A colon polyp real-time segmentation method and system fusing multi-scale high-frequency information

Abstract

The application discloses a colon polyp real-time segmentation method and system fusing multi-scale high-frequency information, and relates to the field of image processing.The method comprises the following steps: acquiring a colon polyp image in real time; inputting the colon polyp image into a trained segmentation model to obtain a segmentation result of the colon polyp image; a high-frequency boundary mining module is used to obtain high-frequency boundary guide information by using edge gradient amplitude of the colon polyp image and a dynamic modulation mechanism; and a high-frequency boundary guide fusion module group is used to fuse a plurality of scale feature maps obtained by the feature extraction module, and to enhance boundary features of the colon polyp in the plurality of scale feature maps according to the high-frequency boundary guide information.The application improves the segmentation precision of small target colon polyps, and effectively reduces the identification probability of false positives (false alarms) and false negatives (missed reports).

Description

Technical Field

[0001] This application relates to the field of image processing, and in particular to a method and system for real-time segmentation of colon polyps that integrates multi-scale high-frequency information. Background Technology

[0002] Colorectal cancer is a leading and deadly cancer worldwide, with most cases originating from colon polyps. Colonoscopy is the gold standard for screening and removing polyps, but in practice, the false negative rate is as high as 20-30%. The main reasons include physician fatigue, difficulty in observing small or flat polyps, the influence of lighting and background factors, and diagnostic subjectivity. Deep learning-based automated polyp image segmentation technology has emerged to assist physicians in real-time, objective, and accurate identification and localization of polyp regions, thereby reducing false negative rates, improving examination quality and efficiency, and contributing to the early prevention and treatment of colorectal cancer. However, limitations such as data scarcity, target complexity, environmental interference, and model generalization restrict the application of colon polyp segmentation in the field of computer vision.

[0003] In the task of segmenting colon polyps, existing methods mainly fall into two categories: attention-based optimization techniques and feature enhancement techniques based on global context modeling. The former guides the model to focus on the polyp region through attention mechanisms, offering the advantage of suppressing background interference. However, these methods rely excessively on local features, lack global modeling capabilities, respond poorly to small polyp features, and often produce incomplete segmentation results for large polyps. The latter utilizes pyramidal feature fusion to expand the receptive field and establish global image dependencies. However, it heavily relies on large-scale labeled datasets, performs poorly in small-sample scenarios, and is sensitive to noise when acquiring deep features, easily losing details. Summary of the Invention

[0004] The purpose of this application is to provide a real-time segmentation method and system for colon polyps that integrates multi-scale high-frequency information, which can improve the segmentation accuracy of small-target colon polyps and effectively reduce the identification probability of false positives (false alarms) and false negatives (false negatives). Furthermore, it can improve the computational efficiency of edge computing devices and servers.

[0005] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a method for real-time segmentation of colonic polyps that integrates multi-scale high-frequency information, including: Real-time acquisition of colon polyp images; The colon polyp image is input into a trained segmentation model to obtain the segmentation result of the colon polyp image; the segmentation model includes a feature extraction module, a high-frequency boundary mining module, and a high-frequency boundary guided fusion module group; the high-frequency boundary guided fusion module group is connected to the feature extraction module and the high-frequency boundary mining module respectively; The high-frequency boundary mining module is used to obtain high-frequency boundary guidance information by using the edge gradient amplitude and dynamic modulation mechanism of the colon polyp image; the high-frequency boundary guidance fusion module group is used to fuse multiple scale feature maps obtained by the feature extraction module, and enhance the boundary features of colon polyps in multiple scale feature maps according to the high-frequency boundary guidance information.

[0006] Secondly, this application provides a real-time segmentation system for colonic polyps that integrates multi-scale high-frequency information, including: The colon polyp image acquisition module is used to acquire colon polyp images in real time. A segmentation result acquisition module for colon polyp images is used to input the colon polyp image into a trained segmentation model to obtain the segmentation result of the colon polyp image; the segmentation model includes a feature extraction module, a high-frequency boundary mining module, and a high-frequency boundary guided fusion module group; the high-frequency boundary guided fusion module group is connected to the feature extraction module and the high-frequency boundary mining module respectively; The high-frequency boundary mining module is used to obtain high-frequency boundary guidance information by using the edge gradient amplitude and dynamic modulation mechanism of the colon polyp image; the high-frequency boundary guidance fusion module group is used to fuse multiple scale feature maps obtained by the feature extraction module, and enhance the boundary features of colon polyps in multiple scale feature maps according to the high-frequency boundary guidance information.

