Coronary target vessel dynamic roadmap registration method and device, electronic equipment and medium

Abstract

The application relates to the technical field of medical image processing, and discloses a coronary target vessel dynamic road map registration method and device, electronic equipment and a medium. The method comprises the following steps: acquiring a reference contrast image sequence of a target vessel region, processing the reference contrast image sequence, and establishing a target reference mapping table; acquiring a non-contrast real-time fluoroscopy image of the target vessel region, and extracting a plurality of real-time Mark points therefrom; determining the real-time cardiac cycle phases of each real-time Mark point in different frames according to the spatiotemporal continuity constraint relationship between the real-time Mark points; determining a plurality of reference Mark points and a plurality of reference vessel masks corresponding to a plurality of real-time cardiac cycle phases based on the target reference mapping table; and based on the plurality of reference Mark points and the plurality of real-time Mark points, using a rigid registration transformation relationship to perform spatial mapping on the plurality of reference vessel masks and superimposing the plurality of reference vessel masks on the non-contrast real-time fluoroscopy image to obtain a vessel dynamic road map. The application reduces the use of contrast and effectively improves the navigation accuracy.

Description

Technical Field

[0001] This invention relates to the field of medical image processing and interventional surgical navigation technology, specifically to a method, device, electronic device, and medium for registering dynamic route maps of coronary target vessels. Background Technology

[0002] In coronary interventional imaging, it is necessary to reconstruct the vessel course and branch location information in real time from fluoroscopic images. Existing dynamic path mapping image processing technologies mainly include: Conventional dynamic vascular mapping: Preoperative / intraoperative vascular images are acquired using contrast agents, and the contrast images are superimposed on real-time fluoroscopic images using ECG gating or image registration techniques. Disadvantages: Motion disturbances from heartbeat and respiration can easily cause overlay position shifts, resulting in larger registration errors.

[0003] Real-time fluoroscopic navigation: relies entirely on X-ray fluoroscopy for image acquisition. Disadvantages: high radiation output during image acquisition, and poor identification of branch vessel images.

[0004] ECG-gated roadmaps: These rely on external ECG signals to filter images of specific cardiac cycles for overlay processing. Disadvantages: Requires additional ECG signal acquisition equipment; phase selection accuracy is easily affected by ECG signal quality.

[0005] Existing dynamic road map image processing solutions have many technical shortcomings. For example, they lack stable, high-contrast local reference points for registration calculations; simple grayscale image registration algorithms are significantly affected by respiratory movements and body position changes; most rely on external ECG hardware to obtain phase information, lacking phase determination methods based on image features themselves, resulting in phase recognition jumps and poor inter-frame consistency; large positioning errors in key areas such as branch vessel openings make it difficult to meet the image processing needs of complex scenarios; and when registration deviations exceed the allowable range, it is necessary to re-acquire angiographic images for calibration, affecting image processing efficiency. Summary of the Invention

[0006] This invention provides a method, apparatus, electronic device, and medium for registering dynamic coronary artery target vessel maps, in order to solve the technical problems of existing dynamic roadmap image processing schemes, such as lack of registration reference, reliance on external equipment for phase recognition, difficulty in adapting to non-angiography processing scenarios, and issues related to registration stability, positioning accuracy, and processing efficiency.

[0007] In a first aspect, the present invention provides a method for registering a coronary target vessel dynamic route map, the method comprising: Obtain a reference angiographic image sequence of the target vascular region; The reference angiography image sequence is processed and a target reference mapping table is established. The target reference mapping table is used to characterize the mapping relationship between cardiac cycle phase, reference vascular mask, reference Mark point and local geometric relationship of Mark point. Acquire angiography-free real-time fluoroscopic images of the target vascular region and extract multiple sets of real-time Mark points from the angiography-free real-time fluoroscopic images; Based on the spatiotemporal continuity constraints between multiple real-time Mark points, the real-time cardiac cycle phase of each real-time Mark point in different frames is determined. Based on the target reference mapping table, multiple reference Mark points and multiple reference vascular masks corresponding to multiple real-time cardiac cycle phases are determined. Based on multiple reference Mark points and multiple real-time Mark points, multiple reference vascular masks are spatially mapped and superimposed onto an angiography-free real-time fluoroscopic image using a preset rigid registration transformation formula to obtain a dynamic roadmap of the target local vascular system. The preset rigid registration transformation formula only performs two-dimensional rigid transformations of translation and rotation, and does not use three or more Mark points to construct affine transformations.

