Dynamic reconstruction method and apparatus
By adjusting the iteration parameters based on the quantized value of the image quality index during the iterative reconstruction of dynamic images, the problem of poor dynamic image quality is solved, and high-quality dynamic reconstruction results are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANGHAI UNITED IMAGING HEALTHCARE
- Filing Date
- 2024-11-25
- Publication Date
- 2026-05-29
AI Technical Summary
The dynamic images reconstructed in the existing technology have the problem of poor image quality.
By iteratively reconstructing each frame of the initial dynamic image, and adjusting the iteration parameters based on the quantized values of the image quality indicators of each frame during each iteration, including the quantized values of region convergence and image resolution, a pre-trained neural network model is used to evaluate the degree of image convergence and optimize the iteration parameters to improve image quality.
This improved the overall quality of dynamically reconstructed images, ensuring consistency in convergence and resolution across frames, and meeting expected performance requirements.
Smart Images
Figure CN122115597A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of medical technology, and in particular to a dynamic reconstruction method and apparatus. Background Technology
[0002] Dynamic imaging is an imaging technique used to observe the process of a drug entering the body from injection until it reaches metabolic equilibrium. It can help determine the drug injection dose and time interval, and is of great significance for pharmacokinetics.
[0003] In related technologies, the dynamic reconstruction process typically divides the total scan into multiple frames, each with a corresponding scan duration, and uses the same reconstruction parameters for each frame. During dynamic reconstruction, these same reconstruction parameters are used to dynamically reconstruct the dynamic scan data of each frame to obtain a dynamic image.
[0004] However, the dynamic images reconstructed in related technologies suffer from poor image quality. Summary of the Invention
[0005] Therefore, it is necessary to provide a dynamic reconstruction method and apparatus to address the aforementioned technical problems, which can improve the image quality of dynamic images.
[0006] Firstly, this application provides a dynamic reconstruction method, including:
[0007] Based on the dynamic scanning data of the target object, obtain the initial dynamic image of the target object;
[0008] The initial dynamic image is reconstructed iteratively, and the iterative parameters for each frame reconstruction are adjusted according to the quantization value of the image quality index of each frame during each iteration.
[0009] Based on the iteration parameters of each frame image in each iteration, the dynamically reconstructed image of the target object is determined.
[0010] In one embodiment, during each iteration of reconstruction, the iteration parameters for reconstructing each frame are adjusted based on the quantized value of the image quality index of each frame, including:
[0011] For a single-iteration reconstruction, the expected image performance is obtained;
[0012] Based on the quantized values of the image quality metrics and the expected performance of each frame, adjust the convergence parameters for the reconstruction of each frame.
[0013] The adjusted convergence parameters are used as the iterative parameters for reconstructing each frame of the image.
[0014] In one embodiment, the image quality index quantization value includes a region convergence quantization value; the method further includes:
[0015] For a single-iteration reconstruction, each frame of the initial dynamic image is segmented to obtain multiple segmented image regions for each frame.
[0016] Based on multiple segmented image regions of each frame, determine the quantization value of the region convergence degree of each frame.
[0017] In one embodiment, image segmentation is performed on each frame of the initial dynamic image to obtain multiple segmented image regions for each frame, including:
[0018] Dimensionality reduction is performed on each frame of the image to obtain the dimension-reduced image corresponding to each frame.
[0019] Image segmentation is performed on the dimensionality-reduced images corresponding to each frame to obtain multiple segmented image regions for each frame.
[0020] In one embodiment, the region convergence quantization value of each frame image is determined based on multiple segmented image regions of each frame image, including:
[0021] Obtain pixel information of multiple segmented image regions in each frame;
[0022] Based on the pixel information of each segmented image region, determine the region value change curve of each segmented image region;
[0023] Based on the region value change curves of each segmented image region, the quantization value of the region convergence degree of each frame image is determined.
[0024] In one embodiment, determining the region value change curve of each segmented image region based on the pixel information of each segmented image region includes:
[0025] Based on the pixel information of each segmented image region, determine the image region value of each segmented image region;
[0026] Based on the image region values of each segmented image region and the image region values of each segmented image region over a historical time period, the region value change curve of each segmented image region is determined.
[0027] In one embodiment, the quantization value of the region convergence degree of each frame image is determined based on the region value change curve of each segmented image region, including:
[0028] The region value change curves of each segmented image region are input into the pre-trained convergence prediction model to obtain the quantified value of the region convergence degree of each segmented image region.
[0029] Based on the quantization value of the region convergence of each segmented image region, the quantization value of the region convergence of each frame image is determined.
[0030] In one embodiment, the training process of the convergent prediction model includes:
[0031] Obtain sample image region information and sample region convergence curves for multiple sample segmentation images corresponding to the dynamic scanning data of the samples;
[0032] The image region information and sample region convergence curve of each segmented image are input into a preset neural network for model training to obtain an initial model;
[0033] The initial model was validated to obtain a convergent prediction model.
[0034] In one embodiment, determining the dynamically reconstructed image of the target object based on the iteration parameters of each frame image in each iteration includes:
[0035] For a single iteration of reconstruction, the convergence of each frame image is determined based on the iteration parameters of each frame image in the iterative reconstruction.
[0036] If all frames converge, then each frame is determined as the dynamically reconstructed image of the target object;
[0037] If the images in each frame do not converge, the iterative reconstruction process is repeated until the images in each frame converge, thus obtaining a dynamically reconstructed image of the target object.
[0038] Secondly, this application also provides a dynamic reconstruction device, comprising:
[0039] The image acquisition module is used to acquire the initial dynamic image of the target object based on the dynamic scanning data of the target object;
[0040] The parameter adjustment module is used to iteratively reconstruct each frame of the initial dynamic image, and adjust the iterative parameters of each frame reconstruction according to the quantization value of the image quality index of each frame during each iteration reconstruction.
[0041] The image determination module is used to determine the dynamically reconstructed image of the target object based on the iteration parameters of each frame image in each iteration of reconstruction.
[0042] Thirdly, embodiments of this application also provide a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps in any of the embodiments of the first aspect described above.
