Deep learning-based automatic segmentation system and method for lung injury region

Abstract

The application provides a kind of deep learning-based lung injury area automatic segmentation system and method, realize the precision of improving lung injury area automatic segmentation.First, the original lung image is de-noised, edge detection and region enhancement, output the processed original lung image;Then, using deep convolutional neural network (CNN) is carried out to the processed original lung image feature extraction and image segmentation, and the segmentation result is optimized in combination with medical atlas, and the output optimization lung injury area segmentation image;Finally, through augmented reality technology and interactive correction function, help doctor real-time adjustment segmentation result, and through morphological operation and smoothing processing, output the final lung injury area segmentation image.The application improves the precision of lung injury area automatic segmentation by multilevel technical means.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of automatic segmentation of lung injury regions, and particularly relates to a lung injury region automatic segmentation system and method based on deep learning. BACKGROUND

[0002] The automatic segmentation technology of lung injury regions has important significance in medical imaging, especially in the early diagnosis and treatment of lung diseases. Currently, traditional lung injury segmentation methods usually rely on manual labeling and artificial judgment, which not only consumes time and effort, but also is easily affected by the experience and subjective factors of doctors, resulting in inaccurate or inconsistent segmentation results. In addition, existing automatic segmentation methods have problems of insufficient accuracy and poor adaptability, especially when dealing with complex lung images, they often cannot effectively segment the subtle lung injury regions.

[0003] In recent years, deep learning technology, especially convolutional neural network (CNN), has made significant progress in medical image analysis. Deep learning methods can automatically learn features from a large number of medical images and perform effective segmentation, but still face problems such as noise influence, image quality difference, and boundary blur. Although some methods try to improve these problems through image preprocessing and post-processing techniques, how to accurately identify and optimize lung injury regions remains a challenge.

[0004] In the prior art, deep learning and convolutional neural network (CNN) are used for image segmentation and optimization. By training a deep network model, features are extracted from the original image and segmented. However, in terms of image noise processing and boundary optimization, it is not fine enough, resulting in low segmentation accuracy of injury regions in complex image backgrounds. In addition, these methods often lack effective post-processing steps, and the segmentation boundary may appear jagged, affecting the accuracy of the diagnosis result. SUMMARY

[0005] To solve the technical problems mentioned in the background art, the present application proposes a lung injury region automatic segmentation system and method based on deep learning.

[0006] To this end, the technical solutions adopted by the present application are as follows:

[0007] A lung injury region automatic segmentation system based on deep learning, characterized in that the system comprises:

[0008] M1, an image processing module, receiving an original lung image from a medical image device, performing denoising processing, edge detection and region enhancement processing on the original lung image, and outputting the processed original lung image;

[0009] M2, a segmentation and optimization module, performs multi-layer convolution operation on the processed original lung image through a deep convolutional neural network, extracts low-level and high-level features, segments the processed original lung image based on the low-level and high-level features, generates a lung injury region segmentation image, optimizes the lung injury region segmentation image using medical atlas knowledge, and outputs an optimized lung injury region segmentation image;

[0010] M3, a visualization and correction module, displays and compares the optimized lung injury region segmentation image and the original lung image, presents the comparison result to the doctor for viewing through augmented reality technology, and outputs a corrected lung injury region segmentation image through interactive correction of the optimized lung injury region segmentation image by the doctor through a virtual interface;

[0011] M4, a post-processing module, performs morphological operation on the corrected lung injury region segmentation image, and smoothes the boundary of the segmentation image through an interpolation algorithm, and outputs a final lung injury region segmentation image.

[0012] Further, the denoising processing is performed through bilateral filtering; the edge detection is realized by using a Canny edge detection algorithm, including gradient calculation, gradient amplitude and edge direction,

[0013] The formula of the gradient calculation is:

[0014]

[0015] wherein, G x is the gradient of the original lung image I in the horizontal direction; G y is the gradient of the original lung image I in the vertical direction; I(x, y) is the pixel value of the original lung image I at position (x, y);

[0016] The formula for calculating the gradient amplitude is:

[0017]

[0018] wherein, G is the gradient amplitude, i.e. the degree of edge visibility;

[0019] The formula for calculating the gradient direction is:

[0020]

[0021] wherein, θ is the gradient direction, indicating the direction of the edge of the original lung image.

