CBCT (cone beam computed tomography) metal artifact suppression method and system based on preoperative prior guidance

Abstract

The invention discloses a CBCT metal artifact suppression method and system based on preoperative prior guidance, and the method comprises the steps: taking second CBCT image data, obtained in real time in an operation, of a target tissue as a reference, and carrying out the rigid registration of first CBCT image data, obtained before the operation, of the target tissue; deforming the first CBCT image data to obtain third image data; the relative positions of the tissue structures in the third image data and the second CBCT image data are consistent; based on the third image data, performing coarse segmentation on the metal implant in the second CBCT image data by adopting a digital subtraction technology; segmenting to obtain first mask data and background prior data of the metal implant; performing true and false positive judgment on the first mask data and performing false positive suppression processing to obtain second mask data; carrying out data restoration on the background prior data by adopting a self-adaptive interpolation grid to obtain projection domain data; and reconstructing by using an FDK reconstruction algorithm to obtain a CBCT image of the target tissue without metal artifacts.

Description

Technical Field

[0001] This application relates to the field of computer image processing, and more specifically, to a method and system for suppressing metal artifacts in CBCT based on preoperative prior guidance. Background Technology

[0002] In medical imaging diagnosis, metallic implants such as orthopedic fixation devices, artificial joints, or dental prostheses often exhibit strong attenuation of X-rays during CT imaging due to the unique physical properties of metals. This attenuation, caused by the high atomic number of the metal, not only leads to missing projection data and streaking artifacts but also causes beam hardening and scattering noise, manifesting as characteristic radial or ring-shaped artifacts on the image. These artifacts severely affect the accurate interpretation of images and clinical diagnosis. Furthermore, the interference of metals with the radio frequency field further degrades image quality.

[0003] To address these artifact problems, existing medical imaging professionals have conducted technical research, such as relying on prior assumptions about known metal masks or employing simple thresholding strategies, followed by reconstruction based on the segmentation results. However, existing artifact correction techniques still have many limitations: the computational process is complex and time-consuming, potentially introducing new image artifacts; the segmentation accuracy is limited for certain metal types or specific implantation locations; achieving an ideal balance between improving the signal-to-noise ratio and maintaining spatial resolution is difficult; and some advanced algorithms rely on specific hardware support or prior knowledge, limiting their application in routine clinical settings. Summary of the Invention

[0004] To address the aforementioned technical problems, this application discloses a method and system for suppressing metal artifacts in CBCT based on preoperative a priori guidance. By intelligently registering preoperative metal-free images with intraoperative CBCT data and utilizing the consistency of the patient's own anatomical structure, precise segmentation and reconstruction of the metal implant are achieved. Specifically, the technical solution of this application is as follows:

[0005] In a first aspect, this application discloses a method for suppressing metal artifacts in CBCT based on preoperative a priori guidance, comprising the following steps:

[0006] Using the second CBCT image data of the target tissue acquired in real time during the operation as a reference, the first CBCT image data of the target tissue acquired before the operation is rigidly registered; so that the first CBCT image data is deformed to obtain the third image data; the relative positions of each tissue structure in the third image data are consistent with those in the second CBCT image data.

[0007] Based on the third image data, digital subtraction angiography is used to coarsely segment the metal implant in the second CBCT image data; the segmentation yields the first mask data of the metal implant and background prior data.

[0008] The first mask data is subjected to true / false positive judgment and false positive suppression processing to obtain the second mask data of the metal implant; the background prior data is repaired by an adaptive interpolation grid to obtain the projection domain data.

[0009] Based on the projection domain data and the second mask data, the CBCT image of the target tissue without metal artifacts is reconstructed using the FDK reconstruction algorithm.

[0010] In some embodiments, the second CBCT image data, in addition to including all tissue structural features of the first CBCT image data, also includes radiation signal features generated by the metal implant.

[0011] In some embodiments, before rigid registration of the first CBCT image data, the method further includes: preprocessing the first CBCT image data and the second CBCT image data respectively; specifically including:

[0012] Correct the target projection data to eliminate inherent hardware errors;

[0013] The corrected target projection data is converted into linear projection data that conforms to the physical model of X-ray attenuation by logarithmic transformation; image noise is adaptively suppressed based on filtering algorithm.

[0014] In some embodiments, the rigid registration of the first CBCT image data of the target tissue acquired preoperatively, using the second CBCT image data of the target tissue acquired in real time during the operation as a reference, specifically includes the following sub-steps:

[0015] Mutual information is used as a similarity metric for registration between the first CBCT image data and the second CBCT image data;

[0016] Based on the stochastic gradient descent optimization algorithm, the image transformation parameters are dynamically adjusted by iteratively calculating the gradient of the mutual information with respect to the image transformation parameters, thereby maximizing the similarity metric.

[0017] Based on the image transformation parameters, the first CBCT image data is subjected to geometric transformation processing; an interpolation algorithm is used to estimate the grayscale value of the transformed image in real time, so that the first CBCT image data and the second CBCT image data are aligned in three-dimensional space.

[0018] In some implementations, the method of using digital subtraction angiography to coarsely segment the metal implant in the second CBCT image data based on the third image data, and obtaining first mask data and background prior data for the metal implant, specifically includes the following sub-steps:

[0019] The third image data is subtracted from the second CBCT image data using digital subtraction angiography to obtain the first mask data of the metal implant.