[0043] Fourthly, embodiments of this application also provide a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the steps in any of the embodiments of the first aspect described above.
[0044] Fifthly, embodiments of this application also provide a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the steps in any of the embodiments of the first aspect described above.
[0045] The aforementioned dynamic reconstruction method and apparatus acquire an initial dynamic image of the target object based on dynamic scanning data, then iteratively reconstruct each frame of the initial dynamic image. During each iteration, the iterative parameters for reconstructing each frame are adjusted based on the quantized image quality index of each frame. Finally, the dynamically reconstructed image of the target object is determined based on the iterative parameters obtained from each iteration. In this method, when dynamically reconstructing the target object's dynamic scanning data, an initial dynamic image of the target object is first acquired, and then each frame of the initial dynamic image is iteratively reconstructed. During each iteration, the image quality index of each frame is quantified, i.e., the quantized image quality index value of each frame is obtained. Based on the quantized image quality index value of each frame, the iterative parameters for reconstructing each frame are adjusted. The dynamically reconstructed image of the target object is then determined based on the adjusted iterative parameters. Essentially, during multiple iterations of dynamic reconstruction, the iterative parameters for reconstructing each frame are continuously optimized and adjusted based on the quantized image quality index value, thereby improving the image quality of each frame and ultimately improving the image quality of the final dynamically reconstructed image. Attached Figure Description
[0046] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0047] Figure 1 This is an internal structural diagram of a computer device in one embodiment;
[0048] Figure 2 This is a flowchart illustrating a dynamic reconstruction method in one embodiment;
[0049] Figure 3 This is a flowchart illustrating the process of adjusting iteration parameters in one embodiment;
[0050] Figure 4 This is a schematic diagram of the process for determining the quantization value of the degree of convergence in a region in one embodiment;
[0051] Figure 5 This is a flowchart illustrating the process of determining the quantization value of the degree of regional convergence in another embodiment;
[0052] Figure 6This is a flowchart illustrating the process of determining the quantization value of the degree of regional convergence in another embodiment;
[0053] Figure 7 This is a schematic diagram of the process for determining a dynamically reconstructed image in one embodiment;
[0054] Figure 8 This is a schematic diagram of the dynamic reconstruction process in one embodiment;
[0055] Figure 9 This is a schematic diagram of a pre-trained network in one embodiment;
[0056] Figure 10 This is a schematic diagram of the dynamic reconstruction device in one embodiment. Detailed Implementation
[0057] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0058] The dynamic reconstruction method provided in this application can be applied to computer devices. The computer device can be a server, and its internal structure diagram can be as follows: Figure 1 As shown. The computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a dynamic reconstruction method. Those skilled in the art will understand that... Figure 1 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0059] Dynamic imaging is an imaging technique used to observe the process of a drug entering the body from injection until it reaches metabolic equilibrium. It can help determine the drug injection dose and time interval, and is of great significance for pharmacokinetics.
[0060] In dynamic reconstruction using PET (Positron Emission Tomography) / SPECT (Single-Photon Emission Computed Tomography), the reconstruction process often divides the total scan into several frames, each with a corresponding scan duration. When the same reconstruction parameters are used for each frame, differences in image acquisition due to varying scan time periods, differences in image convergence levels due to varying scan durations, and differences in pixel values due to noise levels result in variations in the overall convergence of the reconstructed regions across different frames. Therefore, predicting the convergence degree of the reconstructed regions during dynamic reconstruction iterations is beneficial for optimizing and adjusting iterative parameters, thereby improving the image quality of the reconstructed image.
[0061] Based on this, this application provides a dynamic reconstruction method. When dynamically reconstructing the dynamic scanning data of a target object, an initial dynamic image of the target object is first obtained, and then each frame of the initial dynamic image is iteratively reconstructed. In each iterative reconstruction process, the image quality index of each frame is quantified, that is, the quantized value of the image quality index of each frame is obtained, and the iterative parameters of the reconstruction of each frame are adjusted according to the quantized value of the image quality index of each frame. The dynamically reconstructed image of the target object is determined according to the adjusted iterative parameters. In other words, in the multiple iterations of dynamic reconstruction, the iterative parameters of the reconstruction of each frame are continuously optimized and adjusted according to the quantized value of the image quality index of each frame, thereby improving the image quality of each frame and thus improving the image quality of the final dynamically reconstructed image.
[0062] It should be noted that the beneficial effects or technical problems solved by the embodiments of this application are not limited to this one, but may also be other implicit or related problems. For details, please refer to the description of the embodiments below.
[0063] The technical solution of this application and how it solves the above-mentioned technical problems will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will be described below with reference to the accompanying drawings.
[0064] In one exemplary embodiment, such as Figure 2 As shown, a dynamic reconstruction method is provided, which can be applied to... Figure 1The following steps, 201 to 203, are used as an example of computer equipment.
[0065] S201, Based on the dynamic scanning data of the target object, obtain the initial dynamic image of the target object.
[0066] Among them, dynamic scanning data is data obtained by scanning the target object with a PET device or SPECT device. For example, the target object can be continuously tomographically sampled at set time intervals within a predetermined time period by a PET device or SPECT device to obtain dynamic scanning data of the target object.
[0067] The initial dynamic image of the target object refers to the dynamic image obtained by preliminary reconstruction of the target object's dynamic scan data. The initial dynamic image includes multiple frames, each of which is dynamically reconstructed based on the corresponding dynamic scan data.
[0068] Taking data collected over 60 seconds as an example, the 60 seconds of data is dynamically reconstructed at preset time intervals, with each time interval yielding one frame. For instance, if the preset time interval is 10 seconds, dynamic reconstruction of data collected from 0 to 10 seconds yields the first frame, and dynamic reconstruction of data collected from 10 to 20 seconds yields the second frame. Following this method, dynamic reconstruction of 60 seconds of data yields 6 frames, which naturally constitute the dynamic image.
[0069] In this embodiment, the dynamic reconstruction method employs an iterative reconstruction approach. This means that each frame of data undergoes multiple iterations during dynamic reconstruction, and the final dynamically reconstructed image is obtained when the iterations stop. The initial dynamic image in this embodiment can be either a dynamic image obtained through a single iteration or a dynamic image obtained through multiple iterations.
