An image analysis-based evaluation method and system for eye-shaping surgery

Abstract

The application provides an image analysis-based eye plastic surgery evaluation method and system. The method comprises the following steps: preprocessing an image to be evaluated for eye plastic surgery to obtain a pretreated image; simulating real degradation of the pretreated image to obtain a degraded simulated image; performing contrast enhancement and denoising processing on the degraded simulated image to obtain a denoised image; extracting features from the denoised image to obtain an eye feature image, and performing guided reconstruction on the eye feature image to obtain an eye reconstructed image; and inputting the eye reconstructed image into a trained preset evaluation model for evaluation to output an evaluation result. The application not only significantly improves the visual quality of the eye plastic surgery image itself, but also provides a powerful technical tool for precise, objective and efficient evaluation of the surgical effect through the characteristics of certainty, efficiency and quantifiability.

Description

Technical Field

[0001] This invention belongs to the technical field of image processing, and specifically relates to a method and system for evaluating oculoplastic surgery based on image analysis. Background Technology

[0002] The evaluation and follow-up of oculoplastic surgery outcomes heavily rely on high-resolution, accurate medical imaging data. Surgeons need to professionally assess surgical precision, symmetry, swelling reduction, and scar formation by comparing preoperative planning, intraoperative records, and postoperative recovery images. However, in clinical practice, the quality of images used for evaluation is often constrained by multiple complex factors, leading to the loss or distortion of crucial information, directly affecting the accuracy and reliability of the evaluation.

[0003] Existing technologies face the following prominent problems in processing images from oculoplastic surgery: 1. Limitations of traditional image enhancement methods: Common image enhancement algorithms, such as global histogram equalization (HE), contrast linear stretching, and traditional filters (such as median filtering and Gaussian filtering), typically perform global or uniform processing on images. When applied to the eye region, these methods have significant shortcomings: First, they struggle to distinguish between subtle tissue structures (such as meibomian gland contours and micro-incisions) and image noise, often over-amplifying noise or producing artifacts while improving overall contrast. Second, they cannot adaptively handle the differentiated contrast requirements of different areas of the eye (such as skin, sclera, and iris), potentially leading to local overexposure or loss of detail, compromising the realism of tissue texture.

[0004] 2. Shortcomings of learning-based super-resolution methods: In recent years, while single-image super-resolution (SISR) methods based on convolutional neural networks (CNNs) have made progress, they still face challenges in processing medical images, especially surgical images with complex degradation. First, most methods rely on the assumption of "ideal degradation" (such as bicubic downsampling), and their training data does not match the complex degradation processes of real surgical images (such as motion blur, color shift, and JPEG compression), resulting in poor generalization ability of the models in practical applications. Second, CNN models are prone to producing texture artifacts or over-smoothing that do not conform to anatomical structures when reconstructing high-frequency details, and may produce discontinuous stitching marks at image patch boundaries, affecting the interpretation of continuous anatomical boundaries (such as the eyelid margin).

[0005] 3. Adaptability issues of existing diffusion models in medical image processing: Diffusion models, as powerful generative models, excel in natural image generation and editing, but their direct application in professional medical image processing faces obstacles. First, the generation process of standard diffusion models is random; the same input may produce different outputs, which contradicts the repeatability and determinism required for medical diagnosis. Second, general models lack mechanisms for embedding prior medical knowledge (such as anatomical constraints and pathological features), and unconditional generation may deviate from medical realism. Finally, their traditional sampling processes (such as DDPM) typically require thousands of iterations, which are too time-consuming and cannot meet the actual speed requirements of clinical settings. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides an image analysis-based method and system for evaluating oculoplastic surgery, which solves the technical problems in the prior art.

[0007] In a first aspect, the present invention provides the following technical solution: a method for evaluating oculoplastic surgery based on image analysis, comprising: Obtain the target eye plastic surgery image to be evaluated, and preprocess the eye plastic surgery image to be evaluated to obtain a preprocessed image; The preprocessed image is subjected to realistic degradation simulation to obtain a degradation simulation image; The degraded simulated image is subjected to contrast enhancement and denoising processing to obtain a denoised image; Feature extraction is performed on the denoised image to obtain an eye feature image, and guided reconstruction is performed on the eye feature image to obtain an eye reconstructed image; Obtain simulated images for eye plastic surgery evaluation, input the simulated images for eye plastic surgery evaluation into a preset evaluation model for training, input the reconstructed eye images into the trained preset evaluation model for evaluation, and output the evaluation results.