[0008] This invention acquires an angiography-free real-time fluoroscopic image of the target vascular region and extracts multiple sets of real-time Mark points, enabling roadmap generation processing independent of contrast agent input. Furthermore, by utilizing the spatiotemporal continuity constraints between multiple real-time Mark points, the real-time cardiac cycle phase of each real-time Mark point in different frames is determined, eliminating the need for external signal equipment, improving the inter-frame consistency of phase recognition results, and suppressing phase jumps caused by single-frame detection noise. Furthermore, by matching the target reference mapping table to obtain the reference Mark points and reference vascular masks corresponding to the real-time phase, data correspondence between real-time frames and the reference model is achieved. Furthermore, by using multiple reference Mark points and multiple real-time Mark points as local benchmarks for two-dimensional rigid correction, unnecessary scale distortion caused by multi-point affine transformations can be avoided, ultimately outputting a local dynamic roadmap and achieving positional alignment of the reference vascular mask on the real-time fluoroscopic image. Simultaneously, during the correction process, the preset rigid registration transformation formula only performs two-dimensional rigid transformations of translation and rotation, avoiding the use of three or more Mark points to construct affine transformations, reducing unnecessary deformation interference and ensuring the stability of local image registration in the region of interest. Therefore, by implementing this invention, registration misalignment can be effectively eliminated, and local dynamic pathway maps can be output by real-time overlay of vascular pathway maps under angiography-free conditions. Furthermore, the spatiotemporal constraints based on Mark points enhance anti-interference capabilities, thereby improving the stability of pathway map output and increasing computational efficiency.

[0009] In one optional implementation, the reference contrast image sequence is processed, and a target reference mapping table is established, including: Based on the reference contrast image sequence, multiple reference Mark points are extracted and the Mark point motion period curves are constructed; Based on the reference angiography image sequence, determine multiple reference vascular masks and the local geometric relationships between multiple reference Mark points and multiple reference vascular masks. The cardiac motion component and respiratory motion component in the Mark point cardiac cycle curve are separated, and multiple cardiac cycle phases are identified. A target reference mapping table is established based on multiple cardiac cycle phases, multiple reference vascular masks, local geometric relationships of Mark points, and multiple reference Mark points.

[0010] This invention separates the cardiac and respiratory motion components in the heart rate periodic curve of the Mark point, thus isolating the interference between these two motions. This effectively preserves the local geometric relationship between the Mark point and the blood vessel, enabling the construction of a reliable phase mapping relationship. Furthermore, by structurally binding the phase, blood vessel mask, reference point, and geometric constraints, a searchable reference model can be formed. This model supports rapid indexing and matching during real-time processing, providing a complete reference benchmark for subsequent phase matching and accurate registration. This effectively improves the stability of subsequent road map registration and the accuracy of anatomical point positioning.

[0011] In one alternative implementation, multiple reference vascular masks are determined based on a reference angiographic image sequence, including: Extract the contrast agent flow region from the reference contrast image sequence and generate multiple vessel probability maps; Adaptive threshold segmentation is performed on multiple blood vessel probability maps to generate multiple reference blood vessel masks.

[0012] This invention reduces background noise interference by quantizing the probability information of pixels belonging to vascular structures from reference angiographic images and generating multiple vascular probability maps. Furthermore, adaptive threshold segmentation of the multiple vascular probability maps generates multiple reference vascular masks, further reducing imaging noise interference and enabling better differentiation between vascular and background regions. This effectively improves the completeness and accuracy of vascular mask extraction, providing a reliable vascular template for subsequent registration.