[0070] In this embodiment, when the computer device acquires a dynamic reconstructed image of a target object, it can first acquire the dynamic scanning data of the target object, and then use a preset iterative reconstruction algorithm to perform preliminary iterative reconstruction on the dynamic scanning data of the target object to obtain an initial dynamic image of the target object. It should be noted that before performing the preliminary iterative reconstruction, each frame of the initial dynamic image needs to be initialized, for example, by assigning all pixel values of the image to 1. The frames of the initial dynamic image are typically three-dimensional or four-dimensional images. A three-dimensional image represents an image with three spatial dimensions. A four-dimensional image represents an image with three spatial dimensions and a temporal dimension.
[0071] In one embodiment, the dynamic scanning data of the target object can be directly obtained from the PET or SPECT device that scans the target object. For example, the computer device can send a data acquisition request to the PET device that scans the target object. The data acquisition request carries the identification information of the target object. When the PET device receives the data acquisition request from the computer device, it finds the dynamic scanning data of the target object according to the identification information of the target object and sends the dynamic scanning data to the computer device.
[0072] In another embodiment, the dynamic scan data of the target object can be obtained from the server hosting the PACS (Picture Archiving and Communication System). The PACS system stores medical scan data for multiple objects, which can be retrieved when needed. For example, a computer device can send a data retrieval request to the server hosting the PACS system, carrying the identification information of the target object and the device identifier of the scanning device. Upon receiving the data retrieval request, the server hosting the PACS system filters out the corresponding scan data based on the identification information of the target object and the device identifier of the scanning device, and then sends the filtered scan data to the computer device. In this way, the computer device can obtain the dynamic scan data of the target object.
[0073] S202, iteratively reconstruct each frame of the initial dynamic image, and adjust the iterative parameters of each frame reconstruction according to the quantization value of the image quality index of each frame during each iteration;
[0074] In this embodiment, the image quality index quantization values include region convergence quantization values, image quality quantization values, and image resolution quantization values. Region convergence increases with the number of iterations, while image quality and image resolution do not change in this unidirectional way. In image reconstruction, convergence refers to the algorithm's ability to gradually approach the true value during iteration, while region convergence reflects the speed and accuracy with which the algorithm reaches a stable state in various regions of the image. The region convergence quantization value refers to the specific numerical value of the region convergence obtained through actual calculation or measurement, used to intuitively represent the level of region convergence.
[0075] After acquiring the initial dynamic image of the target object, the computer device continues to reconstruct each frame of the initial dynamic image using an iterative algorithm. During each iteration, the image quality index of each frame is quantized to obtain a quantized value. Based on these quantized values, the iterative parameters for each frame's reconstruction are continuously optimized and adjusted. These iterative parameters are determined according to the reconstruction progress of each frame, and the iterative parameters for each frame's reconstruction can be the same or different. The iterative algorithm can be an ordered subset expectation-maximization (OSEM) iterative algorithm, a regularized iterative algorithm, or a deep learning iterative algorithm, etc.
[0076] For example, during each iteration of reconstruction, a pre-defined neural network model can be used to obtain the quantified values of image quality indicators for each frame. Taking the quantified value of regional convergence as an example, each frame can be input into a pre-trained convergence evaluation model, which will analyze each frame and output the quantified value of regional convergence for each frame.
[0077] Furthermore, after obtaining the quantized values of the image quality index for each frame, the computer device can use a preset adjustment strategy to adjust the iterative parameters for reconstructing each frame based on the quantized values of the image quality index for each frame.
[0078] S203, Based on the iteration parameters of each frame image in each iteration, determine the dynamic reconstructed image of the target object.
[0079] After adjusting the iterative parameters for reconstructing each frame of images, the computer device can determine whether the iteration termination condition has been met based on the iterative parameters for reconstructing each frame of images. If it has been met, the frames of images obtained by the current iteration reconstruction can be identified as the dynamic reconstructed images of the target object. If it has not been met, the iterative reconstruction of each frame of images can continue according to the iterative parameters for reconstructing each frame of images until the iteration termination condition is met, and the dynamic reconstructed image of the target object is obtained.
[0080] The dynamic reconstruction method provided in this application involves obtaining an initial dynamic image of the target object based on its dynamic scanning data, then iteratively reconstructing each frame of the initial dynamic image. During each iteration, the iterative parameters for reconstructing each frame are adjusted based on the quantized image quality index of that frame. Finally, the dynamically reconstructed image of the target object is determined based on the iterative parameters obtained from each iteration. In this method, when dynamically reconstructing the target object's dynamic scanning data, an initial dynamic image of the target object is first obtained, and then each frame of the initial dynamic image is iteratively reconstructed. During each iteration, the image quality index of each frame is quantified, i.e., the quantized image quality index value of each frame is obtained. Based on the quantized image quality index value of each frame, the iterative parameters for reconstructing each frame are adjusted. The dynamically reconstructed image of the target object is then determined based on the adjusted iterative parameters. Essentially, during multiple iterations of dynamic reconstruction, the iterative parameters for reconstructing each frame are continuously optimized and adjusted based on the quantized image quality index value, thereby improving the image quality of each frame and ultimately improving the image quality of the final dynamically reconstructed image.
[0081] Based on the above embodiments, an embodiment is provided to illustrate the process of adjusting the iterative parameters for reconstructing each frame image according to the quantized value of the image quality index of each frame image during each iteration of reconstruction.
[0082] In one exemplary embodiment, such as Figure 3 As shown, during each iteration of reconstruction, the iterative parameters for reconstructing each frame are adjusted based on the quantization value of the image quality index of each frame, including:
[0083] S301, for a single iteration of reconstruction, obtain the preset image expected performance.
[0084] The expected image performance can be a specific metric or a gold standard image. For example, the expected image performance could be a 90% convergence rate in the descending aortic vessel region of each frame, or another value; the iteration goal would then be to obtain an image with a 90% convergence rate in the descending aortic vessel region. Alternatively, the expected image performance could be a gold standard image, and the iteration goal could be to continuously approach the gold standard image. Or, the expected image performance could be a spoken target, a sentence, or document information, which is converted into input through a speech conversion module or a text extraction module.