[0008] Compared with the prior art, the beneficial effects of the present invention are as follows: Image quality is comprehensively improved with high detail fidelity: By training a composite degradation model that simulates the real imaging process, the system can effectively reverse various degradations encountered in actual shooting, such as color distortion, motion blur, and compression artifacts, and restore high-definition images that are closer to the colors and textures of real tissues; by combining an adaptive CLAHE and Gaussian filtering fusion strategy, noise is effectively suppressed while significantly enhancing local contrast, making details that are crucial for evaluation, such as skin micro-textures, fine blood vessels, and incision marks, clearly visible; by introducing deep learning-based HED edge detection as a structural condition and using a conditional control network (ControlNet) for strong constraints during diffusion reconstruction, the edges of key anatomical structures such as eyelid margins, double eyelid creases, and corners of the eyes are ensured to be sharp, continuous, and anatomically accurate, effectively avoiding distortion or artifacts; Clinical assessment efficacy is significantly enhanced: the high-quality images provide a reliable foundation for subsequent models, making the measurement of indicators such as binocular symmetry, swelling area and degree, scar color and smoothness more accurate, and transforming subjective visual assessment into objective data support; clear and realistic images help doctors judge the surgical effect more intuitively and confidently, shorten the image reading time, and reduce misjudgment or uncertainty caused by image quality; consistent enhancement processing can be performed on images at each stage before, during and after surgery, making images at different time points comparable and facilitating longitudinal comparison and tracking of the dynamic recovery process; The algorithm is advanced, efficient, and practical: it employs a deterministic sampling method (such as DDIM) to ensure that the system consistently outputs the same enhancement results under the same input and conditions, meeting the stringent repeatability requirements of medical applications. Compared to the thousands of sampling steps of traditional diffusion models, the deterministic sampling strategy adopted in this invention allows for high-quality results within dozens of steps, significantly improving processing efficiency and making it more suitable for actual clinical workflows. The innovative multi-condition (edge, color, texture, and dequantization information) guidance mechanism makes the generation process completely controllable, allowing for targeted adjustments based on specific clinical needs (such as emphasizing the display of blood vessels or smooth skin), offering high flexibility. In short, this invention not only significantly improves the visual quality of oculoplastic surgery images but also provides a powerful technical tool for the accurate, objective, and efficient evaluation of surgical outcomes through its deterministic, efficient, and quantifiable characteristics, possessing broad clinical application prospects and market value.

[0009] Preferably, the step of performing realistic degradation simulation on the preprocessed image to obtain a degradation simulation image includes: Preprocessed images in RGB space Gain shift is applied to each color channel to obtain an offset image. : , ; In the formula, This represents an offset image of one of the RGB channels. This represents the preprocessed image of one of the RGB channels. These are the gain coefficient and offset coefficient of one of the channels, respectively; The offset image is cropped by pixel values ​​to obtain a cropped image. : ; In the formula, For the clipping function, The pixel values ​​of the image after combining the offset images for each channel; Defined as having a length of One-dimensional uniform kernel and Gaussian kernel function : ; In the formula, The standard deviation is Gaussian. This represents the coordinate offset of an element within the convolution kernel; The degradation simulation image is determined based on the random number, one-dimensional uniform kernel, and Gaussian kernel function.

[0010] Preferably, the step of determining the degraded simulated image based on the random number, the one-dimensional uniform kernel, and the Gaussian kernel function includes: Generate a random number between 0 and 1 The cropped image is blurred and degraded based on the random number, a one-dimensional uniform kernel, and a Gaussian kernel function to obtain a blurred image. : ; In the formula, For convolution operations, These are the first fuzzy threshold and the second fuzzy threshold, respectively. Extract the edge map of the preprocessed image Based on the edge map and the blurred image Determine the suppressed image : ; In the formula, It is an inhibitory factor; The suppressed image is converted to HSV space and the H component image is extracted. The H component image is downsampled to obtain several image blocks. The image blocks are then subjected to DCT transformation to obtain the transformation coefficients. For the transformation coefficients Quantization processing is performed to obtain quantization coefficients. : ; In the formula, For the round function, For standard JPEG quantization tables, It is the compression factor; The quantization coefficients are subjected to inverse quantization and inverse DCT transformation, and converted back to RGB space to obtain the target image. The target image is then downsampled using a bicubic interpolation algorithm to obtain a degraded simulated image.

[0011] Preferably, the step of performing contrast enhancement and denoising processing on the degraded simulated image to obtain a denoised image includes: The degraded simulation image is converted to Lab space and the luminance component image is extracted. The luminance component image is then divided into several non-overlapping sub-region images. Determine the grayscale histogram and total number of pixels for each of the sub-regions of the image. The cropping threshold is determined based on the total number of pixels. : ; In the formula, For key cutting parameters, Grayscale; The portion of the grayscale histogram that exceeds the clipping threshold is clipped, and the clipped portion is evenly redistributed to the remaining grayscale levels to obtain a clipping histogram. Determine the grayscale mapping function of the cropped histogram. : ; In the formula, For the round function, Indicates the first The cumulative distribution function corresponding to each clipping histogram This represents the non-zero minimum value of the cumulative distribution function; Four adjacent cropping histograms are selected as the target region. The target region is divided into sixteen smaller regions starting from the center pixel of the four adjacent cropping histograms. According to the type of the smaller regions, they are divided into corner regions, boundary regions, and center regions. Based on pixels The type of the small region determines the denoised image.

[0012] Preferably, the method based on pixels The steps for determining the type of the small region in the denoised image include: If pixel When the data is located in a corner area, the first interpolation method is used to perform interpolation to obtain the first interpolation result. If pixel When the data is in the boundary region, a second interpolation method is used to obtain the second interpolation result. If pixel When the data is in the central region, a third interpolation method is used to obtain the third interpolation result. The interpolation process is repeated for all pixels to obtain the initial enhanced image. : ; ; ; In the formula, These are the grayscale mapping functions corresponding to the center pixels located at the top left, top right, bottom right, and bottom left, respectively. These are the x-coordinates of the center pixels located at the top left and top right, respectively. These are the ordinates of the center pixels located at the top left and bottom left, respectively. The degraded simulated image is subjected to Gaussian filtering to obtain a filtered image. The initial enhanced images are determined respectively. Cumulative histogram Global contrast and the filtered image Cumulative histogram Global contrast : ; ; In the formula, These are the grayscale values ​​of the initial enhanced image and the filtered image, respectively. Based on the initial enhanced image Cumulative histogram Global contrast and the filtered image Cumulative histogram Global contrast Determine the fusion mapping function : ; In the formula, This represents the total number of pixels in the image. The luminance component image is mapped by a fusion mapping function, and the original luminance component image is replaced by the mapped luminance component image to obtain a mapped image. The mapped image is then converted to RGB space to obtain a denoised image.