[0007] According to the specific embodiments provided in this application, this application has the following technical effects: This application provides a method and system for real-time segmentation of colon polyps by fusing multi-scale high-frequency information. It enhances the edge information of colon polyps by dynamically modulating high-frequency boundary guidance information, thereby improving the structure perception capability during the multi-scale feature fusion process in the high-frequency boundary guidance fusion module. This application effectively reduces the probability of false positives and false negatives, improves the segmentation accuracy of small-target colon polyps, and thus enables rapid localization and accurate segmentation of target colon polyps. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is an application environment diagram of a real-time segmentation method for colon polyps that integrates multi-scale high-frequency information according to an embodiment of this application. Figure 2A flowchart illustrating a real-time segmentation method for colon polyps that integrates multi-scale high-frequency information, provided as an embodiment of this application; Figure 3 A schematic diagram of a segmentation model structure for a real-time segmentation method for colon polyps that integrates multi-scale high-frequency information is provided in one embodiment of this application; Figure 4 A schematic diagram of the high-frequency boundary mining module structure of a real-time segmentation method for colon polyps that integrates multi-scale high-frequency information, provided in an embodiment of this application; Figure 5 A schematic diagram of the high-frequency boundary-guided fusion module structure of a real-time segmentation method for colon polyps that integrates multi-scale high-frequency information, provided in an embodiment of this application; Figure 6 A schematic diagram of the functional modules of a real-time colon polyp segmentation system that integrates multi-scale high-frequency information, provided in another embodiment of this application; Figure 7 A schematic diagram of the segmentation results of the colon polyp image obtained in this application; Figure 8 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0010] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0011] To make the objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0012] The real-time segmentation method for colonic polyps that integrates multi-scale high-frequency information provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be set up independently, integrated into server 104, or placed in the cloud or on another server. Terminal 102 can send acquired colon polyp images to server 104, and server 104 inputs the colon polyp images into a trained segmentation model to obtain the segmentation results of the colon polyp images. Server 104 can then provide feedback on the obtained colon polyp image segmentation results to terminal 102.

[0013] The terminal 102 can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, and smart in-vehicle devices. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted devices. The server 104 can be implemented using a standalone server or a server cluster composed of multiple servers, or it can be a cloud server.

[0014] In one exemplary embodiment, such as Figure 2 As shown, a method for real-time segmentation of colonic polyps by fusing multi-scale high-frequency information is provided. This method is executed by a computer device, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method is applied to... Figure 1 Taking server 104 as an example, the explanation includes the following steps 201 to 202. Wherein: Step 201: Acquire images of colon polyps in real time.

[0015] Step 202: Input the colon polyp image into the trained segmentation model to obtain the segmentation result of the colon polyp image. The segmentation model includes a feature extraction module, a high-frequency boundary mining module, and a high-frequency boundary guided fusion module group; the high-frequency boundary guided fusion module group is connected to the feature extraction module and the high-frequency boundary mining module, respectively.

[0016] The high-frequency boundary mining module is used to obtain high-frequency boundary guidance information by using the edge gradient amplitude and dynamic modulation mechanism of the colon polyp image; the high-frequency boundary guidance fusion module group is used to fuse multiple scale feature maps obtained by the feature extraction module, and enhance the boundary features of colon polyps in multiple scale feature maps according to the high-frequency boundary guidance information.

[0017] By implementing steps 201 to 202 above, this application can enhance the structure perception capability of the subsequent high-frequency boundary-guided fusion module during the fusion process through high-frequency boundary guidance information, reduce false positive and false negative predictions, and improve the segmentation accuracy of small target colon polyps.

[0018] In another exemplary embodiment of this application, step 202 is replaced by steps 301 to 303: Step 301: Input the colon polyp image into the high-frequency boundary mining module to obtain high-frequency boundary guidance information.