[0085] In addition, image expected performance can also be measured by other parameters, such as noise level, sharpness, and conformity to recognized standards like the CERN standard. For example, in a SPECT application, image expected performance can be the convergence coefficient (e.g., contrast recovery coefficient CR) and noise level (e.g., background variability BV) of each sphere in the SPECT image of each frame. This can be achieved by distinguishing between the background and the individual spheres in the hot zone based on region segmentation, and setting different thresholds for different regions. Specific threshold settings are shown in Table 1.
[0086] Table 1. Settings for expected image performance in SPECT applications.
[0087]
[0088] S302, adjust the convergence parameters for reconstructing each frame of the image based on the quantized values of the image quality index and the expected performance of the image.
[0089] Among them, the convergence parameters are the parameters related to convergence in the iteration parameters, which may include the number of iterations, iteration coefficients, regularization strength, post-filtering level, etc.
[0090] For example, the convergence parameters for reconstructing each frame of an image can be adjusted based on the quantized values of each image quality index and the expected performance of the image, combined with a preset adjustment strategy.
[0091] Taking the image quality index quantization value as the regional convergence coefficient and the iteration coefficient as the convergence parameter as an example, assuming the expected convergence of a certain region of the image, such as a blood vessel, is 90% (expected image performance), and the quantization value of the regional convergence in the i-th iteration is 89%, and the predicted quantization value of the regional convergence in the (i+1)-th iteration is 91%, then the iteration coefficient can be set to 0.5, so that the final convergence of the blood vessel is 90%. Alternatively, assuming the expected convergence of a certain region of the image, such as a blood vessel, is 90%, and the quantization value of the regional convergence in the i-th iteration is 92%, and the predicted quantization value of the regional convergence in the (i+1)-th iteration is 93%, then the iteration coefficient can be set to 0, and a Gaussian post-filter with a half-width of 2mm can be added, so that the final convergence of the blood vessel is 90%.
[0092] S303, the adjusted convergence parameters are determined as the iterative parameters for the reconstruction of each frame of the image.
[0093] After adjusting the convergence parameters for the reconstruction of each frame, the adjusted convergence parameters are determined as the iterative parameters for the reconstruction of each frame. That is, the iterative parameters for the reconstruction of each frame are the adjusted parameters.
[0094] It should be noted that the expected image performance in this application embodiment can be set according to actual needs. Taking the quantization value of region convergence as an example, if it is desired that the convergence degree of each frame image is consistent, the convergence degree of each region in each frame image can be set to be consistent, or other convergence degrees can be set according to actual needs. In this way, by pre-setting the expected image performance and continuously adjusting the iteration parameters, the final iterative image, that is, the final dynamically reconstructed image, can meet the expected goal. Furthermore, this application can control the convergence degree of different frames images to be consistent, or it can control different frames images to be reconstructed according to different expected convergence degrees. That is, this application is a dynamic reconstruction method that can control the convergence degree of different frames.
[0095] In the dynamic reconstruction method provided in this application, for a single iteration of reconstruction, a preset image expected performance is obtained. Then, based on the quantized values of each image quality index and the image expected performance, the convergence parameters of each frame image reconstruction are adjusted. Finally, the adjusted convergence parameters are determined as the iteration parameters of each frame image reconstruction. This method, by introducing image expected performance, allows for the adjustment of the iteration parameters of each frame image reconstruction based on the image expected performance and the quantized values of the image quality index of each frame, providing an optional method for quickly adjusting iteration parameters.
[0096] Based on any of the above embodiments, an embodiment is provided to illustrate the process of obtaining the quantized value of the region convergence degree in the quantized value of the image quality index.
[0097] In one exemplary embodiment, such as Figure 4 As shown, the image quality index quantization value includes the region convergence degree quantization value; the method also includes:
[0098] S401, for a single iteration of reconstruction, image segmentation is performed on each frame of the initial dynamic image to obtain multiple segmented image regions for each frame.
[0099] In one embodiment, for any given frame image, a preset image segmentation algorithm can be used to segment the frame image, resulting in multiple segmented image regions for that frame image. Similarly, the same method is used for other frames image to obtain multiple segmented image regions for those other frames image, thus obtaining multiple segmented image regions for each frame image. The image segmentation algorithm can be a threshold-based segmentation algorithm, an edge-based segmentation algorithm, a region segmentation algorithm, etc.
[0100] In another embodiment, after dimensionality reduction processing of each frame image, image segmentation can be performed based on the dimensionality-reduced frame images to obtain multiple segmented image regions for each frame image.
[0101] For example, dimensionality reduction processing is performed on each frame image to obtain the dimensionality-reduced image corresponding to each frame image; image segmentation processing is performed on the dimensionality-reduced image corresponding to each frame image to obtain multiple segmented image regions of each frame image.
[0102] As mentioned earlier, each frame in the initial dynamic image is a 3D or 4D image. To improve image segmentation efficiency, dimensionality reduction processing can be performed on each frame to obtain the corresponding 2D image. Taking a 3D image as an example, common medical 3D reconstruction methods can be used to reduce the 3D image to a 2D image. These methods include, but are not limited to, multi-layer reconstruction (MPR), maximum density projection (MIP), minimum density projection (MinIP), surface shading (SSD), volumetric roaming (VRT), surface reconstruction (CPR), and virtual endoscopy (VE). Besides medical 3D reconstruction methods, operations such as orthographic projection, dimensionality reduction techniques such as principal component extraction, or encoder techniques from deep learning can also be used.
[0103] Furthermore, after performing dimensionality reduction on each frame of the image to obtain the corresponding dimensionality-reduced image, the image regions are segmented using the dimensionality-reduced image and other anatomical image information to obtain multiple segmented image regions for each frame. For example, the tool totalsegmentor can be used to segment blood vessel regions, liver regions, brain regions, etc., or other tools can be used to segment lesion regions, etc.
