Imaging data analysis method of X-ray system equipment

Abstract

The invention provides an imaging data analysis method of X-ray system equipment, which relates to the technical field of medical imaging and comprises the following steps: receiving an imaging frame selection area and interested demand information; performing scanning coverage analysis on the imaging frame selection area, and outputting a non-equiangular scanning path; dynamically scanning the imaging target by taking the non-equiangular scanning path as a constraint to obtain a two-dimensional slice sequence; constructing an ROI identification network to analyze and identify the two-dimensional slice sequence to obtain a two-dimensional identification sequence; and cooperatively operating a three-dimensional reconstruction module and a space-time fusion module to respectively carry out space-time reconstruction fusion on the two-dimensional identification sequence and the two-dimensional slice sequence to obtain a dynamic VOI and a dynamic three-dimensional model, and carrying out imaging fusion to obtain dynamic label body data. The technical problem that the dynamic change of the target cannot be comprehensively captured due to redundant scanning paths and lack of accurate data fusion processing in the prior art is solved, and the technical effect of improving the imaging accuracy and the dynamic expressive force is achieved.

Description

Technical Field

[0001] The present application relates to the field of medical imaging technology, and in particular to an imaging data analysis method for an X-ray system device. Background Art

[0002] X-ray imaging technology uses the property of X-rays penetrating objects to generate images by capturing the changes in the transmitted X-rays through the detector. Existing X-ray imaging methods are mainly based on two-dimensional imaging technology, using a fixed scanning path for static image acquisition. These methods often use fixed scanning angles when performing multi-angle dynamic scanning, which has problems such as scanning path redundancy and low efficiency, resulting in long imaging time and a lag in capturing dynamic targets. Secondly, these methods have certain deficiencies when processing data fusion. Due to insufficient alignment accuracy, images overlap or unclear structures appear, especially in multi-angle imaging, and the three-dimensional structure of the target cannot be accurately reconstructed, making it difficult to truly reflect the dynamic changes of the target in different time dimensions, affecting the final imaging quality. Summary of the invention

[0003] The present application provides an imaging data analysis method for an X-ray system device, which solves the technical problem in the prior art that the dynamic changes of the target at different time points cannot be fully captured due to redundant scanning paths and lack of precise data fusion processing, thereby achieving the technical effect of improving the accuracy and dynamic expression of imaging.

[0004] In view of the above problems, the present application provides an imaging data analysis method for an X-ray system device, the method comprising: the X-ray system device receives the imaging frame selection area and the interested demand information transmitted back through an interactive interface, wherein the X-ray system device is composed of a dynamic scanning module, a ROI annotation module, a three-dimensional reconstruction module, a spatiotemporal fusion module and a dynamic visualization module; performing scanning coverage analysis on the imaging frame selection area and outputting a non-conformal scanning path; using the non-conformal scanning path as a constraint, driving the dynamic scanning module to dynamically scan the imaging target to obtain a two-dimensional slice sequence, wherein multiple frames of two-dimensional slices in the two-dimensional slice sequence are attached with multiple scanning time identifiers and multiple scanning angle identifiers; the ROI annotation module is based on The interest demand information is hierarchically modeled to obtain the ROI identification network, and then the ROI identification network is run to analyze and identify the two-dimensional slice sequence to obtain a two-dimensional identification sequence; with the multiple scanning time identifications and the multiple scanning angle identifications as constraints, the three-dimensional reconstruction module and the space-time fusion module are collaboratively run to perform space-time reconstruction and fusion on the two-dimensional identification sequence to obtain a dynamic VOI; with the multiple scanning time identifications and the multiple scanning angle identifications as constraints, the three-dimensional reconstruction module and the space-time fusion module are collaboratively run to perform space-time reconstruction and fusion on the two-dimensional slice sequence to obtain a dynamic three-dimensional model; the dynamic visualization module is run to perform imaging fusion on the dynamic VOI and the dynamic three-dimensional model to obtain dynamic label volume data.

[0005] One or more technical solutions provided in this application have at least the following technical effects or advantages: The imaging frame selection area and information of interest are received through the interactive interface to ensure the pertinence and effectiveness of the imaging data. By performing scanning coverage analysis on the imaging frame selection area, a non-conformal scanning path is output to improve the scanning efficiency and imaging accuracy, and solve the limitations of the traditional conformal scanning path when dealing with irregular shaped targets. Multiple rounds of two-dimensional slice data are obtained through the dynamic scanning module to provide rich data support for subsequent spatiotemporal reconstruction. The two-dimensional slice sequence is analyzed and identified through the hierarchical modeling ROI annotation module to obtain a two-dimensional identification sequence to ensure accurate identification and annotation of key areas and improve the accuracy and reliability of annotation. Through the coordinated operation of the spatiotemporal fusion module and the three-dimensional reconstruction module, the two-dimensional identification sequence and the two-dimensional slice sequence are reconstructed and fused in time and space respectively to generate dynamic VOI and dynamic three-dimensional models to capture the dynamic change information of the imaging target at different time points and angles. Finally, the dynamic VOI and the dynamic three-dimensional model are imaged and fused through the dynamic visualization module to generate dynamic label volume data, providing more intuitive and comprehensive visualization results.

[0006] In summary, this application executes the above steps through the collaborative work of multiple modules of the X-ray imaging system, gradually optimizes the acquisition, processing, analysis and visualization process of imaging data, effectively integrates multi-angle and time series data, significantly improves the quality of X-ray imaging, and the ability to capture the dynamic characteristics of the target, meets the high requirements for spatiotemporal accuracy in complex imaging tasks, and promotes the application innovation of X-ray imaging technology in the medical field.

[0007] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 A schematic flow chart of an imaging data analysis method for an X-ray system device provided in an embodiment of the present application.

[0009] Figure 2 A schematic diagram of a process for obtaining a dynamic VOI in an imaging data analysis method of an X-ray system device provided in an embodiment of the present application.

[0010] Figure 3 A schematic diagram of a flow chart of outputting a non-conformal scanning path in an imaging data analysis method of an X-ray system device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0011] The embodiment of the present application provides an imaging data analysis method for an X-ray system device, thereby solving the technical problem in the prior art that the dynamic changes of the target at different time points cannot be fully captured due to redundant scanning paths and lack of precise data fusion processing, thereby achieving the technical effect of improving the accuracy and dynamic expression of imaging.

[0012] like Figure 1 As shown, an embodiment of the present application provides an imaging data analysis method for an X-ray system device, the method comprising: Step S1: The X-ray system device receives the transmitted imaging frame selection area and interested demand information through the interactive interface, wherein the X-ray system device is composed of a dynamic scanning module, a ROI labeling module, a three-dimensional reconstruction module, a space-time fusion module and a dynamic visualization module.

