Visual field definition strengthening system in gynecological clinical examination, diagnosis and treatment process and implementation method thereof

Abstract

The invention relates to the technical field of gynecological examination, in particular to a visual field definition strengthening system in the gynecological clinical examination, diagnosis and treatment process and an implementation method of the visual field definition strengthening system. The method comprises the following steps: an image acquisition and preprocessing unit acquires a visual field image in a gynecological examination process in real time and preprocesses the visual field image; a feature extraction unit extracts key features from the preprocessed view image; the definition evaluation and optimization unit generates a definition score based on the key features, performs decoupling correction on feature interference caused by mucus scattering and tissue reflection by using the polarization difference image and the polarization average image, generates an optimized definition score, and generates definition optimization parameters through a multi-task learning model; the field-of-view enhancement execution unit enhances the definition of the field-of-view image based on the definition optimization parameter. According to the method, feature correction is carried out based on an imaging physical mechanism, so that the definition score more truly reflects tissue intrinsic details, and a reliable basis is provided for subsequent precision enhancement.

Description

Technical Field

[0001] This invention relates to the field of gynecological examination technology, and more specifically, to a system for enhancing visual clarity during gynecological clinical examination and diagnosis, and its implementation method. Background Technology

[0002] In gynecological clinical examinations and treatments, especially during endoscopic procedures such as colposcopy and hysteroscopy, doctors rely heavily on real-time images to observe key anatomical structures such as the cervix, endometrium, and fallopian tube openings, enabling early identification and diagnosis of diseases such as inflammation, polyps, and precancerous lesions. However, due to the unique physiological environment of the female reproductive tract, the surface of the tissue being examined is often covered with mucus, blood, or secretions. These media cause multiple scattering of light, leading to blurred images, decreased contrast, and loss of detail. Simultaneously, the strong specular reflection (highlights) produced by the moist tissue surface can easily cause local overexposure, obscuring the edges of lesions and minute structures. These interfering factors severely reduce the clarity and identifiability of the examination field, increasing the risk of missed or misdiagnosed lesions.

[0003] Existing image enhancement techniques mostly employ general algorithms (such as histogram equalization and sharpening filtering), lacking an understanding of the physical mechanisms of interference specific to gynecology. This makes it difficult to effectively distinguish and suppress the mixed effects of mucus scattering and tissue reflection, often leading to problems such as noise amplification, color distortion, or over-enhancement. Furthermore, traditional methods typically process images based solely on statistical features, failing to differentiate optimization by considering the anatomical objectives of the current examination (e.g., cervical observation focuses on overall appearance, while uterine cavity examination focuses on texture details), and lacking the ability to perceive the intent of the clinical workflow. More importantly, because interference significantly distorts image features such as brightness, contrast, and edges, directly using it for sharpness evaluation will lead to scoring bias, thus misleading subsequent enhancement strategies. Therefore, this paper presents a system for enhancing visual clarity during gynecological clinical examinations and diagnoses, along with its implementation method. Summary of the Invention

[0004] The purpose of this invention is to provide a system and method for enhancing visual clarity during gynecological clinical examination and diagnosis, in order to solve the problems mentioned in the background art, such as the decline in image quality caused by mucus scattering and tissue reflection, the lack of accurate separation of interference in traditional image enhancement methods, the susceptibility of clarity evaluation to interference distortion, and the failure of the optimization process to combine with the clinical diagnosis and treatment intention, resulting in poor enhancement effect.

[0005] To achieve the above objectives, on the one hand, the present invention provides a system for enhancing visual clarity during gynecological clinical examination and diagnosis, comprising:

[0006] The image acquisition and preprocessing unit acquires visual field images during the gynecological examination in real time and preprocesses the visual field images.

[0007] A feature extraction unit extracts key features from the preprocessed field-of-view image;

[0008] The sharpness evaluation and optimization unit generates a sharpness score based on key features, uses polarization difference images and polarization average images to decouple and correct feature interference caused by mucus scattering and tissue reflection, generates an optimized sharpness score, and generates sharpness optimization parameters through a multi-task learning model.

[0009] A field-of-view enhancement execution unit enhances the clarity of the field-of-view image based on clarity optimization parameters.

[0010] As a further improvement to this technical solution, the image acquisition and preprocessing unit includes an image acquisition module and an image processing module;

[0011] The image acquisition module acquires visual field images during gynecological examinations based on the optical illumination module, and simultaneously acquires two polarized images of the same visual field. and ;

[0012] The image processing module preprocesses the acquired field-of-view images.

[0013] As a further improvement to this technical solution, the feature extraction unit includes a brightness distribution module, a contrast module, an edge sharpness module, and a blur level module;

[0014] The brightness distribution module is used to perform statistical analysis on the overall brightness histogram of the preprocessed visual field image to obtain brightness distribution features that reflect the overall brightness of the image.

[0015] The contrast module is used to analyze the grayscale differences in local areas of the visual field image and extract contrast features that reflect the distinguishability of tissue details in the visual field image.

[0016] The edge sharpness module uses an edge detection method to extract edge sharpness features, which are used to characterize the clarity of the outline of the visual field image.

[0017] The blur level module calculates blur level features using frequency domain analysis to reflect the blurred areas in the field of view image caused by inaccurate focusing.

