SPR image optimization processing method based on image segmentation and edge enhancement

Abstract

The invention discloses an SPR (Surface Plasmon Resonance) image optimization processing method based on image segmentation and edge enhancement, which comprises the following steps: acquiring SPR image data, and preprocessing to generate standardized SPR image data; inputting a structure boundary extraction model constructed based on CGNet, generating a structure boundary label graph, and aligning the structure boundary label graph with the image; gradient amplitude and local entropy mutation detection is executed, and an artifact guide graph is generated; respectively inputting the standardized image into details and context branches of the improved CSDNet, and extracting edge and semantic feature maps; inputting a guide perception gating module, executing structure enhancement and artifact suppression fusion, and generating a fusion feature map; inputting into a multi-scale detail recovery module, and outputting an edge enhanced image; and executing structural similarity and marginal definition scoring based on the original image and the enhanced image, and generating an optimization result. According to the method, synchronous optimization of SPR image edge enhancement and structure maintenance is realized, and the image definition and diagnosis availability are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to an SPR image optimization processing method based on image segmentation and edge enhancement. Background Technology

[0002] In the process of medical image-guided therapy and radiotherapy path planning, SPR images, as the result of volume data compression projection, serve as auxiliary localization and anatomical structure references. In existing technologies, SPR images are usually improved in visual quality through image reconstruction or image enhancement methods, including the use of super-resolution reconstruction networks, edge enhancement filtering algorithms, and feature fusion models based on attention mechanisms. Among these, deep convolutional neural networks such as U-Net, ResNet, and DenseNet are often used for structure extraction and resolution enhancement of SPR images to enhance the visibility of organ boundaries and tissue contours.

[0003] Existing image enhancement methods have significant limitations when processing SPR images: First, the original edges of SPR images are blurred and have low contrast, with structural boundaries and artifact regions overlapping. Traditional enhancement algorithms, while improving image clarity, are prone to false enhancement of non-structural regions, interfering with subsequent diagnosis and localization. Second, existing methods lack effective structure-aware mechanisms and cannot differentiate between anatomical and artifact regions, resulting in weakening of important boundary details or amplification of artifact signals in the enhancement results. Third, fusion strategies generally employ simple stitching or channel weighting methods, without introducing a fusion path control mechanism based on structural labels and artifact guidance information, leading to insufficient edge information recovery and a lack of image quality controllability and structural discrimination capabilities.

[0004] In summary, existing methods have significant shortcomings in SPR image boundary extraction, structure-guided enhancement, and artifact recognition and control, failing to meet the requirements for high-precision edge reconstruction and structure preservation. Therefore, providing an SPR image optimization processing method based on image segmentation and edge enhancement is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose an SPR image optimization processing method based on image segmentation and edge enhancement. This invention utilizes CGNet to extract structural boundary label maps and combines CSDNet with a guided perception gating module to achieve deep fusion of structural enhancement and artifact suppression, which has the advantages of high edge clarity, strong structure preservation, and less artifact interference.

[0006] An SPR image optimization processing method based on image segmentation and edge enhancement according to an embodiment of the present invention includes the following steps: Acquire SPR image data and preprocess it to generate standardized SPR image data. Input the standardized SPR image data into a structural boundary extraction model built on the CGNet architecture to extract anatomical region boundary response maps and generate structural boundary label maps. Align the structural boundary label maps with the standardized SPR image data in the pixel dimension. Perform gradient direction magnitude analysis and local entropy mutation detection based on the standardized SPR image data to locate non-structural mutation regions in the image and generate artifact guide maps. Input the standardized SPR image data into the detail branch and context branch of the improved CSDNet, respectively. The detail branch extracts local edge change feature maps, and the context branch extracts local edge change feature maps. The system extracts multi-scale semantic context feature maps and outputs detail feature maps and context feature maps. The detail feature maps, context feature maps, structural boundary label maps, and artifact guidance maps are input into the guided perception gating module to perform a gated weighted fusion operation based on structural enhancement and artifact suppression, generating a fused feature map. This fused feature map is then input into a multi-scale detail recovery module containing upsampled structures and skip connection units to perform edge region reconstruction and feature restoration, outputting SPR image edge enhancement image data. Based on the standardized SPR image data and the SPR image edge enhancement image data, structural similarity analysis and edge sharpness scoring are performed to generate SPR image optimization results.

[0007] Optionally, the preprocessing process includes performing size normalization, grayscale normalization, and background noise suppression.