[0022] Further, the region enhancement includes calculation of local mean and standard deviation, and the formula is:

[0023]

[0024] wherein, N(x, y) is a neighborhood region centered at pixel (x, y); μ local (x, y) is the mean value of region N(x, y), i.e. the average value of all pixels in the region; σ local (x, y) is the standard deviation of region N(x, y); I(x', y') is the pixel value of pixel (x', y') in the neighborhood;

[0025] According to the local mean and standard deviation, the enhanced value C(x, y) of the original lung image pixel is calculated, and the formula is:

[0026]

[0027] wherein, a is an enhancement coefficient, which controls the intensity of enhancement;

[0028] According to the denoising processing, edge detection and region enhancement, the processed original lung image J is output.

[0029] Further, the operation steps of the segmentation and optimization module include feature extraction, image segmentation and medical knowledge graph optimization,

[0030] The feature extraction uses a deep convolutional neural network to extract low-level and high-level features in the processed original lung image through multi-layer convolution operation. The low-level features include edges, textures and corner points, and the high-level features include shapes, structures and regions. The low-level features y low The processed original lung image is extracted through multi-layer convolution of the convolution kernel, and the formula is:

[0031]

[0032] wherein, J(m, n) is the pixel value of the processed original lung image J at position (m, n); K low (x-m, y-n) is the convolution kernel;

[0033] According to the low-level features, the high-level features y high are extracted through a deep convolutional network, and the formula is:

[0034]

[0035] wherein, y high (x, y) is the output of the high-level features;

[0036] The low-level features and high-level features form the output features f.

[0037] Further, the image segmentation classifies the output features to determine whether they belong to the lung injury region, and the formula is:

[0038]

[0039] wherein W and b are the weight and bias of the deep convolutional network respectively; σ is a Sigmoid activation function, used to compress the output of the deep convolutional network to [0, 1], indicating that the feature belongs to the lung injury region, segmenting the image into a lung injury region;

[0040] The medical knowledge graph optimization optimizes the lung injury region segmentation image by using the medical graph, corrects unreasonable segmentation regions, identifies the error regions in the lung injury region segmentation image according to the medical graph, and optimizes by adjusting the segmentation boundary to output an optimized lung injury region segmentation image Y.

[0041] Further, the superimposed display contrast is realized by an augmented reality technology, and the optimized lung injury region segmentation image Y is superimposed and displayed with the original lung image I, and the formula is:

[0042] I overlay (x,y)=α·I(x,y)+(1-α)·Y(x,y)

[0043] wherein I overlay (x,y) is the superimposed image; α is a weighting coefficient, which controls the contribution proportion of the original lung image and the optimized lung injury region segmentation image, and α ∈ [0, 1], and the display effect is adjusted according to needs.

[0044] Further, the interactive correction specifically means that a doctor selects a region needing correction through an interactive interface, adjusts the lung injury region segmentation image according to the operation of the doctor, and the calculation formula of the corrected lung injury region segmentation image Y corrected is:

[0045]

[0046] wherein R corrected is a coordinate set of the doctor's manual correction region, representing the updated boundary of the lung injury region segmentation image.

[0047] Further, the morphological operation includes dilation and erosion operations, the dilation operation expands the boundary of the corrected lung injury region segmentation image, enlarges the foreground region, and fills small holes,

[0048] the erosion operation shrinks the boundary of the foreground region of the corrected lung injury region segmentation image, and removes small noise regions.

[0049] The smoothing processing is realized by an interpolation algorithm, the interpolation algorithm calculates a smooth curve by weighted average of the boundary pixels to adjust the boundary, and the formula is:

[0050] Y smooth (x,y)=f(Y corrected (x,y))

[0051] wherein Y smooth (x,y) is the boundary smoothed segmentation image; f is a smoothing algorithm adjusted according to the boundary of the segmentation image;

[0052] Through the morphological operation and the smoothing processing, a final lung injury region segmentation image Y smooth is output.

[0053] A deep learning-based automatic lung injury region segmentation method, characterized in that the method comprises the following steps:

[0054] S1, receiving an original lung image from a medical image device and performing denoising, edge detection and region enhancement processing, and outputting a processed original lung image;

[0055] S2, performing multi-layer convolution operation on the processed original lung image through a deep convolutional neural network, extracting low-level and high-level features, segmenting the processed original lung image based on the low-level and high-level features, and optimizing the segmentation result using a medical atlas to generate an optimized lung injury region segmentation image;

[0056] S3, overlapping the optimized lung injury region segmentation image with the original lung image, and a doctor performs interactive correction on the optimized lung injury region segmentation image through a virtual interface to output a corrected lung injury region segmentation image

[0057] S4, performing morphological repair and boundary smoothing on the corrected lung injury region segmentation image to generate a final lung injury region segmentation image.

