscribed rectangle needs to be expanded outward by several pixels so that the second feature area can fully contain the lung cancer feature information. At the same time, after obtaining the corresponding basic feature image, the maximum inter-class variance method can be used to determine the segmentation threshold, and the image can be binarized based on the segmentation threshold to obtain a binary foreground image and a binary background image. The binary foreground image contains the patient's lung features.

[0038] S323, determining a rectangular window with the center position of the binary foreground image as the midpoint, selecting two connected regions with the largest areas in the rectangular window as reference regions, and determining whether the area difference between the two reference regions is less than a first preset value, and whether the centroid coordinates of the two reference regions are less than a second preset value; Specifically, for the binary foreground image, since the lungs include the left lung and the right lung, the two connected regions with the largest area in the image are the lung regions of the patient. At the same time, according to the above judgment rules, it can be determined whether the two connected regions are the lung regions to avoid misidentification.

[0039] S324, if the area difference between the two reference areas is not less than the first preset value and / or the centroid coordinates of the two reference areas are not less than the second preset value, the rectangular window is enlarged and the reference area is reselected; if the area difference between the two reference areas is less than the first preset value and the centroid coordinates of the two reference areas are less than the second preset value, any number of pixel points are selected in the two reference areas as reference points respectively, and diffused outward with the reference points as the center until they diffuse to the edge position of the foreground in the binary foreground image, so as to obtain a diffused image; Specifically, the diffusion process is to diffuse outward along the 8-neighborhood direction of the reference point. At the same time, during the diffusion process, several reference points diffuse outward to form a diffusion area. During the diffusion process, if all the pixel points in the 8-neighborhood direction of the reference point are contained in the diffusion area, the diffusion process of the reference point is stopped. At the same time, during the diffusion process, the edge points of the diffusion area obtained during the diffusion process are identified in real time, and the edge points are used as new reference points to diffuse outward. When the diffusion reaches the edge position of the foreground, the diffusion is stopped, and finally a diffusion image containing human lungs and lung feature information can be obtained.

[0040] S325 , identifying edge points of the diffuse image and storing the edge points in a diffuse edge point set, and removing pixel points in the diffuse edge point set that are not in the second feature area to obtain an edge pixel point set.

[0041] S326, identifying the number of valid pixels of each pixel in the edge pixel set within an 8-neighborhood range, and taking the pixel whose number of valid pixels within the 8-neighborhood range is 1 as an endpoint; Specifically, the chain code tracking algorithm can be used to identify the number of valid pixels within the 8-neighborhood range of each pixel.

[0042] S327, taking the endpoint as the starting point, respectively calculating the Euclidean distance between the endpoint and each pixel point in the edge pixel point set, taking the pixel point with the smallest Euclidean distance as the new starting point and iteratively selecting the starting point and calculating the Euclidean distance to obtain a number of starting points and store the starting points in the starting point set.

[0043] S328, performing minimum multiplication curve fitting on the pixel points in the starting point set to obtain a fitting point set, performing contour update and internal filling on the diffusion image based on the fitting point set to obtain a contour update image, and mapping the contour update image to the basic feature image to obtain an updated feature image; Specifically, through the above diffusion process, the patient's lung feature information can be fully and completely included in the updated feature image.

[0044] S33, performing a second diffusion process on the updated feature image to obtain a lung feature image; Wherein, the step S33 comprises: S331, extracting the highlight area in the updated feature image and performing morphological processing on the highlight area to obtain the original region image, extracting the boundary points of the original region image, taking the boundary points whose grayscale values ​​are greater than the boundary grayscale threshold as diffusion points, storing the diffusion points in a diffusion point set, and storing all pixel points in the updated feature image in a feature point set; Specifically, the highlighted position in the updated feature image is the position where the suspected lung cancer feature is located.

[0045] S332, diffusing outwards with the diffusion point in the diffusion point set as the center, storing the diffused pixel points into the diffused pixel point set, and calculating the average grayscale of each pixel point in the diffused pixel point set.