[0008] Optionally, the process of inputting standardized SPR image data into a structural boundary extraction model constructed based on the CGNet structure, extracting anatomical region boundary response maps, and generating structural boundary label maps includes: A shallow convolutional backbone network is used to perform preliminary feature extraction on standardized SPR image data. Low-level gradient features, texture features and local contrast features of the image are extracted through two consecutive convolutional and normalization operations to generate an initial local feature map. Based on the initial local feature map, several context guidance modules are introduced. Each context guidance module contains a local perception channel and a context modeling channel. The local perception channel uses a 3×3 convolution kernel to extract the edge direction change and gray-level jump features of the pixel neighborhood. The context modeling channel constructs the context information map of the structural region through global pooling and 1×1 convolution to capture the edge consistency features inside the organization region. Dilated convolutions with dilation rates of 2 and 4 are applied to feature maps at different scales within the context modeling channel to construct a multi-scale receptive field. Channel attention weighted fusion is then performed between the local channel and the context channel to generate context-enhanced feature maps. The context-enhanced feature map is input into the boundary response prediction unit. While keeping the image size unchanged, the prediction probability of each pixel being a boundary point is calculated by pixel-wise 1×1 convolution, and the boundary response map is output. A pixel-by-pixel classification operation is performed on the boundary response map. Pixels above the set boundary probability threshold are classified as boundary pixels, and pixels below the threshold are classified as non-boundary pixels. The structural boundary label map is formed by combining the region area, boundary continuity and gradient magnitude distribution. The structural boundary label map is resampled using bilinear interpolation and mapped to pixel size and coordinate system that are completely consistent with the standardized SPR image data.

[0009] Optionally, the generation of the artifact guide map includes: Gradient direction calculation is performed on the standardized SPR image data. The Sobel operator is used to calculate the gradient magnitude maps in the horizontal and vertical directions respectively, and a primary gradient image set containing gradient direction magnitude information is constructed. In the primary gradient image set, calculate the gradient variance and direction change rate of each pixel within the local window, identify regions with abrupt gradient direction changes, and generate candidate artifact region response maps. Local entropy analysis is performed on standardized SPR image data. The distribution of local entropy values ​​in each pixel region is statistically analyzed by sliding a fixed window to locate regions of abrupt entropy changes and generate a local entropy change map. The candidate artifact region response map and the local entropy mutation map are normalized pixel by pixel and superimposed. The joint response map is calculated by weighted fusion to generate a fusion intensity map. The fusion weight is dynamically adjusted according to the gradient fluctuation and entropy mutation degree of each region. A hierarchical threshold segmentation is performed on the fused intensity map. Regions exceeding the set upper threshold are marked as high-intensity artifact regions, and regions below the set lower threshold are marked as background regions. An initial artifact guide map is constructed, and boundary closure and noise region filtering operations are performed on the initial artifact guide map to generate the final artifact guide map, which maintains a one-to-one correspondence with the standardized SPR image data in the pixel dimension.

[0010] Optionally, the improved CSDNet structure includes detail branches and context branches, and has the following characteristics: The detail branch consists of three consecutive depthwise separable convolutional modules. Each layer contains a standard 3×3 depthwise convolution, a pointwise convolution, and a batch normalization operation, which outputs a detail feature map. The context branch consists of four layers of pyramid-shaped semantic coding units. Each layer contains a 5×5 dilated convolution kernel and a channel attention enhancement module. The dilated structure with dilation rates of 2, 4, 6 and 8 is used to achieve multi-scale context expansion of the receptive field. A cross-layer feature residual connection structure is adopted between the outputs of each level of the context branch to enhance the semantic consistency expression capability. By fusing the semantic feature maps of the current layer and the previous layer, a deep context-aware map is formed. The detail branch and the context branch output detail feature maps and context feature maps respectively at the end. The feature dimensions are aligned by a linear transformation with a unified number of channels, and then input into the guided perception gating module.

[0011] Optionally, the guidance sensing gating module has the following structure and operation process: The system receives detail feature maps, context feature maps, structural boundary label maps, and artifact guidance maps as input data, and performs channel dimension expansion and spatial dimension standardization processing respectively to construct an input feature set with a uniform size and number of channels. The guided perception gating module includes a structure enhancement gating unit and an artifact suppression gating unit. The structure enhancement gating unit uses the structure boundary label map as a guide and generates a structure enhancement weight map using pixel-wise multiplication operations, and applies the structure enhancement weight map to the detail feature map. The artifact suppression gating unit uses the artifact guide map as a guide, and uses inverse normalization and the Sigmoid gating function to generate a suppression mask map. The suppression mask map is then multiplied pixel by pixel with the context feature map to suppress feature interference from unstructured regions. After the structural enhancement weight map and the suppression mask map are applied, channel-dimensional weighted fusion is performed on the enhanced detail feature map and the suppression context feature map. The fusion weights are dynamically generated based on the structural region confidence map, and the fused feature map is output. Before output, the fused feature map is superimposed with the shallow feature map of the original normalized SPR image data through residual connections to maintain the consistency between the fused result and the input image structure.

[0012] Optionally, the multi-scale detail restoration module includes two upsampling units and one skip connection unit. Each upsampling unit uses bilinear interpolation and 3×3 convolution to jointly construct a multi-scale feature upsampling image. The skip connection unit connects the intermediate layer feature map of the CSDNet detail branch with the upsampling feature map of the corresponding scale. It compensates for edge detail information through channel alignment and element-wise addition fusion, and outputs edge-enhanced image data with the same resolution as the standardized SPR image data.

[0013] Optionally, the structural similarity analysis includes: using standardized SPR image data as the reference image and SPR image edge enhancement image data as the image to be evaluated, performing a fixed window sliding calculation operation in a unified resolution space, extracting the mean brightness, standard deviation of contrast, and structural gradient vector within each window, calculating the brightness similarity, contrast similarity, and structural similarity components accordingly, and generating a structural similarity index score through weighted fusion for analysis.