[0013] In one optional implementation, based on a target reference mapping table, multiple reference Mark points and multiple reference vessel masks corresponding to multiple real-time cardiac cycle phases are determined, including: Determine whether a corresponding real-time cardiac cycle phase exists in the target reference mapping table; If a corresponding real-time cardiac cycle phase exists in the target reference mapping table, the real-time cardiac cycle phase is used as an index to search the target reference mapping table and determine the corresponding reference Mark point and reference vascular mask. When the real-time cardiac cycle phase is between two adjacent discrete phases in the target reference mapping table, linear interpolation is performed on the coordinates of the reference Mark points and the reference vascular mask of the two adjacent discrete phases, and a reference Mark point and a reference vascular mask matching the current real-time phase are generated.

[0014] This invention distinguishes between two scenarios: a perfectly matched phase and a phase situated between two discrete phases, by determining whether a corresponding real-time cardiac cycle phase exists in the target reference map table. This adapts to real-world scenarios where the actual frame phase may not perfectly coincide with the reference discrete phase. Furthermore, when a corresponding real-time cardiac cycle phase exists in the target reference map table, the corresponding reference Mark point and reference vessel mask can be retrieved based on the real-time cardiac cycle phase, enabling rapid retrieval of existing reference data and ensuring processing efficiency in scenarios with perfectly matched phases. Moreover, when the real-time cardiac cycle phase lies between two adjacent discrete phases in the target reference map table, a reference Mark point and reference vessel mask matching the current real-time phase are generated through linear interpolation. This compensates for the sampling interval of the discrete phases in the reference map table, preventing inter-frame jumps in the roadmap. Therefore, by implementing this invention, it is possible to effectively adapt to any real-time cardiac phase, better compensate for the sampling interval defects of discrete phases, avoid jumps and discontinuities in the roadmap, ensure the continuous and smooth output of reference Mark points and vessel masks, and improve the display stability of dynamic roadmaps.

[0015] In one optional implementation, based on multiple reference Mark points and multiple real-time Mark points, a preset rigid registration transformation formula is used to spatially map multiple reference vascular masks and overlay them onto an angiography-free real-time fluoroscopic image to obtain a local vascular dynamic path map, including: By using multiple reference Mark points and multiple real-time Mark points, the preset rigid registration transformation formula is solved to obtain the rigid registration transformation matrix; Using a rigid registration transformation matrix, multiple reference vascular masks are spatially mapped and superimposed onto an angiography-free real-time fluoroscopic image to obtain a dynamic roadmap of the target local vascular system.

[0016] This invention uses multiple reference mark points and multiple real-time mark points to solve a preset rigid registration transformation formula to obtain a rigid registration transformation matrix, achieving precise spatial alignment of the vascular mask. Furthermore, by combining this rigid registration transformation matrix, the reference vascular information and the real-time perspective image can be accurately fused, improving the spatial matching accuracy of the dynamic roadmap.

[0017] In one optional implementation, a rigid registration transformation matrix is ​​used to spatially map multiple reference vascular masks and overlay them onto a non-contrast real-time fluoroscopic image to obtain a dynamic path map of the target local blood vessels, including: Using a rigid registration transformation matrix, multiple reference vascular masks are translated and rotated in two dimensions to obtain multiple mapped vascular masks. Using the local geometric relationship of Mark points in the target reference mapping table as a constraint verification condition, the relative spatial position deviation between multiple mapped blood vessel masks and multiple real-time Mark points is verified. The validated mapped vascular mask is superimposed onto the angiography-free real-time fluoroscopic image to distinguish the main vessels and branch vessels using differentiated display styles, thus obtaining an initial local vascular dynamic roadmap. In the initial local vascular dynamic roadmap, mark the branch opening location, branch angle, stent expected deployment area or balloon kissing area, and generate the target local vascular dynamic roadmap.