[0058] Compared with the prior art, the advantages of the present application are:

[0059] 1. High image processing and optimization accuracy: the present application adopts bilateral filter denoising technology, which can effectively preserve the detail edges of the lung and the injury region, and through Canny edge detection algorithm and region enhancement, further improves the contrast and clarity of the segmentation region, ensuring that the injury region can be more significantly presented.

[0060] 2. Efficient feature extraction and accurate segmentation: the present application uses deep convolutional neural network (CNN) and combines medical atlas to optimize the segmentation result, effectively abstracts and fuses low-level and high-level features through deep network, not only has strong image understanding ability, but also can identify error areas in segmentation according to medical atlas, and adjusts the boundary, so as to realize accurate lung injury region segmentation.

[0061] 3. Interactive correction function: The present application introduces augmented reality technology, by overlapping the original image and the optimized segmentation image, doctors can interactively correct the segmentation result through the virtual interface, this function ensures that the segmentation result is further optimized under the participation of doctors, improves the reliability of the system.

[0062] 4. Post-processing technology is perfect: through morphological operation and boundary smoothing processing, small errors in the segmentation result are effectively repaired, ensuring that the final image meets the standard of medical images, facilitating clinical application. BRIEF DESCRIPTION OF DRAWINGS

[0063] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0064] Figure 1 Flow chart of the lung injury area automatic segmentation system of the present application;

[0065] Figure 2 Flow chart of the image processing module of the present application;

[0066] Figure 3 Flow chart of the segmentation and optimization module of the present application. DETAILED DESCRIPTION

[0067] To achieve the above purpose, the present application is realized by the following technical solutions, the present application provides a lung injury area automatic segmentation system based on deep learning, the system comprises:

[0068] M1, image processing module, receiving original lung image from medical image device, denoising, edge detection and region enhancement processing to original lung image, outputting processed original lung image;

[0069] Denoising processing is performed on the original lung image by bilateral filtering, which removes high-frequency noise in the original lung image, improves image quality, and ensures the segmentation and enhancement effect, bilateral filtering can not only remove noise, but also retain the edge details of the image,

[0070] Bilateral filtering can effectively retain the edge information of the lung image while removing background noise, and for particularly complex noise, additional denoising processing can be combined with median filtering to further enhance the noise suppression effect, denoising processing removes noise in the original lung image by bilateral filtering, especially retaining the detail edges of the lung and the injury area, providing a clean base image for edge detection and region enhancement;

[0071] The purpose of edge detection is to extract the boundary information of the original lung image to provide structural information for the segmentation task. The Canny edge detection algorithm is adopted to identify the areas with large changes in the image by calculating the gradient amplitude of the original lung image. These areas correspond to the boundaries in the image. The calculation steps include gradient calculation, gradient amplitude and edge direction,

[0072] Gradient calculation formula:

[0073]

[0074] where G x is the gradient of the image I in the horizontal direction, reflecting the change of the image in the horizontal direction; G y is the gradient of the image I in the vertical direction, reflecting the change of the image in the vertical direction; I(x,y) is the pixel value of the image I at position (x,y);

[0075] Gradient amplitude calculation formula:

[0076]

[0077] where G is the gradient amplitude, indicating the change intensity of the image at a certain position, i.e. the degree of edge visibility;

[0078] Gradient direction formula:

[0079]

[0080] where θ is the gradient direction, indicating the direction of the image edge;

[0081] Through edge detection, non-maximum suppression and double threshold processing are completed. The gradient amplitude is suppressed to local maximum value, and redundant edge response is removed, only the most important edge information is retained. Edge connection is performed through threshold, strong edges are marked as boundaries, and weak edges will also be retained as boundaries when certain conditions are met. The Canny algorithm can effectively extract the detailed edges in the image, especially in images with more noise, it can identify the main structural boundaries in the image, such as the outline of the lung and the boundary of the damaged area;