[0020] The background prior data is obtained by subtracting the first mask data from the second CBCT image data.

[0021] In some embodiments, the step of performing a true / false positive determination and false positive suppression processing on the first mask data to obtain the second mask data of the metal implant specifically includes the following sub-steps:

[0022] Based on the physical properties and morphological characteristics of the metal implant, the first mask data is subjected to feature analysis and discrimination of true and false positives.

[0023] The false positive regions identified are subjected to artifact suppression processing, and the second mask data is obtained after processing.

[0024] In some implementations, the background prior data is repaired using an adaptive interpolation grid to obtain projection domain data; specifically, this includes the following sub-steps:

[0025] The missing areas of the background prior data are accurately located using a metal mask; an interpolation grid is constructed using an adaptive interpolation method to repair the missing areas while maintaining the topological structure of the projected data.

[0026] In some implementations, the step of reconstructing a CBCT image of the target tissue without metal artifacts using the FDK reconstruction algorithm based on the projection domain data and the second mask data specifically includes the following sub-steps:

[0027] The projection domain data and the second mask data are subjected to ramp filtering.

[0028] Back projection weights are calculated for the scanning geometric parameters in the filtered projection domain data and the second mask data; and a weighted integral is performed on the filtered projection domain data and the second mask data along the X-ray path to obtain the reconstructed three-dimensional volume data.

[0029] Secondly, this application also discloses a CBCT metal artifact suppression system based on preoperative prior guidance, including:

[0030] The registration module is used to rigidly register the first CBCT image data of the target tissue acquired before surgery with reference to the second CBCT image data of the target tissue acquired in real time during surgery; so that the first CBCT image data is deformed to obtain third image data; wherein the relative positions of each tissue structure in the third image data are consistent with those in the second CBCT image data.

[0031] The segmentation module is used to coarsely segment the metal implant in the second CBCT image data based on the third image data and using digital subtraction angiography; the segmentation obtains the first mask data of the metal implant and background prior data.

[0032] The optimization module is used to perform true / false positive judgment on the first mask data and perform false positive suppression processing to obtain the second mask data of the metal implant;

[0033] The repair module is used to repair the background prior data using an adaptive interpolation grid, and the repaired data is the projection domain data.

[0034] The reconstruction module is used to reconstruct a CBCT image of the target tissue without metal artifacts based on the projection domain data and the second mask data using the FDK reconstruction algorithm.

[0035] Thirdly, this application also discloses a CBCT device, which includes the CBCT metal artifact suppression system based on preoperative prior guidance described in the above embodiments.

[0036] Compared with the prior art, this application has at least one of the following beneficial effects:

[0037] 1. The deep fusion method of preoperative anatomical information and intraoperative dynamic scanning data proposed in this application fully utilizes the homologous nature of preoperative and postoperative imaging data of the same patient, enabling precise identification and localization of non-human tissue structures such as newly added metal implants during surgery. This processing strategy based on individualized prior knowledge not only effectively solves the segmentation error problem caused by the lack of patient-specific information in traditional methods, but also provides a more reliable and accurate solution for suppressing metal artifacts in clinical practice.

[0038] 2. The metal artifact suppression method based on multi-temporal image collaborative analysis proposed in this application uses a high-precision image registration algorithm to spatially align preoperative CBCT data without metal artifacts with intraoperative CBCT scan data, establishing a spatiotemporal correspondence with anatomical consistency.

[0039] 3. The proposed method for precise metal region segmentation based on coarse segmentation optimization significantly improves the segmentation accuracy of metal implants through a two-stage processing framework from coarse segmentation to fine refinement. First, the coarse segmentation result obtained through the above process is used. Then, multi-dimensional features of intraoperative CBCT data are utilized for intelligent identification and removal of false positive regions. Furthermore, the false positive detection algorithm in this application can effectively distinguish between real metal regions and interfering factors such as high-density bone tissue, ultimately outputting accurate metal segmentation results. This progressive optimization strategy not only overcomes the limitations of traditional single-stage segmentation methods but also significantly improves segmentation robustness in complex clinical scenarios, laying a solid foundation for subsequent metal artifact suppression. Attached Figure Description

[0040] The preferred embodiments will now be described in a clear and easy-to-understand manner, in conjunction with the accompanying drawings, to further explain the above-mentioned characteristics, technical features, advantages, and implementation methods of this application.

[0041] Figure 1 This is a flowchart illustrating the steps of an embodiment of the CBCT metal artifact suppression method based on preoperative prior guidance in this application.