[0018] As a further improvement to this technical solution, the sharpness evaluation and optimization unit generates a sharpness score based on key features, including the following steps:

[0019] S1.1, Based on two polarization images and Calculate polarization difference image and polarization average image ;

[0020] S1.2 Utilizing the difference in polarization characteristics between mucus and reflection, the mucus region is detected based on scattering entropy and the reflection region is detected based on polarization difference, respectively. The extracted key features are decoupled and corrected to generate decoupled and corrected key features. The extracted key features include brightness distribution features, contrast features, edge sharpness features and blur degree features.

[0021] S1.3 Normalize the key features after decoupling correction;

[0022] S1.4. Assign preset weight coefficients based on the importance of each decoupled and corrected key feature in image sharpness evaluation;

[0023] S1.5 Calculate the sub-feature scores of the key features respectively. The sub-feature scores of the key features include brightness balance score, contrast score, edge sharpness score and blur score, which reflect the image quality of each single feature dimension.

[0024] S1.6. Weighted fusion of the sub-feature scores of the key features according to their weights to output a comprehensive clarity score. It is used to characterize the overall sharpness level of the current field of view image.

[0025] As a further improvement to this technical solution, in step S1.2, generating the key features after decoupling correction includes the following steps:

[0026] S1.21, Based on polarization difference image Brightness threshold Segmentation and extraction of reflective mask ;

[0027] S1.22, Polarization-averaged image Calculate the local scattering entropy; the local scattering entropy value is higher than the mucus threshold. The area is denoted as the mucus mask. ;

[0028] S1.23. In the overall brightness histogram statistics, reflective masks are excluded. and mucus mask For each region, only the normal region is used to calculate the baseline brightness distribution;

[0029] S1.24, in the mucus mask Within the region, a scattering model is used for contrast compensation, and the original grayscale difference is recovered based on the polarization difference value;

[0030] S1.25, in the reflective mask The region utilizes cross-channel edge information from polarized images to repair edge breaks caused by overexposure.

[0031] As a further improvement to this technical solution, the sharpness evaluation and optimization unit generates sharpness optimization parameters through a multi-task learning model, including the following steps:

[0032] S2.1 Receive the decoupled and corrected key features output by the feature extraction unit, and the comprehensive sharpness score. ;

[0033] S2.2 Utilize a lightweight convolutional neural network to perform real-time analysis on the current field of view image, identify the main anatomical structures, and use the identification results as a workflow status signal vector;

[0034] S2.3. The feature vectors of the key features after decoupling and correction and the workflow state signal vector are fed together as input into the multi-task learning model;

[0035] S2.4. Based on the output of the multi-task learning model, generate the optimization strategy weight vector and the adjustment amount of the basic parameters;

[0036] S2.5. The basic parameter adjustment amount, the optimization strategy weight vector, and the sub-feature scores of each key feature are fused together to obtain the weighted final parameter adjustment amount. Final parameter adjustment amount These are the resolution optimization parameters.

[0037] As a further improvement to this technical solution, in step S2.2, a lightweight convolutional neural network is used to perform real-time analysis of the current field of view image to identify the main anatomical structures, including the following steps:

[0038] S2.21 Receive the preprocessed field-of-view image output by the image acquisition and preprocessing unit, and standardize the preprocessed field-of-view image;

[0039] S2.22. Input the standardized field-of-view image into a lightweight convolutional neural network;

[0040] S2.23. Extract multi-scale feature maps through convolution and pooling operations of a lightweight convolutional neural network;

[0041] S2.24. Input the multi-scale feature map into the classification module to identify the main anatomical structures;

[0042] S2.25. Perform one-hot encoding on the identification results to generate a workflow status signal vector.

[0043] As a further improvement to this technical solution, in S2.4, generating the optimized strategy weight vector specifically involves concatenating the decoupled and corrected key feature vector with the workflow state signal vector along the channel dimension to form a fused feature vector. This fused feature vector is then input into a shared deep neural network encoder to extract a deep shared representation containing the current image quality status and the current clinical task objective. Deep shared representation The input is fed into the clinical intent parsing decoder to generate an optimized strategy weight vector.

[0044] As a further improvement to this technical solution, the visual field enhancement execution unit receives the sharpness optimization parameters generated by the sharpness evaluation and optimization unit, maps the sharpness optimization parameters to an optimization scheme, executes the optimization scheme to generate the enhanced final visual field image, and outputs the final visual field image to the clinical display interface.

[0045] On the other hand, the present invention provides a method for enhancing visual clarity during gynecological clinical examination and diagnosis, based on any one of the above-described systems for enhancing visual clarity during gynecological clinical examination and diagnosis, comprising the following steps:

[0046] S3.1 Real-time acquisition of visual field images during gynecological examinations, and preprocessing of the visual field images;

[0047] S3.2 Extract key features from the preprocessed field-of-view image;

[0048] S3.3. Generate a sharpness score based on key features. Use polarization difference images and polarization average images to decouple and correct feature interference caused by mucus scattering and tissue reflection, generate an optimized sharpness score, and generate sharpness optimization parameters through a multi-task learning model.

[0049] S3.4 Enhance the clarity of the field of view image based on clarity optimization parameters.