[0013] Preferably, the step of extracting features from the denoised image to obtain an eye feature image specifically includes: The denoised image is forward-propagated using a pre-trained HED network. Edge probability maps are generated at multiple scales through the HED network. The edge probability maps are then weighted and solved using a learnable fusion layer to obtain the eye feature image.

[0014] Preferably, the step of guiding reconstruction of the eye feature image to obtain an eye reconstruction image includes: The denoised image is processed using an encoder. Compress to the latent space to obtain the initial image. : ; Within K steps, gradually move towards the initial image. Add Gaussian noise to obtain the noise variable. : ; In the formula, For the first noise variables of the step, For the first The noise scheduling variance of the step, Standard Gaussian noise; Pre-trained denoising U-Net network From the first The noise variables of the first step are started using a pre-trained denoising U-Net network. Perform K-step iterative denoising to obtain the denoised component in the final step. , among which, the Denoising component of step for: ; ; In the formula, For the first The noise scheduling variance of the step, For the first noise variance of the step These are respectively eye feature image, detail information condition, color condition, and texture condition. For injection conditions; The decoder processes the noise reduction components in the final step. Decode the image to obtain a reconstructed eye image.

[0015] Secondly, the present invention provides the following technical solution: an image analysis-based oculoplastic surgery evaluation system, the system comprising: The preprocessing module is used to acquire the target eye plastic surgery image to be evaluated, and to preprocess the eye plastic surgery image to obtain a preprocessed image. The degradation module is used to perform realistic degradation simulation on the preprocessed image to obtain a degradation simulation image; A denoising module is used to perform contrast enhancement and denoising processing on the degraded simulated image to obtain a denoised image; The reconstruction module is used to extract features from the denoised image to obtain an eye feature image, and to perform guided reconstruction on the eye feature image to obtain an eye reconstruction image. The evaluation module is used to acquire simulated images of eye plastic surgery evaluation, input the simulated images of eye plastic surgery evaluation into a preset evaluation model for training, input the reconstructed eye images into the trained preset evaluation model for evaluation, and output the evaluation results.

[0016] Thirdly, the present invention provides the following technical solution: a computer, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the image analysis-based oculoplastic surgery evaluation method described above.

[0017] Fourthly, the present invention provides the following technical solution: a storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the image analysis-based oculoplastic surgery evaluation method described above. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart of an image analysis-based oculoplastic surgery evaluation method provided in Embodiment 1 of the present invention; Figure 2 This is a structural block diagram of the image analysis-based oculoplastic surgery evaluation system provided in Embodiment 2 of the present invention; Figure 3 This is a schematic diagram of the hardware structure of a computer provided for another embodiment of the present invention.

[0020] The embodiments of the present invention will be further described below with reference to the accompanying drawings. Detailed Implementation

[0021] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain embodiments of the present invention, and should not be construed as limiting the present invention.

[0022] Example 1 In Embodiment 1 of the present invention, as Figure 1 As shown, an image analysis-based method for evaluating oculoplastic surgery includes: S1. Obtain the target eye plastic surgery image to be evaluated, and preprocess the eye plastic surgery image to be evaluated to obtain a preprocessed image; Specifically, the images to be evaluated in this oculoplastic surgery procedure can be obtained by using imaging devices installed at different angles and positions to capture postoperative images of the target eye. Then, the images are processed using existing preprocessing steps such as image cropping, rotation, translation, and format conversion to obtain preprocessed images.

[0023] S2. Perform realistic degradation simulation on the preprocessed image to obtain a degradation simulation image.

[0024] Step S2 includes: S21. Preprocessing the image in RGB space Gain shift is applied to each color channel to obtain an offset image. : , ; In the formula, This represents an offset image of one of the RGB channels. This represents the preprocessed image of one of the RGB channels. These are the gain coefficient and offset coefficient of one of the channels, respectively; Specifically, the purpose of gain offset is to simulate color distortion caused by equipment aging, inaccurate white balance, or ambient light. The gain coefficient is randomly sampled within the range of 0.85-1.15 to simulate different degrees of attenuation or enhancement in each channel, and the offset coefficient is randomly sampled within the range of -25 to 25 to simulate overall exposure changes.

[0025] S22. Crop the offset image by pixel values ​​to obtain a cropped image. : ; In the formula, For the clipping function, The pixel values ​​of the image after combining the offset images for each channel.