[0104] S402, determine the quantization value of the region convergence degree of each frame image based on multiple segmented image regions of each frame image.
[0105] For example, a pre-trained convergence prediction model can be used to determine the quantized value of the region convergence of each frame image. For instance, multiple segmented image regions of each frame image can be input into the above convergence prediction model to obtain the quantized value of the region convergence output by the convergence prediction model, which is the quantized value of the region convergence of each frame image.
[0106] The convergence prediction model can be a pre-trained neural network model. Optionally, the convergence prediction model can be constructed using network models such as backpropagation neural networks, recurrent neural networks, deep neural networks, and convolutional neural networks.
[0107] In the dynamic reconstruction method provided in this application embodiment, for a single iteration of reconstruction, image segmentation is performed on each frame of the initial dynamic image to obtain multiple segmented image regions for each frame. Then, based on these multiple segmented image regions, the quantized value of the region convergence degree for each frame is determined. This method provides an optional approach to quickly determine the quantized value of the region convergence degree. By segmenting each frame, multiple segmented image regions can be obtained for each frame. These multiple segmented image regions are then used to determine the quantized value of the region convergence degree for each frame, providing data support for subsequent adjustments to the iteration parameters.
[0108] Based on the above embodiments, an embodiment is provided to illustrate the process of determining the quantization value of the region convergence degree of each frame image based on multiple segmented image regions of each frame image.
[0109] In one exemplary embodiment, such as Figure 5 As shown, based on multiple segmented image regions of each frame, the quantization value of the region convergence degree of each frame image is determined, including:
[0110] S501, obtain pixel information of multiple segmented image regions of each frame image.
[0111] After performing image segmentation on each frame of the image and obtaining multiple segmented image regions for each frame, the computer device acquires the pixel information of each segmented image region to determine the quantization value of the region convergence degree based on the pixel information of each segmented image region.
[0112] S502, determine the region value change curve of each segmented image region based on the pixel information of each segmented image region.
[0113] In one embodiment, determining the region value change curve of each segmented image region based on the pixel information of each segmented image region includes the following steps:
[0114] Step 1: Determine the image region value of each segmented image region based on the pixel information of each segmented image region.
[0115] In this embodiment of the application, the image region value can be the mean, peak, maximum, minimum, or contrast between the region and other regions.
[0116] After acquiring the pixel information of each segmented image region, the computer device determines the image region value of each segmented image region based on the pixel information of each segmented image region and the preset region value calculation formula.
[0117] For example, taking the average value of an image region as an example, for any segmented image region, the average pixel value of the segmented image region is determined according to the pixel information of the segmented image region and the average value calculation formula. Naturally, the average pixel value of the segmented image region is the image region value of the segmented image region.
[0118] Step 2: Based on the image region values of each segmented image region and the image region values of each segmented image region over a historical time period, determine the region value change curve of each segmented image region.
[0119] In this context, the image region values for the historical time period are the image region values of each segmented image region calculated during the historical iterative reconstruction process. For example, if the current iterative reconstruction is the third iterative reconstruction process for each frame image, then the image region values for the historical time period are the image region values determined by the first iterative reconstruction and the image region values determined by the second iterative reconstruction.
[0120] After acquiring the image region values of each segmented image region over a historical time period, the computer device combines the image region values of each segmented image region, plots all the image region values of each segmented image region on a coordinate system, and fits the data points in the coordinate system to obtain the region value change curve of each segmented image region.
[0121] S503, determine the quantization value of the region convergence degree of each frame image based on the region value change curve of each segmented image region.
[0122] After obtaining the region value change curves of each segmented image region, the computer device can input the region value change curves of each segmented image region into a pre-trained convergence prediction model to obtain the quantized value of the region convergence of each frame image.
[0123] The dynamic reconstruction method provided in this application acquires pixel information of multiple segmented image regions in each frame image, then determines the region value change curve of each segmented image region based on the pixel information of each segmented image region, and finally determines the quantized value of the region convergence degree of each frame image based on the region value change curve of each segmented image region. This method, by acquiring pixel information of multiple segmented image regions in each frame image, can determine the region value change curve of each segmented image region based on the pixel information of each segmented image region. The quantized value of the region convergence degree of each frame image can be quickly determined using the region value change curve of each segmented image region, thus improving the efficiency of determining the quantized value of the region convergence degree.
[0124] Based on the above embodiments, an embodiment is provided to illustrate the process of determining the quantization value of the regional convergence degree of each frame image based on the regional value change curve of each segmented image region.
[0125] In one exemplary embodiment, such as Figure 6 As shown, based on the region value variation curves of each segmented image region, the quantization value of the region convergence degree of each frame image is determined, including:
[0126] S601, input the region value change curve of each segmented image region into the pre-trained convergence prediction model to obtain the quantized value of the region convergence degree of each segmented image region.
[0127] In this embodiment of the application, by pre-training a convergence prediction model, the quantification value of the degree of regional convergence of each frame image can be quickly determined by inputting the region value change curve of each segmented image region into the convergence prediction model.
[0128] In one embodiment, the training process of the convergent prediction model includes the following steps:
[0129] Step 1: Obtain the sample image region information and sample region convergence curve of multiple sample segmentation images corresponding to the sample dynamic scanning data.
[0130] Among them, the sample dynamic scan data consists of historical dynamic scan data of multiple objects, which can be obtained from the PACS system.
[0131] Multiple sample segmentation images corresponding to sample dynamic scanning data refer to multiple segmentation images obtained by iteratively reconstructing the sample dynamic scanning data to obtain sample dynamic images, and then performing image segmentation on the sample dynamic images.
[0132] The sample image region information includes the sample region value change curve, sample region identification information, and pixel value information within the sample region. The sample region convergence curve refers to the curve formed by the quantized values of the sample region convergence degree after multiple iterations of each frame in the sample dynamic image.
[0133] In this embodiment, the sample region value change curve in the sample image region information can be obtained in the same way as the region value change curve determined in this application, and the sample region convergence degree quantization value constituting the sample region convergence curve can also be obtained in the same way as the region convergence degree quantization value determined in this application, which will not be repeated here.
