A urinary calculus CT image enhancement processing method

Abstract

The application discloses a kind of urinary calculi CT image enhancement processing methods, it is related to image processing technical field, the method includes: from DICOM file exports initial CT image, and utilizes K sparse angle to initial CT image is sampled, obtains K sampling CT image;With K sparse angle as index, CT artifact image and artifact mapping search are carried out, and obtain sample CT artifact image-sample artifact group set;Construct artifact identifier;Artifact identifier is used to the artifact identification of K sampling CT image, and according to the identification result, K sampling CT image is separated to strip artifact, obtains K separate sampling CT image;K separate sampling CT image is carried out iteration correlation interactive analysis, and obtains target enhanced CT image.The application solves the technical problem that urinary calculi CT image exists artifact in the prior art, cannot clearly identify stone area.The technical effect of enhancing image quality is achieved.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and more specifically to a method for enhancing CT images of urinary calculi. Background Technology

[0002] Urinary calculi are a common disease of the urinary system, and CT images, as a diagnostic tool, are widely used for the detection and localization of stones. However, during low-dose CT scanning, sparse sampling angles often lead to a decrease in image quality and produce severe streak artifacts, affecting the accurate diagnosis of stones.

[0003] Existing CT image enhancement methods typically suffer from several drawbacks when processing low-dose or sparsely sampled CT images: firstly, they cannot effectively identify and remove streak artifacts in the images; secondly, the enhancement effect on urinary calculi areas is unsatisfactory, leading to loss or blurring of details in the calculi area. Especially when the CT image projection angle is small, traditional image enhancement methods often struggle to balance artifact removal with preservation of the calculi area, affecting the effectiveness of clinical applications. Therefore, how to effectively remove artifacts and enhance image quality in calculi areas has become an urgent problem to be solved. Summary of the Invention

[0004] This application provides a method for enhancing CT images of urinary calculi, which addresses the technical problem of artifacts in existing CT images of urinary calculi, making it impossible to clearly identify the calculi area.

[0005] In view of the above problems, this application provides a method for enhancing CT images of urinary calculi, the method comprising:

[0006] Export the initial CT image from the DICOM file, and sample the initial CT image using K sparse angles to obtain K sampled CT images, where K is a positive integer;

[0007] Using the K sparse angles as indices, CT artifact images and artifact mappings are retrieved to obtain a sample CT artifact image-sample artifact group set.

[0008] A tool recognizer is constructed by mapping learning based on sample CT artifact images and sample artifact sets.

[0009] The artifact detector is used to identify artifacts in the K sampled CT images, and strip artifacts are separated from the K sampled CT images based on the identification results to obtain K separated sampled CT images;

[0010] Iterative correlation and interactive analysis is performed on the K separate sampled CT images to obtain the target enhanced CT image.

[0011] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0012] This application exports an initial CT image from a DICOM file and samples it using K sparse angles to obtain K sampled CT images (K being a positive integer). Then, using the K sparse angles as indices, CT artifact images and artifact mappings are retrieved to obtain a set of sample CT artifact images and sample artifact groups. Based on this set, mapping learning is performed to construct an artifact recognizer. The artifact recognizer is then used to identify artifacts in the K sampled CT images, and strip artifacts are separated from the K sampled CT images based on the recognition results to obtain K separated sampled CT images. Finally, iterative correlation and interactive analysis is performed on these K separated sampled CT images to obtain the target enhanced CT image. This achieves the technical effect of improving the quality of CT image enhancement processing. Attached Figure Description

[0013] Appendix Figure 1 This is a schematic diagram of a CT image enhancement processing method for urinary stones provided in an embodiment of the present invention.

[0014] Appendix Figure 2 This is a schematic diagram of the process for obtaining a sample CT artifact image-sample artifact set in a CT image enhancement processing method for urinary stones provided in an embodiment of the present invention. Detailed Implementation

[0015] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims.

[0016] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to these processes, methods, products, or devices.