[0014] Optionally, the edge sharpness score is characterized by using SPR image edge enhancement image data as input image, extracting edge contour response map using multi-scale Laplacian edge response function, constructing a comprehensive scoring function based on gradient magnitude distribution of edge region, edge length continuity and structural boundary label map matching degree, and outputting edge sharpness score result.

[0015] Optionally, the SPR image optimization result is generated based on the structural similarity index score map and the edge sharpness score map. A joint scoring function is used to comprehensively evaluate the structural preservation and edge recognizability of the edge-enhanced image data. A preset scoring threshold is used to determine whether the fused feature map meets the image quality requirements. When the scoring result meets both the structural consistency requirement and the edge sharpness requirement, the SPR image optimization result is output. Otherwise, the parameters of the structural enhancement weight and artifact suppression factor in the guided perception gating module are updated, and the fusion and reconstruction process is re-executed.

[0016] The beneficial effects of this invention are: (1) Achieve accurate extraction of anatomical structure boundaries: This invention adopts a structural boundary extraction model based on CGNet structure to extract boundary response maps and generate structural boundary label maps on standardized SPR image data, which effectively enhances the spatial positioning capability of boundary recognition and provides clear structural guidance for feature fusion.

[0017] (2) Improve the selectivity and stability of the edge enhancement process: This invention introduces an artifact guidance map, which is combined with detail feature map and context feature map input guidance perception gating module to dynamically adjust the structure enhancement factor and artifact suppression factor in the fusion path, thereby enhancing the effective edge region and suppressing the non-structure interference region, which can improve the clarity and credibility of the enhancement result.

[0018] (3) Enhance the quality controllability and structural preservation of edge optimization results: This invention constructs a joint quality assessment mechanism based on structural similarity analysis and edge sharpness scoring, and outputs SPR image optimization results in combination with the scoring results. It has end-to-end closed-loop control capability, ensuring that the image enhancement process maintains the original structural information while improving edge readability. Attached Figure Description

[0019] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is an overall flowchart of an SPR image optimization processing method based on image segmentation and edge enhancement proposed in this invention; Figure 2 This is a schematic diagram of the network structure of the structural boundary extraction model based on the CGNet structure, which is the SPR image optimization processing method based on image segmentation and edge enhancement proposed in this invention. Figure 3 This is a schematic diagram of the feature extraction paths of the detail branch and the context branch in the improved CSDNet structure of the SPR image optimization processing method based on image segmentation and edge enhancement proposed in this invention. Detailed Implementation

[0020] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0021] refer to Figure 1-3 A method for SPR image optimization processing based on image segmentation and edge enhancement includes the following steps: Acquire SPR image data and preprocess it to generate standardized SPR image data. Input the standardized SPR image data into a structural boundary extraction model built on the CGNet architecture to extract anatomical region boundary response maps and generate structural boundary label maps. Align the structural boundary label maps with the standardized SPR image data in the pixel dimension. Perform gradient direction magnitude analysis and local entropy mutation detection based on the standardized SPR image data to locate non-structural mutation regions in the image and generate artifact guide maps. Input the standardized SPR image data into the detail branch and context branch of the improved CSDNet, respectively. The detail branch extracts local edge change feature maps, and the context branch extracts local edge change feature maps. The system extracts multi-scale semantic context feature maps and outputs detail feature maps and context feature maps. The detail feature maps, context feature maps, structural boundary label maps, and artifact guidance maps are input into the guided perception gating module to perform a gated weighted fusion operation based on structural enhancement and artifact suppression, generating a fused feature map. This fused feature map is then input into a multi-scale detail recovery module containing upsampled structures and skip connection units to perform edge region reconstruction and feature restoration, outputting SPR image edge enhancement image data. Based on the standardized SPR image data and the SPR image edge enhancement image data, structural similarity analysis and edge sharpness scoring are performed to generate SPR image optimization results.

[0022] In this embodiment, the preprocessing process includes performing size normalization, grayscale normalization, and background noise suppression.

[0023] In this embodiment, the process of inputting standardized SPR image data into a structural boundary extraction model constructed based on the CGNet structure, extracting anatomical region boundary response maps, and generating structural boundary label maps includes: A shallow convolutional backbone network is used to perform preliminary feature extraction on standardized SPR image data. Low-level gradient features, texture features and local contrast features of the image are extracted through two consecutive convolutional and normalization operations to generate an initial local feature map. Based on the initial local feature map, several context guidance modules are introduced. Each context guidance module contains a local perception channel and a context modeling channel. The local perception channel uses a 3×3 convolution kernel to extract the edge direction change and gray-level jump features of the pixel neighborhood. The context modeling channel constructs the context information map of the structural region through global pooling and 1×1 convolution to capture the edge consistency features inside the organization region. Dilated convolutions with dilation rates of 2 and 4 are applied to feature maps at different scales within the context modeling channel to construct a multi-scale receptive field, enhancing the perception of structural boundaries at different organizational scales. Channel attention weighted fusion is performed between the local channel and the context channel to generate context-enhanced feature maps. The context-enhanced feature map is input into the boundary response prediction unit. While keeping the image size unchanged, the prediction probability of each pixel being a boundary point is calculated by pixel-wise 1×1 convolution. The boundary response map is then output, and the prediction probability is a real value between 0 and 1. A pixel-by-pixel classification operation is performed on the boundary response map. Pixels above the set boundary probability threshold are classified as boundary pixels, and pixels below the threshold are classified as non-boundary pixels. The structural boundary label map is formed by combining the region area, boundary continuity and gradient magnitude distribution. The structural boundary label map is resampled using bilinear interpolation and mapped to the same pixel size and coordinate system as the standardized SPR image data, so that the structural boundary label map and the standardized SPR image data are aligned point by point in the pixel dimension.

