A three-dimensional c-arm three-dimensional image metal segmentation method and system

Abstract

The application relates to the technical field of medical image processing, and particularly discloses a three-dimensional C-arm three-dimensional image metal segmentation method and system. The method is characterized in that a step-by-step processing strategy is adopted: first, a first set of high-low threshold combinations is used to segment three-dimensional image data to obtain a general metal region, and the general metal region is removed from the image after connected domain analysis and verification; then, the image data after removal is segmented by using a second set of high-low threshold combinations to obtain a titanium alloy region and perform analysis and verification; and finally, the regions that pass the two-step verification are combined to form an accurate metal segmentation result. The system correspondingly comprises functional modules for realizing the above steps. Through double-threshold step-by-step processing and a multiple verification mechanism, the segmentation accuracy and anti-interference ability in a multi-material metal coexistence scene are significantly improved, false segmentation caused by artifacts is effectively inhibited, a reliable foundation is provided for subsequent metal artifact removal processing, and therefore the imaging quality of a three-dimensional C-arm is finally improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical image processing, in particular to a three-dimensional C-arm three-dimensional image metal segmentation method and system. BACKGROUND

[0002] In the field of modern medical imaging, mobile three-dimensional C-arm X-ray machines have become a key clinical auxiliary tool in the operating room due to their imaging flexibility. Its three-dimensional imaging principle is consistent with that of computed tomography (CT), based on the attenuation difference of X-rays after penetrating human tissues, through the acquisition of multi-angle two-dimensional projection data, using reconstruction algorithms such as filtered back projection to generate three-dimensional images of human anatomical structures. Compared with CT devices mainly used for preoperative examination, three-dimensional C-arms can provide real-time and flexible three-dimensional imaging during surgery, accurately displaying the anatomical structure of the surgical area. With the rapid development of surgical robot technology and the popularity of minimally invasive surgery, the demand for high-quality three-dimensional C-arm images in the clinic is increasing.

[0003] However, when implants such as orthopedic internal fixators, prostheses, etc. exist in the imaging area, serious metal artifacts will be produced in the reconstructed image, which seriously interferes with diagnosis and surgical guidance. These artifacts are mainly caused by beam hardening, photon hunger, scattering and other physical effects, and exhibit various forms such as stripe-like artifacts, edge halos, dark areas, etc., significantly reducing image quality and affecting the visualization and measurement accuracy of anatomical structures. Metal artifact reduction technology is therefore one of the key links to improve the imaging quality of three-dimensional C-arms.

[0004] Currently, a widely used MAR technical route relies on accurate segmentation of the metal region in the image domain. The process is usually as follows: first, the metal region is extracted by threshold segmentation in the reconstructed three-dimensional image, then it is orthogonally projected into the projection domain to generate a metal mask, and then the metal shadow area in the projection data is repaired by interpolation, and finally the image with artifacts suppressed is obtained by reconstruction. The effect of this route is highly dependent on the accuracy of the initial metal segmentation in the image domain. The existing technology mostly uses an empirical single-threshold segmentation method, that is, a global threshold is set by appropriately floating the CT value (Hu value) of the bone. This method is still acceptable in simple scenarios (such as the presence of only a single high-density metal), but faces serious challenges: on the one hand, the CT values of different metal materials (such as ordinary stainless steel and titanium alloy) differ significantly, and the CT value of ordinary metal is very high, and the CT value of the bright artifact it produces is also high; the CT value of titanium alloy is relatively low, only slightly higher than that of bone. On the other hand, when there are multiple metal implants in the scene, the CT values of the metal implants and their artifacts have a wide distribution range and overlap with each other. The single-threshold method is difficult to accurately capture both high-value metals and low-value titanium alloys. If the threshold is set too high, titanium alloys will be missed, and if the threshold is set too low, the bright artifacts of high-value metals will be misclassified as metal. Although there are some improved methods for adaptively adjusting the single threshold, in complex multi-metal coexistence scenarios, the segmentation errors are still obvious, resulting in poor MAR processing results and failing to meet the high standard requirements of clinical image quality. Therefore, there is an urgent need for a segmentation method that can robustly and accurately handle multi-material metal coexistence scenarios.