[0050] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0051] 1. In this invention, by integrating a switchable linear polarization light source and a polarization-sensitive camera, 0° and 90° polarization images are simultaneously acquired, constructing polarization difference images and polarization average images. Utilizing the fundamental difference in polarization characteristics between mucus multiple scattering and tissue specular reflection, reflective masks and mucus masks are extracted respectively. Based on this, key features such as brightness, contrast, and edges are subjected to regional masking and physical model compensation (e.g., detecting mucus regions based on scattering entropy and implementing contrast restoration, and repairing overexposed edges using polarization channel edge consistency), effectively eliminating the coupling effect of physiological interference on image quality evaluation. Feature correction based on imaging physics mechanisms makes the sharpness score more realistically reflect the intrinsic details of the tissue, providing a reliable basis for subsequent precise enhancement.

[0052] 2. In this invention, a lightweight convolutional neural network is used to identify key anatomical structures such as the cervix, endometrium, and fallopian tube openings in real time, and generate workflow status signals. Combined with decoupled image quality features, a multi-task learning model jointly predicts the base adjustment amount and the optimization strategy weight vector. The final parameter adjustment comprehensively considers the degree of image quality defects (sub-feature scores), the diagnostic focus of the current examination site (e.g., cervical mode emphasizes brightness balance, endometrial mode emphasizes contrast and edges), and clinical intent priority, achieving intelligent enhancement that is "differentiated by location and adjusted according to lesion." This breaks through the limitations of traditional image enhancement algorithms. Figure 1 This approach overcomes the limitations of traditional diagnostic methods, improves doctors' visual experience and diagnostic efficiency at different stages of examination, and has high clinical adaptability and practicality. Attached Figure Description

[0053] Figure 1 This is an overall flowchart of the present invention;

[0054] The meanings of the labels in the diagram are as follows:

[0055] 1. Image acquisition and preprocessing unit; 2. Feature extraction unit; 3. Sharpness evaluation and optimization unit; 4. Field of view enhancement execution unit. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0057] Example 1: Please refer to Figure 1 As shown, a system for enhancing visual clarity during gynecological clinical examinations and diagnoses is provided, including:

[0058] The image acquisition and preprocessing unit 1 acquires visual field images during the gynecological examination in real time and preprocesses the visual field images.

[0059] In this embodiment, the image acquisition and preprocessing unit 1 includes an image acquisition module and an image processing module;

[0060] The image acquisition module acquires visual field images during gynecological examinations based on the optical illumination module. The optical illumination module provides a stable, uniform, and adjustable lighting environment to illuminate the tissue to be examined. Based on this illuminated tissue, the image acquisition module acquires visual field images during the gynecological examination through an endoscopic lens. It also integrates a switchable linear polarization light source (0° and 90° polarization directions) and a polarization-sensitive camera (equipped with orthogonal polarizers) to simultaneously acquire two polarization images of the same visual field. and ;

[0061] The image processing module preprocesses the acquired field-of-view images, including noise reduction, brightness and contrast correction, gamma correction, edge-preserving filtering, and shadow compensation, to obtain higher-quality standardized images.

[0062] Feature extraction unit 2 extracts key features from the preprocessed field-of-view image;

[0063] In this embodiment, the feature extraction unit 2 includes a brightness distribution module, a contrast module, an edge sharpness module, and a blur level module;

[0064] The brightness distribution module is used to statistically analyze the overall brightness histogram of the preprocessed visual field image to obtain brightness distribution features that reflect the overall brightness of the image. Specifically, the preprocessed visual field image is converted into a grayscale image, and then the overall brightness histogram of the image is calculated, that is, the number of pixels appearing in the image at each gray level is counted to obtain the grayscale distribution of the image. Then, the histogram can be normalized to obtain the probability distribution of each gray level, thereby forming a brightness distribution feature vector that reflects the overall brightness of the image.

[0065] The contrast module is used to analyze the grayscale differences in local areas of the visual field image and extract contrast features that reflect the distinguishability of tissue details in the visual field image. Specifically, the preprocessed visual field image is divided into several local windows or sub-regions, and the grayscale variance or grayscale difference index of each local region is calculated to quantify the degree of change in pixel brightness in the region and reflect the recognizability of tissue details. Then, the contrast indices of all local regions are summarized or weighted to form an overall contrast feature vector.

[0066] The edge sharpness module uses an edge detection method to extract edge sharpness features to characterize the clarity of the outline of the visual field image. Specifically, it applies the edge detection operator (Sobel) to the preprocessed visual field image to extract the edge response in the image and obtain the edge intensity information of each pixel. Then, it statistically analyzes or normalizes these edge intensities to form an edge sharpness feature vector.

[0067] The blur level module calculates blur level features using frequency domain analysis to reflect blurred areas in the field of view image caused by inaccurate focusing. Specifically, the blur level module's processing procedure is as follows: First, a two-dimensional fast Fourier transform is performed on the preprocessed field of view image to convert the image from the spatial domain to the frequency domain; then, the proportion of high-frequency components in the spectrum to the total energy is calculated. High-frequency components reflect the details and sharpness of the image, while low-frequency components mainly correspond to smooth areas; the blur level features are obtained by inverse quantization of this proportion.