[0024] In this embodiment, the generation of the artifact guide map includes: Gradient direction calculation is performed on the standardized SPR image data. The Sobel operator is used to calculate the gradient magnitude maps in the horizontal and vertical directions respectively, and a primary gradient image set containing gradient direction magnitude information is constructed. In the primary gradient image set, calculate the gradient variance and direction change rate of each pixel within the local window, identify regions with abrupt gradient direction changes, and generate candidate artifact region response maps. Local entropy analysis is performed on standardized SPR image data. The distribution of local entropy values ​​in each pixel region is statistically analyzed by sliding a fixed window to locate regions of abrupt entropy changes and generate a local entropy change map. The candidate artifact region response map and the local entropy mutation map are pixel-by-pixel normalized and superimposed. A joint response map is calculated using a weighted fusion method to generate a fusion intensity map. The fusion weights are dynamically adjusted based on the gradient volatility and entropy mutation degree of each region. Specifically, during the artifact guidance map construction process, to improve the ability of the fusion intensity map to distinguish non-structural mutation regions, a dynamic weight adjustment mechanism based on gradient volatility and entropy mutation response values ​​is adopted when fusing the candidate artifact region response map and the local entropy mutation map. For each pixel position, the directional gradient magnitude of the pixel position in the candidate artifact region response map is calculated, and the standard deviation of the gradient magnitude is calculated within a fixed window centered on that pixel, defined as... The gradient fluctuation factor is used to characterize the intensity of local directional changes. Simultaneously, the entropy response value of the pixel in the local entropy mutation map is extracted, representing the grayscale complexity and information mutation degree of the region. Based on the normalization results of the gradient fluctuation factor and the entropy response value, a fusion weight allocation function is constructed, mapping the two responses to fusion weight factors according to a proportional relationship. Gradient-dominated regions are assigned higher gradient channel fusion weights, while entropy mutation-dominated regions are assigned higher entropy channel fusion weights. The calculation method of the weight factors ensures that the fusion weights dynamically change within the overall image dimension, accurately adapting to the differences in the dependence intensity of artifact detection features on different structural regions. Finally, a fusion intensity map is generated through a weighted linear combination.

[0025] A hierarchical threshold segmentation is performed on the fused intensity map. Regions exceeding the set upper threshold are marked as high-intensity artifact regions, and regions below the set lower threshold are marked as background regions. An initial artifact guide map is constructed, and boundary closure and noise region filtering operations are performed on the initial artifact guide map to generate the final artifact guide map, which maintains a one-to-one correspondence with the standardized SPR image data in the pixel dimension.

[0026] In this embodiment, the improved CSDNet structure includes detail branches and context branches, and has the following characteristics: The detail branch consists of three consecutive depthwise separable convolutional modules. Each layer contains a standard 3×3 depthwise convolution, pointwise convolution, and batch normalization operations, which are used to extract edge direction change features, low-level texture response features, and local gradient abrupt region features in the SPR image, and output detail feature maps. The context branch consists of four layers of pyramid-shaped semantic coding units. Each layer contains a 5×5 dilated convolution kernel and a channel attention enhancement module. The dilated structure with dilation rates of 2, 4, 6 and 8 is used to achieve multi-scale context expansion of the receptive field. A cross-layer feature residual connection structure is adopted between the outputs of each level of the context branch to enhance the semantic consistency expression capability. By fusing the semantic feature maps of the current layer and the previous layer, a deep context-aware map is formed. The detail branch and the context branch output detail feature maps and context feature maps respectively at the end. The two are aligned in feature dimension by a linear transformation with a unified number of channels and are respectively input into the guided perception gating module for the fusion operation of structure enhancement and artifact suppression. The improved CSDNet structure adopts a shallow and lightweight architecture, avoiding fully connected layers and maintaining the feature map spatial resolution throughout the entire process. This ensures that edge details are not compressed or lost during processing. Compared to the original CSDNet model, it has undergone structural optimization in terms of feature extraction path, receptive field construction method, and edge preservation capability, resulting in the following beneficial effects: By introducing multi-layer depthwise separable convolutional modules into the detail branch, the response capability to abrupt changes in edge direction, low-contrast textures and micro-structural perturbations in SPR images is improved, the edge sensitivity of the feature extraction process is enhanced, and the number of network parameters and computational complexity are reduced.

[0027] The context branch uses multi-level dilated convolution combined with channel attention mechanism to expand contextual features, which significantly enhances the ability to model multi-scale semantics of anatomical structural regions. This enables the model to have stronger semantic consistency and regional continuity expression ability when recognizing the contours of complex organ structures, and effectively suppresses background interference.