[0046] S333, randomly selecting a pixel point from the feature point set as a diffusion cutoff point, scanning the eight-neighborhood pixel points of the diffusion cutoff point, and determining whether the absolute value of the difference between the eight-neighborhood pixel points and the average grayscale of each pixel point in the diffused pixel point set is less than a diffusion threshold.

[0047] S334, if the absolute value of the difference between the average grayscale of each pixel in the 8-neighborhood pixel point and the diffused pixel point set is less than the diffusion threshold, the diffusion cutoff point is stored in the diffused pixel point set; if the absolute value of the difference between the average grayscale of each pixel in the 8-neighborhood pixel point and the diffused pixel point set is not less than the diffusion threshold, the diffusion cutoff point is used as a background point and removed from the feature point set, and the diffusion and judgment process is iteratively repeated until the feature point set is an empty set, and a lung feature image is output; Specifically, through the above diffusion process, the features of the patient's suspected lung cancer can be fully extracted, and the influence of heterogeneous tissue on the segmentation of feature information can be fully reduced.

[0048] S4, obtaining a historical template lung cancer image set, inputting the historical template lung cancer image set into a preset recognition model for training, inputting the lung feature image into the trained preset recognition model for recognition, and outputting a recognition result; Specifically, the preset recognition model here can be the VGG recognition network commonly used in the prior art. The preset recognition model can be trained through the historical template lung cancer image, and the recognition network can be updated. Then, the extracted and segmented lung feature image is input into the trained preset recognition model to output the corresponding recognition result.

[0049] The lung cancer image recognition method provided in the first embodiment of the present invention first obtains the lung image of the patient, performs image correction processing on the lung image to obtain a corrected lung image; then performs iterative denoising processing on the corrected lung image to obtain a denoised lung image; then performs feature image segmentation on the denoised lung image to obtain a lung feature image; finally, obtains a historical template lung cancer image set, inputs the historical template lung cancer image set into a preset recognition model for training, and inputs the lung feature image into the trained preset recognition model for recognition to output a recognition result. The present invention first performs correction processing on the image to avoid the situation where the image appears darker and the features are not obvious, and then performs iterative denoising processing on the image. The correction process can effectively remove the noise and artifacts in the image, and the image is feature segmented at the same time, which can eliminate the interference information in the image, thereby improving the recognition accuracy.

[0050] Embodiment 2 like Figure 2 As shown, in the second embodiment of the present invention, a lung cancer image recognition system is provided, and the system includes: Correction module 1, used for acquiring a lung image of a patient and performing image correction processing on the lung image to obtain a corrected lung image; De-noising module 2, used for iteratively denoising the corrected lung image to obtain a denoised lung image; A segmentation module 3 is used to perform feature image segmentation on the denoised lung image to obtain a lung feature image; The recognition module 4 is used to obtain a historical template lung cancer image set, input the historical template lung cancer image set into a preset recognition model for training, input the lung feature image into the trained preset recognition model for recognition, and output a recognition result.

[0051] The correction module 1 comprises: A division submodule is used to divide the lung image into equal regions to obtain a plurality of sub-images, perform linear transformation on the plurality of sub-images to determine a transformation equation, and extract a first transformation coefficient in the transformation equation and the second transform coefficient , based on the first transform coefficient With the second transform coefficient Determine the cost function : ; In the formula, , Represents the output and input Zhang sub-image, represents the number of sub-images, is the first regularization parameter; Minimization submodule, used to minimize the cost function Perform minimization and output the optimal first transform coefficient and the optimal second transform coefficient : ; ; In the formula, , Respectively , The mean value in the corresponding window, express The number of pixels in the neighborhood window of express Variance within the corresponding window; A correction submodule is used to and the optimal second transform coefficient Image separation and correction are performed to obtain a corrected lung image.