[0042] Figure 2 A comparison diagram showing the effect of the second mask data of the metal implant segmented by the existing technical solution and the CBCT metal artifact suppression method based on preoperative prior guidance provided in this application;

[0043] Figure 3 This is a comparison of the effects of existing technical solutions and the CBCT metal artifact suppression method based on preoperative prior guidance provided in this application, which reconstructs three-dimensional volume data. Detailed Implementation

[0044] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application can also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0045] Furthermore, in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0046] Cone-beam computed tomography (CBCT), a widely used three-dimensional imaging technique in dentistry, radiotherapy, and interventional surgery, offers significant advantages over traditional CT, such as lower radiation dose and faster scanning speed. However, it faces a key challenge in practical clinical applications: when patients have metallic implants, such as dental fillings or orthopedic internal fixation devices, the strong absorption and scattering of X-rays by these high-density materials can lead to severe metal artifacts in the reconstructed images. These artifacts mainly manifest as streaks and dark bands, and their formation mechanism involves three physical processes: first, photon starvation, where the high attenuation of X-rays by metals results in signal loss in the metal region; second, radiation hardening, which alters the X-ray energy spectrum and produces cup-shaped artifacts; and finally, increased noise due to scattered radiation. These artifacts not only degrade image quality, blurring the boundaries of tissues around the metal and distorting CT values, but also severely affect the accuracy of clinical diagnosis. For example, in dental treatment, they may mask periapical lesions, and in radiotherapy, they may interfere with target delineation and dose calculation. Although there are currently various solutions, such as dual-energy CT, iterative reconstruction algorithms, and deep learning methods, these technologies each have their limitations, such as high equipment requirements, high computational complexity, or the need for a large amount of training data. Therefore, developing more effective methods to suppress metal artifacts remains an important research direction for the development of CBCT technology. It requires comprehensive improvements from multiple levels, including physical mechanisms, imaging techniques, and algorithm optimization, to improve the imaging quality and diagnostic value of CBCT in cases involving metal implants.

[0047] Currently, existing metal segmentation methods are mainly divided into two categories, but each has its limitations, as analyzed below.

[0048] The first category is image-domain-based thresholding without prior knowledge. This method employs a traditional image processing workflow: first, preliminary reconstruction of CBCT data containing metal artifacts is performed; then, an empirical thresholding process is used to generate a metal mask in the image domain; finally, the mask is transformed into the projection domain through forward projection. The technical drawbacks of this method are mainly threefold: first, beam hardening causes progressive changes in CT values ​​at the metal boundary, making it difficult for simple thresholding to accurately locate the true boundary; second, cone-beam artifacts generate radial artifacts around the metal, interfering with the accuracy of segmentation; and third, the CT value ranges of some high-density bone tissues overlap with those of metal, increasing the risk of missegmentation. These technical drawbacks have two typical adverse consequences in practical applications: when the segmentation result is too small, residual metal information remains in the projection domain, contaminating the data distribution; when the segmentation result is too large, valuable anatomical structural information is lost. Both situations significantly reduce the quality of subsequent interpolation and restoration, thus affecting the accuracy of the final reconstructed image.

[0049] The second category is projection domain segmentation methods based on supervised deep learning with prior knowledge. This method represents the current technological frontier, and its core idea is to use deep neural networks to directly predict metal masks from CBCT data containing metal artifacts. In the data preparation stage, a professional dataset containing diverse metal implants needs to be constructed, and high-quality labels need to be obtained through expert annotation or pre- and post-operative registration; it should be able to automatically learn the complex feature representations of metals. However, this method also faces severe technical bottlenecks: its performance is highly dependent on a large amount of paired training data, while obtaining clinical data presents three major difficulties: first, data pairing is difficult, as differences in patient positioning and changes in scanning parameters make accurate registration difficult; second, tissue deformation is a problem, as metal implants may cause displacement or deformation of anatomical structures; and finally, annotation costs are high, requiring professional physicians to invest a significant amount of time in meticulous annotation. These factors collectively limit the clinical application and promotion of deep learning methods.

[0050] To address the inherent limitations of traditional projection domain metal artifact suppression methods in terms of metal segmentation accuracy, this application proposes an innovative CBCT metal artifact suppression technique based on preoperative prior knowledge. By intelligently registering preoperative metal-free images with intraoperative CBCT data and leveraging the consistency of the patient's own anatomical structure, precise segmentation of metal implants is achieved. Unlike deep learning methods that rely on large amounts of labeled data, this technique only requires routine preoperative scan data. By combining registered images with intraoperative images, it can accurately identify metal implants while effectively distinguishing easily confused high-density bone tissue.

[0051] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the specific implementation methods of this application will be described below with reference to the accompanying drawings. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings and other implementation methods can be obtained based on these drawings without creative effort.

[0052] Reference manual attached Figure 1 As shown, one embodiment of the CBCT metal artifact suppression method based on preoperative a priori guidance of this application specifically includes the following steps:

[0053] S100, using the second CBCT image data of the target tissue acquired in real time during the operation as a reference, rigid registration is performed on the first CBCT image data of the target tissue acquired before the operation; so that the first CBCT image data is deformed to obtain the third image data; the relative positions of each tissue structure in the third image data are consistent with those in the second CBCT image data.

[0054] The second CBCT image data includes all tissue structure features of the first CBCT image data, as well as the radiation signal features generated by the metal implant.

[0055] After registration, the first CBCT image data is deformed to obtain the third image data. The third image data contains the same image content as the first CBCT image data, but the relative positions of the various tissue structures are changed. The relative positions of the various tissue structures in the third image data are consistent with those in the second CBCT image data.

[0056] Specifically, this application employs a rigid registration method to achieve the registration process from a floating image to a reference image. The registration algorithm divides the image into multiple resolution levels, starting with the lowest resolution and refining layer by layer to avoid getting trapped in local minima. Optionally, the registration is implemented based on the Image Processing Toolbox.