[0028] By constructing cross-layer feature residual connection channels within the context branch, the efficiency of information transmission between different semantic levels is improved, the structural misalignment problem that may occur in deep feature expression is alleviated, and the controllability of context semantic information in terms of spatial positioning accuracy is enhanced.

[0029] The feature map maintains the original spatial resolution in both the detail path and the context path, avoiding the use of fully connected layers and spatial compression structures. This further ensures the ability to restore edge details and structural positions in the SPR image, providing higher-fidelity feature input for fusion operations.

[0030] In this embodiment, the guiding perception gating module is characterized by the following structure and operation process: The system receives detail feature maps, context feature maps, structural boundary label maps, and artifact guidance maps as input data, and performs channel dimension expansion and spatial dimension standardization processing respectively to construct an input feature set with a uniform size and number of channels. The guided perception gating module includes a structure enhancement gating unit and an artifact suppression gating unit. The structure enhancement gating unit uses the structure boundary label map as a guide and generates a structure enhancement weight map using pixel-wise multiplication operations. The structure enhancement weight map is then applied to the detail feature map to enhance the response intensity of the edge region. The artifact suppression gating unit uses the artifact guide map as a guide, and uses inverse normalization and the Sigmoid gating function to generate a suppression mask map. The suppression mask map is then multiplied pixel by pixel with the context feature map to suppress feature interference from unstructured regions. After the structural enhancement weight map and the suppression mask map are applied, channel-dimensional weighted fusion is performed on the enhanced detail feature map and the suppression context feature map. The fusion weights are dynamically generated based on the structural region confidence map, and the fused feature map is output. Before output, the fused feature map is superimposed with a shallow feature map of the original normalized SPR image data through residual connections to enhance information stability and maintain the consistency between the fused result and the input image structure.

[0031] In this embodiment, the multi-scale detail restoration module includes two upsampling units and one skip connection unit. Each upsampling unit uses bilinear interpolation and 3×3 convolution to jointly construct a multi-scale feature upsampling image. The skip connection unit connects the intermediate layer feature map of the CSDNet detail branch with the upsampling feature map of the corresponding scale. It compensates for edge detail information through channel alignment and element-wise addition fusion, and outputs edge-enhanced image data with the same resolution as the standardized SPR image data.

[0032] In this embodiment, the structural similarity analysis is characterized by: using standardized SPR image data as the reference image and SPR image edge enhancement image data as the image to be evaluated, performing a fixed window sliding calculation operation in a unified resolution space, extracting the mean brightness, standard deviation of contrast, and structural gradient vector within each window, calculating the brightness similarity, contrast similarity, and structural similarity components accordingly, and generating a structural similarity index score through weighted fusion for analysis.

[0033] In this embodiment, the edge sharpness scoring is characterized by using SPR image edge enhancement image data as input, employing a multi-scale Laplacian edge response function to extract edge contour response maps, and constructing a comprehensive scoring function based on the gradient magnitude distribution of the edge region, edge length continuity, and structural boundary label map matching degree to output the edge sharpness scoring result. Specifically, during the edge sharpness scoring process of the SPR image edge enhancement image data, a multi-scale Laplacian edge response function is used to construct an edge contour response map to accurately characterize the edge contour intensity and detail continuity at different scales. The multi-scale Laplacian edge response function captures second-order grayscale change information at the edges by performing high-pass edge convolution operations at multiple scales on the input image, maintaining the integrity of the edge response corresponding to different structural sizes. In practice, a scale pyramid sequence is first constructed from the SPR image edge enhancement image data, and three scale versions (low, medium, and high) are generated using a Gaussian blur kernel. The standard Laplacian kernel function is then applied to each scale image. ; Perform a 2D convolution on the image to extract the edge response map at that scale; Subsequently, the edge response maps at each scale are normalized, and pixel-wise maximum response fusion is performed to construct the final multi-scale Laplacian edge contour response map. This response map can simultaneously preserve fine edge textures and large-scale structural boundaries, improving the ability to discriminate the integrity of image details in edge sharpness scoring. Based on the response map, three types of structural indices for the edge region are further calculated: (1) the statistical value of the edge gradient amplitude distribution, which is used to evaluate the overall level of edge intensity; (2) the edge pixel connection length, which is used to evaluate the continuity of the edge curve; and (3) the spatial overlap between the edge response region and the structural boundary label map, which is used to quantify the edge response's ability to preserve the anatomical structure.

[0034] Finally, the above three indicators are input into a weighted scoring function to output an edge sharpness score, which serves as the basis for quantifying the structural edges of the SPR image enhancement effectiveness.

