Dynamic PET image enhancement technology system based on reference point

Abstract

The invention provides a dynamic PET image enhancement technology system based on a reference point, and belongs to the technical field of medical imaging. Comprising an image acquisition module, a data preprocessing module, a reference point identification module, a dynamic texture analysis module, a dynamic enhancement module and an output module, the image acquisition module is responsible for receiving original PET image data; the data preprocessing module is used for preprocessing original PET image data; the reference point identification module identifies an optimal dynamic reference point; the dynamic texture analysis module analyzes texture changes in the image data in real time; the dynamic enhancement module optimizes the quality of the image data; and the output module outputs the optimized image data to a terminal user. According to the invention, focus change can be tracked more accurately, the treatment effect can be evaluated, and the lesion area can be detected in an early stage, so that the diagnostic value and clinical application effect of the image are improved.

Description

Technical Field

[0001] The invention relates to the technical field of medical imaging, and in particular to a dynamic PET image enhancement technology system based on reference points. Background Art

[0002] Positron emission tomography (PET) is an important medical imaging technology that is widely used in tumor detection, heart disease diagnosis, and neurological disease research. PET images provide high-resolution functional imaging information by detecting the distribution of radioactive tracers in the body. However, due to the influence of factors such as patient movement, heartbeat, and breathing, PET images are prone to blur and noise during the acquisition process, which not only affects the clarity and accuracy of the image, but may also lead to misdiagnosis and missed diagnosis. In addition, dynamic PET images face complex texture changes and reference point selection problems in practical applications, requiring more precise image processing technology to improve image quality and diagnostic results.

[0003] At present, commonly used PET image enhancement technologies mainly include image denoising, image registration and texture enhancement methods. However, these traditional methods have the following shortcomings when processing dynamic PET images: First, image denoising methods (such as Gaussian filtering and median filtering) are prone to cause loss of detail information, affecting the clarity and accuracy of the image. Second, the existing image registration technology is difficult to accurately identify and align the best reference points when dealing with dynamic changes, resulting in insufficient image stability. In addition, the texture enhancement method has limited ability to capture dynamic changes and cannot fully reflect subtle changes in the image. These shortcomings limit the effectiveness of existing technologies in improving the quality of dynamic PET images and diagnostic accuracy.

[0004] Therefore, the present application provides a reference point-based dynamic PET image enhancement technology system to meet the needs. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a reference point-based dynamic PET image enhancement technology system to solve the existing problems in processing dynamic PET images, such as image denoising leading to loss of detail information, difficulty in accurately identifying and aligning the best reference points for image registration, and limited ability of texture enhancement methods to capture dynamic changes.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: The reference point-based dynamic PET image enhancement technology system includes an image acquisition module, a data preprocessing module, a reference point recognition module, a dynamic texture analysis module, a dynamic enhancement module and an output module, wherein; The image acquisition module is responsible for receiving original PET image data; The data preprocessing module preprocesses the original PET image data, including removing noise and interference; The reference point identification module automatically analyzes the pre-processed image data and identifies the best dynamic reference point; The dynamic texture analysis module performs real-time analysis on texture changes in image data, specifically including: Texture feature extraction: Use the spatiotemporal convolutional neural network (ST-CNN) to perform spatiotemporal convolution operations to extract dynamically changing texture features in image data; Texture modeling: Based on the extracted dynamically changing texture features, a texture change model is established to analyze the dynamic texture information in the image data; Texture change detection: Use change detection algorithms to analyze texture changes in dynamic texture information and generate change maps; The dynamic enhancement module optimizes the quality of the image data based on the real-time analysis results of the identified optimal dynamic reference points and texture changes; The output module outputs the optimized image data to the terminal user.

[0007] Optionally, the image acquisition module includes: Scanning device: used to connect to the PET scanning device and receive the original PET image data transmitted from the PET scanning device; Data transmission: The received raw PET image data is transmitted through a high-speed data transmission protocol; Data cache: used to temporarily store the received raw PET image data.

[0008] Optionally, the data preprocessing module includes: Noise filtering: Use Gaussian filtering algorithm to filter noise from raw PET image data; Interference removal: Use the adaptive median filtering algorithm to remove interference signals from image data.