[0057] S200, based on the third image data, digital subtraction angiography is used to coarsely segment the metal implant in the second CBCT image data; the segmentation yields the first mask data of the metal implant and background prior data.

[0058] Specifically, the relative positions of each tissue structure in the third image data (data after preoperative registration) are consistent with those in the second CBCT image data (intraoperative data). Furthermore, the third image data only includes the tissue structure of the target tissue and does not include the metal implant. The second CBCT image data, in addition to containing all the tissue structure features of the third image data, also includes the radiation signal features generated by the metal implant.

[0059] Therefore, the third image data (data after preoperative registration) and the second CBCT image data (intraoperative data) share the same origin and fully utilize the information of the same human tissue structure. They also take full advantage of the stability of anatomical structures to achieve coarse segmentation and prior template extraction of non-human metal grafts. Combined with shape and spatial priors, the segmentation results are optimized, providing prior templates and key constraints for the subsequent precise segmentation of metal grafts.

[0060] S300, perform true / false positive judgment and false positive suppression processing on the first mask data to obtain the second mask data of the metal implant; perform data repair on the background prior data using an adaptive interpolation grid to obtain the projection domain data after repair.

[0061] Specifically, due to the influence of surrounding high-density tissues such as bone and calcifications, segmented coarse metal grafts often contain human tissue structures. Directly using these false-positive areas for projection restoration can lead to incorrect interpolation calculations, thus affecting reconstruction quality. Therefore, it is crucial to perform true / false positive assessments on segmented metal grafts, eliminating false-positive portions of non-metallic grafts, and accurately segmenting the metal grafts within the human body for subsequent projection domain restoration.

[0062] In some implementations, the background prior data obtained in the previous step also needs to be repaired using an adaptive interpolation grid to recover the anatomical structure information in the background prior data that is obscured by metal artifacts.

[0063] S400, based on the projection domain data and the second mask data, the CBCT image of the target tissue without metal artifacts is reconstructed using the FDK reconstruction algorithm.

[0064] Specifically, the image with metal artifact suppression is reconstructed using the interpolated projection domain data and the second mask data of the metal implant to remove metal artifacts, employing the FDK algorithm.

[0065] Based on the above embodiments, this application discloses another embodiment of a CBCT metal artifact suppression method based on preoperative prior guidance. Before step S100, rigid registration of the first CBCT image data is performed, the method further includes step S010, which preprocesses the first CBCT image data and the second CBCT image data respectively.

[0066] Specifically, either the first CBCT image data or the second CBCT image data is used as the target projection data. This embodiment uses the preprocessing of the first CBCT image data as an example. The preprocessing process for the second CBCT image data is the same as in this embodiment and will not be repeated here. It includes the following sub-steps:

[0067] S011, Correct the target projection data (first CBCT image data) to eliminate inherent hardware errors.

[0068] In this embodiment, the target projection data (first CBCT image data) undergoes angle-by-angle flat-field correction, dark-field correction, dead-point repair, and internal rotation correction. Flat-field correction uses the detector's response image under uniform X-ray irradiation in an object-free state as a reference. Dark-field correction acquires the detector's background noise image without X-ray irradiation, subtracts the dark-field image from the target projection data, and divides the result by the difference between the flat-field and dark-field images, thereby eliminating inconsistencies in the detector's pixel response and the influence of dark current.

[0069] Optionally, image correction also includes image data cropping. To prevent data from exceeding boundaries, it is cropped to a legal range.

[0070] In this embodiment, data pruning ensures numerical validity, a reasonable threshold range is set to remove outliers, the target projection data is restricted to a physically feasible dynamic range, a percentile truncation method is used to remove outliers that are too high or too low, and linear or nonlinear scaling is performed to adapt the data distribution to the needs of subsequent processing.

[0071] S012, the corrected target projection data (first CBCT image data) is converted into linear projection data that conforms to the physical model of X-ray attenuation by a logarithmic transformation method.

[0072] In this embodiment, to ensure the data conforms to the physical process of X-ray imaging, a logarithmic transformation is performed on the corrected target projection data. To match the projection direction required by the reconstruction algorithm, the Y-axis is reversed to guarantee data quality before reconstruction. Specifically, a negative logarithmic operation is performed on the corrected target projection data to convert the detected photon intensity into an integral value of the linear attenuation coefficient along the X-ray path.

[0073] S103 is based on a filtering algorithm to adaptively suppress image noise.

[0074] In this embodiment, a composite filtering strategy is employed to effectively suppress noise interference. Optionally, a median filter algorithm and a Gaussian filter are used to preprocess the projection domain data to prevent noise from being amplified by the reconstruction algorithm. A low-pass Gaussian filter is used to suppress high-frequency random noise, while anisotropic diffusion filtering is considered to preserve edge features. Advanced algorithms such as wavelet thresholding or nonlocal mean filtering may be used for different noise characteristics.

[0075] Optionally, in other embodiments, step S104 is also included, which involves processing the linear projection data using a physics-based prior energy spectrum hardening correction method to restore the ideal attenuation characteristics of the data.