[0019] Step 302: Input the colon polyp image into the feature extraction module to obtain multiple scale feature maps. For example... Figure 3As shown, the feature extraction module employs the Pyramid VisionTransformer v2 (PVT-v2) model, extracting image features at four levels in stages and processing them through convolutional units to obtain first-scale feature maps, second-scale feature maps, third-scale feature maps, and fourth-scale feature maps. The first-scale and second-scale feature maps are shallow features, focusing on low-level visual information such as edges and simple textures. The third-scale and fourth-scale feature maps capture high-level semantic representations such as polyp morphology and complex texture patterns. Specifically, this application inputs a colon polyp image with height H, width W, and three color channels into the PVT-v2 model, which undergoes block embedding and coding block processing to obtain four-level images. These four-level images are then downsampled through convolutional units to obtain the first-scale feature map. Second-scale feature map Third-scale feature map and fourth-scale feature map . , , and The number of channels corresponding to feature maps of different scales.

[0020] Step 303: Based on the high-frequency boundary guidance information and the multiple scale feature maps, the high-frequency boundary guidance fusion module group is used to obtain the segmentation result of the colon polyp image.

[0021] In another exemplary embodiment of this application, such as Figure 4 As shown, step 301 above is replaced by steps 401 to 402: Step 401: Input the colon polyp image into the bidirectional gradient calculation unit to obtain the normalized edge gradient magnitude.

[0022] Step 402: Based on the normalized edge gradient magnitude and the multiple scale feature maps, a dynamic modulation mechanism is used to obtain high-frequency boundary guidance information.

[0023] In another exemplary embodiment of this application, a colon polyp image is used as input to the bidirectional gradient calculation unit, where the colon polyp image is converted to grayscale to obtain a grayscale image of the colon polyp. To reduce computational complexity, parallel extraction The horizontal and vertical edge gradient magnitudes are calculated, the original edge gradient magnitudes are normalized to obtain the normalized edge gradient magnitudes. The bidirectional gradient calculation unit uses the following formula to obtain the normalized edge gradient magnitude: .

[0024] .

[0025] .

[0026] in, For the vertical Sobel convolution operator, For the horizontal Sobel convolution operator, This is a grayscale image of a colon polyp. This represents the magnitude of the horizontal edge gradient. The magnitude of the vertical edge gradient. This represents the original edge gradient magnitude. The normalized edge gradient magnitude. It represents a minute constant.

[0027] In another exemplary embodiment of this application, the multiple scale feature maps obtained by the feature extraction module and the normalized edge gradient magnitudes are combined. As input to the dynamic modulation mechanism, a convolution operation is performed on the weighted combination of features to reduce the feature size, reduce the number of channels to 1, and obtain the hyperparameters. Using hyperparameters normalized edge gradient magnitude High-frequency boundary guidance information is obtained by performing nonlinear transformation. The purpose of this operation is to utilize hyperparameters. right Dynamic modulation is used to enhance the edges of polyps while filtering out irrelevant interference caused by endoscopic imaging. The dynamic modulation mechanism can be used to obtain high-frequency boundary guidance information using the following formula: .

[0028] .

[0029] .

[0030] .

[0031] in, As the first intermediate feature, As the second intermediate feature, It is the third intermediate feature. It is the fourth intermediate feature. This is the first learnable projection matrix. This is the second learnable projection matrix. This is the third learnable projection matrix. This is the fourth learnable projection matrix. The fifth learnable projection matrix, For the first bias term, For the second bias term, For the third bias term, This is the fourth bias term. This is the fifth bias term. For splicing operations, As the first activation function, For the first Scale feature map This is the scale number. For the second activation function, For high-frequency boundary guidance information, The normalized edge gradient magnitude. This is a hyperparameter.

[0032] In another exemplary embodiment of this application, the plurality of scale feature maps include a first scale feature map, a second scale feature map, a third scale feature map, and a fourth scale feature map; the high-frequency boundary guided fusion module group includes a first high-frequency boundary guided fusion module, a second high-frequency boundary guided fusion module, and a third high-frequency boundary guided fusion module.

[0033] like Figure 3 As shown, step 303 above is replaced by steps 501 to 503: Step 501: Based on the high-frequency boundary guidance information, the third-scale feature map, and the fourth-scale feature map, the third high-frequency boundary guidance fusion module is used to obtain the third-scale fusion feature map.

[0034] Step 502: Based on the third-scale fusion feature map, the high-frequency boundary guidance information, and the second-scale feature map, the second high-frequency boundary guidance fusion module is used to obtain the second-scale fusion feature map.