[0026] S23, Defined as having a length of One-dimensional uniform kernel and Gaussian kernel function : ; In the formula, The standard deviation is Gaussian. This represents the coordinate offset of an element within the convolution kernel; S24. Determine the degradation simulation image based on the random number, one-dimensional uniform kernel, and Gaussian kernel function; Step S24 includes: S241. Generate a random number between 0 and 1. The cropped image is blurred and degraded based on the random number, a one-dimensional uniform kernel, and a Gaussian kernel function to obtain a blurred image. : ; In the formula, For convolution operations, These are the first fuzzy threshold and the second fuzzy threshold, respectively. Specifically, the purpose of blur degradation is to simulate the decrease in sharpness caused by inaccurate focusing or camera / target micro-movements. It is processed using a randomly selected blur kernel, and the first and second blur thresholds here are 0.3 and 0.7, respectively.

[0027] S242. Extract the edge map of the preprocessed image. Based on the edge map and the blurred image Determine the suppressed image : ; In the formula, It is an inhibitory factor; Specifically, the edge map here can be extracted by a pre-trained global nested edge detection network, such as the HED network, which can provide more coherent and semantically rich edges, outperforming traditional operators, with the suppression factor randomly selected in the range of 0.05-0.2.

[0028] S243. Convert the suppressed image to HSV space and extract the H component image. Downsample the H component image to obtain several image blocks. Perform DCT transformation on the image blocks to obtain the transformation coefficients. S244, regarding the transformation coefficients Quantization processing is performed to obtain quantization coefficients. : ; In the formula, For the round function, For standard JPEG quantization tables, It is the compression factor; Specifically, the purpose of this step is to simulate the block artifacts and high-frequency information loss introduced by lossy compression of images during network transmission or storage.

[0029] S245. Perform inverse quantization and inverse DCT transformation on the quantization coefficients and convert them back to RGB space to obtain the target image. Use bicubic interpolation algorithm to downsample the target image to the target size to obtain a degraded simulated image.

[0030] S3. Perform contrast enhancement and denoising processing on the degraded simulated image to obtain a denoised image; Step S3 includes: S31. Convert the degraded simulation image to Lab space and extract the luminance component image, and divide the luminance component image into several non-overlapping sub-region images. S32. Determine the grayscale histogram and the total number of pixels for each sub-region image. The cropping threshold is determined based on the total number of pixels. : ; In the formula, For key cutting parameters, Grayscale; S33. The portion of the grayscale histogram that exceeds the clipping threshold is clipped and the clipped portion is evenly redistributed to the remaining grayscale levels to obtain a clipping histogram. S34. Determine the grayscale mapping function of the cropping histogram. : ; In the formula, For the round function, Indicates the first The cumulative distribution function corresponding to each clipping histogram This represents the non-zero minimum value of the cumulative distribution function; S35. Select four adjacent cropping histograms as the target region. Divide the target region into sixteen small regions starting from the center pixel of the four adjacent cropping histograms. Divide the small regions into corner regions, boundary regions, and center regions according to their type. Specifically, the corner areas are the small areas at the four corner points, the central areas are the four small areas located at the center, and the remaining areas are the boundary areas.

[0031] S36, Based on pixels The type of the small region determines the denoised image.

[0032] Step S36 includes: S361, if pixel points When the data is located in a corner area, the first interpolation method is used to perform interpolation to obtain the first interpolation result. If pixel When the data is in the boundary region, a second interpolation method is used to obtain the second interpolation result. If pixel When the data is in the central region, a third interpolation method is used to obtain the third interpolation result. The interpolation process is repeated for all pixels to obtain the initial enhanced image. : ; ; ; In the formula, These are the grayscale mapping functions corresponding to the center pixels located at the top left, top right, bottom right, and bottom left, respectively. These are the x-coordinates of the center pixels located at the top left and top right, respectively. These are the ordinates of the center pixels located at the top left and bottom left, respectively. S362. Perform Gaussian filtering on the degraded simulated image to obtain a filtered image. The initial enhanced images are determined respectively. Cumulative histogram Global contrast and the filtered image Cumulative histogram Global contrast : ; ; In the formula, These are the grayscale values ​​of the initial enhanced image and the filtered image, respectively. S363, Based on the initial enhanced image Cumulative histogram Global contrast and the filtered image Cumulative histogram Global contrast Determine the fusion mapping function : ; In the formula, This represents the total number of pixels in the image. S364. The luminance component image is mapped by a fusion mapping function and the original luminance component image is replaced by the mapped luminance component image to obtain a mapped image. The mapped image is then converted to RGB space to obtain a denoised image.

[0033] S4. Extract features from the denoised image to obtain an eye feature image, and perform guided reconstruction on the eye feature image to obtain an eye reconstruction image; Specifically, the step of extracting features from the denoised image to obtain an eye feature image is as follows: The denoised image is forward-propagated using a pre-trained HED network. Edge probability maps are generated at multiple scales through the HED network. The edge probability maps are then weighted and solved using a learnable fusion layer to obtain an eye feature image. Specifically, in order to preserve key anatomical structures of the eye (such as eyelid margin, incision line, and skin folds) in subsequent reconstruction, multi-level edges are extracted as conditions.