[0134] Step 2: Input the image region information and sample region convergence curve of each sample segmented image into a preset neural network for model training to obtain an initial model.
[0135] In this embodiment, the image region information of each sample segmented image can be used as the input to the neural network, and the sample region convergence curve can be used as the target to train the neural network and obtain an initial model. For example, the image region information of each sample segmented image is input into the neural network to obtain the predicted region convergence curve output by the neural network. Then, based on the predicted region convergence curve, the sample region convergence curve, and a preset loss function, the loss value of the neural network is determined. Subsequently, the model parameters of the neural network are continuously adjusted based on the loss value until the loss value is less than a preset loss value threshold, thus obtaining the initial model.
[0136] It should be noted that during model training, the input information can include not only image region information, but also iteration count information, scan object information, detector information, scanning protocol information, algorithm information, and data count information. Image region information includes the identification of image regions, such as blood vessels, liver, and lesions, which are distinguished by specific identifiers. Scan object information refers to the information of the scanned patient / phantom / animal, such as age, height, weight, and blood glucose level. Detector information refers to the detector's model and performance. Scanning protocol information refers to the protocol used for scanning. Algorithm information refers to the type and parameters of the algorithm used.
[0137] Step 3: Validate the initial model to obtain a convergent prediction model.
[0138] After obtaining the initial model, cross-validation and other methods can be used to validate the initial model and obtain a convergent prediction model.
[0139] It should also be noted that when training the convergence prediction model, the output of the convergence prediction model is a region convergence curve. When applying the convergence prediction model to determine the quantization value of the region convergence of each segmented image, a curve composed of the quantization values of the region convergence of multiple iterations will also be output. Furthermore, the quantization value of the region convergence corresponding to each iteration can be obtained from the curve based on the number of iterations.
[0140] S602, Based on the quantization value of the region convergence of each segmented image region, determine the quantization value of the region convergence of each frame image.
[0141] After obtaining the quantized values of the regional convergence of each segmented image region, the computer device acquires each segmented image region corresponding to any frame image and determines the quantized values of the regional convergence of each segmented image region as the quantized values of the regional convergence of that frame image. Then, the quantized values of the regional convergence of each frame image are obtained in the same way. In this way, the quantized values of the regional convergence of each frame image can be obtained.
[0142] The quantization value of the region convergence of each segmented image region is determined as the quantization value of the region convergence of each frame image.
[0143] In the dynamic reconstruction method provided in this application, the region value change curves of each segmented image region are input into a pre-trained convergence prediction model to obtain the quantized value of the region convergence degree of each segmented image region. This quantized value is then used to determine the quantized value of the region convergence degree of each frame of the image. By pre-training the convergence prediction model, the quantized value of the region convergence degree of each frame of the image can be quickly obtained by inputting the region value change curves of each segmented image region into the model. This improves the efficiency of determining the quantized value of the region convergence degree and provides data support for subsequent adjustment of iteration parameters.
[0144] Based on the above embodiments, an embodiment is provided to illustrate the process of determining the dynamically reconstructed image of the target object based on the iterative parameters of each frame image in each iteration.
[0145] In one exemplary embodiment, such as Figure 7 As shown, based on the iteration parameters of each frame image in each iteration of reconstruction, the dynamically reconstructed image of the target object is determined, including:
[0146] S701, For a single iteration of reconstruction, determine whether each frame of images has converged based on the iteration parameters of each frame of images in the iterative reconstruction.
[0147] During each iteration of reconstruction, after adjusting the iteration parameters of each frame, it is determined whether each frame has converged, i.e. whether the iteration stopping condition has been met, based on these iteration parameters.
[0148] For example, for any frame image, the iteration parameters of the frame image can be compared with a preset parameter threshold, and then the convergence of the frame image can be determined based on the comparison result between the iteration parameters of the frame image and the preset parameter threshold.
[0149] For example, if the iteration parameters of a frame image meet a preset parameter threshold, then the frame image is determined to have converged; if the iteration parameters of a frame image do not meet the preset parameter threshold, then the frame image is determined to have not converged. Taking a parameter threshold of 1 as an example, if the iteration parameters of a frame image are 1, then the frame image is determined to have converged; if the iteration parameters of a frame image are not 1, then the frame image is determined to have not converged.
[0150] S702, if the images of each frame converge, then each image of the frame is determined as the dynamically reconstructed image of the target object.
[0151] When it is determined that each frame of the image has converged, then each frame of the image is determined as the dynamically reconstructed image of the target object.
[0152] S703, if the images in each frame do not converge, the iterative reconstruction process is repeated until the images in each frame converge, and the dynamic reconstructed image of the target object is obtained.
[0153] When it is determined that the images in each frame have not converged, the iterative reconstruction process is repeated for the images that have not converged until the images converge, and then the images and other images are identified as the dynamically reconstructed images of the target object.
[0154] In the dynamic reconstruction method provided in this application embodiment, for a single iteration of reconstruction, the convergence of each frame image is determined based on the iteration parameters of each frame image in the iterative reconstruction. If the frame images converge, then each frame image is determined as the dynamically reconstructed image of the target object. If the frame images do not converge, the iterative reconstruction process is repeated until the frame images converge, thus obtaining the dynamically reconstructed image of the target object. In this method, the convergence of each frame image is determined by the iteration parameters, i.e., whether the iteration termination condition has been met. If the frame images converge, then the currently converged frame images can be determined as the dynamically reconstructed images. If the frame images do not converge, then the iterative reconstruction process needs to be repeated until the frame images converge, thus obtaining the dynamically reconstructed image. This provides an optional method for quickly obtaining dynamically reconstructed images.
[0155] Additionally, in one embodiment, such as Figure 8 The specific process of dynamic reconstruction is explained as shown.
[0156] (1) Expected performance settings: Expected performance settings are performed before the reconstruction begins. Expected performance can be a certain metric or a gold standard image.
[0157] (2) Reconstruction begins at frame i: Assuming that dynamic reconstruction is sequential, then i represents the reconstruction of the i-th sequential frame. At the beginning of each frame reconstruction, the reconstructed image is initialized, for example, all pixel values of the image are assigned the value 1.