[0068] The sharpness evaluation and optimization unit 3 generates a sharpness score based on key features, uses polarization difference images and polarization average images to decouple and correct feature interference caused by mucus scattering and tissue reflection, generates an optimized sharpness score, and generates sharpness optimization parameters through a multi-task learning model;

[0069] In this embodiment, during gynecological clinical examinations, mucus adhesion can cause semi-transparent coverings in localized areas of the image, leading to light scattering and blurred details; tissue reflection can create bright areas, resulting in local overexposure and loss of detail. These two types of interference significantly distort brightness distribution, contrast, and edge sharpness features, causing deviations in sharpness scores. Traditional methods typically treat these two as a unified image quality degradation phenomenon, employing global enhancement strategies (such as histogram equalization or general deblurring), but these fail to distinguish their physical causes, instead leading to noise amplification or detail distortion. This system highlights reflections (dominated by specular reflection) using polarization difference images and mucus (dominated by multiple scattering) using polarization average images, combining scattering entropy to quantify the degree of local disorder, achieving physical separation and precise localization of the two types of interference regions, providing a reliable basis for subsequent feature correction. Based on the decoupled masks (reflection mask and mucus mask), separate repair strategies are designed: in the reflective area, the broken contour is repaired using polarization cross-channel edge consistency; in the mucus area, grayscale loss is compensated using a scattering model. This partitioned optimization strategy avoids the side effects of global processing and significantly improves the accuracy of detail recovery.

[0070] Sharpness evaluation and optimization unit 3 generates a sharpness score based on key features, including the following steps:

[0071] S1.1, Based on two polarization images and Calculate polarization difference image ( ), and polarization-averaged images ( ), where polarization difference image Polarization-averaged image used to highlight reflective areas Used to highlight the mucus scattering area;

[0072] S1.2 Utilizing the difference in polarization characteristics between slime (multiple scattering) and reflection (specular reflection), slime regions are detected based on scattering entropy and reflection regions are detected based on polarization difference, respectively. The extracted key features are decoupled and corrected to generate decoupled and corrected key features. The extracted key features include brightness distribution features, contrast features, edge sharpness features, and blur degree features. The decoupled and corrected key features include optimized brightness distribution features, contrast features, edge sharpness features, and the original blur degree features.

[0073] The process of generating the decoupling-corrected key features includes the following steps:

[0074] S1.21, Based on polarization difference image Brightness threshold Segmentation and extraction of reflective mask It is used to locate and mark overexposed areas in the field of view caused by tissue reflection;

[0075] Segmentation and extraction of reflective mask The formula is (for polarization difference images) Each pixel in To make a judgment, the polarization difference image At pixel grayscale value ( If greater than or equal to the reflectivity threshold Then in the binary mask The reflective area is marked as 1 (reflective area) if it is not, and 0 (non-reflective area) otherwise; reflective threshold The reflectivity threshold in this embodiment can be preset using statistical experimental data. The value is 30. This formula segments bright areas by setting a brightness threshold on the polarization difference image. When the pixel gray value is greater than the threshold, it is determined to be a reflective area, thereby generating a reflective mask. Its function is to effectively identify overexposed interference areas on the tissue surface caused by specular reflection, preparing for subsequent correction.

[0076] ;

[0077] S1.22, Polarization-averaged image Calculate the local scattering entropy; the local scattering entropy value is higher than the mucus threshold. The area is denoted as the mucus mask. It is used to detect areas blurred by mucus scattering.

[0078] The local scattering entropy is calculated as follows:

[0079] For each pixel in the image Take one of its surroundings (For example Local window And calculate the local scattering entropy value: In the formula, The grayscale level (e.g., 256). grayscale value In local window The probability of occurrence within the window (i.e., frequency divided by the total number of pixels in the window). The entropy value represents the local scattering entropy, reflecting the degree of grayscale disorder in the local area. The entropy value increases due to the multiple scattering effect in the slime-covered area. The purpose is to accurately detect the scattering blurry area caused by slime coverage, thereby isolating this type of interference.

[0080] Generate a slime mask Specifically, this involves (the local scattering entropy value of each pixel) With mucus threshold Compared to the mucus threshold, This is considered a mucus region and marked as 1 in the mask. In this embodiment, the mucus threshold is... (3, determined by experiment)

[0081] ;

[0082] S1.23. In the overall brightness histogram statistics, reflective masks are excluded. and mucus mask The baseline luminance distribution is calculated using only normal regions. This step optimizes the luminance features to avoid distorting the luminance features in interfering regions, making the luminance evaluation more realistically reflect the inherent condition of the tissue.

[0083] S1.24, in the mucus mask Within the region, a scattering model is used for contrast compensation. The original grayscale difference is restored based on the polarization difference value. This step optimizes the contrast features. Based on the linear scattering model of polarization physics mechanism, the direct component information is extracted from the polarization difference image to compensate for the scattering loss in the polarization average image, thereby accurately and efficiently reversing the contrast reduction problem caused by multiple scattering of biological mucosa.

[0084] The scattering model is as follows (this model recovers the direct grayscale information weakened by scattering through polarization difference values, which is used to improve detail discrimination and correct the contrast reduction problem caused by slime):

[0085] ;

[0086] For each pixel within the mucus region, its compensated grayscale value Polarization averaged image The value is obtained by adding a compensation term to the polarization difference value; the compensation term is the same as the polarization difference value. Linear dependence (parameters) This is the gain coefficient. In this embodiment, the offset is... Less than 1, to avoid overcompensation. Set to 0 or a tiny positive value. and (All were calibrated experimentally); because It reflects the direct component information that is not contaminated by scattering (i.e., the portion of the light signal that is directly reflected or transmitted to the camera after illuminating the tissue surface, without undergoing multiple scattering by mucus or other translucent media), and can be used to partially reverse the contrast loss caused by scattering; the compensated image This will be used to recalculate the local contrast characteristics of the region;

[0087] S1.25, in the reflective mask Regions utilize cross-channel edge information from polarization images (such as...) and (Edge Consistency) Repairing edge breaks caused by overexposure. This step optimizes edge clarity features and restores tissue contours lost due to overexposure.