[0052] The correction submodule comprises: An image determination unit is configured to determine an optimal first transform coefficient based on the optimal first transform coefficient. and the optimal second transform coefficient Determine the brightness feature image Image with background reflection : ; ; A correction unit is used to correct the brightness characteristic image Correction processing is performed to obtain a brightness-corrected image : ; ; In the formula, is the grayscale value level of the brightness feature image, is the grayscale mean of the brightness feature image, Respectively represent the length and width of the brightness feature image; A fusion unit is used to transform the brightness correction image With the background reflection image The fusion is performed to obtain the corrected lung image.

[0053] The denoising module 2 comprises: The regeneration submodule is used to use the lung image as an initial iterative image and iteratively regenerate the initial iterative image to obtain an iteratively generated image: ; ; In the formula, , Respectively represent sequence The iterative generation image after iterations is is the relaxation factor, Indicates The set of all projection ray indices corresponding to the projection angles, Indicates The projection coefficient matrix corresponding to the projection rays is: express The corresponding projection ray vector, is the transpose symbol; The difference calculation submodule is used to calculate the sequence The iterative generation image after iterations calculates the first iteration difference : ; Iteration judgment submodule, used to judge the first iteration difference Is it less than the first iteration threshold? If the first iteration difference is less than the first iteration threshold, then The iterative image generated after the iteration is used as the final iterative image. If the first iterative difference If the value is not less than the first iteration threshold, the iteration process is repeated until the final iteration image is output; The denoising submodule is used to iteratively denoise the final iterative image to obtain a denoised lung image.

[0054] The denoising submodule comprises: A conversion unit for converting the final iterative image Convert to image matrix , the image matrix As the initial value and iteratively denoising the initial value to obtain an iterative denoising value: ; ; ; In the formula, , Respectively represent sequence Iterative denoising value after iterations, is the second regularization parameter, Indicates sequence The first variable after iterations, For the first step, represents the initial first variable, , Indicates sequence The second variable after iterations, represents the second step length, , Indicates sequence The image pixel value after iterations, represents the adjustment factor; The difference calculation unit is used to calculate the sequence The iterative denoising value after the iteration calculates the second iteration difference : ; An iteration judgment unit, used to judge the difference between the two iterations Is it less than the second iteration threshold? If the two iteration difference is less than the second iteration threshold, then The iterative denoising value after the iteration is used as the final image matrix and the final image matrix is ​​restored to an image to obtain a denoised lung image. If the two-iteration difference If it is not less than the second iteration threshold, the iteration process is repeated until the final image matrix is ​​output.

[0055] The segmentation module 3 comprises: A filtering submodule, configured to perform convolution filtering on the denoised lung image to obtain a filtered image; A fitting submodule, used for sequentially performing region selection, first diffusion processing and contour fitting on the filtered image to obtain an updated feature image; The diffusion submodule is used to perform a second diffusion process on the updated feature image to obtain a lung feature image.

[0056] The filtering submodule comprises: A convolution unit is used to transform the denoised lung image Perform Gaussian convolution to obtain the scale matrix :

[0057] ; ; ; ; ; In the formula, , , , are the first, second, third, and fourth scale parameters, respectively. represents the Gaussian function, Respectively represent the horizontal and vertical coordinates of the pixels in the denoised lung image, represents the Gaussian parameter, Represents the convolution operation; A decomposition unit, used to perform eigendecomposition on the scale matrix to obtain a first eigenvalue With the second eigenvalue , based on the first eigenvalue With the second eigenvalue Determine the tuning parameters : ; In the formula, is the sensitivity coefficient, is the smoothing coefficient; A filtering unit is used to adjust the parameters The Gaussian parameters in the Gaussian filter are replaced and the denoised lung image is filtered to obtain a filtered image.