[0076] In this embodiment, energy spectrum hardening correction based on physical priors can restore the ideal attenuation characteristics of linear projection data. Specifically, by establishing a multi-material energy spectrum attenuation model, the distribution characteristics of hardening artifacts are estimated using polynomial fitting or iterative optimization methods, and the model parameters are optimized in conjunction with the physical characteristics of beam hardening artifacts. By fitting a set of empirical polynomial parameters, the projection signal affected by ray hardening is corrected back, gradually transforming the target detector signal into high-quality projection data that meets the reconstruction requirements.

[0077] The preprocessing of the first and second CBCT image data in this embodiment is a systematic signal optimization process. Through multi-stage collaborative optimization, the physical accuracy and reconstruction adaptability of the projection data are significantly improved, providing a reliable input foundation for subsequent high-quality imaging.

[0078] This application provides another embodiment of a CBCT metal artifact suppression method based on preoperative prior guidance. In step S100, based on any embodiment of the above method, the first CBCT image data of the target tissue acquired preoperatively is rigidly registered with reference to the second CBCT image data of the target tissue acquired in real-time during surgery. Specifically, this includes the following sub-steps:

[0079] S110, mutual information is used as a similarity metric for registration between the first CBCT image data and the second CBCT image data.

[0080] In this embodiment, mutual information is used as a similarity metric during CBCT image registration. The core idea is to evaluate the alignment degree of two images by utilizing the joint gray-level distribution statistical characteristics. Mutual information is based on the concept of entropy in information theory, calculating the statistical dependency between the gray-level distributions of the first and second CBCT images. The mutual information value reaches its maximum when the two images are spatially aligned. Specifically, the gray-level values ​​of the two images are first subjected to joint histogram statistics, and a joint probability distribution is obtained through normalization. Then, their respective edge probability distributions are calculated, and finally, the mutual information value is derived using the entropy formula.

[0081] S120, based on the stochastic gradient descent optimization algorithm, dynamically adjusts the image transformation parameters by iteratively calculating the gradient of the mutual information with respect to the image transformation parameters, thereby maximizing the similarity metric.

[0082] In this embodiment, the stochastic gradient descent algorithm is used for iterative search of transformation parameters during the optimization process. The gradient of the transformation parameters with respect to mutual information guides the direction of parameter updates. Stochastic gradient descent randomly selects a subset of images for gradient estimation in each iteration, improving computational efficiency. The optimizer adaptively adjusts the learning rate to balance convergence speed and stability, while incorporating a momentum term to avoid getting trapped in local optima. During parameter updates, backpropagation is used to calculate gradients, ensuring that the transformation parameters for each image in the transformation model are accurately optimized.

[0083] During registration, geometric transformations employ either rigid or affine transformation models. Specifically, image transformation parameters include translation, rotation, scaling, and shearing. In each iteration, the transformation model updates its parameters based on the gradient direction calculated by the optimizer, progressively maximizing mutual information to ensure optimal spatial alignment of the two CBCT images in terms of anatomical structures.

[0084] S130, based on the image transformation parameters, perform geometric transformation processing on the first CBCT image data; use an interpolation algorithm to estimate the grayscale value of the transformed image in real time, so that the first CBCT image data and the second CBCT image data are aligned in three-dimensional space.

[0085] In this embodiment, trilinear interpolation is used in the image grayscale interpolation stage to improve registration accuracy. When the first CBCT image data needs to undergo coordinate mapping under the transformation model, trilinear interpolation obtains continuous grayscale value estimates by weighting the calculations among the eight nearest neighbor voxels in three-dimensional space. Compared with nearest neighbor interpolation or bilinear interpolation, trilinear interpolation can better maintain the spatial continuity of the image and reduce the jagged effect introduced by interpolation, thereby improving the accuracy of registration. When calculating mutual information, the smooth interpolation result helps to avoid local oscillations during the optimization process, making the registration process more stable.

[0086] This application provides another embodiment of a CBCT metal artifact suppression method based on preoperative prior guidance. In step S200, based on the third image data, digital subtraction angiography is used to coarsely segment the metal implant in the second CBCT image data; the segmentation yields the first mask data of the metal implant and background prior data. Specifically, it includes the following sub-steps:

[0087] Based on the above embodiments, this embodiment, after registration, ensures that the second CBCT image data containing the metal implant and the third image data without the metal implant have a completely consistent spatial coordinate system.

[0088] S210, using digital subtraction angiography, the third image data is subtracted from the second CBCT image data to obtain the first mask data of the metal implant.

[0089] Specifically, by utilizing the relative positional consistency of each tissue structure in the third image data and the second CBCT image data, digital subtraction angiography is performed to subtract the grayscale value of the corresponding position in the third image data from the grayscale value of the second CBCT image data, thereby obtaining the first mask data reflecting the distribution of the metal implant.

[0090] S220, subtract the first mask data from the second CBCT image data to obtain the background prior data.

[0091] The obtained background prior information appears as blank at the position corresponding to the first mask data. In subsequent steps, the blank area needs to be repaired to obtain complete background prior data.

[0092] This embodiment effectively separates metal implants from surrounding tissues, enabling coarse segmentation of the target tissue and extraction of background prior data. The obtained background prior data exhibits good structural integrity in subsequent processing, providing a reliable data foundation for metal artifact correction and tissue segmentation.

[0093] Optionally, the digital subtraction process employs an edge-preservation-based subtraction algorithm to avoid introducing new artifacts in the metal boundary region. The obtained background prior data, based on the second CBCT image data, removes the metal implant and the artifacts caused by it.