[0009] Optionally, the reference point identification module includes: Image segmentation: Use the region growing algorithm to segment the preprocessed image data and extract the region of interest (ROI). The calculation formula is: ; in, represents the growth area, is the image at position The pixel value at is the preset threshold; Feature extraction: Extract features from the extracted region of interest, including edge, texture and shape features; Reference point selection: The extracted features are classified and evaluated using the support vector machine (SVM) classification algorithm to select the best dynamic reference point.

[0010] Optionally, the spatiotemporal convolutional neural network (ST-CNN) includes: Multi-resolution spatiotemporal convolution: Use multi-resolution convolution kernels to perform convolution operations simultaneously. The calculation formula is: ; in, is the output feature map at time and spatial location The value at is the input feature map at time and spatial location The value at Different resolutions The convolution kernel is offset in time and spatial offset The weight of Different resolutions The bias term, is the number of channels of the input feature map, is the number of different resolutions, Different resolutions The size of the convolution kernel in the spatial and temporal dimensions; Weight sharing mechanism: A weight sharing mechanism is introduced in the spatiotemporal convolution process. The calculation formula is: ; in, Different resolutions The convolution kernel is offset in time and spatial offset The shared weight at Temporal attention mechanism: Combined with the temporal attention mechanism, the convolution weights in the time dimension are dynamically adjusted. The calculation formula is: ; in, is the temporal attention coefficient, which is dynamically adjusted through training to improve the sensitivity to dynamically changing features.

[0011] Optionally, the texture modeling includes: Acquisition of dynamic texture feature sequence: Extract dynamic texture features that change over time in image data to form a dynamic texture feature sequence ,in Indicates at time Extracted texture feature vector; Model building: Use dynamic Bayesian network (DBN) to build a texture change model; Analysis of dynamic texture information: Use the established texture change model to analyze the texture information in the image data that changes over time.

[0012] Optionally, the dynamic Bayesian network (DBN) includes: Fusion of spatiotemporal features: The texture feature vector extracted by the spatiotemporal convolutional neural network (ST-CNN) Fused into the dynamic Bayesian network (DBN), the calculation formula is: ; in, is the fused feature vector, It's time The texture feature vector of is the spatial eigenvector, is the fusion function; Adaptive kernel density estimation: Adaptive kernel density estimation is introduced, and the calculation formula is: ; in, It's time The hidden state of is the weight coefficient, is the kernel function, is the kernel width, is the number of kernel functions; Multi-head attention mechanism: Combined with the multi-head attention mechanism, the calculation formula is: ; in, is the query vector, is the key vector, is a vector of values, is the number of attention heads, is the dimension of the key vector; Multi-scale analysis: A multi-scale analysis mechanism is introduced to handle dynamic texture changes at different time scales. The calculation formula is: ; in, It's time The fused feature vector, is the number of time scales.

[0013] Optionally, the analysis of the dynamic texture information includes: Texture information acquisition: Extract the texture feature sequence that changes over time in the image data to form a dynamic texture feature vector. The calculation formula is: ; in, Represents different texture features; Model prediction: predicting time through the dynamic Bayesian network (DBN) model Texture features of the generated texture feature vector , the calculation formula is: ; in, Indicates that based on the previous The conditional probability distribution of a time point; Analyze texture information over time: by comparing actual texture features and predict texture features , analyze the texture information that changes over time in the image data and generate the texture change analysis results. The calculation formula is: ; in, It's time The texture change vector.

[0014] Optionally, the texture change detection includes: Get texture change vector: Get texture change vector by comparing actual texture features with predicted texture features ; Change detection: Mahalanobis distance is used as a change detection algorithm to analyze texture changes in dynamic texture information. The calculation formula is: ; in, It's time The change detection value of is the covariance matrix; Generate change map: Visualize the detected texture changes and generate a change map for image enhancement and diagnosis. The calculation formula is: ; in, is the change graph over time and spatial location The value at is the variation graph generating function.

[0015] Optionally, the dynamic enhancement module includes: Reference point adjustment: based on the best dynamic reference point and the texture change vector The reference point of the image data is adjusted, and the calculation formula is: ; in, It's time The adjusted image data, It's time Original image data, and is the offset of the reference point adjustment; Texture enhancement: Leveraging the results of texture change detection The texture enhancement of image data is calculated as follows: ; in, It's time The enhanced image data, is the texture enhancement coefficient; Multi-scale fusion: Combine the multi-scale analysis results to fuse image data at different scales. The calculation formula is: ; in, It's time The final optimized image data, It's time In the The enhanced image data at each scale is It is The weight of the scale, is the number of scales.