[0035] In this embodiment, the SPR image optimization result is generated based on the structural similarity index score map and the edge sharpness score map. A joint scoring function is used to comprehensively evaluate the structural preservation and edge recognizability of the edge-enhanced image data. A preset scoring threshold is used to determine whether the fused feature map meets the image quality requirements. When the scoring result simultaneously meets the structural consistency and edge sharpness requirements, the SPR image optimization result is output; otherwise, the structural enhancement weights and artifact suppression factors in the guided perception gating module are updated, and the fusion and reconstruction process is re-executed. Specifically: 1. Calculate the structural similarity score Its value is derived from the average of the structural similarity index score maps of the standardized SPR image data and the SPR image edge enhancement image data within the area defined by the structural boundary label map, and is used to quantify the fidelity of the anatomical structure. Second, calculate the edge sharpness index. The edge response map is extracted using the multi-scale Laplacian edge response function, and the gradient magnitude distribution is statistically analyzed through a local window in the response map. The edge intensity at different scales is then weighted and averaged to measure the overall edge sharpness level. Third, introduce structural alignment index. The pixel overlap between the structural boundary label map and the edge response map is used to evaluate whether the enhanced edge accurately falls on the boundary of the anatomical region. Input the above three indicators into the joint scoring function: ; Among them, the weighting coefficient , , The summation is 1, and it can be adaptively adjusted according to the complexity of the image structure and the sparsity of the anatomical region. Based on preset quality thresholds Determine whether the output of the scoring function meets the image optimization criteria: like If the current edge enhancement image data quality meets the optimization requirements, the result is output as the SPR image optimization result. like Then the feedback-driven mechanism will update the parameters in the perception gating module; The feedback update strategy is as follows: Structural enhancement weights in structural enhancement paths Update using the following formula: ; Artifact suppression factor in the artifact suppression pathway Update using the following formula: ; in, , This is the learning rate coefficient. To prevent division by zero error, Indicates taking The maximum value; After the parameters are updated, the fusion feature map generation and multi-scale detail reconstruction process is re-executed until the scoring results meet the image optimization requirements, ensuring a comprehensive improvement in structural consistency and edge quality in the final output image.

[0036] Example 1: To verify the feasibility of this invention in practice, it was applied to treatment plan verification and real-time localization during treatment. Because SPR images inherently have blurred edges and significant low-contrast features, conventional enhancement methods cannot highlight key edges while maintaining structural accuracy, affecting physicians' judgment of organ interfaces and precise localization of radiotherapy targets. This invention, based on image segmentation and edge enhancement techniques, inputs standardized SPR images into a niche segmentation network based on CGNet to accurately extract organ boundary response maps and generate structural boundary label maps. Artifact guidance maps are generated through gradient direction magnitude analysis and local entropy mutation detection. An improved CSDNet network is constructed, with detail and context branches extracting features in parallel. Enhancement and suppression weights are adjusted in the guidance perception gating module using structural labels and artifact maps, respectively. Finally, image reconstruction is completed through a multi-scale detail recovery module, and a joint scoring function is used for closed-loop evaluation of structural similarity and edge sharpness, ensuring that the optimized results are both clear and realistic.

[0037] In clinical applications, radiation physicists feed the patient's SPR projection image into the processing flow of this invention before developing a treatment plan. After processing, the image edge sharpness is significantly improved, organ contour visibility is enhanced, and radiologists can verify treatment positioning accuracy in real time. For 10 different cases of the same site, conventional super-resolution reconstruction, U-Net-based edge enhancement methods, and the method of this invention were compared. All enhanced images were evaluated using a joint scoring function, and the results show that this invention outperforms other methods in terms of structural similarity index and edge sharpness score.

[0038] To verify the beneficial effects, the improvements of various methods in three metrics—PSNR, SSIM, and edge sharpness index—were compared. Conventional super-resolution methods achieved an average PSNR improvement of 2.3 dB, an SSIM improvement of 0.04, and an edge sharpness index improvement of 12%; the U-Net method achieved an average PSNR improvement of 3.8 dB, an SSIM improvement of 0.06, and an edge sharpness index improvement of 18%; while the method of this invention achieved an average PSNR improvement of 5.2 dB, an SSIM improvement of 0.09, and an edge sharpness index improvement of 27%. These data demonstrate that the method of this invention can highlight the edges of real structures while suppressing artifacts, providing a reliable image foundation for subsequent registration and target localization.

[0039] In complex tissue regions, such as the lung field junction and the liver-diaphragm junction, the method of this invention effectively preserves fine folds and organ edge curvature information by aligning multi-scale Laplacian edge responses with structural labels; while conventional methods often result in discontinuous or misaligned edges in these areas. Based on the combined image segmentation and enhancement effects in various cases, the method of this invention achieved a positive feedback rate of over 90% in physician subjective ratings, demonstrating its good clinical applicability and promotional value.

[0040] Table 1: Image quality comparison before and after SPR optimization (based on 10 patients) Case number Raw PSNR (dB) Original SSIM Original Margin Index (%) The PSNR (dB) of this invention This invention, SSIM Edge index (%) of this invention Similarity Enhancement (SSIM Increment) Edge boost (%) 1 24.6 0.72 45 29.8 0.81 58 0.09 13 2 23.9 0.7 43 29.0 0.79 55 0.09 12 3 25.1 0.74 47 30.3 0.83 61 0.09 14 4 24.3 0.71 44 29.6 0.8 59 0.09 15 5 23.7 0.69 42 28.9 0.78 54 0.09 12 6 24.8 0.73 46 30.1 0.82 60 0.09 14 7 24.2 0.71 44 29.4 0.8 57 0.09 13 8 23.5 0.68 41 28.7 0.77 53 0.09 12 9 25.4 0.75 48 30.6 0.84 62 0.09 14 10 24.0 0.7 43 29.2 0.79 56 0.09 13 As shown in Table 1, the method of this invention improves signal-to-noise ratio, structural fidelity, and edge sharpness. First, the processed PSNR value is approximately 5.2 dB higher than the original image in all patients, specifically from 23.5 dB–25.4 dB to 28.7 dB–30.6 dB. This magnitude indicates a significant reduction in image noise and a substantial enhancement in detail restoration. Second, the structural similarity index (SSIM) increases from 0.68–0.75 to 0.77–0.84, maintaining an increment of 0.09 in all cases. This indicates that boundary continuity and overall structural integrity are reliably guaranteed, and the enhanced image is highly consistent with the standard SPR image in terms of anatomical structure.