[0088] The specific process for repairing edge breakage caused by overexposure is as follows: Two polarization images are extracted separately using the Sobel operator. and Edge: , Define the edge consistency function ( In the formula, (To be a very small constant to prevent division by zero errors); in reflective areas Inside, if a pixel is in If the edge response is weak but the edge consistency is high, then the edge is considered to be real but suppressed by overexposure and should be enhanced or repaired. The repaired edge map will be used to generate more accurate edge sharpness features.

[0089] , ;

[0090] In the formula, As the edge consistency threshold, based on experimental statistics, in this embodiment, It is 0.8; This is the edge intensity map after repair.

[0091] S1.3 Normalize the key features after decoupling correction;

[0092] S1.4. Assign preset weight coefficients based on the importance of each decoupled and corrected key feature in image sharpness evaluation;

[0093] S1.5 Calculate the sub-feature scores of the key features respectively. The sub-feature scores of the key features include brightness balance score, contrast score, edge sharpness score and blur score, which reflect the image quality of each single feature dimension.

[0094] Among them, the brightness uniformity score for: ;

[0095] Contrast Rating for: ;

[0096] Edge sharpness score for: ;

[0097] Ambiguity score for: ;

[0098] In the formula, The mean gray level of the field-of-view image is derived from regions excluding those marked by reflective and slime masks. This is the ideal brightness value (the empirical value is the median of the grayscale range). The adjustment constant for controlling the decay rate (determined experimentally, set to 50 in this embodiment) is used. The global grayscale standard deviation of the field of view image is derived from the field of view image after compensation and recovery by the scattering model in the normal region or mucus region (the compensation and recovery by the scattering model is specifically step S1.24). The adjustment constant for controlling the growth rate (determined experimentally, and set to 20 in this embodiment) is used. This is the edge intensity map after repair. This represents the total number of edge pixels in the field of view image, used for normalization to prevent the number of edges from affecting the average intensity. The energy of the high-frequency components (usually defined as the portion outside the central region of the spectrum, i.e., the sum of squares of the spectral amplitudes) is the energy of the high-frequency components in the spectrum obtained by performing a Fast Fourier Transform on the visual field image. The total energy of the entire spectrum. The coordinates are two-dimensional coordinates in the spectrum, representing the frequency domain position of the image after the Fourier transform;

[0099] S1.6. Weighted fusion of the sub-feature scores of the key features according to their weights to output a comprehensive clarity score. This is used to characterize the overall sharpness level of the current field-of-view image;

[0100] Among them, the overall clarity score for:

[0101] ;

[0102] In the formula, This is a weighting coefficient for brightness uniformity. The weighting factor for contrast. This is a weighting coefficient for edge sharpness. The weighting coefficient for ambiguity. , , , , All determined by experiments;

[0103] In this embodiment, the score is used to quantify the quality level of the current field-of-view image in four dimensions: brightness balance, contrast, edge sharpness, and blur. By calculating the sub-feature scores and fusing them into a comprehensive sharpness score, the overall image sharpness can be intuitively reflected. In the optimization process, the system will sort the defects of each dimension according to the score results. The lower the score value, the more serious the problem in that dimension, and the higher the corresponding adjustment priority.

[0104] Furthermore, the sharpness evaluation and optimization unit 3 generates sharpness optimization parameters through a multi-task learning model, including the following steps:

[0105] S2.1 Receive the decoupled and corrected key features output by feature extraction unit 2, and the comprehensive sharpness score. ;

[0106] S2.2 Utilize a lightweight convolutional neural network (CNN) to perform real-time analysis on the current field of view image, identify the main anatomical structures (including the cervix, endometrium, and fallopian tube openings), and use the identification results as a workflow status signal vector;

[0107] Traditional image enhancement systems employ a one-size-fits-all, fixed optimization strategy, which cannot adapt to the drastically different image quality requirements of various anatomical sites (such as the cervix, uterine cavity, and fallopian tubes) in gynecological clinical examinations. This system uses the clinical diagnostic intent (workflow status) as a key input and achieves scene-aware adaptive optimization through a multi-task learning model. Most existing technologies only mechanically enhance low-level features of the image itself (such as low contrast and low brightness). In contrast, this system uses a lightweight CNN to identify the main anatomical structures in the current field of view in real time and fuses this workflow status signal with image features. A deep learning model then infers which quality dimensions should be prioritized for enhancement (e.g., assigning higher weights to contrast and edge sharpness in uterine cavity mode). This transforms the optimization strategy from blindly generalized to targeted and precise, truly matching the diagnostic needs of doctors.

[0108] Real-time analysis of the current field-of-view image using a lightweight convolutional neural network (CNN) to identify key anatomical structures includes the following steps:

[0109] S2.21 Receive the preprocessed field-of-view image output by the image acquisition and preprocessing unit 1, and standardize the preprocessed field-of-view image to adjust it to a standard size (e.g., 224×224 pixels) and number of channels suitable for CNN input;

[0110] S2.22. Input the standardized field-of-view image into a lightweight convolutional neural network (CNN). This structure includes 3 to 5 convolutional layers, with the number of kernels in each layer ranging from 16, 32 to 64. After convolution, batch normalization layers and ReLU activation layers are connected in sequence. Max pooling layers are added after some convolutional layers to reduce the feature map size and extract multi-scale features. At the end of the network, one or two fully connected layers map the multi-scale features to the main anatomical structure categories as output. The entire CNN can be trained from scratch or lightly pre-trained on datasets such as ImageNet to accelerate convergence and supports the standardization of input image size 224×224×3.