[0016] Compared with the prior art, the present invention has at least the following beneficial effects: In the above scheme, through the image acquisition module and the data preprocessing module, it is possible to ensure that high-quality original image data is obtained from the PET scanning device, and through the Gaussian filtering and adaptive median filtering algorithms, noise and interference are removed to provide high-quality input data, which provides a reliable foundation for subsequent image processing and enhancement. By automatically identifying the best dynamic reference points and performing real-time analysis of texture changes in image data, the dynamic detail performance and diagnostic accuracy of PET images are significantly improved, and lesion changes can be tracked more accurately, treatment effects can be evaluated, and lesion areas can be detected early, thereby improving the diagnostic value of images and clinical application effects.

[0017] By using spatiotemporal convolutional neural networks to extract texture features, dynamic Bayesian networks to model and analyze texture changes, and Mahalanobis distance to detect texture changes, the accuracy of image analysis can be effectively improved, ensuring that the optimal reference point is always located during dynamic changes, significantly improving the image enhancement effect and diagnostic accuracy, it also reduces human intervention, enhances the system's automation and intelligence level, and improves the overall system performance and diagnostic accuracy.

[0018] Image quality is optimized through three steps: reference point adjustment, texture enhancement, and multi-scale fusion, ensuring image alignment and stability, enhancing detail expression and clarity, and fusing enhanced images at different scales to generate a high-quality final image. This comprehensive optimization method enables the system to maintain high-performance image quality under various dynamic changes, significantly improving the detail expression and diagnostic accuracy of dynamic PET images, and providing more reliable and accurate image data for subsequent medical analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings, which are incorporated herein and constitute a part of the specification, illustrate embodiments of the invention and, together with the description, further serve to explain the principles of the invention and to enable those skilled in the relevant art to make and use the invention.

[0020] Figure 1 It is a schematic diagram of the functional modules of the dynamic PET image enhancement technology system based on reference points; Figure 2 Schematic diagram of the dynamic texture analysis module of the reference point-based dynamic PET image enhancement technology system. DETAILED DESCRIPTION

[0021] The following is a detailed description of the reference point-based dynamic PET image enhancement technology system provided by the present invention in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0022] It should be noted that the references to "one embodiment", "embodiment", "exemplary embodiments", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0023] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0024] It will be understood that the meanings of “on,” “over,” and “above” in the present invention should be interpreted in the broadest manner, so that “on” not only means “directly on” something, but also includes the meaning of being “on” something with intervening features or layers therebetween, and “on” or “over” not only means “on” or “above” something, but also includes the meaning of being “on” or “above” something with no intervening features or layers therebetween.

[0025] Additionally, spatially relative terms such as "under," "beneath," "lower," "above," "upper," and the like may be used herein for descriptive convenience to describe the relationship of one element or feature to another element or features, as shown in the accompanying drawings. Spatially relative terms are intended to encompass different orientations of the device in use or operation in addition to the orientation depicted in the accompanying drawings. The device may be oriented in other ways, and the spatially relative descriptors used herein may be similarly interpreted accordingly.

[0026] like Figure 1 and Figure 2 As shown, an embodiment of the present invention provides a reference point-based dynamic PET image enhancement technology system, including an image acquisition module, a data preprocessing module, a reference point recognition module, a dynamic texture analysis module, a dynamic enhancement module and an output module, wherein; The image acquisition module is responsible for receiving the original PET image data; The data preprocessing module preprocesses the raw PET image data, including removing noise and interference, to provide high-quality input data for subsequent processing; The reference point recognition module automatically analyzes the pre-processed image data and identifies the best dynamic reference points; The dynamic texture analysis module analyzes the texture changes in image data in real time to improve the dynamic details and diagnostic accuracy of the image, including: Texture feature extraction: Use the spatiotemporal convolutional neural network (ST-CNN) to perform spatiotemporal convolution operations to extract dynamically changing texture features in image data; Texture modeling: Based on the extracted dynamically changing texture features, a texture change model is established to analyze the dynamic texture information in the image data; Texture change detection: Use change detection algorithms to analyze texture changes in dynamic texture information and generate change maps; The dynamic enhancement module optimizes the quality of image data based on the identification of the best dynamic reference points and real-time analysis of texture changes; The output module outputs the optimized image data to the end user; Through the above content, the dynamic detail expression and diagnostic accuracy of PET images have been significantly improved, which can more accurately track lesion changes, evaluate treatment effects, and detect lesion areas at an early stage, thereby improving the diagnostic value and clinical application effect of images.