[0013] Specifically, the X-ray system equipment is mainly composed of functional modules such as dynamic scanning module, ROI labeling module, 3D reconstruction module, spatiotemporal fusion module and dynamic visualization module. These modules contain hardware and software to complete specific functions. Among them, the dynamic scanning module is responsible for dynamically scanning the imaging target to obtain imaging data at different times and angles. The ROI labeling module is used to label the region of interest (ROI) in the imaging according to specific needs. The 3D reconstruction module is responsible for building a 3D model of the imaging target based on the acquired 2D imaging data. The spatiotemporal fusion module is responsible for fusing imaging data at different times and spaces so that the imaging results contain information about time and space. The dynamic visualization module is responsible for visualizing the processed imaging data so that users can see the results intuitively.

[0014] The interactive interface of the X-ray system equipment is a window for the user to interact with the equipment. The user inputs the imaging frame selection area and the information of interest through the interactive interface, and the X-ray system equipment receives the user input information. Among them, the imaging frame selection area refers to the specific area on the imaging target selected by the user through the interactive interface. This area is the focus area for subsequent scanning, analysis and other operations. For example, when examining a specific part of a certain organ in medicine, the area selected by the user on the imaging interface is the imaging frame selection area. The information of interest is the user's special requirements for the imaging results or the imaging process, such as the medical demand for attention to certain lesion characteristics.

[0015] By receiving the returned imaging frame selection area and interested requirement information, the target area and required direction of subsequent operations are determined, providing clear input information for the entire imaging process.

[0016] Step S2: performing scanning coverage analysis on the imaging frame selection area and outputting a non-conformal scanning path.

[0017] Specifically, the non-conformal scanning path is opposite to the conformal scanning path. It is a path in which the scanning angle is non-uniformly distributed during the scanning process, and the scanning angle can be flexibly adjusted according to the shape and needs of the object. After receiving the imaging frame selection area, the X-ray system equipment performs a scanning coverage analysis on this area. This analysis process is carried out based on factors such as the shape and size of the imaging frame selection area and the set information of interest. Through this analysis, a non-conformal scanning path is output, which can cover the imaging frame selection area more efficiently, improve the pertinence and effectiveness of the scan, avoid unnecessary scanning operations, save time and improve imaging quality.

[0018] Step S3: using the non-conformal scanning path as a constraint, driving the dynamic scanning module to dynamically scan the imaging target to obtain a two-dimensional slice sequence, wherein multiple frames of two-dimensional slices in the two-dimensional slice sequence are attached with multiple scanning time identifiers and multiple scanning angle identifiers.

[0019] Specifically, a two-dimensional slice sequence is a collection of a series of two-dimensional images obtained by dynamically scanning an imaging target through a dynamic scanning module, and contains two-dimensional slice data of the same target at different time points and angles. Each frame of a two-dimensional slice in a two-dimensional slice sequence is marked with a unique scanning time identifier and a scanning angle identifier. Among them, the scanning time identifier is used to record the specific scanning time of the slice. The scanning angle identifier is used to record the specific scanning angle of the slice.

[0020] The non-conformal scanning path obtained in step S2 is used as a constraint condition to drive the dynamic scanning module (such as an X-ray emitter, a detector, etc.) to perform multiple rounds of dynamic scanning on the imaging target according to this path. For example, in medical imaging equipment, the X-ray emitter emits X-rays according to a set angle and path, and the detector receives the X-rays after penetrating the object and converts them into electrical signals, which are then processed to obtain a two-dimensional slice image. Due to dynamic scanning, multiple two-dimensional slices will be obtained at different times and angles, and each slice is accompanied by a scanning time mark and a scanning angle mark.

[0021] By performing dynamic scanning with a non-conformal scanning path as a constraint, a two-dimensional slice sequence with time and angle information is acquired, providing a rich data basis for subsequent analysis, identification, and reconstruction.

[0022] Step S4: the ROI annotation module performs hierarchical modeling according to the interested demand information, obtains the ROI identification network, and then runs the ROI identification network to analyze and identify the two-dimensional slice sequence to obtain a two-dimensional identification sequence.

[0023] Specifically, hierarchical modeling refers to multi-level modeling of ROI (region of interest) based on the information of interest requirements, so as to more finely identify and annotate the region of interest. For example, multiple interest identification models are scheduled according to the user's various types of interest requirements, the multiple models are connected in parallel, and then the output of multiple model output configurations is fused to build an ROI identification network. The two-dimensional identification sequence is a two-dimensional slice sequence identified by the ROI identification network analysis, which includes multi-dimensional annotation results.

[0024] The ROI annotation module first performs hierarchical modeling based on the interested demand information received in step S1 and constructs an ROI identification network. This ROI identification network is used to identify and analyze ROI (region of interest). After the ROI identification network is built, it is run to analyze and identify the two-dimensional slice sequence. In this process, the identification network will identify the region of interest in each two-dimensional slice according to the pre-set rules and models to obtain a two-dimensional identification sequence.

[0025] By constructing a ROI identification network, the region of interest is accurately identified in the two-dimensional slice sequence, providing input data with specific identification for subsequent spatiotemporal reconstruction fusion.

[0026] Step S5: Taking the multiple scanning time identifiers and the multiple scanning angle identifiers as constraints, the three-dimensional reconstruction module and the time-space fusion module are operated in coordination to perform time-space reconstruction and fusion on the two-dimensional identifier sequence to obtain a dynamic VOI.

[0027] Specifically, dynamic VOI, or volume of interest, is a three-dimensional data structure that contains the region of interest and has temporal and spatial information after spatiotemporal reconstruction and fusion. For example, when the information of interest is the lungs or heart, dynamic VOI is used to describe the contraction and relaxation of the heart cavity and the expansion and contraction of the lungs.

[0028] The 3D reconstruction module and the spatiotemporal fusion module work together with the constraints of multiple scanning time labels and multiple scanning angle labels in the 2D identification sequence. The 3D reconstruction module constructs a 3D structure based on the 2D identification sequence, and the spatiotemporal fusion module fuses these 3D structures with time and space labels to make them more coherent in time and space. For example, in the detection of the bone growth process, the 2D identification slices at different time points are 3D reconstructed, and then the 3D structures at different times are fused together through the spatiotemporal fusion module to obtain a dynamic VOI. By obtaining a dynamic VOI, the dynamic changes of the region of interest can be better displayed.

[0029] Step S6: using the multiple scanning time identifiers and the multiple scanning angle identifiers as constraints, the three-dimensional reconstruction module and the spatiotemporal fusion module are collaboratively operated to perform spatiotemporal reconstruction and fusion on the two-dimensional slice sequence to obtain a dynamic three-dimensional model.