[0111] S2.23. Extract multi-scale feature maps by using convolution and pooling operations of a lightweight convolutional neural network (CNN) to preserve the texture and edge information of the main anatomical structures;

[0112] S2.24. Input the multi-scale feature map into the classification module to identify the main anatomical structures, including the cervix, endometrium, and fallopian tube openings.

[0113] S2.25. Perform one-hot encoding on the recognition results to generate a workflow status signal vector: The workflow status signal vector is: [1,0,0] represents the cervical pattern, [0,1,0] represents the uterine cavity pattern, and [0,0,1] represents the fallopian tube pattern;

[0114] S2.3. The feature vectors of the decoupled and corrected key features and the workflow state signal vector are fed into the multi-task learning model as input. The architecture of the multi-task learning model is an end-to-end neural network structure. The input is the decoupled and corrected key feature vectors and the one-hot encoded workflow state signal. Common feature representations are extracted through several shared convolutional or fully connected layers, and then divided into multiple task branches: one branch is used for the sharpness regression task to output a comprehensive sharpness score, one branch is used for the interference discrimination task to identify reflective or slime areas, and one branch is used for the parameter prediction task to output the parameter adjustment amount of brightness, contrast, edge sharpness and blur. Each task branch is jointly trained through shared features and task-specific layers to achieve multi-dimensional field-of-view image sharpness optimization.

[0115] S2.4. Based on the output of the multi-task learning model, an optimization strategy weight vector and basic parameter adjustment amount are generated. This optimization strategy weight vector, obtained from the multi-task learning model, reflects the relative importance priority of the four optimization dimensions—brightness, contrast, edge sharpness, and blurriness—under the current clinical workflow. For example, in cervical mode, the weight vector might be [brightness: 0.5, contrast: 0.3, edge: 0.1, blurriness: 0.1], emphasizing overall appearance; while in uterine mode, the vector becomes [brightness: 0.2, contrast: 0.4, edge: 0.4, blurriness: 0.0]. In uterine mode, more attention is usually paid to endometrial texture and opening details, therefore contrast and edge features have higher weights, while blurriness correction has lower weights. The basic parameter adjustment amount is a general, unweighted optimization suggestion, including brightness adjustment values. Contrast adjustment value Edge adjustment value Deblurring adjustment value It is a standard adjustment scheme given by the model based on the quality of the image features (brightness, contrast, etc.). For example, if the model detects that the image is too dark overall, it will output a basic adjustment value to increase the brightness.

[0116] Furthermore, the generation of the optimization strategy weight vector specifically involves concatenating the decoupled and corrected key feature vector with the workflow state signal vector along the channel dimension to form a fused feature vector. This fused feature vector is then input into a shared deep neural network encoder (a multi-layer fully connected network) to extract a deep shared representation containing the current image quality status and the current clinical task objective. Deep shared representation The input is fed into the clinical intent parsing decoder to generate an optimized policy weight vector (this clinical intent parsing decoder is a simple fully connected network followed by a softmax activation layer; this clinical intent parsing decoder will share representations). The weights are mapped to a 4-dimensional vector through one or two fully connected layers, corresponding to the four dimensions of brightness, contrast, edge sharpness, and blur. This 4-dimensional vector is then normalized using the Softmax function to ensure that the sum of the four weights is 1. This results in a standard, interpretable probability distribution of the optimization strategy weight vector, which includes weights for brightness adjustment, contrast adjustment, edge adjustment, and blur adjustment.

[0117] The shared deep neural network encoder architecture is a multi-layer fully connected network. The input is a concatenated fused feature vector, which includes the decoupled and corrected key feature vector and the current workflow state signal vector. After passing through multiple fully connected layers (each layer is followed by normalization and non-linear activation functions (such as ReLU)) to progressively extract high-dimensional features and generate a deep shared representation, the intermediate layers contain dimensionality reduction layers or residual connections to enhance feature representation capabilities. The output is a deep shared representation, which is typically a low-dimensional continuous vector, such as 128-dimensional or 256-dimensional, used to uniformly encode image quality and clinical task information.

[0118] S2.5. The basic parameter adjustment amount, the optimization strategy weight vector, and the sub-feature scores of each key feature are fused together to obtain the weighted final parameter adjustment amount. Final parameter adjustment amount These are the resolution optimization parameters;

[0119] Final parameter adjustment amount for:

[0120] ;

[0121] In the formula, Indicates the weight of brightness adjustment. Indicates the weight of contrast adjustment. Indicates the weight of edge adjustment. Indicates the weight for ambiguity adjustment;