[0082] After edge detection, the lung edge and damaged area identified in the image can be used as the target area for further processing. Region enhancement mainly improves the visibility of low-contrast areas, making the damaged area more prominent. In this process, edge information can be used as a reference for enhancing the target area. The enhancement algorithm preferentially processes areas close to the edge, which usually contain the most important structural information. By enhancing the contrast, region enhancement provides a clearer visual effect for the segmentation of the damaged area;

[0083] Region enhancement is performed after edge detection, utilizing edge information to define the lung edges and lesion areas and enhance contrast. This includes calculating the local mean and standard deviation, using the following formula:

[0084]

[0085]

[0086] Where N(x,y) is the neighborhood region centered at pixel (x,y), encompassing pixels within a certain surrounding range; μ local (x,y) represents the mean of region N(x,y), which is the average value of all pixels in that region; σ local (x,y) represents the standard deviation of region N(x,y), indicating the degree of variation in pixel values ​​within that region; I(x′,y′) represents the pixel value of the neighboring pixel (x′,y′).

[0087] The enhancement value C(x,y) of the image pixels is calculated based on the local mean and standard deviation, using the following formula:

[0088]

[0089] Where α is the enhancement coefficient, which controls the intensity of the enhancement;

[0090] The original lung image is preprocessed using denoising, edge detection, and region enhancement techniques, thus providing a high-quality input image for subsequent segmentation and optimization modules.

[0091] The M2 segmentation and optimization module uses a deep convolutional neural network to perform multi-layer convolution operations on the processed original lung image, extracting low-level and high-level features. Based on these features, it segments the processed original lung image to generate a segmented image of the lung injury region. Using medical atlas knowledge, it optimizes the segmented image of the lung injury region and outputs the optimized segmented image of the lung injury region.

[0092] The main task of the segmentation and optimization module is to use a deep convolutional neural network (CNN) to extract features and segment the preprocessed raw lung image, and to optimize the preliminary segmentation results based on medical knowledge graphs, ultimately generating a high-precision segmented image of the lung injury region. The technical steps include feature extraction, image segmentation, and medical knowledge graph optimization.

[0093] In the feature extraction stage, deep convolutional neural networks (CNNs) extract low-level and high-level features from an image through multiple convolutional operations. Low-level features include edges, textures, and corners in the image, while high-level features include shapes, structures, and regions in the image. low Image extraction is performed using multiple convolutional layers with convolutional kernels, as shown in the formula:

[0094]

[0095] where J(m, n) is the pixel value of the processed original lung image J at position (m, n); K low (x-m, y-n) is the convolution kernel;

[0096] According to the low-level features, high-level features y high are extracted through the deep convolutional network, and the formula is:

[0097]

[0098] where y high (x, y) is the output of the high-level features;

[0099] The final output feature f includes low-level features and high-level features;

[0100] Image segmentation classifies the features to determine whether they belong to the lung injury area, and the formula is:

[0101]

[0102] where W and b are the weights and biases of the deep convolutional network, respectively, and σ is the Sigmoid activation function, which compresses the network output to [0, 1], indicating that the feature belongs to the lung injury area, is the lung injury area image segmentation result;

[0103] Medical knowledge graph optimization uses medical graph knowledge to optimize the lung injury area segmentation result and correct unreasonable segmentation areas. The medical graph provides anatomical structure, pathological information and common location of the lung injury area. According to the medical graph, the wrong areas in the segmentation result are identified and optimized by adjusting the segmentation boundary, and the final optimized segmentation image Y is output.

[0104] M3, visualization and correction module, the optimized lung injury area segmentation image is overlaid and displayed with the original lung image, the comparison result is presented to the doctor for viewing through augmented reality technology, the doctor interacts with the optimized lung injury area segmentation image through a virtual interface, and outputs the corrected lung injury area segmentation image;

[0105] The main task of the visualization and correction module is to overlay and display the segmentation result with the original image through augmented reality technology, and provide an interactive interface to allow the doctor to manually correct it. This module ensures that the segmentation result is corrected by the doctor to obtain more accurate lung injury area labeling, thereby improving the reliability of automatic segmentation. The technical steps include overlay display and interactive correction,

[0106] The optimized segmentation image Y is overlaid with the original lung image I to help the doctor visually see the difference between the segmentation result and the original image. The image content is superimposed through augmented reality technology to obtain a new image for display, so that the doctor can more clearly evaluate and adjust. The formula is:

[0107] I overlay (x,y)=α·I(x,y)+(1-α)·Y(x,y)