[0058] The fitting submodule includes: A first region unit is used to determine a histogram of the filtered image, extract connected regions in the histogram whose grayscale values ​​are greater than a preset grayscale value, calculate the circularity of each of the connected regions, and use the image region corresponding to the connected region with the largest circularity as a first feature region; A second area unit is used to identify the maximum circumscribed rectangular area of ​​the first feature area, expand the edge of the maximum circumscribed rectangular area outward by a preset number of pixels to obtain a second feature area, map the second feature area to the filtered image to obtain a basic feature image, and binarize the basic feature image to obtain a binary foreground image; A window unit, used for determining a rectangular window with the center position of the binary foreground image as the midpoint, selecting two connected areas with the largest areas in the rectangular window as reference areas, and determining whether the area difference between the two reference areas is less than a first preset value and whether the centroid coordinates of the two reference areas are less than a second preset value; A reference unit, configured to expand the rectangular window and reselect the reference area if the area difference between the two reference areas is not less than a first preset value and / or the centroid coordinates of the two reference areas are not less than a second preset value, and to select any number of pixel points in the two reference areas as reference points respectively if the area difference between the two reference areas is less than the first preset value and the centroid coordinates of the two reference areas are less than the second preset value, and diffuse outward with the reference points as the center until they diffuse to the edge position of the foreground in the binary foreground image to obtain a diffused image; a culling unit, configured to identify edge points of the diffuse image and store the edge points in a diffuse edge point set, and to cull pixel points in the diffuse edge point set that are not in the second feature area, so as to obtain an edge pixel point set; A pixel identification unit, used to identify the number of valid pixels of each pixel point in the edge pixel point set within an 8-neighborhood range, and take the pixel point whose number of valid pixels within the 8-neighborhood range is 1 as an endpoint; A distance calculation unit is used to calculate the Euclidean distance between the endpoint and each pixel point in the edge pixel point set, taking the endpoint as the starting point, taking the pixel point with the smallest Euclidean distance as the new starting point, and iteratively selecting the starting point and calculating the Euclidean distance to obtain a plurality of starting points and store the plurality of starting points in the starting point set; A fitting unit is used to perform minimum multiplication curve fitting on the pixel points in the starting point set to obtain a fitting point set, perform contour update and internal filling on the diffusion image based on the fitting point set to obtain a contour update image, and map the contour update image to the basic feature image to obtain an updated feature image.

[0059] The diffusion submodule comprises: A processing unit, configured to extract a highlight area in the updated feature image and perform morphological processing on the highlight area to obtain an original region image, extract boundary points of the original region image, use boundary points with grayscale values ​​greater than a boundary grayscale threshold as diffusion points, store the diffusion points in a diffusion point set, and store all pixel points in the updated feature image in a feature point set; A diffusion unit, used for diffusing outwards with a diffusion point in the diffusion point set as the center, storing the diffused pixel points into the diffused pixel point set, and calculating the average grayscale of each pixel point in the diffused pixel point set; A cutoff unit, used for arbitrarily selecting a pixel point from the feature point set as a diffusion cutoff point, scanning eight neighboring pixel points of the diffusion cutoff point, and determining whether an absolute value of a difference between the eight neighboring pixel points and an average grayscale of each pixel point in the diffused pixel point set is less than a diffusion threshold; The iterative diffusion unit is used to store the diffusion cutoff point in the diffused pixel set if the absolute value of the difference between the average grayscale of each pixel in the 8-neighborhood pixel point and the diffused pixel point set is less than the diffusion threshold; if the absolute value of the difference between the average grayscale of each pixel in the 8-neighborhood pixel point and the diffused pixel point set is not less than the diffusion threshold, the diffusion cutoff point is used as a background point and is removed from the feature point set, and the diffusion and judgment process is iteratively repeated until the feature point set is an empty set, and a lung feature image is output.

[0060] In some other embodiments of the present invention, the embodiments of the present invention provide the following technical solution: a computer, comprising a memory 102, a processor 101, and a computer program stored in the memory 102 and executable on the processor 101, wherein the processor 101 implements the lung cancer image recognition method as described above when executing the computer program.

[0061] Specifically, the processor 101 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiment of the present invention.