[0094] In other embodiments, the method further includes the step of: S230, optimizing the segmentation results of the first mask data and the background prior data; specifically, the first mask data is processed by adaptive thresholding to remove noise interference and retain significant metallic signal regions. Optionally, the digital subtraction process uses double-precision floating-point operations to ensure calculation accuracy, while introducing normalization processing to eliminate possible brightness differences. For the metallic mask obtained after subtraction, small holes are filled by morphological closing operations, and isolated noise points are removed by connected component analysis to ensure the integrity of the metallic region.

[0095] This application provides another embodiment of a CBCT metal artifact suppression method based on preoperative prior guidance. Based on any embodiment of the above method, step S300 includes step S310, which involves determining true and false positives in the first mask data and performing false positive suppression processing to obtain the second mask data for the metal implant. Specifically, it includes the following sub-steps:

[0096] S310, based on the physical characteristics and morphological features of the metal implant, perform feature analysis and discrimination of true and false positives on the first mask data.

[0097] In this embodiment, the first mask data may contain true positive regions, i.e., the mask of the actual metal implant, and false positive regions, i.e., the mask caused by metal artifacts. In some implementations, a feature extraction model is constructed based on the physical characteristics of the metal object, such as density characteristics. By analyzing key indicators such as CT value distribution, edge gradient features, and three-dimensional spatial continuity, a discrimination criterion for the metal region is established.

[0098] In other implementations, an adaptive threshold segmentation combined with a region growing algorithm is used to initially screen for true positive regions based on the characteristic that metals exhibit extremely high CT values ​​in CT images. Simultaneously, morphological feature analysis is introduced to calculate geometric parameters such as volume, surface area, and sphericity of each connected region of the metal implant, and these parameters are matched and verified against the physical characteristics of the expected metal implant.

[0099] S320, the false positive regions obtained by discrimination are subjected to artifact suppression processing, and the second mask data is obtained after processing.

[0100] In one embodiment of this example, based on the false positive regions identified in step S310, false positive suppression in the projection data first involves establishing a projection domain mask based on prior location information. For a specific projection region, an iterative reconstruction algorithm is employed, and spatial weights are applied during the reconstruction process to suppress artifact effects.

[0101] In other embodiments, the method also includes combining metal projection trajectory analysis with image inpainting techniques to interpolate and compensate for the affected projection area. Optionally, it also includes introducing energy spectrum information to distinguish between real metal signals and high-density tissue artifacts through dual-energy or multi-energy decomposition.

[0102] This embodiment effectively distinguishes genuine metal implants from high-density artifacts such as bone calcification and contrast agent deposition. The final output second mask data effectively eliminates artifact interference while preserving the complete morphology of the metal implant, providing precise localization data for subsequent metal artifact correction. (See attached instruction manual.) Figure 2 As shown, Figure 2 This image compares the performance of the second mask data of the metal implant segmented using existing technologies and the preoperative prior-guided CBCT metal artifact suppression method provided in this application. The data uses the reconstruction results of the pedicle screw in the projection domain and image domain as an example. From left to right: the original acquired second CBCT image data; the effect of the metal implant obtained using existing technology one: the image domain-based threshold-free segmentation method; the effect of the metal implant obtained using existing technology two: the projection-domain segmentation-based metal artifact reduction algorithm (PDS-MAR); and a schematic diagram comparing the metal implant segmented using the preoperative prior-guided CBCT metal artifact suppression method of this application.

[0103] This application provides another embodiment of a CBCT metal artifact suppression method based on preoperative prior guidance. Based on any embodiment of the above method, step S300 includes step S320, which involves using an adaptive interpolation grid to repair the background prior data, resulting in projection domain data. Specifically, it includes the following sub-steps:

[0104] S321, the missing area of ​​the background prior data is accurately located using a metal mask.

[0105] S322 uses an adaptive interpolation method to construct an interpolation grid, which restores the missing areas while maintaining the topological structure of the projected data.

[0106] In this embodiment, a metal mask is used to locate the defective region affected by metal artifacts within the projection space. An adaptive interpolation method is then used to establish a topological connection relationship of non-uniform sampling points at the boundary of the defective region, forming an interpolation mesh with geometric adaptability. The mesh density is dynamically adjusted according to the gradient characteristics of the projection data, with denser sampling points used in the edge regions to ensure detail preservation.

[0107] During interpolation, reliable neighborhood information determined by morphological dilation operations is incorporated, while projection continuity constraints are introduced to ensure the integrity of the data structure. For complex defective regions, a multi-scale hierarchical repair strategy is implemented, first restoring the overall structure at a large scale, and then gradually refining local features.

[0108] In other implementations, the specific steps are as follows: In the mask preprocessing stage, connected component analysis is performed on the input binarized metal mask, and all independent metal regions are identified using an 8-neighborhood detection algorithm. A 3×3 circular structure is used to perform morphological dilation on the mask, with the dilation radius dynamically adjusted according to the projection resolution, which is set to 2-5 pixels. The outer boundary contours of each connected component are accurately extracted through set difference operations, i.e., calculating the difference between the dilated mask and the original mask.