[0122] The final parameter adjustments include brightness adjustment coefficient, contrast gain coefficient, edge enhancement coefficient, and deblurring parameters. This final parameter adjustment is a personalized, weighted final execution plan that considers not only the quality of the visual field image itself but also the current clinical intent (workflow status) and the current scores of various features. Different examination sites (cervix, uterine cavity, fallopian tubes) have different emphasis requirements for visual field image quality: Cervical examination: greater focus on overall appearance and color uniformity (high weight for brightness); Uterine cavity examination: greater focus on endometrial texture and details of minute lesions (high weight for contrast and edge sharpness). If only the general parameter adjustments output by the model are used, then indiscriminate optimization will be performed for all scenarios, failing to best assist doctors in making the current key diagnosis. Sub-feature scores quantitatively reflect the current image's status level in various quality dimensions, and the system needs to determine the optimization intensity based on the current status. For example, if the contrast score of the current image is already high (e.g., 0.9), then even if the weight of the uterine mode indicates that contrast should be emphasized, the system will only make minor adjustments or maintain it, because it is already good enough, and over-optimization will introduce noise; conversely, if the blur score of the current image is very low (the image is very blurry), and the parameter adjustment also suggests deblurring, then the system will use a stronger deblurring filter.

[0123] The field-of-view enhancement execution unit 4 enhances the clarity of the field-of-view image based on the clarity optimization parameters;

[0124] In this embodiment, the visual field enhancement execution unit 4 receives the sharpness optimization parameters generated by the sharpness evaluation and optimization unit 3, maps the sharpness optimization parameters to an optimization scheme, executes the optimization scheme to generate the enhanced final visual field image, and outputs the final visual field image to the clinical display interface.

[0125] The optimization scheme is as follows: Wiener filtering is applied to the image based on the deblurring parameters to reverse the defocus blur; then, the coefficients are adjusted according to the brightness and contrast, and the overall brightness and darkness levels and grayscale differences between tissues are improved through linear transformation; finally, based on the edge enhancement coefficient, adaptive unsharpening masking technology is used to sharpen the tissue contours and texture details, and the intensity of each operation is weighted, fused and adaptively adjusted according to the specific scenario of gynecological examination (such as uterine examination mode).

[0126] The optimization scheme is mapped to a series of specific device instructions and algorithm parameters: for example, deblurring parameters are converted into instructions controlling the kernel size (e.g., 5x5 pixels) and noise power ratio (NSR = 0.02~0.1) of the real-time Wiener filter; contrast gain coefficients are converted into adjusting the slope of the programmable lookup table (LUT) curve, mapped to a dynamic range (e.g., 12-bit data of 0-4096); and edge enhancement coefficients are converted into controlling the weights and strengths of the convolution kernel (e.g., high-pass filter kernel) of the spatial convolution filter. Finally, all processed image data is output to a clinical monitor via a high-definition video interface (e.g., HDMI or SDI) in a standard medical video format (e.g., 1080p, 30fps), thus completing the entire process from optimization parameters to physical image enhancement.

[0127] Example 2: The difference between Example 2 and Example 1 is that this example introduces the clarity enhancement method used in the visual field clarity enhancement system during gynecological clinical examination and diagnosis.

[0128] A method for enhancing visual clarity during gynecological clinical examination and diagnosis, based on the aforementioned system for enhancing visual clarity during gynecological clinical examination and diagnosis, includes the following steps:

[0129] S3.1 Real-time acquisition of visual field images during gynecological examinations, and preprocessing of the visual field images;

[0130] S3.2 Extract key features from the preprocessed field-of-view image;

[0131] S3.3. Generate a sharpness score based on key features. Use polarization difference images and polarization average images to decouple and correct feature interference caused by mucus scattering and tissue reflection, generate an optimized sharpness score, and generate sharpness optimization parameters through a multi-task learning model.

[0132] S3.4 Enhance the clarity of the field of view image based on clarity optimization parameters.

[0133] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A system for enhancing visual clarity during gynecological clinical examination and diagnosis, characterized in that, include: Image acquisition and preprocessing unit (1), wherein the image acquisition and preprocessing unit (1) acquires visual field images during the gynecological examination in real time and preprocesses the visual field images; Feature extraction unit (2), which extracts key features from the preprocessed visual field image; The sharpness evaluation and optimization unit (3) generates a sharpness score based on key features, uses polarization difference image and polarization average image to decouple and correct feature interference caused by mucus scattering and tissue reflection, generates an optimized sharpness score, and generates sharpness optimization parameters through a multi-task learning model. The field of view enhancement execution unit (4) enhances the clarity of the field of view image based on the clarity optimization parameters.

2. The visual clarity enhancement system during gynecological clinical examination and diagnosis as described in claim 1, characterized in that: The image acquisition and preprocessing unit (1) includes an image acquisition module and an image processing module; The image acquisition module acquires visual field images during gynecological examinations based on the optical illumination module, and simultaneously acquires two polarized images of the same visual field. and ; The image processing module preprocesses the acquired field-of-view images.

3. The visual field enhancement system for gynecological clinical examination and diagnosis as described in claim 2, characterized in that: The feature extraction unit (2) includes a brightness distribution module, a contrast module, an edge sharpness module, and a blur degree module; The brightness distribution module is used to perform statistical analysis on the overall brightness histogram of the preprocessed visual field image to obtain brightness distribution features that reflect the overall brightness of the image. The contrast module is used to analyze the grayscale differences in local areas of the visual field image and extract contrast features that reflect the distinguishability of tissue details in the visual field image. The edge sharpness module uses an edge detection method to extract edge sharpness features, which are used to characterize the clarity of the outline of the visual field image. The blur level module calculates blur level features using frequency domain analysis to reflect the blurred areas in the field of view image caused by inaccurate focusing.