[0062] Among them, the memory 102 may include a large-capacity memory 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 disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory 102 may include a removable or non-removable (or fixed) medium. Where appropriate, the memory 102 may be inside or outside the data processing device. In a specific embodiment, the memory 102 is a non-volatile memory. In a specific embodiment, the memory 102 includes a read-only memory (ROM) and a random access memory (RAM). Where appropriate, 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 (FLASH), or a combination of two or more of these. Under appropriate circumstances, the RAM can be a static random access memory (SRAM) or a dynamic random access memory (DRAM), wherein the DRAM can be a fast page mode dynamic random access memory (FPMDRAM), an extended date out dynamic random access memory (EDODRAM), a synchronous dynamic random access memory (SDRAM), etc.

[0063] The memory 102 may be used to store or cache various data files that need to be processed and / or used for communication, as well as possible computer program instructions executed by the processor 101 .

[0064] The processor 101 implements the above-mentioned lung cancer image recognition method by reading and executing the computer program instructions stored in the memory 102 .

[0065] In some embodiments, the computer may further include a communication interface 103 and a bus 100. Figure 3 As shown, the processor 101, the memory 102, and the communication interface 103 are connected via a bus 100 and communicate with each other.

[0066] The communication interface 103 is used to implement communication between the modules, devices, units and / or equipment in the embodiment of the present invention. The communication interface 103 can also implement data communication with other components such as: external devices, image / data acquisition equipment, databases, external storage, and image / data processing workstations.

[0067] The bus 100 includes hardware, software or both, and couples the components of the computer device to each other. The bus 100 includes but is not limited to at least one of the following: a data bus, an address bus, a control bus, an expansion bus, and a local bus. By way of example and not limitation, 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. Where appropriate, bus 100 may include one or more buses. Although embodiments of the present invention describe and illustrate a particular bus, the present invention contemplates any suitable bus or interconnect.

[0068] The computer can execute the lung cancer image recognition method of the present invention based on the acquired lung cancer image recognition system, thereby realizing lung cancer image recognition.

[0069] In some further embodiments of the present invention, in combination with the above-mentioned lung cancer image recognition method, the embodiments of the present invention provide the following technical solutions: a storage medium having a computer program stored thereon, and the computer program implements the above-mentioned lung cancer image recognition method when executed by a processor.

[0070] Those skilled in the art will appreciate that the logic and / or steps represented in the flowchart or otherwise described herein, for example, may be considered as an ordered list of executable instructions for implementing logical functions, and may be specifically implemented in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For purposes of this specification, "computer-readable medium" may be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0071] More specific examples of readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0072] 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-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or a combination thereof: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0073] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described 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.

[0074] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A lung cancer image recognition method, characterized in that: include: Acquiring a lung image of a patient, and performing image correction processing on the lung image to obtain a corrected lung image; Performing iterative denoising processing on the corrected lung image to obtain a denoised lung image; Performing feature image segmentation on the denoised lung image to obtain a lung feature image; A historical template lung cancer image set is obtained, the historical template lung cancer image set is input into a preset recognition model for training, and the lung feature image is input into the trained preset recognition model for recognition to output a recognition result.

2. The lung cancer image recognition method according to claim 1, characterized in that: The step of performing image correction processing on the lung image to obtain a corrected lung image comprises: The lung image is divided into equal regions to obtain a plurality of sub-images, a linear transformation is performed on the plurality of sub-images to determine a transformation equation, and a first transformation coefficient in the transformation equation is extracted. and the second transform coefficient , based on the first transform coefficient With the second transform coefficient Determine the cost function : ； In the formula, , Represents the output and input Zhang sub-image, represents the number of sub-images, is the first regularization parameter; For the cost function Perform minimization and output the optimal first transform coefficient and the optimal second transform coefficient : ； ； In the formula, , Respectively , The mean value in the corresponding window, express The number of pixels in the neighborhood window of express Variance within the corresponding window; Based on the optimal first transform coefficient and the optimal second transform coefficient Image separation and correction are performed to obtain a corrected lung image.