[0108] where I overlay (x,y) is the overlaid image displayed to the doctor, combining the original image and the segmentation image; alpha is the weighting coefficient, controlling the contribution proportion of the original image and the segmentation image, alpha∈[0,1], so as to adjust the display effect as needed;

[0109] Through the overlaid display, the doctor can visually see the difference between the segmentation result and the original image, especially in the distribution and morphology of the lung injury area, so as to help the doctor make more accurate judgments;

[0110] On the basis of the overlaid display, an interactive interface is provided to allow the doctor to modify the segmentation image through a virtual interface. The specific operation is,

[0111] The doctor selects the area to be modified through the interactive interface, and the system automatically updates the segmentation result of the area. The image is adjusted according to the doctor's operation. The modification process involves merging the boundary of the doctor's selected area with the segmentation image. The area adjusted manually by the doctor is marked as R corrected , and the modified segmentation image Y corrected The calculation formula is:

[0112]

[0113] where R corrected is the coordinate set of the doctor's manually modified area, representing the updated boundary of the lung injury area;

[0114] Through interactive modification, the doctor can adjust the segmentation result in real time to ensure that the segmentation area is more consistent with the actual injury area. This process further improves the accuracy of the automatic segmentation result, especially in complex or ambiguous boundary conditions.

[0115] M4, post-processing module, performs morphological operation on the modified lung injury area segmentation image, and smoothes the boundary of the segmentation image through interpolation algorithm, outputs the final lung injury area segmentation image,

[0116] The task of the post-processing module is to perform final optimization on the corrected segmentation image, to ensure that the final output segmentation result has higher precision and meets the standard of medical images. The key goal of post-processing is to repair any small segmentation errors, smooth the segmentation boundary, and finally provide an image that meets clinical requirements. The technical steps include morphological operations and boundary smoothing.

[0117] The purpose of morphological operation is to eliminate small holes, noise and region missing in the segmentation image, to ensure the integrity of the segmentation region. Morphological operation includes dilation and erosion operation for repairing small region error,

[0118] Dilation operation expands the boundary of the segmentation image, making the foreground region in the image larger, filling small gaps. By selecting a structure element to process the image, the center of the structure element is aligned with each pixel in the image, and the pixel value is updated to the maximum value in the neighborhood,

[0119] Erosion operation shrinks the boundary of the foreground region, removing small noise regions in the image. By selecting a structure element to process the image, the pixel value is updated to the minimum value in the neighborhood,

[0120] Dilation and erosion are used alternately to repair small region errors in the image. By first dilating and then eroding, it can effectively fill the holes in the region while removing small noise points, ensuring the continuity and integrity of the segmentation region.

[0121] The purpose of boundary smoothing is to make the boundary of the segmentation image smoother, avoiding too sharp and irregular edges. Through boundary smoothing, the contour of the segmentation image will be more natural, conforming to the actual morphology of medical images. Interpolation algorithm is used for boundary smoothing, the specific operation is,

[0122] The interpolation algorithm calculates a smooth curve to adjust the boundary by weighted average of the boundary pixels, the formula is:

[0123] Y smooth (x,y)=f(Y corrected (x,y))

[0124] Where Y smooth [x,y) is the segmentation image after boundary smoothing; f is the smoothing algorithm, which adjusts according to the boundary of the segmentation image;

[0125] Boundary smoothing ensures that the boundary in the segmentation image is more natural and smooth, removing overly sharp edges, making the contour of the lesion region more consistent with the actual morphology of medical images, enhancing the clinical applicability of the segmentation image;

[0126] The final output of the post-processing module is the lung injury area segmentation image after morphological repair and boundary smoothing, which will be saved as the final result and provided to the clinician for diagnosis and subsequent treatment.

[0127] The application discloses an automatic lung injury area segmentation method based on deep learning.

[0128] S1, receiving an original lung image from a medical image device and performing denoising, edge detection and region enhancement processing, and outputting the processed original lung image;

[0129] S2, performing multi-layer convolution operation on the processed original lung image through a deep convolutional neural network, extracting low-level and high-level features, segmenting the processed original lung image based on the low-level and high-level features, and optimizing the segmentation result by using a medical atlas to generate an optimized lung injury area segmentation image;

[0130] S3, overlapping the optimized lung injury area segmentation image with the original lung image, and the doctor performs interactive correction on the optimized lung injury area segmentation image through a virtual interface, and outputs a corrected lung injury area segmentation image

[0131] S4, performing morphological repair and boundary smoothing on the corrected lung injury area segmentation image to generate a final lung injury area segmentation image.