[0030] Specifically, the dynamic three-dimensional model is a three-dimensional model that includes the entire imaging target and has time and space information, which is obtained by performing spatiotemporal reconstruction and fusion on a two-dimensional slice sequence.

[0031] Similarly, with multiple scanning time identifiers and multiple scanning angle identifiers as constraints, the three-dimensional reconstruction module and the space-time fusion module work together. The difference here from step S5 is that the operation object is a two-dimensional slice sequence rather than a two-dimensional identification sequence. The three-dimensional reconstruction module constructs a three-dimensional model based on the two-dimensional slice sequence, and the space-time fusion module fuses the three-dimensional model with time and space identifiers to make it more complete in time and space. For example, in the imaging of the entire organ, the two-dimensional slices at different times and angles are three-dimensionally reconstructed and space-time fused to obtain a dynamic three-dimensional model. This dynamic three-dimensional model can fully reflect the dynamic changes of the imaging target.

[0032] Step S7: running the dynamic visualization module to perform imaging fusion on the dynamic VOI and the dynamic three-dimensional model to obtain dynamic label volume data.

[0033] Specifically, the dynamic label volume data is a fusion result of the dynamic VOI and the dynamic three-dimensional model. The dynamic visualization module performs imaging fusion on the dynamic VOI obtained in step S5 and the dynamic three-dimensional model obtained in step S6. For example, in medical image display, the dynamic VOI with detailed lesion information and the dynamic three-dimensional model of the entire organ are fused together to obtain the dynamic label volume data, so that the imaging results can be displayed more comprehensively and intuitively.

[0034] Further, such as Figure 2 As shown, step S5 includes: Step S51: the three-dimensional reconstruction module receives and generates an initial three-dimensional model according to the two-dimensional identification sequence and multiple scanning angle identifications.

[0035] Step S52: the spatiotemporal fusion module receives and performs a dynamic consistency evaluation on the initial three-dimensional model according to the multiple scanning time identifiers to obtain first model error information.

[0036] Step S53: the 3D reconstruction module performs compensation reconstruction on the initial 3D model according to the model error information to generate a first 3D model.

[0037] Step S54: the space-time fusion module receives and performs a dynamic consistency evaluation on the first three-dimensional model according to the multiple scanning time identifiers to obtain second model error information.

[0038] Step S55: Similarly, the 3D reconstruction module and the spatiotemporal fusion module are alternately operated to perform spatiotemporal reconstruction and fusion on the 2D marker sequence until the model error information output by the spatiotemporal fusion module is lower than a preset error threshold, and the dynamic VOI is output.

[0039] Specifically, the initial three-dimensional model is a three-dimensional model constructed by the three-dimensional reconstruction module based on two-dimensional slices at different angles at the same time, which represents the structure of the region of interest of the imaging target at each time point. The three-dimensional reconstruction module receives a two-dimensional identification sequence and multiple scanning angle identifications as input, and uses a three-dimensional modeling tool to first extract contour information from the two-dimensional identification sequence, and then splices and fits these contour information in three-dimensional space according to the scanning angle identification to form an initial three-dimensional model. For example, when imaging bones, the contours of the bones are extracted from two-dimensional slices at different angles, and then these contours are combined in three-dimensional space using scanning angle identification to construct an initial three-dimensional model of the bones. Through the three-dimensional reconstruction module, multiple initial three-dimensional models corresponding to multiple time points can be constructed.

[0040] Since the two-dimensional slices at different time points are not exactly the same, there are certain degrees of errors in the multiple initial three-dimensional models constructed. Therefore, the spatiotemporal fusion module is required to evaluate the consistency of the three-dimensional model at different time points, calculate the errors of the model at different time points, and generate the first model error information. This error information reflects the degree of deviation of the model in the time dimension. When performing dynamic consistency evaluation, the volume change rate, shape similarity coefficient, etc. can be selected as evaluation indicators. The volume change rate is used to measure the volume fluctuation of the model at different time points. If in a normal dynamic process, the model volume should have a certain reasonable range of change, exceeding this range may indicate an error; the shape similarity coefficient is to determine whether the model is consistent by calculating the similarity between the model shapes at different time points. For the calculation of the volume change rate, the volume can be calculated according to the three-dimensional model under different time marks, and then the change ratio of the volume at adjacent time points is calculated. For example, in the dynamic imaging of the heart, if the initial three-dimensional model represents the state of the heart in diastole, and the model corresponding to the next time mark represents the state of the heart in systole, then under normal circumstances, the volume of the heart will be reduced by a certain proportion. The spatiotemporal fusion module determines the first model error information by calculating this volume change ratio and comparing it with the pre-set normal range. For the calculation of shape similarity coefficients, some shape descriptor algorithms can be used, such as algorithms based on moment features or contour features. For example, the curvature change of the model contour under different time stamps is calculated. If the curvature suddenly changes greatly at a certain time point, it may indicate that the model shape is abnormal. By calculating the first model error information of the initial three-dimensional model, the degree of deviation of the initial three-dimensional model in the time dimension can be accurately reflected, providing a clear improvement direction for subsequent compensation reconstruction.

[0041] The three-dimensional reconstruction module performs compensatory reconstruction on the initial three-dimensional model according to the first model error information. In the compensatory reconstruction, the initial three-dimensional model can be regarded as a deformed model, and some parameters or node positions of the model can be adjusted to make it more consistent with the expected state. For example, the shape of a certain organ part of the initial three-dimensional model in the time dimension does not conform to the actual situation. The area that needs to be adjusted is determined according to the model error information, and the model mesh vertices corresponding to this area are used as adjustable nodes, and the energy minimization algorithm is used to adjust the positions of these nodes. For example, using an energy minimization algorithm based on elastic deformation, the direction and distance that each node needs to move are calculated according to the model error information, so that the model satisfies the physical law of elastic deformation during the adjustment process, that is, the total energy of the model is minimized. By compensating and reconstructing the initial three-dimensional model, multiple first three-dimensional models that have been corrected once can be obtained, making them closer to the actual situation in the time dimension and reducing the model error.

[0042] The spatiotemporal fusion module receives the first 3D model and multiple scanning time tags again for dynamic consistency evaluation, and uses the same dynamic consistency evaluation indicators as mentioned above, such as volume change rate and shape similarity coefficient, and the corresponding calculation method to obtain the second model error information. The second model error information can be used to determine whether the previous round of compensation reconstruction is effective and whether the next round of compensation reconstruction is needed, thereby providing data support for the error convergence of the entire spatiotemporal reconstruction fusion process.