[0027] The image acquisition module includes: Scanning device: used to connect to the PET scanning device and receive the original PET image data transmitted from the PET scanning device; Data transmission: The received raw PET image data is transmitted through a high-speed data transmission protocol; Data cache: used to temporarily store the received raw PET image data to ensure the integrity and stability of the data during transmission; Through the above content, it is ensured that high-quality original image data is obtained from the PET scanning device, which provides a reliable foundation for subsequent data preprocessing and image enhancement, thereby improving the performance and diagnostic accuracy of the overall system.

[0028] The data preprocessing module includes: Noise filtering: Use the Gaussian filtering algorithm to filter the noise of the original PET image data. The calculation formula is: ; in, is in position The Gaussian filter value, is the standard deviation of the Gaussian function, which determines the width of the filter, is the coordinate position in the image; Interference removal: Adaptive median filtering algorithm is used to remove interference signals in image data. The calculation formula is: ; in, Indicates surrounding position Window, is the pixel value within the window; Through the above content, the high quality of input data is ensured. The Gaussian filter reduces noise by smoothing the image, while the adaptive median filter removes interference signals by calculating the median of the local window, which significantly improves the reliability and accuracy of subsequent image enhancement and analysis, and provides clearer and more accurate image data.

[0029] The reference point identification module includes: Image segmentation: Use the region growing algorithm to segment the preprocessed image data and extract the region of interest (ROI). The calculation formula is: ; in, represents the growth area, is the image at position The pixel value at is the preset threshold; Feature extraction: Extract features from the extracted region of interest, including edge, texture and shape features; The calculation formula of edge features is: ; The calculation formula of texture features is: ; ; ; The calculation formula of shape characteristics is: ; ; Reference point selection: The extracted features are classified and evaluated using the support vector machine (SVM) classification algorithm to select the best dynamic reference point; The calculation formula of the support vector machine (SVM) classification algorithm is: ; in, is the classification decision function, is the Lagrange multiplier, is the category label of the training sample, is the kernel function, is the support vector, is the bias term, sgn is the sign function used to determine the classification result; Kernel Function Using the radial basis function (RBF) kernel, the calculation formula is: ; in, is the value of the RBF kernel function, is the support vector, is the input vector, is the parameter of the RBF kernel function, usually , is the kernel width parameter, express and The Euclidean distance between Through the above content, the accuracy of image analysis is effectively improved, ensuring that the optimal reference point is always located during dynamic changes, significantly improving the image enhancement effect and diagnostic accuracy, reducing human intervention, and improving the system's automation and intelligence level.

[0030] The spatiotemporal convolutional neural network (ST-CNN) includes: Multi-resolution spatiotemporal convolution: Use multi-resolution convolution kernels to perform convolution operations simultaneously. The calculation formula is: ; in, is the output feature map at time and spatial location The value at is the input feature map at time and spatial location The value at Different resolutions The convolution kernel is offset in time and spatial offset The weight of Different resolutions The bias term, is the number of channels of the input feature map, is the number of different resolutions, Different resolutions The size of the convolution kernel in the spatial and temporal dimensions; Weight sharing mechanism: A weight sharing mechanism is introduced in the spatiotemporal convolution process. The calculation formula is: ; in, Different resolutions The convolution kernel is offset in time and spatial offset The shared weight at Temporal attention mechanism: Combined with the temporal attention mechanism, the convolution weights in the time dimension are dynamically adjusted. The calculation formula is: ; in, is the temporal attention coefficient, which is dynamically adjusted through training to improve the sensitivity to dynamically changing features; Through the above content, the spatiotemporal convolutional neural network can more accurately extract the dynamically changing texture features in PET image data through multi-resolution spatiotemporal convolution, weight sharing mechanism and temporal attention mechanism. It can not only capture details of different scales, but also dynamically adjust the weight of the time dimension to improve sensitivity and stability to dynamic changes. Using spatiotemporal convolutional neural network for texture feature extraction helps to improve the dynamic detail performance and diagnostic accuracy of images, and provide high-quality feature data for reference point recognition and image enhancement, thereby significantly improving the overall performance and clinical application value of the system.