[0132] The application discloses an automatic lung injury area segmentation system and method based on deep learning, which combines an image processing module, a segmentation and optimization module, a visualization and correction module and a post-processing module, aims to improve the accuracy and reliability of lung injury area segmentation, first, processes the original lung image through double-sided filter denoising, Canny edge detection and region enhancement technology to improve image quality, then uses deep convolutional neural network (CNN) to segment the image and combines medical atlas for optimization to improve the accuracy of segmentation result, finally, realizes interactive correction of doctors through augmented reality technology, and uses morphological operation and smoothing processing to optimize the segmentation result, to ensure that the final output of the lung injury area segmentation image meets the medical standards.

[0133] In summary, the application has the advantages that advanced image processing technology is adopted to improve the accuracy and clarity of the segmentation result, especially in processing complex images; secondly, the segmentation result is optimized in combination with a medical atlas, so that the segmentation is more in line with clinical requirements and the practical application value is improved; thirdly, the augmented reality technology and interactive correction function are introduced to ensure that the segmentation result can be further optimized by doctors, and the reliability of the system is increased; finally, the morphological operation and boundary smoothing processing of the post-processing module ensure the smoothness and integrity of the segmented image, and finally the high-precision automatic segmentation of the lung injury area is realized, which has important clinical application prospects.

[0134] The above merely describes a specific implementation of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A deep learning-based automatic segmentation system for lung injury regions, characterized by, The system comprises: M1, an image processing module, receiving an original lung image from a medical image device, performing denoising processing, edge detection and region enhancement processing on the original lung image, and outputting a processed original lung image; M2, a segmentation and optimization module, performing multi-layer convolution operation on the processed original lung image through a deep convolutional neural network, extracting low-level and high-level features, segmenting the processed original lung image based on the low-level and high-level features, generating a lung injury region segmentation image, optimizing the lung injury region segmentation image using a medical atlas, and outputting an optimized lung injury region segmentation image; M3, a visualization and correction module, superimposing and displaying the optimized lung injury region segmentation image and the original lung image for comparison, presenting the comparison result to a doctor for viewing through augmented reality technology, and outputting a corrected lung injury region segmentation image through the doctor's interactive correction of the optimized lung injury region segmentation image through a virtual interface; The image segmentation classifies the extracted low-level and high-level features to determine whether they belong to a lung injury region, and the formula is: ； wherein, and are weights and biases of the deep convolutional network, respectively; is a sigmoid activation function that compresses the output of the deep convolutional network to between 0 and 1, indicating that the feature belongs to the lung injury region, segmenting the image into lung injury regions; are the extracted low-level and high-level features; The optimization of the medical atlas optimizes the lung injury region segmentation image using the medical atlas, corrects unreasonable segmentation regions, identifies incorrect regions in the lung injury region segmentation image according to the medical atlas, and optimizes the lung injury region segmentation image Y by adjusting the segmentation boundary; The superimposed display contrast is realized by an augmented reality technology, and the optimized lung damage region segmentation image Y is superimposed on the original lung image X to obtain a superimposed image YX. The superimposed display is performed, and the formula is: ； wherein, is the superimposed image; is the original lung image is the pixel value at position (x, y); is the weighting coefficient, which controls the contribution ratio of the original lung image and the optimized lung injury region segmentation image, adjusts the display effect as needed; The interactive correction specifically includes the doctor selecting a region that needs to be corrected through an interactive interface and adjusting the optimized lung injury region segmentation image according to the doctor's operation; The corrected lung lesion region segmentation image The calculation formula is: wherein, is a set of coordinates of the physician's manual correction region, representing the updated boundary of the lung injury region segmentation image; M4, a post-processing module, performing morphological operations on the corrected lung injury region segmentation image and smoothing the boundary of the segmentation image through an interpolation algorithm, and outputting a final lung injury region segmentation image. 2.The deep learning-based lung injury region automatic segmentation system according to claim 1, wherein, The denoising processing is performed through bilateral filtering; the edge detection is realized by using a Canny edge detection algorithm, including gradient calculation, gradient amplitude and gradient direction, The formula for calculating the gradient is: ； wherein, is the original lung image a gradient in the horizontal direction; is the original lung image a gradient in the vertical direction; The formula for calculating the gradient amplitude is: ； wherein, is the gradient magnitude, i.e. the degree of sharpness of the edge; The formula for calculating the gradient direction is: ； wherein, is the gradient direction, representing the direction of the edge of the original lung image. 3.The deep learning-based lung injury region automatic segmentation system according to claim 2, characterized in that, The region enhancement includes calculation of local mean and standard deviation, and the formula is: ； ； wherein, is a neighborhood region centered at pixel (x, y); is the mean value of the region , i.e. the average value of all pixels of the region; is the standard deviation of the region ; is the pixel value of a pixel within the neighborhood; According to the local mean and standard deviation, the enhancement value of the pixel of the original lung image is calculated; According to the denoising processing, edge detection and region enhancement, the processed original lung image J is outputted. 4.The deep learning-based lung injury region automatic segmentation system according to claim 3, characterized in that, The operation steps of the segmentation and optimization module include feature extraction, image segmentation and optimization of the medical atlas, The feature extraction uses a deep convolutional neural network to extract low-level and high-level features in the processed original lung image through multi-layer convolution operation, the low-level features including edges, textures and corner points, and the high-level features including shapes, structures and regions The low-level and high-level features are extracted through multi-layer convolution of the processed original lung image by a convolution kernel According to the low-level features, high-level features are extracted by a deep convolutional network ; The low-level features and high-level features make up output features .