[0043] The above-mentioned model compensation reconstruction and dynamic consistency evaluation operations are performed alternately to continuously correct the error of the three-dimensional model in the time dimension. Each time, the three-dimensional reconstruction module compensates and reconstructs the model according to the model error information obtained by the time-space fusion module, and the time-space fusion module performs dynamic consistency evaluation on the new model to obtain new error information until the model error information output by the time-space fusion module is lower than the preset error threshold. This preset error threshold is a pre-set value used to measure whether the error of the model is acceptable. When the model error information is lower than this threshold, it means that the accuracy of the model in time and space has met the requirements, and the iterative optimization process can be stopped. The three-dimensional model that meets the preset error threshold is output as a dynamic VOI. This dynamic VOI contains the shape, distribution and dynamic change information of the region of interest in three-dimensional space, providing high-quality three-dimensional data for subsequent analysis and application.

[0044] Further, such as Figure 3 As shown, step S2 includes: Step S21: performing spatial mapping according to the imaging frame selection area to obtain an imaging space range.

[0045] Step S22: dispatching the dynamic scanning module to perform full-range equiangular scanning on the imaging target falling within the imaging space range to generate a local three-dimensional volume.

[0046] Step S23: Generate a conformal coverage scanning path by performing scanning coverage analysis on the local three-dimensional volume.

[0047] Step S24: performing redundant scanning path optimization on the equiangular coverage scanning path, and outputting the non-equiangular scanning path.

[0048] Specifically, the imaging frame selection area selected by the user through the interactive interface is received, and the coordinate conversion algorithm is used to convert the imaging frame selection area into an imaging space range that can be recognized by the device, that is, the coordinate system and space range inside the X-ray system device. Through spatial mapping, the imaging space range is determined to provide an accurate space range for subsequent scanning and processing, ensuring the pertinence and effectiveness of the scan.

[0049] The dynamic scanning module is scheduled to perform full-range equiangular scanning of the imaging target. In this scanning mode, the scanning angles are evenly distributed and will cover the entire imaging space range. The dynamic scanning module scans the imaging targets that fall within the imaging space range according to the pre-set equiangular scanning parameters (such as scanning angle interval, starting angle, etc.). For example, in a medical CT scanning device, the X-ray emitter rotates around the human body at a set angle (such as every 1 degree) to emit X-rays, and the detector receives the X-rays that penetrate the human body and converts them into electrical signals, which are processed to generate a local three-dimensional volume. This local three-dimensional volume is the preliminary three-dimensional data of the imaging target within the imaging space range, which provides a data basis for subsequent scanning coverage analysis.

[0050] Use tools such as shape analysis algorithms to perform scan coverage analysis on local 3D bodies. Usually, the local 3D body is an irregularly shaped object (such as various parts of the human body). The shape analysis algorithm will first determine the approximate shape features of the local 3D body (such as convex hulls, pits, etc.), and then plan the scan path based on these features and the requirements of equiangular scanning. For example, if the local 3D body is a cylinder with a depression, the algorithm will consider the position and shape of the depression and plan a scan path that can equiangularly cover the cylinder including the depression, that is, the equiangular coverage scan path. This equiangular coverage scan path provides an initial path for subsequent redundant scan path optimization.

[0051] Since the scanning angle of conformal scanning is fixed, when scanning imaging targets, especially those with irregular shapes, the scanning areas usually overlap, resulting in unnecessary waste of time and resources. Therefore, it is necessary to perform redundant scanning path optimization on the conformal coverage scanning path to remove unnecessary and repeated scanning path parts, so as to obtain a more efficient non-conformal scanning path. By evaluating the scanning effect of the conformal coverage scanning path, it is determined which areas already have sufficient scanning data and do not need conformal scanning, so that this part of the path is marked as a redundant path and removed or adjusted. By obtaining a non-conformal scanning path, the scanning efficiency can be improved, unnecessary scanning operations can be reduced, and at the same time, effective scanning of the imaging target can be guaranteed.

[0052] Further, step S23 includes: Step S231: extracting the three-dimensional boundary constraints of the imaging by performing voxel distribution analysis on the local three-dimensional volume.

[0053] Step S232: Generate a minimum envelope range according to the imaging three-dimensional boundary constraint fitting.

[0054] Step S233: performing geometric shape feature recognition on the minimum envelope range, and decomposing the minimum envelope range according to the recognition result to obtain N regular areas and M irregular areas, wherein N and M are positive integers.

[0055] Step S234: extracting boundary characteristics of the M irregular regions to obtain M boundary curvatures and M boundary shape change rates.

[0056] Step S235: traverse the scanning interval table using the M boundary curvatures and the M boundary shape change rates to obtain M initial scanning intervals, wherein the scanning interval table stores a plurality of sample curvatures, a plurality of sample shape change rates and a plurality of sample scanning intervals in an associated manner.

[0057] Step S236: extracting a target scanning interval by serializing the M initial scanning intervals.

[0058] Step S237: Based on the target scanning interval, generate the equiangular coverage scanning path adapted to the global local three-dimensional volume.

[0059] Specifically, tools such as voxel statistical algorithms are used to perform voxel distribution analysis on the local three-dimensional volume to understand the structural characteristics of the local three-dimensional volume in space, and then the three-dimensional boundary constraints of the imaging can be accurately extracted to provide key information about the boundaries of the local three-dimensional volume for subsequent operations. Voxel distribution analysis includes calculating the density distribution of voxels (the smallest unit in three-dimensional space, similar to pixels in two-dimensional images) in different regions, and counting the distribution range of voxels in the directions of each coordinate axis. Through these calculations and statistics, the boundary voxels of the local three-dimensional volume can be identified. For example, in a local three-dimensional volume of medical imaging, through voxel distribution analysis, it can be found that the density of voxels in the area close to the bone surface suddenly changes, and these changed areas can be determined as part of the three-dimensional boundary constraints of the imaging.

[0060] The minimum envelope range is the minimum bounding box generated by fitting the imaging three-dimensional boundary constraints, which is used to cover the imaging target. The minimum envelope range is generated by fitting the imaging three-dimensional boundary constraints through a geometric fitting algorithm. For example, by performing a minimum convex hull calculation on the boundary points determined by the imaging three-dimensional boundary constraints (if the imaging target is approximately a convex shape) or using a more complex fitting algorithm (for irregular shapes), multiple discrete boundary points of the imaging three-dimensional boundary constraints are fitted to find a three-dimensional shape that can minimally contain these boundary points. The range defined by this shape is the minimum envelope range. By obtaining the minimum envelope range, the processing object of the subsequent scanning path planning of the imaging target can be simplified.