[0031] Texture modeling includes: Acquisition of dynamic texture feature sequence: Extract dynamic texture features that change over time in image data to form a dynamic texture feature sequence ,in Indicates at time Extracted texture feature vector; Model building: Use dynamic Bayesian network (DBN) to build a texture change model; Analysis of dynamic texture information: Using the established texture change model, the texture information in the image data that changes over time is analyzed to provide a reliable basis for subsequent image enhancement and diagnosis; Through the above content, the time dependence and complex dynamic relationship of texture features can be effectively captured and modeled, the ability to analyze dynamic changes in image data is improved, and a more reliable and accurate basis is provided for subsequent image enhancement and diagnosis, thereby significantly improving the overall performance of the system and its clinical application value.

[0032] Dynamic Bayesian Network (DBN) includes: Fusion of spatiotemporal features: The texture feature vector extracted by the spatiotemporal convolutional neural network (ST-CNN) Fused into the dynamic Bayesian network (DBN), the calculation formula is: ; in, is the fused feature vector, It's time The texture feature vector of is the spatial eigenvector, is the fusion function; Fusion Function The calculation formula is: ; in, and is the fusion coefficient, satisfying ; Adaptive kernel density estimation: Adaptive kernel density estimation is introduced, and the calculation formula is: ; in, It's time The hidden state of is the weight coefficient, is the kernel function, is the kernel width, is the number of kernel functions; Multi-head attention mechanism: Combined with the multi-head attention mechanism, the calculation formula is: ; in, is the query vector, is the key vector, is a vector of values, is the number of attention heads, is the dimension of the key vector; Multi-scale analysis: A multi-scale analysis mechanism is introduced to handle dynamic texture changes at different time scales. The calculation formula is: ; in, It's time The fused feature vector, is the number of time scales; Through the above content, the dynamic Bayesian network can better combine the dynamically changing texture features extracted by ST-CNN, improve the ability to analyze dynamic changes in PET image data, and provide a more reliable and accurate basis for subsequent image enhancement and diagnosis.

[0033] The analysis of dynamic texture information includes: Texture information acquisition: Extract the texture feature sequence that changes over time in the image data to form a dynamic texture feature vector. The calculation formula is: ; in, Represents different texture features; Model prediction: predicting time through the dynamic Bayesian network (DBN) model Texture features of the generated texture feature vector , the calculation formula is: ; in, Indicates that based on the previous The conditional probability distribution of time points; Analyze texture information over time: by comparing actual texture features and predict texture features , analyze the texture information that changes over time in the image data and generate the texture change analysis results. The calculation formula is: ; in, It's time The texture change vector of Through the above steps, the analysis of dynamic texture information can utilize the established texture change model to analyze in detail the texture information that changes over time in the image data, providing a reliable basis for subsequent image enhancement and diagnosis.

[0034] Texture change detection includes: Get texture change vector: Get texture change vector by comparing actual texture features with predicted texture features ; Change detection: Mahalanobis distance is used as a change detection algorithm to analyze texture changes in dynamic texture information. The calculation formula is: ; in, It's time The change detection value of is the covariance matrix; Generate change map: Visualize the detected texture changes and generate a change map for image enhancement and diagnosis. The calculation formula is: ; in, is the change graph over time and spatial location The value at is the change graph generation function, based on the change detection value and spatial location Generate a change graph; Through the above content, the use of Mahalanobis distance for texture change detection can fully consider the covariance matrix of the data and accurately quantify the correlation and difference between multidimensional features. The Mahalanobis distance is more robust and accurate when processing multidimensional data with different scales and correlations. The present invention can more effectively capture subtle changes in texture change detection of dynamic PET images, improve the accuracy and robustness of change detection, and thus provide a more reliable basis for image enhancement and diagnosis.