[0005] Therefore, the prior art still needs further development. SUMMARY

[0006] The purpose of the present application is to overcome the above technical deficiencies and provide a three-dimensional C-arm three-dimensional image metal segmentation method and system to solve the problems existing in the prior art.

[0007] To achieve the above technical purpose, according to the first aspect of the present application, the present application provides a three-dimensional C-arm three-dimensional image metal segmentation method, comprising:

[0008] S100, performing first threshold segmentation processing on the three-dimensional image data to obtain a first metal region; performing connected component analysis and verification on the first metal region, and removing the first metal region meeting the conditions from the three-dimensional image based on the verification result;

[0009] S200, performing second threshold segmentation processing on the cleaned three-dimensional image data to obtain a second metal region;

[0010] S300, performing connected component analysis and verification on the second metal region; merging the first metal region and the second metal region after verification to generate a final metal segmentation result.

[0011] Specifically, the first threshold segmentation processing uses a set of high and low threshold values, wherein the low threshold value is used for preliminary segmentation, and the high threshold value is used for connected domain verification.

[0012] Specifically, the second threshold segmentation processing uses another set of high and low threshold values, wherein the low threshold value is used for preliminary segmentation, and the high threshold value is used for connected domain verification.

[0013] Specifically, the connected domain analysis includes 3D connected domain detection for identifying and extracting connected metal regions in the image.

[0014] Specifically, the connected domain analysis and verification includes at least one of volume verification and intensity verification.

[0015] Specifically, the volume verification is to determine whether the number of voxels of the connected domain is greater than a preset volume threshold.

[0016] Specifically, the intensity verification is to determine whether the maximum voxel value in the connected region is greater than a preset intensity threshold.

[0017] Specifically, in the first threshold segmentation processing, the first metal region includes a general metal region.

[0018] Specifically, in the second threshold segmentation processing, the second metal region includes a titanium alloy region.

[0019] Specifically, the method is executed on a GPU to accelerate processing.

[0020] Specifically, the three-dimensional image data is acquired by a mobile three-dimensional C-arm X-ray machine and reconstructed by a filtered back-projection algorithm.

[0021] Specifically, the high threshold value for the second threshold segmentation processing has a value greater than the low threshold value for the first threshold segmentation processing.

[0022] According to a second aspect of the present application, a three-dimensional C-arm three-dimensional image metal segmentation system is provided, comprising:

[0023] A first segmentation module is configured to perform first threshold segmentation processing on three-dimensional image data to obtain a first metal region.

[0024] A first verification and removal module is configured to perform connected domain analysis and verification on the first metal region, and remove the first metal region meeting the conditions from the three-dimensional image based on the verification result.

[0025] A second segmentation module is configured to perform second threshold segmentation processing on the three-dimensional image data after removal to obtain a second metal region.

[0026] A second checking module is configured to perform connected component analysis and checking on the second metal region.

[0027] A merging and output module is configured to merge the first metal region and the second metal region after checking, and output a final metal segmentation result.

[0028] Specifically, the threshold values used by the first segmentation module and the second segmentation module can be adaptively adjusted according to the metal material characteristics.

[0029] Specifically, the system further comprises a GPU acceleration module configured to enable GPU parallel computing during the segmentation and checking processes.

[0030] Specifically, the final metal segmentation result output by the merging and output module is used for subsequent metal artifact reduction (MAR) processing.

[0031] Beneficial effects:

[0032] Compared with the prior art, the three-dimensional C-arm three-dimensional image metal segmentation method and system provided by the application have a series of significant beneficial effects, mainly in improving segmentation accuracy, enhancing robustness, and improving overall process efficiency.