[0109] This application provides another embodiment of a CBCT metal artifact suppression method based on preoperative prior guidance. In step S400, based on the projection domain data and the second mask data, a CBCT image of the target tissue without metal artifacts is reconstructed using the FDK reconstruction algorithm. Specifically, this includes the following sub-steps:

[0110] S410, the projection domain data and the second mask data are subjected to ramp filtering processing.

[0111] Specifically, projection data filtering: The repaired projection data P(u,v) is subjected to ramp filtering, and the filtered projection data is denoted as P. * (u,v).

[0112] The design of the ramp filter comprehensively considers the balance between high-frequency signal enhancement and noise suppression, optimizing the modulation transfer function by applying an appropriate window function in the frequency domain. An adaptive parameter adjustment mechanism is introduced during the filtering process, dynamically adjusting the filter intensity based on the signal-to-noise ratio of the projected data. This achieves both high-frequency information enhancement and noise suppression.

[0113] S420, calculate the back projection weights for the filtered projection domain data and the scanning geometric parameters in the second mask data; and perform a weighted integral along the X-ray path on the filtered projection domain data to obtain the reconstructed three-dimensional volume data.

[0114] Specifically, the filtered projection data enters the back projection stage, where the back projection weight of each voxel is precisely calculated based on the CBCT scan geometry parameters. These weights take into account the divergence characteristics of the X-ray beam, the nonlinear response of the detector, and the geometric calibration of the rotation trajectory.

[0115] During the back projection process, a distance-driven method is used to perform weighted integration along each X-ray path, and continuous spatial sampling is achieved through trilinear interpolation to ensure accurate accumulation of voxel values.

[0116] In other implementations, backprojection weights are calculated based on scanning geometry parameters, such as the X-ray source-detector distance and rotation radius. The filtered projection data is then weighted and integrated along the X-ray path to reconstruct the 3D volume data.

[0117] Where R is the distance from the X-ray source to the center of rotation, and u,v are the detector coordinates.

[0118] After processing with the aforementioned optimization algorithm, the final reconstructed image exhibits excellent imaging quality, possessing not only outstanding spatial resolution and tissue contrast but also effectively eliminating common image interferences such as metal artifacts and beam hardening artifacts. This achieves high-definition, low-noise, and artifact-free clinical diagnostic-grade imaging. The reconstructed three-dimensional volume data clearly displays fine anatomical structures, meeting the needs of precise diagnosis and treatment planning. While maintaining the image signal-to-noise ratio, it significantly improves detail resolution, providing clinicians with more accurate and reliable imaging evidence. (See attached instruction manual.) Figure 3 As shown. Figure 3 This image shows a comparison of the reconstructed 3D volume data obtained from existing technical solutions and the CBCT metal artifact suppression method based on preoperative prior guidance provided in this application. The data uses the reconstruction effect of pedicle screws in the image domain as an example. From left to right: the original acquired second CBCT image data; the reconstruction effect obtained using existing technology one: the image domain-based threshold-free segmentation method; the reconstruction effect obtained using existing technology two: the projection-domain segmentation-based metal artifact reduction algorithm (PDS-MAR); and a schematic diagram of the 3D volume data reconstructed using the CBCT metal artifact suppression method based on preoperative prior guidance provided in this application.

[0119] Based on the same concept, this application also discloses a CBCT metal artifact suppression system based on preoperative a priori guidance. The system is used to implement the steps described in any of the above method embodiments. Specifically, one embodiment of the CBCT metal artifact suppression system based on preoperative a priori guidance of this application includes:

[0120] The registration module is used to rigidly register the first CBCT image data of the target tissue acquired before surgery with reference to the second CBCT image data of the target tissue acquired in real time during surgery; so that the first CBCT image data is deformed to obtain third image data; wherein the relative positions of each tissue structure in the third image data are consistent with those in the second CBCT image data.

[0121] The segmentation module is used to coarsely segment the metal implant in the second CBCT image data based on the third image data using digital subtraction angiography; the segmentation obtains the first mask data of the metal implant and background prior data.

[0122] An optimization module is used to perform true / false positive judgment and false positive suppression processing on the first mask data to obtain the second mask data of the metal implant.

[0123] The repair module is used to repair the background prior data using an adaptive interpolation grid, and the repaired data is the projection domain data.

[0124] The reconstruction module is used to reconstruct a CBCT image of the target tissue without metal artifacts based on the projection domain data and the second mask data using the FDK reconstruction algorithm.

[0125] Based on the same concept, this application also discloses a CBCT device, which includes the CBCT metal artifact suppression system based on preoperative prior guidance described in the above embodiments.

[0126] The CBCT metal artifact suppression method, system and device based on preoperative prior guidance in this application have the same technical concept, and the technical details of the embodiments of the three are applicable to each other. In order to reduce repetition, they will not be repeated here.

[0127] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of program modules is merely an example. In practical applications, the above functions can be assigned to different program modules as needed, that is, the internal structure of the device can be divided into different program units or modules to complete all or part of the functions described above. The program modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one processing unit. The integrated unit can be implemented in hardware or as a software program unit. Furthermore, the specific names of the program modules are only for easy differentiation and are not intended to limit the scope of protection of this application.