[0017] Examples, as shown in the appendix Figure 1 As shown, this application provides a method for enhancing CT images of urinary calculi, wherein the method includes:

[0018] S1: Export the initial CT image from the DICOM file, and sample the initial CT image using K sparse angles to obtain K sampled CT images, where K is a positive integer;

[0019] In one possible embodiment, DICOM (Digital Imaging and Communications in Medicine) is a standard format in the field of medical imaging used for storing, transmitting, and managing medical image data. A DICOM file contains image data, patient information, examination details, and other information. The initial CT image refers to the raw CT scan image extracted from the DICOM file, typically containing a cross-sectional image of the patient's urinary system. The sparse angle refers to selecting a small number of projection angles for image sampling during the CT scan. Traditional CT scans use more angles to acquire image data, while sparse angles sample with fewer scanning angles, simulating low-dose CT.

[0020] Preferably, the initial CT image is first exported from the DICOM file to ensure the acquisition of raw medical imaging data. Then, the initial CT image is sampled by selecting K sparse angles. This operation reduces the number of sampling angles while retaining sufficient projection data to ensure the basic integrity of the image information. Here, K is the number of selected angles. The purpose of this step is to provide clean, limited, but representative input data for subsequent image enhancement and artifact suppression, simulating the situation of low-dose scanning, thereby ensuring that the entire processing flow conforms to the characteristics of low-dose CT images.

[0021] S2: Using the K sparse angles as indexes, perform CT artifact image and artifact mapping retrieval to obtain the sample CT artifact image-sample artifact group set;

[0022] In one possible embodiment, the CT artifact image refers to an image containing artifacts or unrealistic structures that appear due to insufficient sampling angle, equipment error, or algorithm limitations. In low-dose or sparse-angle scans, stripe artifacts are the most common, appearing as regular stripes or unnatural structures in the image. Artifacts are the artifact portions identified and extracted from the CT artifact image. In this step, a mapping retrieval of CT artifact images and artifacts is performed using K sparse angles as indices to obtain the sample CT artifact image-sample artifact set. Each sample CT artifact image-sample artifact set includes one sample CT artifact image and its corresponding sample artifact.

[0023] Each sample CT artifact image in the sample artifact group is paired with its corresponding artifact information (such as the shape and location of strip artifacts). This set is used to provide reference and training data for subsequent artifact recognition.

[0024] Furthermore, such as Figure 2As shown, using the K sparse angles as indices, CT artifact images and artifact mappings are retrieved to obtain a sample CT artifact image-sample artifact group set. Step S2 in this embodiment further includes:

[0025] K sparse angles are used as indices to retrieve CT artifact images and artifact mappings in big data, and an initial set of sample CT artifact images and sample artifact groups is obtained.

[0026] The initial sample CT artifact image-sample artifact group set is retrieved and certified for quality based on two dimensions: CT artifact image similarity and artifact similarity. If the certification is successful, the initial sample CT artifact image-sample artifact group set is used as the sample CT artifact image-sample artifact group set.

[0027] Furthermore, step S2 in this embodiment of the application also includes:

[0028] M sample CT artifact images and sample artifact groups are selected from the initial sample CT artifact image-sample artifact group set using a random, non-replacement method, where M is a positive integer;

[0029] Using M sample CT artifact images and sample artifact groups as M cluster centers, cluster gain analysis is performed on the initial sample CT artifact image and sample artifact group set from two dimensions: CT artifact image similarity and artifact similarity, to obtain the sample cluster gain quantity.

[0030] Determine whether the clustering gain of the sample is less than or equal to a preset clustering gain threshold. If so, the authentication is successful.

[0031] Furthermore, using M sample CT artifact images-sample artifact groups as M cluster centers, cluster gain analysis is performed on the initial sample CT artifact image-sample artifact group set from two dimensions: CT artifact image similarity and artifact similarity, to obtain the sample cluster gain. Step S2 in this embodiment further includes:

[0032] Obtain the sample clustering gain analysis function, wherein the sample clustering gain analysis function is:

[0033] ;

[0034] in, This is the sample clustering gain. for In the cluster centers, the first CT artifact images of clustered samples with cluster centers - a set of sample artifact groups. For the first CT artifact images of samples from each cluster center. For the first CT artifact images of clustered samples with cluster centers - the first sample artifact group set CT artifact images of individual clusters - CT artifact images of sample artifact groups For the first Sample artifacts at each cluster center, For the first CT artifact images of clustered samples with cluster centers - the first sample artifact group set CT artifact images of individual clusters - sample artifacts of sample artifact groups, To balance the weights of CT artifact image similarity and artifact similarity;

[0035] Based on M cluster centers, the initial sample CT artifact images-sample artifact group set is clustered to obtain M clustered sample CT artifact images-sample artifact group sets;

[0036] The cluster gain is obtained by performing cluster gain analysis on the M cluster centers and the M cluster sample CT artifact image-sample artifact group set using the sample cluster gain analysis function.