[0033] The most core beneficial effect of the application is that it significantly improves the segmentation accuracy in complex scenes where ordinary metals and titanium alloys coexist. By adopting a step-by-step double-threshold processing strategy, the method first optimizes the segmentation and removal of ordinary metals with high CT values and their characteristics, and then specifically segments titanium alloys with different characteristics, effectively avoiding the problems of missed segmentation and missegmentation caused by a single threshold value that cannot adapt to multiple materials. The method can more completely and accurately separate all metal implants from background tissues and artifacts, providing a high-quality and reliable input basis for subsequent metal artifact reduction processing.

[0034] Another important beneficial effect of the application is its excellent robustness. The method incorporates multiple checking mechanisms, including using a high threshold to check the intensity of the connected component and using a volume threshold to check the area. These checking steps can effectively identify and filter out false segmentation regions formed by metal artifacts (especially bright artifacts produced by ordinary metals) or image noise, greatly reducing the occurrence of false segmentation. Even in images with severe artifact interference, the method can maintain stable and reliable segmentation performance, reducing the over-reliance on the accuracy of the initial threshold setting, enhancing the anti-interference ability and generalization ability of the method.

[0035] In addition, the present application creates favorable conditions for accurate segmentation of titanium alloy by first segmenting and removing the general metal region. This step eliminates the large interference of high-density metal on adjacent pixel values and the resulting artifacts, avoids the misjudgment of general metal artifacts as targets when segmenting titanium alloy, and enables the second step of segmentation to be performed on relatively "clean" image data, thereby further improving the overall accuracy of the final segmentation result. By setting the key threshold relationship of "titanium alloy high threshold greater than metal low threshold", the present application constructs a cascading filtering mechanism at the algorithm level. This mechanism can automatically and effectively discard the bright artifacts produced by general metal during the titanium alloy segmentation stage without additional complex operations, ensuring both segmentation accuracy and algorithm efficiency. At the same time, the present application is designed for GPU-accelerated parallel computing and can efficiently process large-scale three-dimensional body data generated by three-dimensional C-arm systems, meeting the stringent requirements of real-time or quasi-real-time imaging for processing speed in clinical procedures and ensuring smooth and efficient workflow. In summary, through its innovative design, the present application ultimately achieves the fundamental purpose of improving metal segmentation quality and significantly improving the visual effect and practical value of the final three-dimensional image of the three-dimensional C-arm system. BRIEF DESCRIPTION OF DRAWINGS

[0036] Figure 1 is a flowchart of the three-dimensional C-arm three-dimensional image metal segmentation method provided in the embodiments of the present application;

[0037] Figure 2 is a system composition diagram of the three-dimensional C-arm three-dimensional image metal segmentation system provided in the embodiments of the present application. DETAILED DESCRIPTION

[0038] In order to better understand the technical solutions of the present application, the technical solutions of the present application will be described clearly and completely below in conjunction with the drawings of the present application. Based on the embodiments in the present application, other similar embodiments obtained by those skilled in the art without making creative efforts should fall within the scope of protection of the present application.

[0039] It can be understood that when there are general metals with high Hu values and titanium alloys with slightly low Hu values in the image, neither the single threshold method based on experience nor the single threshold method based on adaptive adjustment can well segment the metals and titanium alloys at the same time. The present application considers the Hu value characteristics of general metals, titanium alloys and bones in the reconstructed image, as well as the Hu value and morphological characteristics of metal artifacts, and thus designs an algorithm to improve the robustness of the metal segmentation algorithm.

[0040] Further, the common metal Hu value is high, and its artifact Hu value presents both bright and dark forms. When metal segmentation is performed, the dark artifact does not affect the metal segmentation and can be ignored. For the bright artifact, its value is usually lower than the metal and gradually spreads from high to low. In view of the characteristics that the common metal Hu value is high and the bright artifact Hu value of the common metal is also high, a set of high and low threshold values (high ThreshMetal and low ThreshMetal) are designed, and the low threshold value is used to segment the common metal region, and the high threshold value is used to check whether the region is a common metal region. After segmentation using the set of high and low threshold values, the common metal region is emptied from the image domain. Then, in view of the characteristics that the titanium alloy Hu value is low, another set of high and low threshold values (high ThreshTai and low ThreshTai) are designed. Similarly, the low threshold value is used for segmentation, and the high threshold value is used for checking. Through such a design, when the first set of threshold values is segmented, a better common metal segmentation result can be obtained. When the second set of threshold values is segmented, the artifact region of the common metal can be filtered through checking. Through the combination of the two sets of high and low threshold values, the robustness of the metal segmentation algorithm is improved. In addition to the combination of high and low threshold values, the area of the segmented region is also checked. Therefore, the robustness of the algorithm is further improved.