[0041] Regarding edge sharpness, the method of this invention improves the original 41%–48% to 53%–62%, with an average improvement of approximately 13%. The improvements in Cases 4 and 9 are particularly significant, reaching 15% and 14% respectively, fully validating the ability of this invention to enhance minute edge details in complex structural regions (SPR image optimization comparison table). Furthermore, as can be seen from the "Similarity Improvement" and "Edge Improvement" columns, the method of this invention exhibits highly consistent robustness across different cases and organ sites, meeting the dual clinical requirements for image quality and structural accuracy.

[0042] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for SPR image optimization processing based on image segmentation and edge enhancement, characterized in that, Includes the following steps: Acquire SPR image data and preprocess it to generate standardized SPR image data. Input the standardized SPR image data into a structural boundary extraction model built on the CGNet architecture to extract anatomical region boundary response maps and generate structural boundary label maps. Align the structural boundary label maps with the standardized SPR image data in the pixel dimension. Perform gradient direction magnitude analysis and local entropy mutation detection based on the standardized SPR image data to locate non-structural mutation regions in the image and generate artifact guide maps. Input the standardized SPR image data into the detail branch and context branch of the improved CSDNet, respectively. The detail branch extracts local edge change feature maps, and the context branch extracts local edge change feature maps. The system extracts multi-scale semantic context feature maps and outputs detail feature maps and context feature maps. The detail feature maps, context feature maps, structural boundary label maps, and artifact guidance maps are input into the guided perception gating module to perform a gated weighted fusion operation based on structural enhancement and artifact suppression, generating a fused feature map. This fused feature map is then input into a multi-scale detail recovery module containing upsampled structures and skip connection units to perform edge region reconstruction and feature restoration, outputting SPR image edge enhancement image data. Based on the standardized SPR image data and the SPR image edge enhancement image data, structural similarity analysis and edge sharpness scoring are performed to generate SPR image optimization results.

2. The SPR image optimization processing method based on image segmentation and edge enhancement according to claim 1, characterized in that, The preprocessing process includes performing size normalization, grayscale normalization, and background noise suppression.

3. The SPR image optimization processing method based on image segmentation and edge enhancement according to claim 2, characterized in that, The process of inputting standardized SPR image data into a structural boundary extraction model built on the CGNet structure to extract anatomical region boundary response maps and generate structural boundary label maps includes: A shallow convolutional backbone network is used to perform preliminary feature extraction on standardized SPR image data. Low-level gradient features, texture features and local contrast features of the image are extracted through two consecutive convolutional and normalization operations to generate an initial local feature map. Based on the initial local feature map, several context guidance modules are introduced. Each context guidance module contains a local perception channel and a context modeling channel. The local perception channel uses a 3×3 convolution kernel to extract the edge direction change and gray-level jump features of the pixel neighborhood. The context modeling channel constructs the context information map of the structural region through global pooling and 1×1 convolution to capture the edge consistency features inside the organization region. Dilated convolutions with dilation rates of 2 and 4 are applied to feature maps at different scales within the context modeling channel to construct a multi-scale receptive field. Channel attention weighted fusion is then performed between the local channel and the context channel to generate context-enhanced feature maps. The context-enhanced feature map is input into the boundary response prediction unit. While keeping the image size unchanged, the prediction probability of each pixel being a boundary point is calculated by pixel-wise 1×1 convolution, and the boundary response map is output. A pixel-by-pixel classification operation is performed on the boundary response map. Pixels above the set boundary probability threshold are classified as boundary pixels, and pixels below the threshold are classified as non-boundary pixels. The structural boundary label map is formed by combining the region area, boundary continuity and gradient magnitude distribution. The structural boundary label map is resampled using bilinear interpolation and mapped to pixel size and coordinate system that are completely consistent with the standardized SPR image data.