[0037] In one possible implementation, firstly, K sparse angles are selected as indices, which can simulate image data under low-dose CT scan conditions. Then, using these sparse angles, corresponding CT images are retrieved from a large dataset. These images may contain varying degrees of artifacts during their generation. These retrieved images, along with artifact information (e.g., stripe artifacts, structural artifacts, etc.), constitute an initial sample CT artifact image-sample artifact set. This set provides data support and an analytical foundation for subsequent image enhancement and artifact removal. This mapping retrieval effectively filters artifact-related samples from a large dataset, laying a solid foundation for subsequent artifact identification and removal.

[0038] Furthermore, to obtain a richer sample pool and avoid a small sample coverage that could negatively impact the training of the subsequent artifact recognizer, it is necessary to perform sample quality verification on the initial sample CT artifact image-sample artifact group set. If the verification is successful, meaning the initial sample CT artifact image-sample artifact group set contains a sufficient variety of sample types, then this initial sample CT artifact image-sample artifact group set can be used as the final sample CT artifact image-sample artifact group set. This provides data support for the subsequent training of the artifact recognizer.

[0039] If the authentication fails, that is, when the initial sample CT artifact image-sample artifact group set contains not enough sample types, it is necessary to retrieve data samples from the big data again using the K sparse angles as indexes, and re-authenticate the quality of the retrieved data samples.

[0040] Preferably, when authenticating the quality of retrieved samples, identification is primarily based on two dimensions: CT artifact image similarity and artifact similarity. Authentication is successful when the sample clustering gain, i.e., the overall similarity between the initial sample CT artifact image and the sample artifact group set, is low. CT artifact image similarity refers to the similarity between artifact regions and other known artifact images in a CT image. This similarity is typically evaluated by calculating the distance between image features (e.g., Euclidean distance). Higher similarity indicates a greater similarity between the artifact pattern in the current image and the artifact images in the sample CT artifact image-sample artifact group at the cluster center. Artifact similarity focuses on the degree of matching between artifact features in the image and sample artifacts in the sample CT artifact image-sample artifact group at the cluster center. Artifact similarity calculation involves analyzing the morphology, intensity, and other features of artifacts in the image, comparing the similarity between the current artifact and the reference artifact, and thus determining whether the types are similar.

[0041] In one possible implementation, random sampling without replacement is a commonly used random sampling method where samples are not replaced after being selected, so each sample has only one chance to be selected. Specifically, M sample CT artifact images / sample artifact groups are randomly selected from the initial set of sample CT artifact images / sample artifact groups, and the same sample CT artifact images / sample artifact groups are not selected again after each selection. The M randomly selected sample CT artifact images / sample artifact groups are used as M cluster centers. In clustering algorithms, cluster centers refer to the central points representing a certain class. In this step, the M selected sample CT artifact images / sample artifact groups are used as the initial centers for clustering. These centers are used in subsequent cluster analysis to help determine the similarity and distribution between samples. Cluster gain analysis refers to using image features (such as CT artifact image similarity and artifact similarity) to evaluate the clustering effect during the clustering process. The sample clustering gain refers to the increase in similarity between samples or the improvement in overall classification performance after clustering, measured by analyzing the clustering results. It reflects the effectiveness of the clustering operation. A higher sample clustering gain indicates that the initial CT artifact image-sample artifact group set has too few types and high similarity, which is not conducive to the subsequent training of the artifact recognizer.

[0042] Preferably, the preset clustering gain threshold is a standard value pre-set by those skilled in the art to determine whether the number and types of the sample set have achieved the expected effect. If the sample clustering gain is less than or equal to this threshold, the sample quality is considered to be high enough. When the sample clustering gain is less than the preset clustering gain threshold, it indicates that the sample dispersion is good and can pass the certification, proceeding to the next step of processing.