[0035] Step 503: Based on the second scale fusion feature map, the high-frequency boundary guidance information, and the first scale feature map, the first high-frequency boundary guidance fusion module is used to obtain the first scale fusion feature map; the first scale fusion feature map is the segmentation result of the colon polyp image.

[0036] In another exemplary embodiment of this application, high-frequency boundary guidance information is used as input to the high-frequency boundary guidance fusion module to enhance the structure perception capability of the subsequent high-frequency boundary guidance fusion module in the process of fusing feature maps of multiple scales. The purpose is to reduce false positive and false negative predictions and improve the segmentation accuracy of colon polyp segmentation boundaries.

[0037] like Figure 5As shown, the third high-frequency boundary guided fusion module, the second high-frequency boundary guided fusion module, and the first high-frequency boundary guided fusion module all use the following formulas to obtain the third fusion feature map, the second fusion feature map, and the first fusion feature map: .

[0038] .

[0039] .

[0040] .

[0041] .

[0042] in, This is the scale number. For the first Scale-fused feature maps Indicates upsampling, For the second activation function, This is a reverse attention operation. This is a Gaussian filter convolution operation. For the first Scale feature map For the first The feature map obtained after the scale feature map undergoes reverse attention processing. For the first The feature map obtained after the scale feature map is processed by Gaussian filtering. For the first First-scale intermediate fused feature map. For splicing operations, For convolution operations, For high-frequency boundary guidance information, No. Scale feature map For the first Second-scale intermediate fused feature map. For the first Scale-fused feature maps For spatial attention, For channel attention.

[0043] Specifically, in this application, the output of the high-frequency boundary guided fusion module is compared with that of the previous high-frequency boundary guided fusion module. (i=2,3) or the fourth-scale feature map after convolution operation. (i=4) Features are extracted using reverse attention and Gaussian filter convolution operations respectively, with the aim of emphasizing the detailed regions that are ignored by deep networks.

[0044] Utilizing high-frequency boundary guidance information This guides the fusion of feature maps obtained after inverse attention and Gaussian filtering convolution operations, achieving detail restoration and spatial boundary localization at the feature level, thereby obtaining the first... First-scale intermediate fused feature map.

[0045] The first First-scale intermediate fusion feature map By applying spatial attention and channel attention mechanisms and reducing the number of channels to 1, we obtain... .

[0046] In another exemplary embodiment of this application, a loss function is used to supervise the training of the third-scale fusion feature map, the second-scale fusion feature map, and the first-scale fusion feature map output by the high-frequency boundary-guided fusion module. The training process of the segmentation model specifically includes: Using the third-scale fusion feature map, the second-scale fusion feature map, and the first-scale fusion feature map as predicted values, and the colon polyp image as the ground truth, the loss value of the segmentation model is obtained based on the predicted values ​​and the ground truth.

[0047] The parameters of the segmentation model are adjusted based on the loss value of the segmentation model to obtain a trained segmentation model.

[0048] In another exemplary embodiment of this application, the loss value of the segmentation model is calculated using the following formula: .

[0049] .

[0050] .

[0051] in, The loss value of the segmentation model is... This is the scale number. As the first weighting coefficient, This is the second weighting coefficient. The binary cross-entropy of a pixel aims to differentiate between predicted and true values, emphasizing accurate individual classification for each pixel. To measure the degree of overlap between predicted and actual values, the similarity of the entire set is determined. For the first Scale-fused feature maps For the true value, This represents the total number of pixels in a single image. This represents the predicted probability value for the j-th pixel. This represents the true label of the j-th pixel. It is a small constant.

[0052] Based on the same inventive concept, as shown in Figure 6, this application also provides a real-time segmentation system for colonic polyps that integrates multi-scale high-frequency information, the system comprising: The colon polyp image acquisition module 601 is used to acquire colon polyp images in real time.

[0053] The colon polyp image segmentation result acquisition module 602 is used to input the colon polyp image into a trained segmentation model to obtain the segmentation result of the colon polyp image; the segmentation model includes a feature extraction module, a high-frequency boundary mining module, and a high-frequency boundary guided fusion module group; the high-frequency boundary guided fusion module group is connected to the feature extraction module and the high-frequency boundary mining module respectively.