[0034] Step S4 includes: S41. Use an encoder to process the denoised image. Compress to the latent space to obtain the initial image. : ; S42. Within K steps, gradually move towards the initial image. Add Gaussian noise to obtain the noise variable. : ; In the formula, For the first noise variables of the step, For the first The noise scheduling variance of the step, Standard Gaussian noise; S43, Pre-trained denoising U-Net network From the first The noise variables of the first step are started using a pre-trained denoising U-Net network. Perform K-step iterative denoising to obtain the denoised component in the final step. , among which, the Denoising component of step for: ; ; In the formula, For the first The noise scheduling variance of the step, For the first noise variance of the step These are respectively eye feature image, detail information condition, color condition, and texture condition. For injection conditions; Specifically, regarding the injection conditions here, the eye feature image is the image condition obtained in step S4, the detail information condition is to perform JPEG inverse quantization on the input image in an attempt to recover some of the mid-to-high frequency information lost due to compression, which can be seen as the reverse steps of steps S243-S244, the color condition is to apply a slight color enhancement to the input image, which can be seen in step S21, where the gain coefficient is between 1.0 and 1.1 and the offset coefficient is between 0 and 10, and the texture condition specifically uses the Tile branch of ControlNet to process the input image to guide the generation of more natural and realistic local textures.

[0035] It should be noted that in the actual U-Net Transformer block, window attention is used instead of global attention to improve efficiency and better model local dependencies. This is crucial for restoring details such as skin texture. During each denoising layer, the window position is periodically shifted so that different windows can exchange information, thereby avoiding block artifacts and capturing the global context.

[0036] S44, Denoising the final component using the decoder. Decode the image to obtain a reconstructed eye image.

[0037] S5. Obtain a simulated image of eye plastic surgery evaluation, input the simulated image of eye plastic surgery evaluation into a preset evaluation model for training, input the reconstructed eye image into the trained preset evaluation model for evaluation, and output the evaluation result. The preset evaluation model uses a deep learning regression model (such as a CNN-based regression network). By inputting the reconstructed eye image into the model, the corresponding evaluation results can be output, such as the overall aesthetic score of the eye after surgery (which requires training with expert scoring data), recovery status, etc.

[0038] The image analysis-based evaluation method for oculoplastic surgery provided in Embodiment 1 of this invention: Image quality is comprehensively improved with high detail fidelity: By training a composite degradation model that simulates the real imaging process, the system can effectively reverse various degradations encountered in actual shooting, such as color distortion, motion blur, and compression artifacts, and restore high-definition images that are closer to the colors and textures of real tissues; by combining an adaptive CLAHE and Gaussian filtering fusion strategy, noise is effectively suppressed while significantly enhancing local contrast, making details that are crucial for evaluation, such as skin micro-textures, fine blood vessels, and incision marks, clearly visible; by introducing deep learning-based HED edge detection as a structural condition and using a conditional control network (ControlNet) for strong constraints during diffusion reconstruction, the edges of key anatomical structures such as eyelid margins, double eyelid creases, and corners of the eyes are ensured to be sharp, continuous, and anatomically accurate, effectively avoiding distortion or artifacts; Clinical assessment efficacy is significantly enhanced: the high-quality images provide a reliable foundation for subsequent models, making the measurement of indicators such as binocular symmetry, swelling area and degree, scar color and smoothness more accurate, and transforming subjective visual assessment into objective data support; clear and realistic images help doctors judge the surgical effect more intuitively and confidently, shorten the image reading time, and reduce misjudgment or uncertainty caused by image quality; consistent enhancement processing can be performed on images at each stage before, during and after surgery, making images at different time points comparable and facilitating longitudinal comparison and tracking of the dynamic recovery process; The algorithm is advanced, efficient, and practical: it employs a deterministic sampling method (such as DDIM) to ensure that the system consistently outputs the same enhancement results under the same input and conditions, meeting the stringent repeatability requirements of medical applications. Compared to the thousands of sampling steps of traditional diffusion models, the deterministic sampling strategy adopted in this invention allows for high-quality results within dozens of steps, significantly improving processing efficiency and making it more suitable for actual clinical workflows. The innovative multi-condition (edge, color, texture, and dequantization information) guidance mechanism makes the generation process completely controllable, allowing for targeted adjustments based on specific clinical needs (such as emphasizing the display of blood vessels or smooth skin), offering high flexibility. In short, this invention not only significantly improves the visual quality of oculoplastic surgery images but also provides a powerful technical tool for the accurate, objective, and efficient evaluation of surgical outcomes through its deterministic, efficient, and quantifiable characteristics, possessing broad clinical application prospects and market value.

[0039] Example 2 like Figure 2 As shown, in Embodiment 2 of the present invention, an image analysis-based oculoplastic surgery evaluation system is provided, the system comprising: Preprocessing module 1 is used to acquire the target eye plastic surgery image to be evaluated, and to preprocess the eye plastic surgery image to obtain a preprocessed image. Degradation module 2 is used to perform realistic degradation simulation on the preprocessed image to obtain a degradation simulation image; The denoising module 3 is used to perform contrast enhancement and denoising processing on the degraded simulated image to obtain a denoised image; Reconstruction module 4 is used to extract features from the denoised image to obtain an eye feature image, and to perform guided reconstruction on the eye feature image to obtain an eye reconstructed image. Evaluation module 5 is used to acquire simulated images of eye plastic surgery evaluation, input the simulated images of eye plastic surgery evaluation into a preset evaluation model for training, input the reconstructed eye images into the trained preset evaluation model for evaluation, and output evaluation results.