[0043] Optionally, the sample clustering gain analysis function is used to analyze the overall similarity of the initial sample CT artifact image-sample artifact group set from two dimensions: CT artifact image similarity and artifact similarity. The cosine similarity calculation formula is used to calculate the similarity between each initial sample CT artifact image-sample artifact group and each cluster center in the initial sample CT artifact image-sample artifact group set, and then assigns it to the class containing the cluster center with the highest similarity, thereby obtaining the M clustered sample CT artifact image-sample artifact group sets. The CT artifact images and their artifact information in each cluster set have high similarity on certain features (such as CT artifact image similarity and artifact similarity).

[0044] The purpose of clustering is to reduce data complexity and improve the efficiency of subsequent analysis by grouping samples. Through clustering, we can identify subsets with similar characteristics from the entire sample set, and then perform more detailed processing and analysis within these subsets. Ultimately, the resulting set of M clustered sample CT artifact images—sample artifact groups—provides high-quality, low-redundancy sample data for subsequent artifact identification and removal, making the processing more accurate and efficient.

[0045] Then, the sample clustering gain analysis function is used to quantify the overall similarity between the M cluster centers and the M cluster sample CT artifact image-sample artifact group set, and the sample clustering gain is obtained.

[0046] S3: Based on the mapping learning of sample CT artifact images and sample artifact sets, an artifact recognizer is constructed.

[0047] Furthermore, based on the mapping learning of sample CT artifact images and sample artifact sets, an artifact recognizer is constructed. Step S3 in this embodiment of the application also includes:

[0048] Each sample artifact in the sample CT artifact image-sample artifact set is labeled. The convolutional neural network-based framework is trained using the sample CT artifact image-sample artifact set. The learning process is supervised using the labeled sample artifacts. By minimizing the loss function, the trained artifact recognizer is obtained.

[0049] Furthermore, the loss function is:

[0050] ;

[0051] in, The loss amount is the output of the loss function. For sample artifacts, The artifacts output by the artifact detector. Let i be the sample artifact group of the i-th sample CT artifact image-sample artifact group in the sample CT artifact image-sample artifact group set. The artifact is the output of the artifact recognizer after recognizing the CT artifact image of the i-th sample CT artifact image-sample artifact group. N is the number of sample CT artifact images-sample artifact groups in the sample CT artifact image-sample artifact group set, and N is an integer less than or equal to 3.

[0052] In one embodiment, the artifact detector is a trained neural network model specifically designed to identify artifacts in CT images. It can distinguish between artifacts and normal structures in an image, and the identified artifacts can be further processed and removed to improve image quality. Convolutional neural networks (CNNs) are deep learning models widely used in image recognition tasks. By extracting image features through convolutional and pooling layers, they can effectively process image data and extract important information. The loss function measures the difference between the predicted result and the actual label. Its goal is to minimize the loss function so that the model's output is closer to the true label. In this context, the loss function measures the error between the artifact detector's output and the actual artifact label. By adjusting the model's weights, the output value (error) of the loss function is minimized. This goal is typically achieved through optimization algorithms such as gradient descent.

[0053] Preferably, sample artifacts are first extracted from a set of sample CT artifact images—sample artifact groups—and each sample artifact is labeled. This labeling operation provides supervision information for subsequent training, i.e., identifying which regions are artifacts and which are normal regions. Then, based on a Convolutional Neural Network (CNN) framework, these labeled sample CT artifact image-sample artifact group sets are trained. The CNN extracts artifact features from the original images through multiple convolutional and pooling operations and learns how to distinguish artifacts from normal structures. During training, the labeled sample artifacts provide supervision signals, helping the network learn the correct artifact recognition method. By minimizing a loss function (e.g., mean squared error loss), the training process continuously adjusts the parameters of the CNN network to improve the recognition accuracy of the artifact recognizer. The loss function measures the difference between the predicted artifact and the actual label; minimizing this loss means that the network can better recognize and process artifacts in the image.

[0054] Ultimately, the trained artifact detector will be able to accurately identify artifact regions in an image and provide support for subsequent image inpainting and enhancement. The core function of this process is to build an effective artifact recognition model, which can significantly improve the effect and accuracy of image artifact removal.