[0158] (3) Obtaining intermediate images during the iterative process: During the reconstruction process, the intermediate images during the iterative process are generally three-dimensional or four-dimensional images. Three-dimensional images generally represent images with three spatial dimensions. Four-dimensional images generally represent images with three spatial dimensions and a time dimension.
[0159] (4) Image dimensionality reduction: The image is processed to reduce its dimensionality.
[0160] (5) Image region segmentation: Obtain the dimension-reduced image and segment the image region using the dimension-reduced image and other anatomical image information.
[0161] (6) Calculate region values: Calculate image region values, such as mean, peak, maximum, minimum, contrast between this region and other regions, etc.
[0162] (7) Calculate the regional value change curve: Statistically analyze the changes in regional values in each iteration.
[0163] (8) Predicting convergence coefficients using artificial intelligence methods: The convergence degree of the next iteration or the convergence curve for all iterations can be predicted based on single or combined information, for subsequent parameter adjustment. For example: input iteration number information, input trend of region values in the current and previous iterations, input image region identifier information, input pixel value information within the image region, input scan object information, input detector information, input scanning protocol information, input algorithm information, input data count information, etc. By combining this information, the convergence coefficient for several future iterations or all iterations can be predicted. Artificial intelligence network parameters can be obtained through pre-training. Examples of pre-trained network design include (e.g.) Figure 9 As shown in the figure, the pre-training input consists of the region convergence curves under different iterations obtained from different count data reconstructions, as well as image region information, scanned object information, detector information, scanning protocol information, and algorithm information; the target is the region convergence curves obtained from multiple iterations (such as 50 iterations) obtained from different count data reconstructions.
[0164] (9) Adjusting iteration parameters: The iteration parameters can be convergence-related parameters, such as the number of iterations, iteration coefficients, regularization strength, post-filtering level, etc.
[0165] (10) Output image: After all conditions are met, the final image is output.
[0166] In this embodiment, during the iteration process, the image dimensionality is reduced, and the convergence degree of the regional image is predicted by deep learning methods, combining the image after dimensionality reduction with the regional image value change curve. This is beneficial for optimizing and adjusting the iteration parameters. At the same time, the convergence degree of different regions is quantitatively evaluated, and the iteration parameters of each frame are determined by the predicted convergence degree. This can automatically adjust the iteration parameters of dynamic reconstruction, providing more comparable images between frames for dynamic reconstruction image research.
[0167] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.
[0168] Based on the same inventive concept, this application also provides a dynamic reconstruction apparatus for implementing the dynamic reconstruction method described above. The solution provided by this apparatus is similar to the implementation described in the above method; therefore, the specific limitations in one or more embodiments of the dynamic reconstruction apparatus provided below can be found in the limitations of the dynamic reconstruction method described above, and will not be repeated here.
[0169] In one exemplary embodiment, such as Figure 10 As shown, a dynamic reconstruction device 1 is provided, comprising: an image acquisition module 10, a parameter adjustment module 20, and an image determination module 30, wherein:
[0170] The image acquisition module 10 is used to acquire an initial dynamic image of the target object based on the dynamic scanning data of the target object;
[0171] The parameter adjustment module 20 is used to iteratively reconstruct each frame of the initial dynamic image, and adjust the iterative parameters of each frame reconstruction according to the quantization value of the image quality index of each frame during each iteration reconstruction.
[0172] The image determination module 30 is used to determine the dynamically reconstructed image of the target object based on the iteration parameters of each frame image in each iteration of reconstruction.
[0173] In one embodiment, the parameter adjustment module 20 is further configured to:
[0174] For a single iteration of reconstruction, the preset expected image performance is obtained; based on the quantized value of the image quality index and the expected image performance of each frame, the convergence parameters of each frame reconstruction are adjusted; the adjusted convergence parameters are determined as the iteration parameters of each frame reconstruction.
[0175] In one embodiment, the image quality index quantization value includes a region convergence degree quantization value; the dynamic reconstruction device 1 further includes:
[0176] The image segmentation module is used to segment each frame of the initial dynamic image for a single iteration of reconstruction, and obtain multiple segmented image regions for each frame.
[0177] The convergence degree determination module is used to determine the quantized value of the region convergence degree of each frame image based on multiple segmented image regions of each frame image.
[0178] In one embodiment, the image segmentation module is further configured to:
[0179] Dimensionality reduction is performed on each frame of the image to obtain the corresponding dimensionality-reduced image; image segmentation is then performed on the corresponding dimensionality-reduced image to obtain multiple segmented image regions for each frame.
[0180] In one embodiment, the convergence determination module is further configured to:
[0181] Obtain pixel information of multiple segmented image regions in each frame; determine the region value change curve of each segmented image region based on the pixel information of each segmented image region; determine the quantization value of the region convergence degree of each frame image based on the region value change curve of each segmented image region.
[0182] In one embodiment, the convergence determination module is further configured to:
[0183] Based on the pixel information of each segmented image region, determine the image region value of each segmented image region; based on the image region value of each segmented image region and the image region value of each segmented image region over a historical time period, determine the region value change curve of each segmented image region.
[0184] In one embodiment, the convergence determination module is further configured to:
[0185] The region value change curves of each segmented image region are input into a pre-trained convergence prediction model to obtain the quantized value of the region convergence of each segmented image region; based on the quantized value of the region convergence of each segmented image region, the quantized value of the region convergence of each frame image is determined.
[0186] In one embodiment, the dynamic reconstruction device 1 further includes:
[0187] The data acquisition module is used to acquire sample image region information and sample region convergence curves of multiple sample segmentation images corresponding to the sample dynamic scanning data.
[0188] The model training module is used to input the image region information and sample region convergence curve of each sample segmented image into a preset neural network for model training to obtain an initial model;
[0189] The model validation module is used to validate the initial model and obtain a converged prediction model.
[0190] In one embodiment, the image determination module 30 is further configured to:
[0191] For a single iteration of reconstruction, the convergence of each frame image is determined based on the iteration parameters of each frame image in the iterative reconstruction. If each frame image converges, it is determined as the dynamic reconstructed image of the target object. If each frame image does not converge, the iterative reconstruction process is repeated until each frame image converges, and the dynamic reconstructed image of the target object is obtained.