[0061] In medical imaging, the imaging target is the human body. Due to the natural complexity of the human body structure, the minimum envelope range needs to be distinguished in order to generate a uniform coverage scanning path in a targeted manner. The minimum envelope range can be identified by using a shape classification algorithm to identify geometric shape characteristics, and the minimum envelope range is divided into N regular areas (such as a generally flat chest area, an ellipsoidal head area, etc.) and M irregular areas (such as the curved boundaries of bones such as ribs and knee joints). Among them, N and M are positive integers, respectively referring to the number of regular areas and irregular areas identified. Exemplarily, the minimum envelope range can be identified by calculating some geometric feature parameters of the minimum envelope range, such as the length, width and height ratio, curvature distribution, etc., and then comparing them with the pre-defined feature parameter ranges of various geometric shapes. If the length, width and height ratio is close to 1:1:1 and the curvature distribution is relatively uniform, it may be identified as a regular area in the shape of a cube; if the shape parameters do not meet the definition of any common geometric shape, it is determined to be an irregular area.

[0062] Boundary characteristics of M irregular regions are extracted to extract the special properties of their boundaries, mainly the boundary curvature and boundary shape change rate. Boundary curvature reflects the degree of curvature of the boundary curve at a certain point, and the boundary shape change rate describes the speed of change of the boundary shape within a certain range. This process can use the curvature calculation algorithm and the shape change rate calculation algorithm. For example, for the boundary curve of the irregular region, the boundary curvature is calculated by the curvature calculation formula in differential geometry. For the change of the boundary shape, the shape change rate is calculated by comparing the shape differences of adjacent boundary segments and according to a certain distance range. By calculating the boundary curvature and shape change rate, M boundary curvatures and M boundary shape change rates are obtained. These boundary curvatures and shape change rates can reflect the complexity and shape change characteristics of the boundary of the irregular region, and provide a basis for the subsequent determination of the scanning interval of the irregular region.

[0063] The scanning interval table is traversed using M boundary curvatures and M boundary shape change rates, and the calculated boundary curvature and shape change rate of each irregular area are matched with the sample curvature and sample shape change rate in the scanning interval table to obtain M initial scanning intervals. The scanning interval table is a pre-established associative storage table, which stores multiple sample curvatures, sample shape change rates and corresponding multiple sample scanning intervals.

[0064] According to the positional relationship of the irregular area within the minimum envelope, M initial scanning intervals are sequenced to obtain the target scanning interval.

[0065] Based on the target scanning interval, the path planning algorithm is used to determine the parameters such as the equiangular scanning angle and step length in different areas (including regular areas and irregular areas), and generate an equiangular coverage scanning path that adapts to the global local 3D volume. This path can fully cover the entire local 3D volume in an equiangular manner based on the characteristics of different areas of the local 3D volume, providing a basic path for the subsequent optimization of redundant scanning paths.

[0066] Further, step S24 includes: Step S241: preset segment angle parameters, and use the segment angle parameters as segmentation basis to divide the equiangular coverage scanning path into K segments of local paths.

[0067] Step S242: Calculate the coverage of the K local paths and output K coverage areas.

[0068] Step S243: performing adjacent path cross comparison on the K segments of local paths according to the K coverage areas to obtain K coverage overlaps of the K segments of local paths.

[0069] Step S244: preset an overlap threshold, and divide the K local paths into a high-redundancy path segment set and a critical path segment set according to the overlap threshold and K coverage overlaps.

[0070] Step S245: performing path elimination optimization on the high-redundancy path segment set through coverage integrity check, so as to obtain a low-redundancy path segment set from the high-redundancy path segment set.

[0071] Step S246: performing curve fitting on the low-redundancy path segment set and the key path segment set to generate the non-conformal scanning path.

[0072] Specifically, the segment angle parameter is a preset angle value used to divide the equiangular coverage scanning path according to the angle. This parameter determines the angle range of each local path. The segment angle parameter is preset, and then the equiangular coverage scanning path is divided into K local paths based on this parameter. Among them, K is a positive integer, which represents the number of divided local paths. For example, if the equiangular coverage scanning path is a 360° circular scanning path, and the preset segment angle parameter is 45°, then this path can be divided into 360° / 45°=8 (K=8) local paths.

[0073] Use the spatial geometry calculation algorithm to calculate the coverage of K local paths and obtain K coverage areas, which provide a data basis for the subsequent cross-comparison of adjacent paths. The calculation process can determine the coverage area of ​​each local path by calculating the intersection of the ray and the imaging space based on the starting point, end point, scanning angle range and scanning characteristics of the imaging device (such as the diffusion range of the scanning ray). For example, in a medical imaging device, each local path is an X-ray scanning path at a specific angle. By calculating the propagation path and penetration range of the X-ray in human tissue, the coverage area of ​​each local path can be determined.

[0074] According to the K coverage areas, the adjacent paths of K segments of local paths are cross-compared, and the K coverage overlaps of the K segments of local paths are calculated. The coverage overlap refers to the percentage of the coverage area of ​​a certain section of the road that is also covered by the coverage area of ​​other sections of the road. The larger the coverage percentage, the higher the overlap. For example, for two adjacent local paths, their coverage areas are regarded as two sets. By calculating the intersection area of ​​the two sets, and then dividing the intersection area by the set area, the coverage overlap can be obtained.

[0075] The overlap threshold is a pre-set value used to determine whether a local path belongs to a high-redundancy path. When the coverage overlap of a local path is greater than this threshold, it is considered a high-redundancy path. The overlap threshold is preset, and then the K segments of local paths are divided into a high-redundancy path segment set and a critical path segment set based on this threshold and K coverage overlaps. For example, if the overlap threshold is set to 0.5, when the coverage overlap of a certain local path is greater than 0.5, it is classified into the high-redundancy path segment set; otherwise, it is classified into the critical path segment set.

[0076] The coverage integrity check is performed on the high-redundancy path fragment set to ensure that the entire scanning path can still fully cover the imaging target after the redundant paths are eliminated. First, the coverage contribution of each path in the high-redundancy path fragment set to different areas of the imaging target is analyzed. The higher the overlap, the lower the contribution. Then, according to certain rules (such as preferentially eliminating the path with the least impact on the overall coverage), the path is eliminated to obtain the low-redundancy path fragment set, which reduces redundancy while ensuring the integrity of the scan.

[0077] The curve fitting algorithm is used to perform curve fitting on the low-redundancy path segment set and the critical path segment set. For example, discrete path points in the low-redundancy path segment set and the critical path segment set are used as input, and a smooth curve that best fits these points is found through the least squares curve fitting algorithm. This curve is the generated non-conformal scanning path.