[0040] Among them, degradation module 2 is used for: Preprocessed images in RGB space Gain shift is applied to each color channel to obtain an offset image. : , ; In the formula, This represents an offset image of one of the RGB channels. This represents the preprocessed image of one of the RGB channels. These are the gain coefficient and offset coefficient of one of the channels, respectively; The offset image is cropped by pixel values ​​to obtain a cropped image. : ; In the formula, For the clipping function, The pixel values ​​of the image after combining the offset images for each channel; Defined as having a length of One-dimensional uniform kernel and Gaussian kernel function : ; In the formula, The standard deviation is Gaussian. This represents the coordinate offset of an element within the convolution kernel; The degradation simulation image is determined based on the random number, one-dimensional uniform kernel, and Gaussian kernel function.

[0041] The degradation module 2 is further used for: Generate a random number between 0 and 1 The cropped image is blurred and degraded based on the random number, a one-dimensional uniform kernel, and a Gaussian kernel function to obtain a blurred image. : ; In the formula, For convolution operations, These are the first fuzzy threshold and the second fuzzy threshold, respectively. Extract the edge map of the preprocessed image Based on the edge map and the blurred image Determine the suppressed image : ; In the formula, It is an inhibitory factor; The suppressed image is converted to HSV space and the H component image is extracted. The H component image is downsampled to obtain several image blocks. The image blocks are then subjected to DCT transformation to obtain the transformation coefficients. For the transformation coefficients Quantization processing is performed to obtain quantization coefficients. : ; In the formula, For the round function, For standard JPEG quantization tables, It is the compression factor; The quantization coefficients are subjected to inverse quantization and inverse DCT transformation, and converted back to RGB space to obtain the target image. The target image is then downsampled using a bicubic interpolation algorithm to obtain a degraded simulated image.

[0042] The noise reduction module 3 is used for: The degraded simulation image is converted to Lab space and the luminance component image is extracted. The luminance component image is then divided into several non-overlapping sub-region images. Determine the grayscale histogram and total number of pixels for each of the sub-regions of the image. The cropping threshold is determined based on the total number of pixels. : ; In the formula, For key cutting parameters, Grayscale; The portion of the grayscale histogram that exceeds the clipping threshold is clipped, and the clipped portion is evenly redistributed to the remaining grayscale levels to obtain a clipping histogram. Determine the grayscale mapping function of the cropped histogram. : ; In the formula, For the round function, Indicates the first The cumulative distribution function corresponding to each clipping histogram This represents the non-zero minimum value of the cumulative distribution function; Four adjacent cropping histograms are selected as the target region. The target region is divided into sixteen smaller regions starting from the center pixel of the four adjacent cropping histograms. According to the type of the smaller regions, they are divided into corner regions, boundary regions, and center regions. Based on pixels The type of the small region determines the denoised image.

[0043] The noise reduction module 3 is further used for: If pixel When the data is located in a corner area, the first interpolation method is used to perform interpolation to obtain the first interpolation result. If pixel When the data is in the boundary region, a second interpolation method is used to obtain the second interpolation result. If pixel When the data is in the central region, a third interpolation method is used to obtain the third interpolation result. The interpolation process is repeated for all pixels to obtain the initial enhanced image. : ; ; ; In the formula, These are the grayscale mapping functions corresponding to the center pixels located at the top left, top right, bottom right, and bottom left, respectively. These are the x-coordinates of the center pixels located at the top left and top right, respectively. These are the ordinates of the center pixels located at the top left and bottom left, respectively. The degraded simulated image is subjected to Gaussian filtering to obtain a filtered image. The initial enhanced images are determined respectively. Cumulative histogram Global contrast and the filtered image Cumulative histogram Global contrast : ; ; In the formula, These are the grayscale values ​​of the initial enhanced image and the filtered image, respectively. Based on the initial enhanced image Cumulative histogram Global contrast and the filtered image Cumulative histogram Global contrast Determine the fusion mapping function : ; In the formula, This represents the total number of pixels in the image. The luminance component image is mapped by a fusion mapping function, and the original luminance component image is replaced by the mapped luminance component image to obtain a mapped image. The mapped image is then converted to RGB space to obtain a denoised image.

[0044] The reconstruction module 4 is used for: The denoised image is forward-propagated using a pre-trained HED network. Edge probability maps are generated at multiple scales through the HED network. The edge probability maps are then weighted and solved using a learnable fusion layer to obtain the eye feature image.

[0045] The reconstruction module 4 is further used for: The denoised image is processed using an encoder. Compress to the latent space to obtain the initial image. : ; Within K steps, gradually move towards the initial image. Add Gaussian noise to obtain the noise variable. : ; In the formula, For the first noise variables of the step, For the first The noise scheduling variance of the step, Standard Gaussian noise; Pre-trained denoising U-Net network From the first The noise variables of the first step are started using a pre-trained denoising U-Net network. Perform K-step iterative denoising to obtain the denoised component in the final step. , among which, the Denoising component of step for: ; ; In the formula, For the first The noise scheduling variance of the step, For the first noise variance of the step These are respectively eye feature image, detail information condition, color condition, and texture condition. For injection conditions; The decoder processes the noise reduction components in the final step. Decode the image to obtain a reconstructed eye image.

[0046] In other embodiments of the present invention, the present invention provides the following technical solution: a computer, including a memory 102, a processor 101, and a computer program stored in the memory 102 and executable on the processor 101, wherein the processor 101 executes the computer program to implement the image analysis-based oculoplastic surgery evaluation method as described above.

[0047] Specifically, the processor 101 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of the present invention.