[0054] The high-frequency boundary mining module is used to obtain high-frequency boundary guidance information by using the edge gradient amplitude and dynamic modulation mechanism of the colon polyp image; the high-frequency boundary guidance fusion module group is used to fuse multiple scale feature maps obtained by the feature extraction module, and enhance the boundary features of colon polyps in multiple scale feature maps according to the high-frequency boundary guidance information.

[0055] This application proposes the first method for colon polyp segmentation utilizing multi-scale high-frequency (traditionally considered noise in medical imaging) edge information. Specifically, it employs a bidirectional gradient computation unit and a dynamic modulation mechanism to selectively enhance diagnostically relevant boundary features while suppressing noise. Based on a high-frequency boundary-guided fusion module group, it uses high-frequency boundary guidance information to guide the fusion of feature maps at multiple scales, thereby reducing false positive and false negative predictions. Furthermore, extensive experimental results demonstrate that the proposed method outperforms most existing models on several challenging datasets, achieving new state-of-the-art performance in colon polyp segmentation tasks, such as… Figure 7 The diagram shows the segmentation result of a colon polyp image obtained using this application. Furthermore, this application significantly improves the segmentation effect of colon polyp images while maintaining low parameter and computational costs, thereby quickly locating and accurately segmenting the target polyp, achieving the best results among current methods in this field.

[0056] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 8As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores images of colon polyps. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a real-time colon polyp segmentation method that integrates multi-scale, high-frequency information.

[0057] Those skilled in the art will understand that Figure 8 The structures shown are merely block diagrams of some structures related to the present application and do not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0058] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0059] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0060] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application.

[0061] In conclusion, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for real-time segmentation of colonic polyps that integrates multi-scale high-frequency information, characterized in that, The method includes: Real-time acquisition of colon polyp images; The colon polyp image is input into a trained segmentation model to obtain the segmentation result of the colon polyp image; the segmentation model includes a feature extraction module, a high-frequency boundary mining module, and a high-frequency boundary guided fusion module group; the high-frequency boundary guided fusion module group is connected to the feature extraction module and the high-frequency boundary mining module respectively; The high-frequency boundary mining module is used to obtain high-frequency boundary guidance information by using the edge gradient magnitude and dynamic modulation mechanism of the colon polyp image; the high-frequency boundary guidance fusion module group is used to fuse multiple scale feature maps obtained by the feature extraction module, and enhance the boundary features of colon polyps in multiple scale feature maps according to the high-frequency boundary guidance information. The high-frequency boundary mining module is also connected to the feature extraction module; the colon polyp image is input into the high-frequency boundary mining module to obtain high-frequency boundary guidance information, specifically including: The colon polyp image is input into a bidirectional gradient calculation unit to obtain the normalized edge gradient magnitude; Based on the normalized edge gradient magnitude and the multiple scale feature maps, a dynamic modulation mechanism is used to obtain high-frequency boundary guidance information. The multiple scale feature maps include a first scale feature map, a second scale feature map, a third scale feature map, and a fourth scale feature map; the high-frequency boundary guidance information is obtained using the following formula: ； ； ； ； in, As the first intermediate feature, As the second intermediate feature, It is the third intermediate feature. It is the fourth intermediate feature. This is the first learnable projection matrix. This is the second learnable projection matrix. This is the third learnable projection matrix. This is the fourth learnable projection matrix. The fifth learnable projection matrix, For the first bias term, For the second bias term, For the third bias term, This is the fourth bias term. This is the fifth bias term. For splicing operations, As the first activation function, For the first Scale feature map This is the scale number. For the second activation function, For high-frequency boundary guidance information, The normalized edge gradient magnitude. For hyperparameters; The high-frequency boundary guided fusion module group includes a first high-frequency boundary guided fusion module, a second high-frequency boundary guided fusion module, and a third high-frequency boundary guided fusion module; Based on the high-frequency boundary guidance information and the multiple scale feature maps, the high-frequency boundary guidance fusion module group is used to obtain the segmentation result of the colon polyp image, specifically including: Based on the high-frequency boundary guidance information, the third-scale feature map, and the fourth-scale feature map, the third high-frequency boundary guidance fusion module is used to obtain the third-scale fusion feature map; Based on the third-scale fusion feature map, the high-frequency boundary guidance information, and the second-scale feature map, the second high-frequency boundary guidance fusion module is used to obtain the second-scale fusion feature map; Based on the second-scale fusion feature map, the high-frequency boundary guidance information, and the first-scale feature map, the first high-frequency boundary guidance fusion module is used to obtain the first-scale fusion feature map; the first-scale fusion feature map is the segmentation result of the colon polyp image.