[0078] Through the above steps, the overlap of the equiangular coverage scanning path is analyzed, and then the redundant scanning path optimization is performed to obtain the non-equiangular scanning path, which not only improves the scanning efficiency but also ensures the high quality of the imaging data.

[0079] Further, step S4 includes: Step S41: the ROI annotation module extracts the interest requirement priority and multiple interest requirement types from the interest requirement information.

[0080] Step S42: Directedly dispatching a plurality of annotation function units from the annotation function library of the ROI annotation module according to the plurality of interested demand types.

[0081] Step S43: After configuring the computing power weights of the multiple labeling function units according to the priority of the interested requirements, the multiple labeling function units are connected in parallel, and a feature integration unit is configured at the output ends of the multiple labeling function units to obtain the ROI identification network.

[0082] Step S44: input the first frame of two-dimensional slices in the two-dimensional slice sequence into the ROI identification network, and after the multiple annotation result slices are obtained through the annotation processing of the multiple annotation function units in the ROI identification network, the multiple annotation result slices are spatially integrated by the feature integration unit to obtain a first multi-dimensional annotation result.

[0083] Step S45: Similarly, the ROI identification network is run to analyze and identify the two-dimensional slice sequence to obtain the two-dimensional identification sequence, wherein the two-dimensional identification sequence includes multiple multi-dimensional annotation results corresponding to the multiple frames of two-dimensional slices.

[0084] Specifically, the priority of the requirement of interest is obtained from the requirement of interest information, indicating the order of importance or processing of different requirements of interest. The type of requirement of interest is the different types of requirements contained in the requirement of interest information. The ROI annotation module extracts the priority of the requirement of interest and multiple types of requirement of interest from the requirement of interest information. For example, for the requirement of interest information stored in a specific format (such as XML or JSON format), the information parsing algorithm reads the file content, identifies the value of the field marked as priority as the priority of the requirement of interest, and classifies different types of requirement content into multiple types of requirement of interest. By extracting the priority of the requirement of interest and the type of requirement of interest, the priority order and requirement type required for subsequent operations are clarified, providing a basis for subsequent operations such as annotation function scheduling.

[0085] The annotation function library is a collection of units with different annotation functions stored in the ROI annotation module. These annotation function units can realize different annotation functions. According to multiple types of requirements of interest, the annotation function units that can realize the corresponding requirements are searched and called from the annotation function library of the ROI annotation module to prepare for the construction of the ROI identification network.

[0086] Computing power weight configuration refers to the proportion of computing resources allocated to different annotation function units according to the priority of the demand of interest. The annotation function unit with a high priority will be allocated more computing power to ensure that it can complete the annotation task more efficiently. The computing power weights of multiple annotation function units are configured according to the priority of the demand of interest. For example, if there are three annotation function units A, B, and C, and their corresponding priorities of interest are high, medium, and low, respectively, the resource allocation algorithm may allocate computing power to A, B, and C at a ratio of 50%, 30%, and 20%. Then these annotation function units are connected in parallel, and a feature integration unit is configured at the output end to obtain the ROI identification network. The feature integration unit is used to integrate the output results of multiple annotation function units to generate the final annotation result. By allocating computing resources according to the priority of the demand and integrating multiple annotation function units to obtain the ROI identification network, not only the accuracy of the annotation is improved, but also the comprehensiveness and completeness of the annotation are ensured.

[0087] Input the first frame of two-dimensional slices in the two-dimensional slice sequence into the ROI identification network. First, multiple annotation function units perform annotation processing on the first frame of two-dimensional slices according to their respective functions and allocated computing power to obtain multiple annotation result slices. Then, the feature integration unit spatially integrates these annotation result slices, such as using a fusion algorithm to merge the information in different annotation results according to spatial position, and finally obtains the first multi-dimensional annotation result. This first multi-dimensional annotation result contains the annotation information of the first frame of two-dimensional slices by multiple annotation function units.

[0088] The remaining multi-frame 2D slices in the 2D slice sequence are input into the ROI identification network one by one for annotation processing and spatial integration to obtain the corresponding multiple multi-dimensional annotation results. The multiple multi-dimensional annotation results of these multi-frame 2D slices are integrated in the order of the multi-frame 2D slices in the 2D slice sequence to obtain a 2D identification sequence, which completely records the annotation results of the 2D slice sequence and provides high-quality data support for the subsequent 3D model construction.

[0089] Furthermore, before step S42, the following steps are also included: Step S42-1: Obtain H types of sample demand types by aggregating historical demands of interest.

[0090] Step S42-2: Perform network data call according to the H types of sample requirements to obtain H standard annotation function models.

[0091] Step S42-3: Perform local data call according to the H types of sample requirements to obtain H historical annotation data sets.

[0092] Step S42-4: Use the H historical annotation data sets to optimize the parameters of the H standard annotation function models to obtain H sample function units.

[0093] Step S42-5: associate and store the H sample requirement types and the H sample function units to complete the pre-construction of the annotation function library.

[0094] Specifically, historical interesting demands refer to the interesting demands accumulated in the past. By collecting, organizing and summarizing historical interesting demands, various types of interesting demands that have been processed before can be obtained, so as to discover patterns and commonalities therefrom. Representative demand types are extracted from the historical interesting demands to obtain H types of sample demand types. Among them, H is a positive integer, which refers to the number of sample demand types. Exemplarily, a clustering algorithm can be used to cluster similar demands together from the database of historical interesting demands, and then each cluster is analyzed and labeled to determine H types of sample demand types. By determining representative sample demand types, a basis for demand types is provided for subsequent data calls and the construction of annotation function libraries.

[0095] Using network request and data retrieval algorithms, network data calls are made according to H sample demand types to obtain H standard annotation function models. These standard annotation function models are pre-built models that can meet the annotation function requirements of specific sample demand types. For example, a query request is constructed according to the sample demand type and sent to a dedicated model library server (such as an artificial intelligence model library in the cloud). The server searches and returns a standard annotation function model that meets the requirements based on the request content. Through network data calls, standard annotation function models that can meet the sample demand types can be obtained, providing an initial model foundation for subsequent parameter optimization.

[0096] According to H types of sample requirements, local data is called to obtain H historical annotation data sets from local storage devices (such as local databases, etc.). These historical annotation data sets are previously annotated data sets, which contain annotation information related to the sample requirement type and can be used to optimize the standard annotation function model. This process can use the local database query algorithm. For example, in the database of local annotation data, based on the sample requirement type as the query condition, the historical annotation data set matching it is retrieved. Through local data calls, local historical annotation data sets used to optimize the standard annotation function model are obtained.