3. The lung cancer image recognition method according to claim 2, characterized in that: The optimal first transform coefficient and the optimal second transform coefficient The steps of performing image separation and correction to obtain a corrected lung image include: Based on the optimal first transform coefficient and the optimal second transform coefficient Determine the brightness feature image Image with background reflection : ； ； For the brightness feature image Correction processing is performed to obtain a brightness-corrected image : ； ； In the formula, is the grayscale value level of the brightness feature image, is the grayscale mean of the brightness feature image, Respectively represent the length and width of the brightness feature image; The brightness correction image With the background reflection image The fusion is performed to obtain the corrected lung image.

4. The lung cancer image recognition method according to claim 1, characterized in that: The step of iteratively denoising the corrected lung image to obtain a denoised lung image comprises: The lung image is used as an initial iterative image, and the initial iterative image is iteratively regenerated to obtain an iteratively generated image: ； ； In the formula, , Respectively represent sequence The iterative generation image after iterations is is the relaxation factor, Indicates The set of all projection ray indices corresponding to the projection angles, Indicates The projection coefficient matrix corresponding to the projection rays is: express The corresponding projection ray vector, is the transpose symbol; Based on sequence The iterative generation image after the iterations is calculated as the first iteration difference : ； Determine the first iterative difference Is it less than the first iteration threshold? If the first iteration difference is less than the first iteration threshold, then The iterative image generated after the iteration is used as the final iterative image. If the first iterative difference If the value is not less than the first iteration threshold, the iteration process is repeated until the final iteration image is output; Iterative denoising is performed on the final iterative image to obtain a denoised lung image.

5. The lung cancer image recognition method according to claim 4, characterized in that: The step of iteratively denoising the final iterative image to obtain a denoised lung image comprises: The final iteration image Convert to image matrix , the image matrix As the initial value and iteratively denoising the initial value to obtain an iterative denoising value: ； ； ； In the formula, , Respectively represent sequence Iterative denoising value after iterations, is the second regularization parameter, Indicates sequence The first variable after iterations, For the first step, represents the initial first variable, , Indicates sequence The second variable after iterations, represents the second step length, , Indicates sequence The image pixel value after iterations, represents the adjustment factor; Based on sequence The iterative denoising value after the iteration calculates the second iteration difference : ； Determine the two-iteration difference Is it less than the second iteration threshold? If the two iteration difference is less than the second iteration threshold, then The iterative denoising value after the iteration is used as the final image matrix and the final image matrix is ​​restored to an image to obtain a denoised lung image. If the two-iteration difference If it is not less than the second iteration threshold, the iteration process is repeated until the final image matrix is ​​output.

6. The lung cancer image recognition method according to claim 1, characterized in that: The step of performing feature image segmentation on the denoised lung image to obtain a lung feature image comprises: Performing convolution filtering on the denoised lung image to obtain a filtered image; performing region selection, first diffusion processing and contour fitting on the filtered image in sequence to obtain an updated feature image; The updated feature image is subjected to a second diffusion process to obtain a lung feature image.

7. The lung cancer image recognition method according to claim 6, characterized in that: The step of performing convolution filtering on the denoised lung image to obtain a filtered image comprises: The denoised lung image Perform Gaussian convolution to obtain the scale matrix : ； ； ； ； ； In the formula, , , , are the first, second, third, and fourth scale parameters, respectively. represents the Gaussian function, Respectively represent the horizontal and vertical coordinates of the pixels in the denoised lung image, represents the Gaussian parameter, Represents the convolution operation; Perform eigendecomposition on the scale matrix to obtain the first eigenvalue With the second eigenvalue , based on the first eigenvalue With the second eigenvalue Determine the tuning parameters : ； In the formula, is the sensitivity coefficient, is the smoothing coefficient; The adjustment parameters The Gaussian parameters in the Gaussian filter are replaced and the denoised lung image is filtered to obtain a filtered image.