[0048] The memory 102 may include a large-capacity memory for data or instructions. For example, and not limitingly, the memory 102 may include a hard disk drive (HDD), a floppy disk drive, a solid-state drive (SSD), flash memory, an optical disk drive, a magneto-optical disk drive, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory 102 may include removable or non-removable (or fixed) media. Where appropriate, the memory 102 may be internal or external to a data processing device. In a particular embodiment, the memory 102 is non-volatile memory. In a particular embodiment, the memory 102 includes read-only memory (ROM) and random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable read-only memory (PROM), an erasable read-only memory (EPROM), an electrically erasable read-only memory (EEPROM), an electrically alterable read-only memory (EAROM), or flash memory, or a combination of two or more of these. Where appropriate, the RAM can be Static Random-Access Memory (SRAM) or Dynamic Random-Access Memory (DRAM). DRAM can be Fast Page Mode Dynamic Random Access Memory (FPMDRAM), Extended Data Out Dynamic Random Access Memory (EDODRAM), Synchronous Dynamic Random-Access Memory (SDRAM), etc.

[0049] The memory 102 can be used to store or cache various data files that need to be processed and / or used for communication, as well as possible computer program instructions executed by the processor 101.

[0050] The processor 101 reads and executes the computer program instructions stored in the memory 102 to implement the above-described image analysis-based oculoplastic surgery evaluation method.

[0051] In some embodiments, the computer may further include a communication interface 103 and a bus 100. For example, Figure 3 As shown, the processor 101, memory 102, and communication interface 103 are connected through bus 100 and complete communication with each other.

[0052] The communication interface 103 is used to enable communication between the various modules, devices, units, and / or equipment in the embodiments of the present invention. The communication interface 103 can also enable data communication with other components such as external devices, image / data acquisition devices, databases, external storage, and image / data processing workstations.

[0053] Bus 100 includes hardware, software, or both, that couples components of a computer device together. Bus 100 includes, but is not limited to, at least one of the following: data bus, address bus, control bus, expansion bus, and local bus. For example, and not as a limitation, bus 100 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 100 may include one or more buses. Although specific buses are described and illustrated in the embodiments of the present invention, the present invention is contemplated by any suitable bus or interconnect.

[0054] The computer can execute the image analysis-based oculoplastic surgery evaluation method of the present invention based on the acquired image analysis-based oculoplastic surgery evaluation system, thereby realizing image analysis-based oculoplastic surgery evaluation.

[0055] In some further embodiments of the present invention, in conjunction with the above-described image analysis-based oculoplastic surgery evaluation method, the present invention provides the following technical solution: a storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the above-described image analysis-based oculoplastic surgery evaluation method.

[0056] Those skilled in the art will understand that the logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can mean any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0057] More specific examples of readable media (a non-exhaustive list) include: electrical connections (electronic devices) with one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0058] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[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] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A method for evaluating oculoplastic surgery based on image analysis, characterized in that, include: Obtain the target eye plastic surgery image to be evaluated, and preprocess the eye plastic surgery image to be evaluated to obtain a preprocessed image; The preprocessed image is subjected to realistic degradation simulation to obtain a degradation simulation image; The degraded simulated image is subjected to contrast enhancement and denoising processing to obtain a denoised image; Feature extraction is performed on the denoised image to obtain an eye feature image, and guided reconstruction is performed on the eye feature image to obtain an eye reconstructed image; Obtain simulated images for eye plastic surgery evaluation, input the simulated images for eye plastic surgery evaluation into a preset evaluation model for training, input the reconstructed eye images into the trained preset evaluation model for evaluation, and output the evaluation results.

2. The image analysis-based oculoplastic surgery evaluation method according to claim 1, characterized in that, The step of performing realistic degradation simulation on the preprocessed image to obtain a degradation simulation image includes: Preprocessed images in RGB space Gain shift is applied to each color channel to obtain an offset image. : ， ； In the formula, This represents an offset image of one of the RGB channels. This represents the preprocessed image of one of the RGB channels. These are the gain coefficient and offset coefficient of one of the channels, respectively; The offset image is cropped by pixel values ​​to obtain a cropped image. : ； In the formula, For the clipping function, The pixel values ​​of the image after combining the offset images for each channel; Defined as having a length of One-dimensional uniform kernel and Gaussian kernel function : ； In the formula, The standard deviation is Gaussian. This represents the coordinate offset of an element within the convolution kernel; The degradation simulation image is determined based on the random number, one-dimensional uniform kernel, and Gaussian kernel function.

3. The image analysis-based oculoplastic surgery evaluation method according to claim 2, characterized in that, The step of determining the degraded simulation image based on the random number, one-dimensional uniform kernel, and Gaussian kernel function includes: Generate a random number between 0 and 1 The cropped image is blurred and degraded based on the random number, a one-dimensional uniform kernel, and a Gaussian kernel function to obtain a blurred image. : ； In the formula, For convolution operations, These are the first fuzzy threshold and the second fuzzy threshold, respectively. Extract the edge map of the preprocessed image Based on the edge map and the blurred image Determine the suppressed image : ； In the formula, It is an inhibitory factor; The suppressed image is converted to HSV space and the H component image is extracted. The H component image is downsampled to obtain several image blocks. The image blocks are then subjected to DCT transformation to obtain the transformation coefficients. For the transformation coefficients Quantization processing is performed to obtain quantization coefficients. : ； In the formula, For the round function, For standard JPEG quantization tables, It is the compression factor; The quantization coefficients are subjected to inverse quantization and inverse DCT transformation, and converted back to RGB space to obtain the target image. The target image is then downsampled using a bicubic interpolation algorithm to obtain a degraded simulated image.