[0041] It can be understood that the present application protects a three-dimensional C-arm three-dimensional image metal segmentation method, the core of which is to adopt a step-by-step double-threshold segmentation strategy to process metals of different materials in turn. In specific implementation, first, a first threshold value is used to segment three-dimensional image data (such as volume data obtained by filtering back projection reconstruction from C-arm acquisition). The threshold value for common metal should be set relatively high, for example, the low threshold value can be set to 3000 HU, and the high threshold value can be set to 6000 HU, to preliminarily segment the common metal region with a high Hu value. Subsequently, 3D connected domain analysis is performed on the segmented region to identify independent metal regions, and checking is performed: the checking conditions include whether the voxel number (volume) of the connected domain is greater than a preset threshold value (such as 100 voxels) and whether the maximum Hu value in the region is higher than the high threshold value. The connected domain that passes the checking is determined as a real metal region, and its voxel value is cleared from the original image to eliminate its interference on subsequent segmentation. Next, a second threshold value is used to segment the image data after the common metal is removed. The threshold value for titanium alloy should be set to be higher than the bone but significantly lower than the common metal, for example, the low threshold value can be set to 1500 HU. The high threshold value for titanium alloy must be greater than the low threshold value (3000 HU) of the aforementioned metal, for example, it can be set to 3500 HU. Similarly, connected domain analysis and checking (the checking conditions are similar to the aforementioned, but the threshold parameters are different) are performed on the titanium alloy region. Finally, the metal regions that pass the two-step segmentation and checking are merged, and the final metal segmentation result is output.

[0042] It can be understood that the method effectively avoids mutual interference of ordinary metal and titanium alloy caused by Hu value difference through step-by-step processing and clearing strategy; through connected domain verification, false segmentation caused by artifacts and noise is excluded, and the accuracy of metal segmentation is significantly improved, thereby providing a high-quality basis for subsequent MAR processing.

[0043] The application will be further described below in conjunction with the drawings and preferred embodiments.

[0044] Please refer to Figure 1 The application provides a three-dimensional C-shaped arm three-dimensional image metal segmentation method, comprising:

[0045] S100, first threshold segmentation processing is performed on three-dimensional image data to obtain a first metal region; connected domain analysis and verification are performed on the first metal region, and the first metal region meeting the conditions is removed from the three-dimensional image based on the verification result.

[0046] Specifically, the first threshold segmentation processing uses a group of high threshold values and low threshold values, wherein the low threshold values are used for preliminary segmentation, and the high threshold values are used for connected domain verification.

[0047] It should be further explained that in implementation, the low threshold value (such as 3000HU) is used for preliminary segmentation to capture all voxels that may belong to ordinary metal; the high threshold value (such as 6000HU) is used for verification, and only the region with a maximum Hu value exceeding the high threshold value in the connected domain is retained.

[0048] It can be understood that the application adopts a combination of high and low threshold values, which not only ensures complete capture of the ordinary metal region (low threshold value), but also excludes the interference of high Hu value artifacts through high threshold value verification, thereby improving the reliability of ordinary metal segmentation.

[0049] Specifically, in the first threshold segmentation processing, the first metal region includes an ordinary metal region.

[0050] It should be further explained that the application preferably uses the first metal region as the ordinary metal region. In implementation, ordinary metal refers to metal materials such as stainless steel and cobalt-chromium alloy with a very high Hu value (such as >3000HU).

[0051] S200, second threshold segmentation processing is performed on the removed three-dimensional image data to obtain a second metal region.