4. The SPR image optimization processing method based on image segmentation and edge enhancement according to claim 3, characterized in that, The generation of the artifact guide map includes: Gradient direction calculation is performed on the standardized SPR image data. The Sobel operator is used to calculate the gradient magnitude maps in the horizontal and vertical directions respectively, and a primary gradient image set containing gradient direction magnitude information is constructed. In the primary gradient image set, calculate the gradient variance and direction change rate of each pixel within the local window, identify regions with abrupt gradient direction changes, and generate candidate artifact region response maps. Local entropy analysis is performed on standardized SPR image data. The distribution of local entropy values ​​in each pixel region is statistically analyzed by sliding a fixed window to locate regions of abrupt entropy changes and generate a local entropy change map. The candidate artifact region response map and the local entropy mutation map are normalized pixel by pixel and superimposed. The joint response map is calculated by weighted fusion to generate a fusion intensity map. The fusion weight is dynamically adjusted according to the gradient fluctuation and entropy mutation degree of each region. A hierarchical threshold segmentation is performed on the fused intensity map. Regions exceeding the set upper threshold are marked as high-intensity artifact regions, and regions below the set lower threshold are marked as background regions. An initial artifact guide map is constructed, and boundary closure and noise region filtering operations are performed on the initial artifact guide map to generate the final artifact guide map, which maintains a one-to-one correspondence with the standardized SPR image data in the pixel dimension.

5. The SPR image optimization processing method based on image segmentation and edge enhancement according to claim 4, characterized in that, The improved CSDNet structure includes detail branches and context branches, and has the following characteristics: The detail branch consists of three consecutive depthwise separable convolutional modules. Each layer contains a standard 3×3 depthwise convolution, a pointwise convolution, and a batch normalization operation, which outputs a detail feature map. The context branch consists of four layers of pyramid-shaped semantic coding units. Each layer contains a 5×5 dilated convolution kernel and a channel attention enhancement module. The dilated structure with dilation rates of 2, 4, 6 and 8 is used to achieve multi-scale context expansion of the receptive field. A cross-layer feature residual connection structure is adopted between the outputs of each level of the context branch to enhance the semantic consistency expression capability. By fusing the semantic feature maps of the current layer and the previous layer, a deep context-aware map is formed. The detail branch and the context branch output detail feature maps and context feature maps respectively at the end. The feature dimensions are aligned by a linear transformation with a unified number of channels, and then input into the guided perception gating module.

6. The SPR image optimization processing method based on image segmentation and edge enhancement according to claim 5, characterized in that, The guided perception gating module has the following structure and operation process: Receive detail feature maps, context feature maps, structural boundary label maps, and artifact guide maps as input data, and perform channel dimension expansion and spatial dimension standardization processing respectively to construct an input feature set with a uniform size and number of channels; The guided perception gating module includes a structure enhancement gating unit and an artifact suppression gating unit. The structure enhancement gating unit uses the structure boundary label map as a guide and generates a structure enhancement weight map using pixel-wise multiplication operations, and applies the structure enhancement weight map to the detail feature map. The artifact suppression gating unit uses the artifact guide map as a guide, and uses inverse normalization and the Sigmoid gating function to generate a suppression mask map. The suppression mask map is then multiplied pixel by pixel with the context feature map to suppress feature interference from unstructured regions. After the structural enhancement weight map and the suppression mask map are applied, channel-dimensional weighted fusion is performed on the enhanced detail feature map and the suppression context feature map. The fusion weights are dynamically generated based on the structural region confidence map, and the fused feature map is output. Before output, the fused feature map is superimposed with the shallow feature map of the original normalized SPR image data through residual connections to maintain the consistency between the fused result and the input image structure.

7. The SPR image optimization processing method based on image segmentation and edge enhancement according to claim 6, characterized in that, The multi-scale detail restoration module includes two upsampling units and one skip connection unit. Each upsampling unit uses bilinear interpolation and 3×3 convolution to jointly construct a multi-scale feature upsampled image. The skip connection unit connects the intermediate layer feature map of the CSDNet detail branch with the upsampled feature map of the corresponding scale. It compensates for edge detail information through channel alignment and element-wise addition fusion, and outputs edge-enhanced image data with the same resolution as the standardized SPR image data.

8. The SPR image optimization processing method based on image segmentation and edge enhancement according to claim 7, characterized in that, The structural similarity analysis includes: using standardized SPR image data as the reference image and SPR image edge enhancement image data as the image to be evaluated, performing a fixed window sliding calculation operation in a uniform resolution space, extracting the mean brightness, standard deviation of contrast, and structural gradient vector within each window, calculating the brightness similarity, contrast similarity, and structural similarity components accordingly, and generating a structural similarity index score through weighted fusion for analysis.

9. The SPR image optimization processing method based on image segmentation and edge enhancement according to claim 8, characterized in that, The edge sharpness score uses SPR image edge enhancement image data as input image, extracts edge contour response map using multi-scale Laplacian edge response function, constructs comprehensive scoring function based on gradient magnitude distribution of edge region, edge length continuity and structural boundary label map matching degree, and outputs edge sharpness score result.

10. The SPR image optimization processing method based on image segmentation and edge enhancement according to claim 9, characterized in that, The SPR image optimization result is generated based on the structural similarity index score map and the edge sharpness score map. A joint scoring function is used to comprehensively evaluate the structural preservation and edge recognizability of the edge-enhanced image data. Based on the preset scoring threshold, it is determined whether the fused feature map meets the image quality requirements. When the scoring result meets both the structural consistency requirement and the edge sharpness requirement, the SPR image optimization result is output. Otherwise, the parameters of the structural enhancement weight and artifact suppression factor in the guided perception gating module are updated, and the fusion and reconstruction process is re-executed.

Images