[0128] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

Claims

1. A method for suppressing metal artifacts in CBCT based on preoperative prior guidance, characterized in that, Includes the following steps: Using the second CBCT image data of the target tissue acquired in real time during the operation as a reference, rigid registration is performed on the first CBCT image data of the target tissue acquired before the operation; The first CBCT image data is deformed to obtain the third image data; the relative positions of the tissue structures in the third image data are consistent with those in the second CBCT image data. Based on the third image data, digital subtraction angiography is used to coarsely segment the metal implant in the second CBCT image data. The first mask data and background prior data of the metal implant are obtained by segmentation; The first mask data is subjected to a true / false positive judgment and a false positive suppression process to obtain the second mask data of the metal implant. The background prior data is repaired using an adaptive interpolation grid, and the repaired data is the projection domain data. Based on the projection domain data and the second mask data, the CBCT image of the target tissue without metal artifacts is reconstructed using the FDK reconstruction algorithm.

2. The CBCT metal artifact suppression method based on preoperative prior guidance as described in claim 1, characterized in that: The second CBCT image data includes all tissue structure features of the first CBCT image data, as well as the radiation signal features generated by the metal implant.

3. The CBCT metal artifact suppression method based on preoperative prior guidance as described in claim 1, characterized in that, Before rigid registration of the first CBCT image data, the method further includes: preprocessing the first CBCT image data and the second CBCT image data respectively; specifically including: Correct the target projection data to eliminate inherent hardware errors; The corrected target projection data is converted into linear projection data that conforms to the physical model of X-ray attenuation by logarithmic transformation; image noise is adaptively suppressed based on filtering algorithm.

4. The CBCT metal artifact suppression method based on preoperative prior guidance as described in claim 1, characterized in that, The method described above uses the second CBCT image data of the target tissue acquired in real time during the operation as a reference to perform rigid registration on the first CBCT image data of the target tissue acquired before the operation. Specifically, it includes the following sub-steps: Mutual information is used as a similarity metric for registration between the first CBCT image data and the second CBCT image data; Based on the stochastic gradient descent optimization algorithm, the image transformation parameters are dynamically adjusted by iteratively calculating the gradient of the mutual information with respect to the image transformation parameters, thereby maximizing the similarity metric. Based on the image transformation parameters, the first CBCT image data is subjected to geometric transformation processing; an interpolation algorithm is used to estimate the grayscale value of the transformed image in real time, so that the first CBCT image data and the second CBCT image data are aligned in three-dimensional space.

5. The CBCT metal artifact suppression method based on preoperative prior guidance as described in claim 1, characterized in that, Based on the third image data, digital subtraction angiography is used to coarsely segment the metal implant in the second CBCT image data; the segmentation yields the first mask data of the metal implant and background prior data; specifically, it includes the following sub-steps: The third image data is subtracted from the second CBCT image data using digital subtraction angiography to obtain the first mask data of the metal implant. The background prior data is obtained by subtracting the first mask data from the second CBCT image data.

6. The CBCT metal artifact suppression method based on preoperative prior guidance as described in claim 1, characterized in that, The process of determining true and false positives and suppressing false positives in the first mask data to obtain the second mask data for the metal implant specifically includes the following sub-steps: Based on the physical properties and morphological characteristics of the metal implant, the first mask data is subjected to feature analysis and discrimination of true and false positives. The false positive regions identified are subjected to artifact suppression processing, and the second mask data is obtained after processing.

7. The CBCT metal artifact suppression method based on preoperative prior guidance as described in claim 1, characterized in that, The process of using an adaptive interpolation grid to repair the background prior data, resulting in projection domain data, specifically includes the following sub-steps: The missing areas of the background prior data are accurately located using a metal mask; an interpolation grid is constructed using an adaptive interpolation method to repair the missing areas while maintaining the topological structure of the projected data.

8. The CBCT metal artifact suppression method based on preoperative prior guidance as described in claim 1, characterized in that, The method of reconstructing a CBCT image of the target tissue without metal artifacts using the FDK reconstruction algorithm based on the projection domain data and the second mask data specifically includes the following sub-steps: The projection domain data and the second mask data are subjected to ramp filtering. Calculate the back projection weights for the filtered projection domain data and the scan geometry parameters in the second mask data; The filtered projection domain data and the second mask data are then weighted and integrated along the X-ray path to obtain the reconstructed three-dimensional volume data.

9. A CBCT metal artifact suppression system based on preoperative a priori guidance, characterized in that, include: The registration module is used to rigidly register the first CBCT image data of the target tissue acquired before surgery with reference to the second CBCT image data of the target tissue acquired in real time during surgery. The first CBCT image data is deformed to obtain the third image data; wherein the relative positions of the tissue structures in the third image data are consistent with those in the second CBCT image data; The segmentation module is used to coarsely segment the metal implant in the second CBCT image data based on the third image data and using digital subtraction angiography; the segmentation obtains the first mask data of the metal implant and background prior data. The optimization module is used to perform true / false positive judgment on the first mask data and perform false positive suppression processing to obtain the second mask data of the metal implant; The repair module is used to repair the background prior data using an adaptive interpolation grid, and the repaired data is the projection domain data. The reconstruction module is used to reconstruct a CBCT image of the target tissue without metal artifacts based on the projection domain data and the second mask data using the FDK reconstruction algorithm.

10. A CBCT device, characterized in that, The CBCT device includes the CBCT metal artifact suppression system based on preoperative prior guidance as described in claim 9.

Images