[0035] Dynamic enhancement modules include: Reference point adjustment: based on the best dynamic reference point and the texture change vector The reference point of the image data is adjusted, and the calculation formula is: ; in, It's time The adjusted image data, It's time Original image data, and is the offset of the reference point adjustment; Texture enhancement: Leveraging the results of texture change detection The texture enhancement of image data is calculated as follows: ; in, It's time The enhanced image data, is the texture enhancement coefficient; Multi-scale fusion: Combine the multi-scale analysis results to fuse image data at different scales. The calculation formula is: ; in, It's time The final optimized image data, It's time In the The enhanced image data at each scale is It is The weight of the scale, is the number of scales; Through the above content, combined with the real-time analysis results of the best dynamic reference points and texture changes, the image quality is optimized in three steps: reference point adjustment, texture enhancement and multi-scale fusion. The image is effectively aligned and stabilized, the detail expression and clarity are enhanced, and enhanced images at different scales are fused to generate a high-quality final image. The detail expression and diagnostic accuracy of dynamic PET images are significantly improved, providing more reliable and accurate image data for subsequent medical analysis.

[0036] The working principle provided by the present invention is as follows: the original data of the PET scanning device is received through the image acquisition module, and the noise is filtered and the interference is removed through the data preprocessing module to improve the image data quality; the reference point recognition module automatically recognizes the best dynamic reference point through image segmentation, feature extraction and support vector machine classification algorithm; the dynamic texture analysis module uses the spatiotemporal convolutional neural network (ST-CNN) to extract dynamic texture features, and establishes a texture change model through the dynamic Bayesian network (DBN), analyzes the texture change in real time, and generates a change map using the change detection algorithm; the dynamic enhancement module adjusts the reference point and enhances the texture according to the recognized reference point and the real-time analysis result, and combines the multi-scale analysis to perform image fusion, and finally optimizes the image quality. The optimized image data is output to the terminal user through the output module. Through the collaborative work of the above modules, the system significantly improves the dynamic detail performance and diagnostic accuracy of the PET image, can accurately track the change of the lesion, evaluate the treatment effect, and detect the lesion area at an early stage, thereby improving the diagnostic value and clinical application effect of the image.

[0037] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0038] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A reference point-based dynamic PET image enhancement technology system, characterized in that: It includes an image acquisition module, a data preprocessing module, a reference point recognition module, a dynamic texture analysis module, a dynamic enhancement module and an output module, wherein; The image acquisition module is responsible for receiving original PET image data; The data preprocessing module preprocesses the original PET image data, including removing noise and interference; The reference point identification module automatically analyzes the pre-processed image data and identifies the best dynamic reference point; The dynamic texture analysis module performs real-time analysis on texture changes in image data, specifically including: Texture feature extraction: Use spatiotemporal convolutional neural networks to perform spatiotemporal convolution operations to extract dynamically changing texture features in image data; Texture modeling: Based on the extracted dynamically changing texture features, a texture change model is established to analyze the dynamic texture information in the image data; Texture change detection: Use change detection algorithms to analyze texture changes in dynamic texture information and generate change maps; The dynamic enhancement module optimizes the quality of the image data based on the real-time analysis results of the identified optimal dynamic reference points and texture changes; The output module outputs the optimized image data to the terminal user.

2. The reference point-based dynamic PET image enhancement technology system according to claim 1, characterized in that: The image acquisition module comprises: Scanning device: used to connect to the PET scanning device and receive the original PET image data transmitted from the PET scanning device; Data transmission: The received raw PET image data is transmitted through a high-speed data transmission protocol; Data cache: used to temporarily store the received raw PET image data.

3. The reference point-based dynamic PET image enhancement technology system according to claim 2, characterized in that: The data preprocessing module comprises: Noise filtering: Use Gaussian filtering algorithm to filter noise from raw PET image data; Interference removal: Use the adaptive median filtering algorithm to remove interference signals from image data.

4. The reference point-based dynamic PET image enhancement technology system according to claim 1, characterized in that: The reference point identification module comprises: Image segmentation: Use the region growing algorithm to segment the preprocessed image data and extract the region of interest. The calculation formula is: ； in, represents the growth area, is the image at position The pixel value at is the preset threshold; Feature extraction: Extract features from the extracted region of interest, including edge, texture and shape features; Reference point selection: The extracted features are classified and evaluated using the support vector machine classification algorithm to select the best dynamic reference point.