[0052] Specifically, the second threshold segmentation processing uses another group of high threshold values and low threshold values, wherein the low threshold values are used for preliminary segmentation, and the high threshold values are used for connected domain verification.

[0053] It should be further explained that the second threshold segmentation of the application preferably uses another set of high and low thresholds. In practice, the low threshold (e.g. 1500HU) of the titanium alloy segmentation is used for preliminary segmentation, and the high threshold (e.g. 3500HU) is used for verification.

[0054] It can be understood that, in view of the low Hu value of the titanium alloy, a special threshold set is set, and the remaining artifacts after the ordinary metal is removed are filtered out through high threshold verification, which ensures the accuracy of the titanium alloy segmentation.

[0055] Specifically, in the second threshold segmentation process, the second metal region includes a titanium alloy region.

[0056] It should be further explained that the second metal region of the application is preferably a titanium alloy region. In practice, the titanium alloy refers to a metal material with a relatively low Hu value (e.g. 1500-3500HU).

[0057] S300, performing connected component analysis and verification on the second metal region; merging the first metal region and the second metal region after verification to generate a final metal segmentation result.

[0058] Specifically, the connected component analysis includes 3D connected component detection, which is used to identify and extract the connected metal region in the image.

[0059] It should be further explained that the application preferably uses 3D connected component detection for connected component analysis. In practice, the 26 connectivity criterion is used to label the binarized voxels to identify the spatially connected metal region.

[0060] It can be understood that the 3D connected component detection can accurately identify the independent metal implant in the three-dimensional space, which provides a basis for subsequent verification and removal, and avoids the error merging of different metal objects.

[0061] Specifically, the connected component analysis and verification includes at least one of volume verification and intensity verification.

[0062] It should be further explained that the application preferably includes volume verification and / or intensity verification for connected component verification. In practice, the volume verification judges whether the number of connected component voxels is greater than the minimum volume threshold (e.g. 100 voxels), and the intensity verification judges whether the maximum Hu value in the connected component is greater than the corresponding high threshold.

[0063] It can be understood that the volume verification filters out small particle artifacts formed by noise, and the intensity verification ensures that the remaining region has the high Hu value characteristics of the real metal. The double verification mechanism greatly improves the robustness of the segmentation result.

[0064] Specifically, the volume verification is to determine whether the number of voxels in the connected domain is greater than a preset volume threshold.

[0065] It needs to be further explained that the volume verification of the application is preferably to determine whether the number of voxels is greater than a preset volume threshold. In implementation, the volume threshold is set according to the minimum size of a typical metal implant (such as 50-200 voxels), and the connected domain with a number of voxels lower than the threshold is removed.

[0066] It can be understood that the above scheme effectively removes scattered point noise and small artifacts, and avoids their influence on the final segmentation result.

[0067] Specifically, the intensity verification is to determine whether the maximum voxel value in the connected region is greater than a preset intensity threshold.

[0068] It needs to be further explained that the intensity verification of the application is preferably to determine whether the maximum voxel value in the connected domain is greater than a preset intensity threshold. In implementation, the intensity threshold is the high threshold used in the first threshold segmentation process.

[0069] It can be understood that the above scheme ensures that each retained connected domain has the high-density feature of real metal, and further prevents high-brightness artifacts from being misjudged as metal.

[0070] Specifically, the method is executed on a GPU to accelerate processing.

[0071] It needs to be further explained that the method of the application is preferably executed on a GPU. In implementation, the threshold segmentation, connected domain marking and other computationally intensive tasks are deployed to the GPU for parallel processing by using a parallel computing framework such as CUDA or Open CL.

[0072] It can be understood that the above scheme greatly accelerates the processing process of three-dimensional large data, and meets the high requirement of real-time imaging on processing speed in surgery.

[0073] Specifically, the three-dimensional image data is acquired by a mobile three-dimensional C-arm X-ray machine and obtained by a filtered back-projection algorithm.