[0192] Each module in the aforementioned dynamic reconstruction device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.
[0193] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:
[0194] Based on the dynamic scanning data of the target object, obtain the initial dynamic image of the target object;
[0195] The initial dynamic image is reconstructed iteratively, and the iterative parameters for each frame reconstruction are adjusted according to the quantization value of the image quality index of each frame during each iteration.
[0196] Based on the iteration parameters of each frame image in each iteration, the dynamically reconstructed image of the target object is determined.
[0197] The implementation principles and technical effects of each step in the processor embodiment of this application are similar to those of the dynamic reconstruction method described above, and will not be repeated here.
[0198] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:
[0199] Based on the dynamic scanning data of the target object, obtain the initial dynamic image of the target object;
[0200] The initial dynamic image is reconstructed iteratively, and the iterative parameters for each frame reconstruction are adjusted according to the quantization value of the image quality index of each frame during each iteration.
[0201] Based on the iteration parameters of each frame image in each iteration, the dynamically reconstructed image of the target object is determined.
[0202] The implementation principles and technical effects of each step in the computer program executed by the processor in this embodiment are similar to those of the dynamic reconstruction method described above, and will not be repeated here.
[0203] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:
[0204] Based on the dynamic scanning data of the target object, obtain the initial dynamic image of the target object;
[0205] The system iteratively reconstructs each frame of the initial dynamic image, and adjusts the iterative parameters of each frame's reconstruction based on the quantization value of the region convergence of each frame during each iteration. The iterative parameters for each frame are different for each iteration.
[0206] Based on the iteration parameters of each frame image in each iteration, the dynamically reconstructed image of the target object is determined.
[0207] The implementation principles and technical effects of each step in the computer program executed by the processor in this embodiment are similar to those of the dynamic reconstruction method described above, and will not be repeated here.
[0208] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.
[0209] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.
[0210] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0211] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. A dynamic reconstruction method, characterized in that, The method includes: Based on the dynamic scanning data of the target object, obtain the initial dynamic image of the target object; The initial dynamic image is reconstructed iteratively, and the iterative parameters for reconstructing each frame are adjusted according to the quantization value of the image quality index of each frame during each iteration. Based on the iteration parameters of each frame image in each iteration, the dynamically reconstructed image of the target object is determined.
2. The method according to claim 1, characterized in that, The step of adjusting the iterative parameters for reconstructing each frame of the image based on the quantized value of the image quality index of each frame during each iteration of reconstruction includes: For a single-iteration reconstruction, the expected image performance is obtained; Based on the quantized values of the image quality index of each frame and the expected performance of the image, the convergence parameters for the reconstruction of each frame are adjusted. The adjusted convergence parameters are determined as the iterative parameters for the reconstruction of each frame of the image.
3. The method according to claim 1 or 2, characterized in that, The image quality index quantization value includes a region convergence degree quantization value; the method further includes: For a single iteration of reconstruction, each frame of the initial dynamic image is segmented to obtain multiple segmented image regions for each frame. Based on multiple segmented image regions of each frame image, a quantized value of the region convergence degree of each frame image is determined.
4. The method according to claim 3, characterized in that, The step of segmenting each frame of the initial dynamic image to obtain multiple segmented image regions for each frame includes: The dimensions of each frame image are reduced to obtain the dimension-reduced image corresponding to each frame image; Image segmentation processing is performed on the dimensionality-reduced images corresponding to each frame to obtain multiple segmented image regions for each frame.
5. The method according to claim 3, characterized in that, The step of determining the region convergence quantization value of each frame image based on multiple segmented image regions of each frame image includes: Obtain pixel information of multiple segmented image regions in each frame of the image; Based on the pixel information of each segmented image region, determine the region value change curve of each segmented image region; Based on the region value change curves of each segmented image region, the quantization value of the region convergence degree of each frame image is determined.
6. The method according to claim 5, characterized in that, The step of determining the region value change curve of each segmented image region based on the pixel information of each segmented image region includes: Based on the pixel information of each segmented image region, determine the image region value of each segmented image region; Based on the image region values of each segmented image region and the image region values of each segmented image region over historical time periods, the region value change curve of each segmented image region is determined.
7. The method according to claim 5, characterized in that, The step of determining the quantization value of the region convergence degree of each frame image based on the region value change curve of each segmented image region includes: The region value change curves of each segmented image region are input into a pre-trained convergence prediction model to obtain a quantitative value of the region convergence degree of each segmented image region. The quantization value of the region convergence of each frame image is determined based on the quantization value of the region convergence of each segmented image region.
8. The method according to claim 7, characterized in that, The training process of the convergent prediction model includes: Obtain sample image region information and sample region convergence curves for multiple sample segmentation images corresponding to the dynamic scanning data of the samples; The image region information and sample region convergence curve of each of the segmented images are input into a preset neural network for model training to obtain an initial model; The initial model is validated to obtain the convergent prediction model.
9. The method according to claim 1 or 2, characterized in that, The step of determining the dynamically reconstructed image of the target object based on the iteration parameters reconstructed in each frame of the image includes: For a single iteration of reconstruction, it is determined whether each frame of the image has converged based on the iteration parameters of each frame in the iterative reconstruction. If the images in each frame converge, then each image in each frame is determined as the dynamically reconstructed image of the target object; If the images in each frame do not converge, the iterative reconstruction process is repeated until the images in each frame converge, thus obtaining the dynamic reconstructed image of the target object.
10. A dynamic reconstruction device, characterized in that, The device includes: The image acquisition module is used to acquire an initial dynamic image of the target object based on the dynamic scanning data of the target object; The parameter adjustment module is used to iteratively reconstruct each frame of the initial dynamic image, and adjust the iterative parameters of each frame reconstruction according to the quantization value of the image quality index of each frame during each iteration reconstruction. The image determination module is used to determine the dynamically reconstructed image of the target object based on the iteration parameters of each frame image in each iteration of reconstruction.