5. The deep learning-based automatic lung injury region segmentation system according to claim 4, wherein, The morphological operation includes dilation operation and erosion operation, the dilation operation expands the boundary of the corrected lung injury region segmentation image, enlarges the foreground region, and fills small holes, The erosion operation shrinks the boundary of the foreground region of the corrected lung injury region segmentation image, and removes small noise regions, The smoothing processing is realized through an interpolation algorithm, the interpolation algorithm calculates a smooth curve by weighting and averaging the boundary pixels to adjust the boundary, and the formula is: ； wherein, is the smoothed segmented image; is a smoothing algorithm adjusted according to the boundaries of the segmented image; By the morphological operation and the smoothing processing, an output final lung injury region segmentation image .

6. The method for automatic segmentation of lung injury regions based on deep learning, characterized in that, The method comprises the following steps: S1, receiving a raw lung image from a medical image device and performing denoising, edge detection and region enhancement processing, and outputting a processed raw lung image; S2, performing multi-layer convolution operation on the processed raw lung image through a deep convolutional neural network, extracting low-level and high-level features, segmenting the processed raw lung image based on the low-level and high-level features, and optimizing the segmentation result by using a medical atlas to generate an optimized lung injury region segmentation image; S3, overlapping the optimized lung injury region segmentation image and the raw lung image for display, and a doctor performing interactive correction on the optimized lung injury region segmentation image through a virtual interface, and outputting a corrected lung injury region segmentation image; The image segmentation is classified by the extracted low-level and high-level features to determine whether it belongs to the lung injury region, and the formula is: ； wherein, and are weights and biases of the deep convolutional network, respectively; is a sigmoid activation function used to compress the output of the deep convolutional network to between 0 and 1, indicating that the feature belongs to the lung injury region, segmenting the image into lung injury regions; are the extracted low-level and high-level features; The optimization of the medical atlas is to optimize the lung injury region segmentation image by using the medical atlas, correct unreasonable segmentation regions, identify the error regions in the lung injury region segmentation image according to the medical atlas, and optimize by adjusting the segmentation boundary to output the optimized lung injury region segmentation image Y; The superimposed display contrast is realized by an augmented reality technology, and the optimized lung damage region segmentation image Y is superimposed on the original lung image X to obtain a superimposed image YX. The superimposed display is performed, and the formula is: ； wherein, is the superimposed image; is a weighting coefficient, which controls the contribution proportion of the original lung image and the optimized lung injury region segmentation image, , adjusts the display effect as needed; The interactive correction specifically includes selecting the region to be corrected by the doctor through the interactive interface, and adjusting the optimized lung injury region segmentation image according to the doctor's operation; The corrected lung lesion region segmentation image The calculation formula is: wherein, is a set of coordinates of the physician's manual correction region, representing the updated boundary of the lung injury region segmentation image; S4, performing morphological repair and boundary smoothing on the corrected lung injury region segmentation image to generate a final lung injury region segmentation image.

Images