[0097] H historical annotation data sets are used to adjust and optimize the parameters of H standard annotation function models to obtain H sample function units. By adjusting the parameters of the model, the model can be better adapted to specific sample demand types. This process can use model training algorithms, such as gradient descent algorithms. The historical annotation data sets are divided into training sets and validation sets. The standard annotation function model is trained with the training set. The parameters of the model are continuously adjusted with the gradient descent algorithm. Then, the adjusted model is verified with the validation set until the model performance reaches a satisfactory level, and the optimized sample function unit is obtained. By adjusting and optimizing the parameters of H standard annotation function models, sample function units optimized for specific sample demand types are obtained, which improves the accuracy and efficiency of the annotation function.

[0098] H sample requirement types and H sample function units are stored in association to complete the pre-construction of the annotated function library. This process can use database storage technology, such as table structure storage in a relational database. For example, create an annotated function library table, in which one column stores the sample requirement type and the other column stores the corresponding sample function unit. In this way, when the sample function unit needs to be called according to the sample requirement type, the corresponding sample function unit can be quickly obtained by querying this table.

[0099] The above steps make targeted adjustments and optimizations to the standard annotation function model through the local historical annotation data set, improve the model's adaptability to current annotation needs, and build an annotation function library, providing a model basis for the directional scheduling of annotation function units.

[0100] Further, step S7 includes: Step S71: after spatially aligning the dynamic VOI and the dynamic three-dimensional model, the dynamic VOI is displayed on the dynamic three-dimensional model by using pseudo-color annotation fusion according to the spatial distribution of the dynamic VOI, so as to obtain initial label volume data.

[0101] Step S72: In the initial labeled volume data, the dynamic VOI and the dynamic three-dimensional model are fused frame by frame in time series to obtain the dynamic labeled volume data.

[0102] Specifically, when acquiring dynamic label volume data, the dynamic VOI and the dynamic three-dimensional model are first spatially aligned, and tools such as spatial coordinate transformation algorithms can be used. For example, by determining the key feature points of the dynamic VOI and the dynamic three-dimensional model, and then calculating the required translation, rotation, and scaling transformation parameters based on the coordinate relationship of these feature points, the two can be spatially aligned. Then, according to the spatial distribution of the dynamic VOI, the dynamic VOI is displayed by pseudo-color annotation fusion in the dynamic three-dimensional model. This process can be implemented using graphics rendering and color mapping algorithms. For example, according to the different attributes of the VOI (such as density, intensity, etc.), it is mapped to different pseudo-colors, and then the VOI with pseudo-color annotations is fused into the rendering of the dynamic three-dimensional model, and finally the initial label volume data is obtained, so that the information of the dynamic VOI can be intuitively displayed in the dynamic three-dimensional model, and the data basis is provided for the subsequent frame-by-frame time series fusion.

[0103] In the initial labeled volume data, the dynamic VOI and the dynamic 3D model are fused frame by frame in time series. For each frame of data, the relevant data of the dynamic VOI and the dynamic 3D model are fused according to the pre-set fusion rules (such as weighted average, logical operation, etc.). For example, in medical imaging, for each time point of the dynamic VOI and the dynamic 3D model, the data of the two are fused according to a certain weight to obtain the fused result of each frame, and finally the dynamic labeled volume data is obtained.

[0104] Through the above steps, dynamic label volume data is generated, which ensures the continuity and consistency of dynamic changes and provides users with more intuitive and comprehensive visualization results, thereby helping users to better understand and analyze imaging data.

[0105] In summary, the imaging data analysis method of an X-ray system device provided in the embodiment of the present application has the following technical effects: The imaging frame selection area and information of interest are received through the interactive interface to ensure the pertinence and effectiveness of the imaging data. By performing scanning coverage analysis on the imaging frame selection area, a non-conformal scanning path is output to improve the scanning efficiency and imaging accuracy, and solve the limitations of the traditional conformal scanning path when dealing with irregular-shaped targets. The dynamic scanning module uses the non-conformal scanning path as a constraint to obtain multiple rounds of two-dimensional slice data, providing rich data support for subsequent spatiotemporal reconstruction. The ROI annotation module is obtained through hierarchical modeling to analyze and identify the two-dimensional slice sequence, obtain a two-dimensional identification sequence, ensure the accurate identification and annotation of key areas, and improve the accuracy and reliability of the annotation. Through the coordinated operation of the spatiotemporal fusion module and the three-dimensional reconstruction module, the two-dimensional identification sequence and the two-dimensional slice sequence are reconstructed and fused in time and space respectively to generate dynamic VOI and dynamic three-dimensional models to capture the dynamic change information of the imaging target at different time points and angles. Finally, the dynamic VOI and the dynamic three-dimensional model are imaged and fused through the dynamic visualization module to generate dynamic label volume data, providing more intuitive and comprehensive visualization results.

[0106] In general, the embodiments of the present application execute the above steps through the collaborative work of multiple modules of the X-ray imaging system, gradually optimize the acquisition, processing, analysis and visualization process of imaging data, effectively integrate multi-angle and time series data, significantly improve the quality of X-ray imaging, and the ability to capture the dynamic characteristics of the target, meet the high requirements for spatiotemporal accuracy in complex imaging tasks, and bring more reliable and comprehensive imaging solutions to fields such as medical diagnosis.

[0107] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An imaging data analysis method for an X-ray system device, characterized in that: The method comprises: The X-ray system device receives the returned imaging frame selection area and interested demand information through the interactive interface, wherein the X-ray system device is composed of a dynamic scanning module, a ROI annotation module, a three-dimensional reconstruction module, a time-space fusion module and a dynamic visualization module; Performing scanning coverage analysis on the imaging frame selection area and outputting a non-conformal scanning path; Using the non-conformal scanning path as a constraint, driving the dynamic scanning module to dynamically scan the imaging target to obtain a two-dimensional slice sequence, wherein multiple frames of two-dimensional slices in the two-dimensional slice sequence are attached with multiple scanning time identifiers and multiple scanning angle identifiers; The ROI annotation module performs hierarchical modeling according to the interested demand information, obtains the ROI identification network, and then runs the ROI identification network to analyze and identify the two-dimensional slice sequence to obtain a two-dimensional identification sequence; Taking the multiple scanning time identifiers and the multiple scanning angle identifiers as constraints, the three-dimensional reconstruction module and the time-space fusion module are operated in coordination to perform time-space reconstruction and fusion on the two-dimensional identifier sequence to obtain a dynamic VOI; Taking the multiple scanning time identifiers and the multiple scanning angle identifiers as constraints, the three-dimensional reconstruction module and the time-space fusion module are operated in coordination to perform time-space reconstruction and fusion on the two-dimensional slice sequence to obtain a dynamic three-dimensional model; The dynamic visualization module is run to perform imaging fusion on the dynamic VOI and the dynamic three-dimensional model to obtain dynamic label volume data.