4. The visual clarity enhancement system during gynecological clinical examination and diagnosis as described in claim 3, characterized in that: The sharpness evaluation and optimization unit (3) generates a sharpness score based on key features, including the following steps: S1.1, Based on two polarization images and Calculate polarization difference image and polarization average image ; S1.2 Utilizing the difference in polarization characteristics between mucus and reflection, the mucus region is detected based on scattering entropy and the reflection region is detected based on polarization difference, respectively. The extracted key features are decoupled and corrected to generate decoupled and corrected key features. The extracted key features include brightness distribution features, contrast features, edge sharpness features and blur degree features. S1.3 Normalize the key features after decoupling correction; S1.

4. Assign preset weight coefficients based on the importance of each decoupled and corrected key feature in image sharpness evaluation; S1.5 Calculate the sub-feature scores of the key features respectively. The sub-feature scores of the key features include brightness balance score, contrast score, edge sharpness score and blur score, which reflect the image quality of each single feature dimension. S1.

6. Weighted fusion of the sub-feature scores of the key features according to their weights to output a comprehensive clarity score. It is used to characterize the overall sharpness level of the current field of view image.

5. The visual clarity enhancement system during gynecological clinical examination and diagnosis as described in claim 4, characterized in that: In step S1.2, generating the decoupling-corrected key features includes the following steps: S1.21, Based on polarization difference image Brightness threshold Segmentation and extraction of reflective mask ; S1.22, Polarization-averaged image Calculate the local scattering entropy; the local scattering entropy value is higher than the mucus threshold. The area is denoted as the mucus mask. ; S1.

23. In the overall brightness histogram statistics, reflective masks are excluded. and mucus mask For each region, only the normal region is used to calculate the baseline brightness distribution; S1.24, in the mucus mask Within the region, a scattering model is used for contrast compensation, and the original grayscale difference is recovered based on the polarization difference value; S1.25, in the reflective mask The region utilizes cross-channel edge information from polarized images to repair edge breaks caused by overexposure.

6. The visual field enhancement system for gynecological clinical examination and diagnosis as described in claim 4, characterized in that: The sharpness evaluation and optimization unit (3) generates sharpness optimization parameters through a multi-task learning model, including the following steps: S2.1 Receive the decoupled and corrected key features output by the feature extraction unit (2), and the comprehensive sharpness score. ; S2.2 Utilize a lightweight convolutional neural network to perform real-time analysis on the current field of view image, identify the main anatomical structures, and use the identification results as a workflow status signal vector; S2.

3. The feature vectors of the key features after decoupling and correction and the workflow state signal vector are fed together as input into the multi-task learning model; S2.

4. Based on the output of the multi-task learning model, generate the optimization strategy weight vector and the adjustment amount of the basic parameters; S2.

5. The basic parameter adjustment amount, the optimization strategy weight vector, and the sub-feature scores of each key feature are fused together to obtain the weighted final parameter adjustment amount. Final parameter adjustment amount These are the resolution optimization parameters.

7. The visual clarity enhancement system during gynecological clinical examination and diagnosis as described in claim 6, characterized in that: In step S2.2, a lightweight convolutional neural network is used to perform real-time analysis of the current field-of-view image and identify the main anatomical structures, including the following steps: S2.21 Receive the preprocessed field-of-view image output by the image acquisition and preprocessing unit (1) and standardize the preprocessed field-of-view image; S2.

22. Input the standardized field-of-view image into a lightweight convolutional neural network; S2.

23. Extract multi-scale feature maps through convolution and pooling operations of a lightweight convolutional neural network; S2.

24. Input the multi-scale feature map into the classification module to identify the main anatomical structures; S2.

25. Perform one-hot encoding on the identification results to generate a workflow status signal vector.

8. The visual clarity enhancement system during gynecological clinical examination and diagnosis as described in claim 7, characterized in that: In S2.4, generating the optimization strategy weight vector specifically involves concatenating the decoupled and corrected key feature vector with the workflow state signal vector along the channel dimension to form a fused feature vector. This fused feature vector is then input into a shared deep neural network encoder to extract a deep shared representation containing the current image quality status and the current clinical task objective. Deep shared representation The input is fed into the clinical intent parsing decoder to generate an optimized strategy weight vector.

9. The visual clarity enhancement system during gynecological clinical examination and diagnosis according to claim 1, characterized in that: The visual field enhancement execution unit (4) receives the clarity optimization parameters generated by the clarity evaluation and optimization unit (3), maps the clarity optimization parameters to an optimization scheme, executes the optimization scheme to generate the enhanced final visual field image, and outputs the final visual field image to the clinical display interface.

10. A method for enhancing visual clarity during gynecological clinical examination and diagnosis, based on the visual clarity enhancement system for gynecological clinical examination and diagnosis as described in any one of claims 1-9, characterized in that, Includes the following steps: S3.1 Real-time acquisition of visual field images during gynecological examinations, and preprocessing of the visual field images; S3.2 Extract key features from the preprocessed field-of-view image; S3.

3. Generate a sharpness score based on key features. Use polarization difference images and polarization average images to decouple and correct feature interference caused by mucus scattering and tissue reflection, generate an optimized sharpness score, and generate sharpness optimization parameters through a multi-task learning model. S3.4 Enhance the clarity of the field of view image based on clarity optimization parameters.

Images