[0055] S4: Use the artifact recognizer to identify artifacts in the K sampled CT images, and perform strip artifact separation on the K sampled CT images according to the recognition results to obtain K separated sampled CT images;

[0056] In one embodiment, the K separated sampled CT images refer to the images after the strip-shaped artifacts in the K sampled CT images have been separated from the images. The separated images, free of artifacts, provide clearer CT image content. First, a pre-trained artifact recognizer is used to identify artifacts in the K sampled CT images. The artifact recognizer automatically detects potential artifact regions in each image. The role of the artifact recognizer is to identify the location of artifact regions in each CT image; these regions are usually labeled as artifacts for subsequent processing. After recognition, the strip-shaped artifacts in each sampled CT image are separated based on the artifact recognition results. Strip-shaped artifacts typically appear as lines or stripes and can affect the doctor's interpretation of the image in CT images. To improve image quality, the system uses image processing techniques (such as filtering, denoising, and image restoration) to separate these artifacts from the image.

[0057] Finally, K separated sampled CT images are output, which are CT images after artifacts have been removed or separated. These images will more closely resemble the actual lesion images and reduce interference from artifacts, thus providing clearer image data for further image analysis and diagnosis. The core function of this process is to significantly improve the quality of CT images through artifact identification and separation, making subsequent image enhancement or analysis more accurate.

[0058] S5: Perform iterative correlation and interactive analysis on the K separated sampled CT images to obtain the target enhanced CT image.

[0059] Furthermore, iterative correlation and interactive analysis is performed on the K separated sampled CT images to obtain the target enhanced CT image. In this embodiment, step S5 further includes:

[0060] Feature extraction is performed on the K separate sampled CT images to obtain a feature set of K separate sampled CT images;

[0061] Randomly extract a first set of separated sampled CT image features and a second set of separated sampled CT image features without replacement from the K sets of separated sampled CT image features;

[0062] Iterative correlation and interaction analysis is performed on the first set of separated sampled CT image features and the second set of separated sampled CT image features to obtain the first iterative correlation and interaction set of separated sampled CT image features.

[0063] The first iterative correlation interaction separated sampled CT image feature set is subjected to iterative correlation interaction analysis with the third separated sampled CT image feature set randomly extracted without replacement from the K separated sampled CT image feature sets to obtain the second iterative correlation interaction separated sampled CT image feature set.

[0064] After multiple iterations of correlation and interaction analysis, the feature set of the K-1th iteration correlation and interaction separated sampled CT image is obtained. The feature set is then analyzed with the Kth separated sampled CT image using a forward feedback network to obtain the target enhanced CT image.

[0065] Furthermore, iterative correlation and interaction analysis is performed on the first and second separated sampled CT image feature sets to obtain the first iterative correlation and interaction separated sampled CT image feature set. Step S5 in this embodiment further includes:

[0066] Perform inner product mapping analysis on the first and second separated sampled CT image feature sets to obtain the first iterative correlation interaction feature similarity set;

[0067] The first iterative association interaction feature similarity set is normalized, and the result of the normalization is filled into the initially empty matrix to obtain the first iterative association interaction matrix;

[0068] The first iterative correlation interaction matrix and the first iterative correlation interaction feature similarity set are convolved to obtain the first iterative correlation interaction separated sampled CT image feature set.

[0069] In one possible embodiment, since the K separately sampled CT images originate from K different sparse angles, the information scales reflected in the images differ. Smaller sparse angles result in images with larger, coarser information, but they reflect more contours. Conversely, larger sparse angles yield higher image detail, but contain excessively redundant information. Therefore, by performing iterative correlation and interaction analysis on the K separately sampled CT images, images sampled at different sparse angles are fused. Images with smaller sparse angles capture more large-scale structures, while images with larger sparse angles capture more details. This fusion of information at different scales effectively improves image detail and quality.

[0070] First, from K separate sampled CT image feature sets, a first separate sampled CT image feature set and a second separate sampled CT image feature set are randomly extracted without replacement. These two feature sets represent feature data from two sets of CT images, respectively. This operation provides two sets of benchmark data for subsequent feature association analysis. Next, iterative association and interaction analysis is performed on these two feature sets. Through iteration, the feature sets are continuously optimized and adjusted, allowing image features to gradually influence each other in the two sets of data, achieving a higher feature matching degree. The purpose of this process is to improve the similarity and accuracy of image features, especially in CT images of urinary stones, enabling the effective extraction of detailed features of the stones.