2. The imaging data analysis method of an X-ray system device according to claim 1, characterized in that: Taking the multiple scanning time identifiers and the multiple scanning angle identifiers as constraints, the three-dimensional reconstruction module and the time-space fusion module are operated in coordination to perform time-space reconstruction and fusion on the two-dimensional identifier sequence to obtain a dynamic VOI, the method comprising: The three-dimensional reconstruction module receives and generates an initial three-dimensional model according to the two-dimensional identification sequence and multiple scanning angle identifications; The spatiotemporal fusion module receives and performs a dynamic consistency evaluation on the initial three-dimensional model according to the multiple scanning time identifiers to obtain first model error information; The three-dimensional reconstruction module performs compensation reconstruction on the initial three-dimensional model according to the model error information to generate a first three-dimensional model; The spatiotemporal fusion module receives and performs a dynamic consistency evaluation on the first three-dimensional model according to the multiple scanning time identifiers to obtain second model error information; By analogy, the 3D reconstruction module and the spatiotemporal fusion module are alternately operated to perform spatiotemporal reconstruction and fusion on the 2D marker sequence until the model error information output by the spatiotemporal fusion module is lower than a preset error threshold, and the dynamic VOI is output.

3. The imaging data analysis method of an X-ray system device according to claim 1, characterized in that: Performing scanning coverage analysis on the imaging frame selection area and outputting a non-conformal scanning path, the method comprising: Performing spatial mapping according to the imaging frame selection area to obtain an imaging spatial range; Scheduling the dynamic scanning module to perform full-range equiangular scanning on the imaging target falling within the imaging space range to generate a local three-dimensional volume; Generate a conformal coverage scanning path by performing scanning coverage analysis on the local three-dimensional volume; A redundant scan path optimization is performed on the conformal coverage scan path to output the non-conformal scan path.

4. The imaging data analysis method of an X-ray system device according to claim 3, characterized in that: By performing scanning coverage analysis on the local three-dimensional volume to generate a conformal coverage scanning path, the method comprises: Extracting three-dimensional boundary constraints of imaging by performing voxel distribution analysis on the local three-dimensional volume; Generating a minimum envelope range according to the imaging three-dimensional boundary constraint fitting; Performing geometric shape feature recognition on the minimum envelope range, and decomposing the minimum envelope range according to the recognition result to obtain N regular areas and M irregular areas, wherein N and M are positive integers; Extracting boundary characteristics of the M irregular regions to obtain M boundary curvatures and M boundary shape change rates; Using the M boundary curvatures and the M boundary shape change rates to traverse a scanning interval number table to obtain M initial scanning intervals, wherein the scanning interval number table stores a plurality of sample curvatures, a plurality of sample shape change rates, and a plurality of sample scanning intervals in association with each other; By serializing the M initial scanning intervals, a target scanning interval is extracted; Based on the target scanning interval, the conformal coverage scanning path adapted to the global local three-dimensional volume is generated.

5. The imaging data analysis method of an X-ray system device according to claim 4, characterized in that: Performing redundant scanning path optimization on the conformal coverage scanning path and outputting the non-conformal scanning path, the method comprising: Preset segment angle parameters, and use the segment angle parameters as segmentation basis to divide the equiangular coverage scanning path into K segments of local paths; Calculate the coverage of the K local paths and output K coverage areas; Performing a cross comparison of adjacent paths on the K segments of local paths according to the K coverage areas to obtain K coverage overlaps of the K segments of local paths; Preset an overlap threshold, and divide the K local paths into a high-redundancy path segment set and a critical path segment set according to the overlap threshold and K coverage overlaps; Performing path elimination and optimization on the high-redundancy path segment set by coverage integrity check, so as to obtain a low-redundancy path segment set from the high-redundancy path segment set; Curve fitting is performed on the low-redundancy path segment set and the critical path segment set to generate the non-conformal scanning path.

6. The imaging data analysis method of an X-ray system device according to claim 1, characterized in that: The ROI annotation module performs hierarchical modeling according to the interested demand information to obtain a ROI identification network, and then runs the ROI identification network to analyze and identify the two-dimensional slice sequence to obtain a two-dimensional identification sequence. The method includes: The ROI annotation module extracts the interest requirement priority and multiple interest requirement types from the interest requirement information; Directedly dispatching a plurality of annotation function units from the annotation function library of the ROI annotation module according to the plurality of interested demand types; After configuring the computing power weights of the multiple labeling function units according to the priority of the interested requirements, the multiple labeling function units are connected in parallel, and a feature integration unit is configured at the output ends of the multiple labeling function units to obtain the ROI identification network; Inputting a first frame of two-dimensional slices in the two-dimensional slice sequence into the ROI identification network, and after the multiple annotation function units in the ROI identification network have annotated and processed to obtain multiple annotation result slices, the feature integration unit spatially integrates the multiple annotation result slices to obtain a first multi-dimensional annotation result; By analogy, the ROI identification network is run to analyze and identify the two-dimensional slice sequence to obtain the two-dimensional identification sequence, wherein the two-dimensional identification sequence includes multiple multi-dimensional annotation results corresponding to the multiple frames of two-dimensional slices.

7. The imaging data analysis method of an X-ray system device according to claim 6, characterized in that: Prior to directing and scheduling a plurality of annotation function units from the annotation function library of the ROI annotation module according to the plurality of interest requirement types, the method further comprises: By aggregating historical interesting demands, H sample demand types are obtained; Perform online data call according to the H sample demand types to obtain H standard annotation function models; Perform local data call according to the H sample demand types to obtain H historical annotated data sets; Using the H historical annotation data sets to optimize the H standard annotation function models, to obtain H sample function units; The H sample requirement types and the H sample function units are stored in association with each other to complete the pre-construction of the annotation function library.

8. The imaging data analysis method of an X-ray system device according to claim 1, characterized in that: The dynamic visualization module is run to perform imaging fusion on the dynamic VOI and the dynamic three-dimensional model to obtain dynamic label volume data. The method includes: After spatial alignment processing is performed on the dynamic VOI and the dynamic three-dimensional model, the dynamic VOI is displayed on the dynamic three-dimensional model by using pseudo-color annotation fusion according to the spatial distribution of the dynamic VOI to obtain initial label volume data; In the initial labeled volume data, the dynamic VOI and the dynamic three-dimensional model are fused frame by frame in time series to obtain the dynamic labeled volume data.

Images