[0074] It needs to be further explained that the three-dimensional image data of the application is preferably acquired by a mobile three-dimensional C-arm and obtained by filtered back-projection. In implementation, the C-arm rotates around the patient to acquire hundreds of two-dimensional projection images, and then uses a FBP or iterative reconstruction algorithm to generate three-dimensional volume data.

[0075] It can be understood that, as shown in the flow chart, threshold segmentation, connected component detection, connected component verification and other methods are combined in the present application to segment the metal region from the 3D image. In the following, the present inventors refer to the 3D image as volume, and each point in the volume as voxel. The 3D image data is large, and the algorithm of the present inventors is run on GPU to accelerate.

[0076] It can be understood that threshold segmentation is to use a threshold, and the voxels below the threshold are set to 0, while the voxels above the threshold remain the original value. 3D connected component detection is an extension of 2D connected component detection on 3D images. The present application directly uses an open source method to realize 3D connected component detection.

[0077] It needs to be further explained that in the connected component analysis step, the present inventors traverse the connected component to obtain the number of voxels and the maximum value. The number of voxels multiplied by the volume of a single voxel is the volume of the connected component. In actual use, the present inventors directly use the number of voxels for convenience.

[0078] It can be understood that the verification of the connected component is very simple, which directly compares the volume of the connected component with the volume threshold, and simultaneously compares the maximum value of the connected component with the maximum value threshold. Both of them pass, and the connected component is retained.

[0079] Further, the metal threshold segmentation is used to obtain the metal region connected component; then all the voxels corresponding to the metal connected component region in the volume are cleared; then the titanium alloy threshold is used to obtain the titanium alloy region connected component; finally, the two connected components are merged to obtain the final metal segmentation result.

[0080] Further, in the design, the metal low threshold is slightly lower than the titanium alloy high threshold. In this way, when the titanium alloy high threshold is used for verification, the artifacts of ordinary metal cannot pass the verification, so that the algorithm can accurately segment the metal.

[0081] Specifically, the high threshold value used for the second threshold segmentation processing is greater than the low threshold value used for the first threshold segmentation processing.

[0082] Further, the second high threshold value used for segmenting the titanium alloy is set to be greater than the first low threshold value used for segmenting the ordinary metal. Through this threshold relationship design, when the titanium alloy region is segmented and verified, the voxel value of the ordinary metal and its high value artifact region is lower than the first low threshold value, and therefore is necessarily lower than the second high threshold value, so that it cannot pass the intensity verification of the second high threshold value and can be effectively filtered. This design fundamentally avoids the inclusion of ordinary metal artifacts in the titanium alloy segmentation result, and significantly improves the segmentation accuracy.

[0083] It can be understood that in the preferred embodiment of the present application, the key of the threshold parameter lies in its relative numerical relationship: that is, the high threshold of titanium alloy (for example, 3500HU) must be greater than the low threshold of metal (for example, 3000HU). As shown in the step S100, the region whose voxel value is higher than the low threshold of metal has been segmented and removed. In the subsequent steps S200 and S300, when the candidate region is checked using the high threshold of titanium alloy, those metal artifacts near the ordinary metal which remain after the step S100, because their Hu values are lower than the low threshold of metal, must also be lower than the high threshold of titanium alloy, and when the high threshold of titanium alloy is checked, they will be filtered. In this way, the final merged metal segmentation result can accurately contain the real ordinary metal and titanium alloy, while the false inclusion of artifacts is minimized. Figure 1

[0084] Please refer to Figure 2 The present application provides another embodiment, which provides a three-dimensional C-arm three-dimensional image metal segmentation system, comprising:

[0085] A first segmentation module 100 is configured to perform first threshold segmentation processing on the three-dimensional image data to obtain a first metal region;

[0086] A first checking and removing module 200 is configured to perform connected domain analysis and checking on the first metal region, and remove the first metal region meeting the conditions from the three-dimensional image based on the checking result;

[0087] A second segmentation module 300 is configured to perform second threshold segmentation processing on the three-dimensional image data after removal to obtain a second metal region;

[0088] A second checking module 400 is configured to perform connected domain analysis and checking on the second metal region;

[0089] A merging and output module 500 is configured to merge the first metal region and the second metal region after checking, and output a final metal segmentation result.