[0071] During iterative analysis, inner product mapping analysis is also performed to measure the similarity between feature sets. For example, the cosine similarity formula can be used for inner product mapping analysis. Through inner product operations, the similarity between two sets of features can be quantitatively evaluated, providing a basis for feature optimization. Then, the similarity results are normalized to ensure that the differences between different features are reasonably controlled, preventing bias in subsequent analyses.

[0072] Furthermore, by filling the matrix with the normalized similarity results and performing convolution operations, the effectiveness of feature extraction and analysis is further enhanced. Convolution operations not only facilitate feature extraction but also create finer correlations between different image features, thereby providing more accurate input data for the final image enhancement.

[0073] Ultimately, iterative correlation and interaction analysis refers to the repeated processing of image features through multiple steps during the processing. Each iteration updates the features based on the results of the previous round, gradually optimizing the extraction of image features. This analysis method can continuously improve the accuracy and effect of the target image through continuous adjustments. After obtaining the feature set of the K-1th iteration correlation and interaction separated sampling CT image, it is analyzed with the Kth separated sampling CT image using a feedforward feedback network to generate the target enhanced CT image. This process continuously optimizes the image features through a feedback mechanism, ultimately resulting in a target enhanced CT image that has undergone enhancement processing, artifact removal, and detail optimization.

[0074] Preferably, the framework constructed based on a feedforward neural network is trained under supervised conditions by acquiring multiple sample iteratively correlated and interactively separated sampled CT image feature sets, multiple sample separated sampled CT images, and corresponding multiple sample enhanced CT images as training data until convergence, thus obtaining the trained feedforward network. Then, the trained feedforward network is used to analyze the K-1th iteratively correlated and interactively separated sampled CT image feature set and the Kth separated sampled CT image to obtain the target enhanced CT image. This achieves the technical effect of obtaining CT images sampled from different sparse angles, optimizing image quality through iterative analysis and deep learning, and ultimately obtaining high-quality target enhanced CT images, thus improving the detection accuracy and image quality of urinary stones.

[0075] In summary, the embodiments of this application have at least the following technical effects:

[0076] 1. This application simulates low-dose CT scans by sampling the initial CT image at K sparse angles, avoiding image artifacts caused by excessive doses in traditional methods. Simultaneously, the artifact detector effectively identifies and removes stripe artifacts in the image, improving the clarity and accuracy of the final image.

[0077] 2. By obtaining a sample CT artifact image-sample artifact set from CT images and artifact mapping, and performing mapping learning based on this set, an artifact recognizer is constructed, thereby improving the accuracy of artifact recognition. Furthermore, through iterative correlation and interaction analysis, image features can be continuously optimized in each round of processing, making the final target-enhanced CT image more accurate and clear.

[0078] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0079] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0080] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A method for enhancing CT images of urinary calculi, characterized in that, The method includes: Export the initial CT image from the DICOM file, and sample the initial CT image using K sparse angles to obtain K sampled CT images, where K is a positive integer; Using the K sparse angles as indices, CT artifact images and artifact mappings are retrieved to obtain a sample CT artifact image-sample artifact group set. A tool recognizer is constructed by mapping learning based on sample CT artifact images and sample artifact sets. The artifact detector is used to identify artifacts in the K sampled CT images, and strip artifacts are separated from the K sampled CT images based on the identification results to obtain K separated sampled CT images; Iterative correlation and interaction analysis is performed on the K separate sampled CT images to obtain the target enhanced CT image; Specifically, using the K sparse angles as indices, CT artifact images and artifact mappings are retrieved to obtain a sample CT artifact image-sample artifact set, including: K sparse angles are used as indices to retrieve CT artifact images and artifact mappings in big data, and an initial set of sample CT artifact images and sample artifact groups is obtained. The initial sample CT artifact image-sample artifact group set is retrieved and certified for quality based on two dimensions: CT artifact image similarity and artifact similarity. If the certification is successful, the initial sample CT artifact image-sample artifact group set is used as the sample CT artifact image-sample artifact group set. M sample CT artifact images and sample artifact groups are selected from the initial sample CT artifact image-sample artifact group set using a random, non-replacement method, where M is a positive integer; Using M sample CT artifact images and sample artifact groups as M cluster centers, cluster gain analysis is performed on the initial sample CT artifact image and sample artifact group set from two dimensions: CT artifact image similarity and artifact similarity, to obtain the sample cluster gain quantity. Determine whether the clustering gain of the sample is less than or equal to a preset clustering gain threshold. If so, the authentication is successful.