8. The lung cancer image recognition method according to claim 6, characterized in that: The step of sequentially performing region selection, diffusion processing and contour fitting on the filtered image to obtain an updated feature image comprises: Determine a histogram of the filtered image, extract connected areas in the histogram whose grayscale values ​​are greater than a preset grayscale value, calculate the circularity of each of the connected areas, and use the image area corresponding to the connected area with the largest circularity as the first feature area; Identify the maximum circumscribed rectangular area of ​​the first feature area, expand the edge of the maximum circumscribed rectangular area outward by a preset number of pixels to obtain a second feature area, map the second feature area to the filtered image to obtain a basic feature image, and binarize the basic feature image to obtain a binary foreground image; Determine a rectangular window with the center position of the binary foreground image as the midpoint, select two connected areas with the largest area in the rectangular window as reference areas, and determine whether the area difference between the two reference areas is less than a first preset value and whether the centroid coordinates of the two reference areas are less than a second preset value; If the area difference between the two reference areas is not less than a first preset value and / or the centroid coordinates of the two reference areas are not less than a second preset value, the rectangular window is enlarged and the reference area is reselected; if the area difference between the two reference areas is less than the first preset value and the centroid coordinates of the two reference areas are less than the second preset value, any number of pixel points are selected in the two reference areas as reference points respectively, and diffused outward with the reference points as the center until they diffuse to the edge position of the foreground in the binary foreground image, so as to obtain a diffused image; Identifying edge points of the diffuse image and storing the edge points in a diffuse edge point set, and removing pixel points in the diffuse edge point set that are not in the second feature area to obtain an edge pixel point set; Identify the number of valid pixels of each pixel in the edge pixel point set within an 8-neighborhood range, and take the pixel point whose number of valid pixels in the 8-neighborhood range is 1 as an endpoint; Taking the endpoint as the starting point, respectively calculating the Euclidean distance between the endpoint and each pixel point in the edge pixel point set, taking the pixel point with the smallest Euclidean distance as the new starting point, and iteratively selecting the starting point and calculating the Euclidean distance to obtain a number of starting points and store the starting points in the starting point set; Minimum multiplication curve fitting is performed on the pixel points in the starting point set to obtain a fitting point set, contour updating and internal filling are performed on the diffusion image based on the fitting point set to obtain a contour update image, and the contour update image is mapped to the basic feature image to obtain an updated feature image.

9. The lung cancer image recognition method according to claim 6, characterized in that: The step of performing a second diffusion process on the updated feature image to obtain a lung feature image comprises: Extracting the highlight area in the updated feature image and performing morphological processing on the highlight area to obtain the original area image, extracting the boundary points of the original area image, taking the boundary points whose grayscale values ​​are greater than the boundary grayscale threshold as diffusion points, storing the diffusion points in a diffusion point set, and storing all the pixel points in the updated feature image in a feature point set; Diffusion is performed outwards with the diffusion point in the diffusion point set as the center, and the pixel points obtained by diffusion are stored in the diffused pixel point set, and the average grayscale of each pixel point in the diffused pixel point set is calculated; Randomly select a pixel point from the feature point set as a diffusion cutoff point, scan the eight neighboring pixel points of the diffusion cutoff point, and determine whether the absolute value of the difference between the eight neighboring pixel points and the average grayscale of each pixel point in the diffused pixel point set is less than a diffusion threshold; If the absolute value of the difference between the average grayscale of each pixel in the 8-neighborhood pixel point and the diffused pixel point set is less than the diffusion threshold, the diffusion cutoff point is stored in the diffused pixel point set. If the absolute value of the difference between the average grayscale of each pixel in the 8-neighborhood pixel point and the diffused pixel point set is not less than the diffusion threshold, the diffusion cutoff point is used as a background point and removed from the feature point set. The diffusion and judgment process is iteratively repeated until the feature point set is an empty set, and the lung feature image is output.

10. A lung cancer image recognition system, characterized in that: The system comprises: A correction module, used for acquiring a lung image of a patient and performing image correction processing on the lung image to obtain a corrected lung image; A denoising module, used for iteratively denoising the corrected lung image to obtain a denoised lung image; A segmentation module, used for performing feature image segmentation on the denoised lung image to obtain a lung feature image; The recognition module is used to obtain a historical template lung cancer image set, input the historical template lung cancer image set into a preset recognition model for training, input the lung feature image into the trained preset recognition model for recognition, and output a recognition result.

Images