4. The image analysis-based oculoplastic surgery evaluation method according to claim 1, characterized in that, The step of performing contrast enhancement and denoising processing on the degraded simulated image to obtain a denoised image includes: The degraded simulation image is converted to Lab space and the luminance component image is extracted. The luminance component image is then divided into several non-overlapping sub-region images. Determine the grayscale histogram and total number of pixels for each of the sub-regions of the image. The cropping threshold is determined based on the total number of pixels. : ； In the formula, For key cutting parameters, Grayscale; The portion of the grayscale histogram that exceeds the clipping threshold is clipped, and the clipped portion is evenly redistributed to the remaining grayscale levels to obtain a clipping histogram. Determine the grayscale mapping function of the cropped histogram. : ； In the formula, For the round function, Indicates the first The cumulative distribution function corresponding to each clipping histogram This represents the non-zero minimum value of the cumulative distribution function; Four adjacent cropping histograms are selected as the target region. The target region is divided into sixteen smaller regions starting from the center pixel of the four adjacent cropping histograms. According to the type of the smaller regions, they are divided into corner regions, boundary regions, and center regions. Based on pixels The type of the small region determines the denoised image.

5. The image analysis-based oculoplastic surgery evaluation method according to claim 4, characterized in that, The according to pixel points The steps for determining the type of the small region in the denoised image include: If pixel When the data is located in a corner area, the first interpolation method is used to perform interpolation to obtain the first interpolation result. If pixel When the data is in the boundary region, a second interpolation method is used to obtain the second interpolation result. If pixel When the data is in the central region, a third interpolation method is used to obtain the third interpolation result. The interpolation process is repeated for all pixels to obtain the initial enhanced image. : ； ； ； In the formula, These are the grayscale mapping functions corresponding to the center pixels located at the top left, top right, bottom right, and bottom left, respectively. These are the x-coordinates of the center pixels located at the top left and top right, respectively. These are the ordinates of the center pixels located at the top left and bottom left, respectively. The degraded simulated image is subjected to Gaussian filtering to obtain a filtered image. The initial enhanced images are determined respectively. Cumulative histogram Global contrast and the filtered image Cumulative histogram Global contrast : ； ； In the formula, These are the grayscale values ​​of the initial enhanced image and the filtered image, respectively. Based on the initial enhanced image Cumulative histogram Global contrast and the filtered image Cumulative histogram Global contrast Determine the fusion mapping function : ； In the formula, This represents the total number of pixels in the image. The luminance component image is mapped by a fusion mapping function, and the original luminance component image is replaced by the mapped luminance component image to obtain a mapped image. The mapped image is then converted to RGB space to obtain a denoised image.

6. The image analysis-based oculoplastic surgery evaluation method according to claim 1, characterized in that, The specific steps for extracting features from the denoised image to obtain an eye feature image are as follows: The denoised image is forward-propagated using a pre-trained HED network. Edge probability maps are generated at multiple scales through the HED network. The edge probability maps are then weighted and solved using a learnable fusion layer to obtain the eye feature image.

7. The image analysis-based oculoplastic surgery evaluation method according to claim 1, characterized in that, The step of guiding reconstruction of the eye feature image to obtain an eye reconstruction image includes: The denoised image is processed using an encoder. Compress to the latent space to obtain the initial image. : ； Within K steps, gradually move towards the initial image. Add Gaussian noise to obtain the noise variable. : ； In the formula, For the first noise variables of the step, For the first The noise scheduling variance of the step, Standard Gaussian noise; Pre-trained denoising U-Net network From the first The noise variables of the first step are started using a pre-trained denoising U-Net network. Perform K-step iterative denoising to obtain the denoised component in the final step. , among which, the Denoising component of step for: ； ； In the formula, For the first The noise scheduling variance of the step, For the first noise variance of the step These are respectively eye feature image, detail information condition, color condition, and texture condition. For injection conditions; The decoder processes the noise reduction components in the final step. Decode the image to obtain a reconstructed eye image.

8. An image analysis-based evaluation system for oculoplastic surgery, characterized in that, The system includes: The preprocessing module is used to acquire the target eye plastic surgery image to be evaluated, and to preprocess the eye plastic surgery image to obtain a preprocessed image. The degradation module is used to perform realistic degradation simulation on the preprocessed image to obtain a degradation simulation image; A denoising module is used to perform contrast enhancement and denoising processing on the degraded simulated image to obtain a denoised image; The reconstruction module is used to extract features from the denoised image to obtain an eye feature image, and to perform guided reconstruction on the eye feature image to obtain an eye reconstruction image. The evaluation module is used to acquire simulated images of eye plastic surgery evaluation, input the simulated images of eye plastic surgery evaluation into a preset evaluation model for training, input the reconstructed eye images into the trained preset evaluation model for evaluation, and output the evaluation results.

9. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the image analysis-based oculoplastic surgery evaluation method as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium stores a computer program that, when executed by a processor, implements the image analysis-based oculoplastic surgery evaluation method as described in any one of claims 1 to 7.

Images