[0090] It needs to be further explained that the system includes five modules: the first segmentation module is configured to load the three-dimensional volume data and apply the first set of thresholds for segmentation; the first checking and removing module performs 3D connected domain analysis and checking on the segmentation result, and clears the region voxels confirmed as ordinary metal; the second segmentation module applies the second set of thresholds for titanium alloy segmentation on the processed data; the second checking module performs similar connected domain checking on the titanium alloy segmentation result; the merging and output module merges the metal regions passing the checking in two steps into the final segmentation result, and outputs to the MAR algorithm or other downstream modules. Each module can be integrated in the image processing workstation of the C-arm system, and the GPU hardware resources are accelerated for calculation through software calling.​

[0091] It can be understood that the system realizes all advantages of the above method in a modular form, provides a complete solution, and facilitates integration into existing medical imaging equipment to improve the metal artifact processing capability thereof.

[0092] Specifically, the threshold values used by the first segmentation module and the second segmentation module can be adaptively adjusted according to the metal material characteristics.

[0093] It should be further explained that the system can adaptively adjust the threshold values according to the metal material characteristics. In implementation, the system can include a preset threshold lookup table, and automatically select or fine-tune the specific values of the high and low threshold values according to the common materials of different metal implants (obtained from medical records or during surgery).

[0094] It can be understood that the above scheme enhances the adaptability of the system to different patients and different implants, and further improves the segmentation accuracy and the generalization ability of the algorithm.

[0095] Specifically, the system further includes a GPU acceleration module for enabling GPU parallel computing during the execution of the segmentation and verification processes.

[0096] It should be further explained that the system includes a GPU acceleration module. In implementation, the module is responsible for managing GPU memory allocation, starting parallel computing kernels (such as CUDA Kernel for threshold segmentation and connected component labeling), and realizing hardware acceleration of the entire segmentation process.

[0097] It can be understood that the above scheme ensures that the system can meet the stringent real-time requirements of the clinic and realize efficient processing.

[0098] Specifically, the final metal segmentation result output by the merging output module is used for subsequent metal artifact reduction (MAR) processing.

[0099] It should be further explained that the output of the system is used for subsequent MAR processing. In implementation, the final metal segmentation result is passed to the MAR algorithm in the form of a binary image mask (Mask) to guide the projection domain interpolation or image domain correction.

[0100] In a preferred embodiment, the present application also provides an electronic device, which comprises:

[0101] The computer device can be a server, a terminal, or any other electronic device with necessary computing and / or processing capabilities in a broad sense. In an embodiment, the computer device can include a processor, a memory, a network interface, a communication interface, and the like connected by a system bus. The processor of the computer device can be configured to provide necessary computing, processing, and / or control capabilities. The memory of the computer device can include a non-volatile storage medium and an internal memory. The non-volatile storage medium can store an operating system, a computer program, and the like therein or thereon. The internal memory can provide an environment for running the operating system and the computer program in the non-volatile storage medium. The network interface and the communication interface of the computer device can be configured to connect and communicate with external devices through a network. The computer program, when executed by the processor, performs the steps of the method of the present application.

[0102] The present application can be implemented as a computer readable storage medium having a computer program stored thereon, which, when executed by a processor, causes the steps of the method of the embodiments of the present application to be performed. In an embodiment, the computer program is distributed over a plurality of computer devices or processors coupled by a network, such that the computer program is stored, accessed, and executed by one or more computer devices or processors in a distributed manner. A single method step / operation, or two or more method steps / operations, can be performed by a single computer device or processor, or by two or more computer devices or processors. One or more method steps / operations can be performed by one or more computer devices or processors, and one or more other method steps / operations can be performed by one or more other computer devices or processors. One or more computer devices or processors can perform a single method step / operation, or perform two or more method steps / operations.