2. The method for enhancing CT images of urinary calculi as described in claim 1, characterized in that, Iterative correlation and interaction analysis is performed on the K separated sampled CT images to obtain the target enhanced CT image, including: Feature extraction is performed on the K separate sampled CT images to obtain a feature set of K separate sampled CT images; Randomly extract a first set of separated sampled CT image features and a second set of separated sampled CT image features without replacement from the K sets of separated sampled CT image features; Iterative correlation and interaction analysis is performed on the first set of separated sampled CT image features and the second set of separated sampled CT image features to obtain the first iterative correlation and interaction set of separated sampled CT image features. The first iterative correlation interaction separated sampled CT image feature set is subjected to iterative correlation interaction analysis with the third separated sampled CT image feature set randomly extracted without replacement from the K separated sampled CT image feature sets to obtain the second iterative correlation interaction separated sampled CT image feature set. After multiple iterations of correlation and interaction analysis, the feature set of the K-1th iteration correlation and interaction separated sampled CT image is obtained. The feature set is then analyzed with the Kth separated sampled CT image using a forward feedback network to obtain the target enhanced CT image.

3. The method for enhancing CT images of urinary calculi as described in claim 2, characterized in that, Iterative correlation and interaction analysis is performed on the first and second separated sampled CT image feature sets to obtain the first iteratively correlated and interactive separated sampled CT image feature set, including: Perform inner product mapping analysis on the first and second separated sampled CT image feature sets to obtain the first iterative correlation interaction feature similarity set; The first iterative association interaction feature similarity set is normalized, and the result of the normalization is filled into the initially empty matrix to obtain the first iterative association interaction matrix; The first iterative correlation interaction matrix and the first iterative correlation interaction feature similarity set are convolved to obtain the first iterative correlation interaction separated sampled CT image feature set.

4. The method for enhancing CT images of urinary calculi as described in claim 1, characterized in that, Using M sample CT artifact images and sample artifact groups as M cluster centers, cluster gain analysis is performed on the initial sample CT artifact image and sample artifact group set from two dimensions: CT artifact image similarity and artifact similarity. The sample cluster gain is obtained, including: Obtain the sample clustering gain analysis function, wherein the sample clustering gain analysis function is: ； in, This is the sample clustering gain. for In the cluster centers, the first CT artifact images of clustered samples with cluster centers - a set of sample artifact groups. For the first CT artifact images of samples from each cluster center. For the first CT artifact images of clustered samples with cluster centers - the first sample artifact group set CT artifact images of individual clusters - CT artifact images of sample artifact groups For the first Sample artifacts at each cluster center, For the first CT artifact images of clustered samples with cluster centers - the first sample artifact group set CT artifact images of individual clusters - sample artifacts of sample artifact groups, To balance the weights of CT artifact image similarity and artifact similarity; Based on M cluster centers, the initial sample CT artifact images-sample artifact group set is clustered to obtain M clustered sample CT artifact images-sample artifact group sets; The cluster gain is obtained by performing cluster gain analysis on the M cluster centers and the M cluster sample CT artifact image-sample artifact group set using the sample cluster gain analysis function.

5. The method for enhancing CT images of urinary calculi as described in claim 1, characterized in that, Based on the mapping learning of sample CT artifact images and sample artifact sets, an artifact recognizer is constructed, including: Each sample artifact in the sample CT artifact image-sample artifact set is labeled. The convolutional neural network-based framework is trained using the sample CT artifact image-sample artifact set. The learning process is supervised using the labeled sample artifacts. By minimizing the loss function, the trained artifact recognizer is obtained.

6. The method for enhancing CT images of urinary calculi as described in claim 5, characterized in that, The loss function is: ； in, The loss amount is the output of the loss function. For sample artifacts, The artifacts output by the artifact detector. Let i be the sample artifact group of the i-th sample CT artifact image-sample artifact group in the sample CT artifact image-sample artifact group set. The artifact is the output of the artifact recognizer after recognizing the CT artifact image of the i-th sample CT artifact image-sample artifact group. N is the number of sample CT artifact images-sample artifact groups in the sample CT artifact image-sample artifact group set, and N is an integer less than or equal to 3.

Images