5. The reference point-based dynamic PET image enhancement technology system according to claim 4, characterized in that: The spatiotemporal convolutional neural network comprises: Multi-resolution spatiotemporal convolution: Use multi-resolution convolution kernels to perform convolution operations simultaneously. The calculation formula is: ； in, is the output feature map at time and spatial location The value at is the input feature map at time and spatial location The value at Different resolutions The convolution kernel is offset in time and spatial offset The weight of Different resolutions The bias term, is the number of channels of the input feature map, is the number of different resolutions, Different resolutions The size of the convolution kernel in the spatial and temporal dimensions; Weight sharing mechanism: A weight sharing mechanism is introduced in the spatiotemporal convolution process. The calculation formula is: ； in, Different resolutions The convolution kernel is offset in time and spatial offset The shared weight at Temporal attention mechanism: Combined with the temporal attention mechanism, the convolution weights in the time dimension are dynamically adjusted. The calculation formula is: ； in, is the temporal attention coefficient, which is dynamically adjusted through training to improve the sensitivity to dynamically changing features.

6. The reference point-based dynamic PET image enhancement technology system according to claim 5, characterized in that: The texture modeling includes: Acquisition of dynamic texture feature sequence: Extract dynamic texture features that change over time in image data to form a dynamic texture feature sequence ,in Indicates at time Extracted texture feature vector; Model building: Use dynamic Bayesian network to build texture change model; Analysis of dynamic texture information: Use the established texture change model to analyze the texture information in the image data that changes over time.

7. The reference point-based dynamic PET image enhancement technology system according to claim 6, characterized in that: The dynamic Bayesian network includes: Fusion of spatiotemporal features: The texture feature vector extracted by the spatiotemporal convolutional neural network Fused into the dynamic Bayesian network, the calculation formula is: ； in, is the fused feature vector, It's time The texture feature vector of is the spatial eigenvector, is the fusion function; Adaptive kernel density estimation: Adaptive kernel density estimation is introduced, and the calculation formula is: ； in, It's time The hidden state of is the weight coefficient, is the kernel function, is the kernel width, is the number of kernel functions; Multi-head attention mechanism: Combined with the multi-head attention mechanism, the calculation formula is: ； in, is the query vector, is the key vector, is a vector of values, is the number of attention heads, is the dimension of the key vector; Multi-scale analysis: A multi-scale analysis mechanism is introduced to handle dynamic texture changes at different time scales. The calculation formula is: ； in, It's time The fused feature vector, is the number of time scales.

8. The reference point-based dynamic PET image enhancement technology system according to claim 7, characterized in that: The analysis of the dynamic texture information includes: Texture information acquisition: Extract the texture feature sequence that changes over time in the image data to form a dynamic texture feature vector. The calculation formula is: ； in, Represents different texture features; Model prediction: predicting time through dynamic Bayesian network model Texture features of the generated texture feature vector , the calculation formula is: ； in, Indicates that based on the previous The conditional probability distribution of a time point; Analyze texture information over time: by comparing actual texture features and predict texture features , analyze the texture information that changes over time in the image data and generate the texture change analysis results. The calculation formula is: ； in, It's time The texture change vector.

9. The reference point-based dynamic PET image enhancement technology system according to claim 8, characterized in that: The texture change detection comprises: Get texture change vector: Get texture change vector by comparing actual texture features with predicted texture features ; Change detection: Mahalanobis distance is used as a change detection algorithm to analyze texture changes in dynamic texture information. The calculation formula is: ； in, It's time The change detection value of is the covariance matrix; Generate change map: Visualize the detected texture changes and generate a change map for image enhancement and diagnosis. The calculation formula is: ； in, is the change graph over time and spatial location The value at is the variation graph generating function.

10. The reference point-based dynamic PET image enhancement technology system according to claim 1, characterized in that: The dynamic enhancement module comprises: Reference point adjustment: based on the best dynamic reference point and the texture change vector The reference point of the image data is adjusted, and the calculation formula is: ； in, It's time The adjusted image data, It's time Original image data, and is the offset of the reference point adjustment; Texture enhancement: Leveraging the results of texture change detection The texture enhancement of image data is calculated as follows: ； in, It's time The enhanced image data, is the texture enhancement coefficient; Multi-scale fusion: Combine the multi-scale analysis results to fuse image data at different scales. The calculation formula is: ； in, It's time The final optimized image data, It's time In the The enhanced image data at each scale is It is The weight of the scale, is the number of scales.

Images