[0103] It will be appreciated by those skilled in the art that the method steps of the present application can be instructed by a computer program to relevant hardware such as a computer device or a processor, which can be stored in a non-transitory computer readable storage medium, and which, when executed, causes the steps of the present application to be performed. Depending on the circumstances, any reference herein to a memory, storage, database, or other medium can include non-volatile and / or volatile memory. Examples of non-volatile memory include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), flash memory, magnetic tapes, floppy disks, magneto-optical data storage devices, optical data storage devices, hard disks, solid-state disks, and the like. Examples of volatile memory include random access memory (RAM), external cache memory, and the like.

[0104] The technical features described above can be combined arbitrarily. Although all possible combinations of the technical features are not described, any combination of the technical features should be considered to be covered by the present specification, as long as there is no contradiction in such a combination.

[0105] The specific embodiments of the application described above are not to be construed as limiting the scope of the present application. Any other corresponding changes and modifications of the technical concept according to the present application should be included in the scope of the protection of the present application.

Claims

1. A method for metal segmentation of a three-dimensional C-arm image, characterized in that, Includes the following steps: S100. Perform a first threshold segmentation process on the three-dimensional image data to obtain a first metal region; The first metal region is subjected to connected component analysis and verification, and the first metal region that meets the conditions is removed from the three-dimensional image based on the verification results; S200. Perform a second threshold segmentation process on the cleared 3D image data to obtain the second metal region; S300: Perform connected component analysis and verification on the second metal region; merge the verified first metal region and the second metal region to generate the final metal segmentation result; The first threshold segmentation process uses a set of high thresholds and low thresholds, where the low thresholds are used for initial segmentation and the high thresholds are used for connected component verification. The second threshold segmentation process uses another set of high and low thresholds, where the low threshold is used for initial segmentation and the high threshold is used for connected component verification. The high threshold used for the second threshold segmentation process is greater than the low threshold used for the first threshold segmentation process.

2. The method according to claim 1, characterized in that, The connected component analysis includes 3D connected component detection, used to identify and extract connected metallic regions in an image.

3. The method according to claim 2, characterized in that, The connected component analysis and verification includes at least one of volume verification and strength verification.

4. The method according to claim 3, characterized in that, The volume check is used to determine whether the number of voxels in the connected component is greater than a preset volume threshold.

5. The method according to claim 3, characterized in that, The intensity verification is to determine whether the maximum voxel value in the connected region is greater than a preset intensity threshold.

6. The method according to claim 1, characterized in that, In the first threshold segmentation process, the first metal region includes a normal metal region.

7. The method according to claim 1, characterized in that, In the second threshold segmentation process, the second metal region includes a titanium alloy region.

8. The method according to claim 1, characterized in that, The method is executed on a GPU to accelerate processing.

9. The method according to claim 1, characterized in that, The three-dimensional image data was acquired by a mobile three-dimensional C-arm X-ray machine and reconstructed using a filtered back projection algorithm.

10. A three-dimensional C-arm three-dimensional image metal segmentation system, characterized in that, The three-dimensional C-arm three-dimensional image metal segmentation method according to any one of claims 1-9 includes: The first segmentation module is used to perform a first threshold segmentation process on the three-dimensional image data to obtain a first metal region; The first verification and clearing module is used to perform connected component analysis and verification on the first metal region, and clear the first metal region that meets the conditions from the three-dimensional image based on the verification results. The second segmentation module is used to perform a second threshold segmentation process on the cleared 3D image data to obtain the second metal region. The second verification module is used to perform connected component analysis and verification on the second metal region; The merge output module is used to merge the verified first metal region and the second metal region, and output the final metal segmentation result.

11. The system according to claim 10, characterized in that, The thresholds used by the first and second segmentation modules are adaptively adjusted based on the characteristics of the metal material.

12. The system according to claim 10, characterized in that, The system also includes a GPU acceleration module for enabling GPU parallel computing during the segmentation and verification process.

13. The system according to claim 10, characterized in that, The final metal segmentation result output by the merging output module is used for subsequent metal artifact removal (MAR) processing.

Images