2. The method for real-time segmentation of colonic polyps by fusing multi-scale high-frequency information according to claim 1, characterized in that, The colon polyp image is input into a trained segmentation model to obtain the segmentation result of the colon polyp image, specifically including: The colon polyp image is input into the high-frequency boundary mining module to obtain high-frequency boundary guidance information; The colon polyp image is input into the feature extraction module to obtain multiple scale feature maps; Based on the high-frequency boundary guidance information and the multiple scale feature maps, the high-frequency boundary guidance fusion module group is used to obtain the segmentation result of the colon polyp image.

3. The method for real-time segmentation of colonic polyps by fusing multi-scale high-frequency information according to claim 1, characterized in that, The normalized edge gradient magnitude is obtained using the following formula: ； ； ； in, For the vertical Sobel convolution operator, For the horizontal Sobel convolution operator, This is a grayscale image of a colon polyp. This represents the magnitude of the horizontal edge gradient. The magnitude of the vertical edge gradient. This represents the original edge gradient magnitude. The normalized edge gradient magnitude. It is a small constant.

4. The method for real-time segmentation of colonic polyps by fusing multi-scale high-frequency information according to claim 1, characterized in that, The third-scale fused feature map, the second-scale fused feature map, and the first-scale fused feature map are obtained using the following formulas: ； ； ； ； ； in, This is the scale number. For the first Scale-fused feature maps Indicates upsampling, For the second activation function, This is a reverse attention operation. This is a Gaussian filter convolution operation. For the first Scale feature map For the first The feature map obtained after the scale feature map undergoes reverse attention processing. For the first The feature map obtained after the scale feature map is processed by Gaussian filtering. For the first First-scale intermediate fused feature map. For splicing operations, For convolution operations, For high-frequency boundary guidance information, No. Scale feature map For the first Second-scale intermediate fused feature map. For the first Scale-fused feature maps For spatial attention, For channel attention.

5. The method for real-time segmentation of colonic polyps by fusing multi-scale high-frequency information according to claim 1, characterized in that, The training process of the segmentation model specifically includes: The third-scale fusion feature map, the second-scale fusion feature map, and the first-scale fusion feature map are used as predicted values, and the colon polyp image is used as the ground truth value. Based on the predicted value and the ground truth value, the loss value of the segmentation model is obtained. The parameters of the segmentation model are adjusted based on the loss value of the segmentation model to obtain a trained segmentation model.

6. The method for real-time segmentation of colonic polyps by fusing multi-scale high-frequency information according to claim 1, characterized in that, The loss value of the segmentation model is calculated using the following formula: ； ； ； in, The loss value of the segmentation model is... This is the scale number. As the first weighting coefficient, This is the second weighting coefficient. The binary cross-entropy of pixels. The degree of overlap between predicted and actual values, For the first Scale-fused feature maps For the true value, This represents the total number of pixels in a single image. This represents the predicted probability value for the j-th pixel. This represents the true label of the j-th pixel. It is a small constant.

7. A real-time segmentation system for colonic polyps that integrates multi-scale high-frequency information, employing the real-time segmentation method for colonic polyps that integrates multi-scale high-frequency information as described in any one of claims 1-6, wherein the system comprises: The colon polyp image acquisition module is used to acquire colon polyp images in real time. A segmentation result acquisition module for colon polyp images is used to input the colon polyp image into a trained segmentation model to obtain the segmentation result of the colon polyp image; the segmentation model includes a feature extraction module, a high-frequency boundary mining module, and a high-frequency boundary guided fusion module group; the high-frequency boundary guided fusion module group is connected to the feature extraction module and the high-frequency boundary mining module respectively; The high-frequency boundary mining module is used to obtain high-frequency boundary guidance information by using the edge gradient amplitude and dynamic modulation mechanism of the colon polyp image; the high-frequency boundary guidance fusion module group is used to fuse multiple scale feature maps obtained by the feature extraction module, and enhance the boundary features of colon polyps in multiple scale feature maps according to the high-frequency boundary guidance information.

Images