[0018] This invention utilizes a rigid registration transformation matrix to perform two-dimensional spatial translation and rotation mapping on multiple reference vascular masks. By performing only translation and rotation transformations without introducing scale deformation, a vascular mask adapted to the real-time image coordinate system can be obtained, effectively ensuring the accurate and reliable position of the mapped vascular mask. Furthermore, the relative spatial position deviation between multiple mapped vascular masks and multiple real-time Mark points is verified using the local geometric relationship of Mark points in the target reference mapping table as a constraint verification condition, reducing erroneous registration output caused by false detections of Mark points in a single frame. Furthermore, by distinguishing the display of main vessels and branch vessels, different vascular structures can be easily distinguished intuitively. Furthermore, by supplementing the output with key structural position and angle annotation information, the output content of the roadmap is enriched.

[0019] In one optional implementation, the coronary target vessel dynamic route map registration method further includes: Based on the target local vascular dynamic path map, the registration residual value is calculated; If the registration residual value is greater than the preset residual threshold, return to the step of obtaining the reference angiographic image sequence of the target vascular region, and iterate repeatedly until the registration residual value is less than or equal to the preset residual threshold to obtain the target local vascular dynamic roadmap; or, if the registration residual value is greater than the preset residual threshold, reduce the display reliability of the target local vascular dynamic roadmap.

[0020] When the registration deviation exceeds the allowable range, this invention can avoid unreliable road maps with excessive output deviation by updating the reference data source for recalculation or lowering the confidence level of the output results, thereby improving the reliability of the evaluation registration results.

[0021] In a second aspect, the present invention provides a coronary target vessel dynamic route map registration device, the device comprising: The acquisition module is used to acquire a reference angiographic image sequence of the target vascular region; A module is established to process the reference angiographic image sequence and establish a target reference mapping table, wherein the target reference mapping table is used to characterize the mapping relationship between cardiac cycle phase, reference vascular mask, reference Mark point and local geometric relationship of Mark point; The extraction module is used to acquire an angiography-free real-time fluoroscopic image of the target vascular region and extract multiple sets of real-time Mark points from the angiography-free real-time fluoroscopic image. The cardiac cycle phase determination module is used to determine the real-time cardiac cycle phase of each real-time Mark point in different frames based on the spatiotemporal continuity constraint relationship between the multiple real-time Mark points. The reference vessel mask determination module is used to determine multiple reference Mark points and multiple reference vessel masks corresponding to the multiple real-time cardiac cycle phases based on the target reference mapping table. The mapping module is used to spatially map the multiple reference vascular masks based on the multiple reference Mark points and the multiple real-time Mark points using a preset rigid registration transformation formula and superimpose them onto the angiography-free real-time fluoroscopic image to obtain a target local vascular dynamic roadmap. The preset rigid registration transformation formula only performs two-dimensional rigid transformations of translation and rotation, and does not use three or more Mark points to construct affine transformations.

[0022] Thirdly, the present invention provides an electronic device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the coronary target vessel dynamic path map registration method of the first aspect or any corresponding embodiment described above.

[0023] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the coronary target vessel dynamic path registration method of the first aspect or any corresponding embodiment thereof. Attached Figure Description

[0024] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0025] Figure 1 This is a schematic diagram of an application scenario according to an embodiment of the present invention; Figure 2This is a schematic diagram of the first process of the coronary target vessel dynamic route map registration method according to an embodiment of the present invention; Figure 3 This is a specific processing procedure of the local dynamic path map registration method for coronary target vessels according to an embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the pre-setting of balloon Mark points and the establishment of the initial angiographic reference model according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the second process of the coronary target vessel dynamic route map registration method according to an embodiment of the present invention; Figure 6 This is a schematic diagram of the detection and location modeling of imaging mark points for interventional devices; Figure 7 This is a schematic diagram of the third process of the coronary target vessel dynamic route map registration method according to an embodiment of the present invention; Figure 8 This is a schematic diagram of the fourth process of the coronary target vessel dynamic route map registration method according to an embodiment of the present invention; Figure 9 This is a diagram illustrating the effect of overlaying dynamic road maps according to the present invention. Figure 10 This is a schematic diagram of precise branch entry location based on the registration of the main branch Mark point; Figure 11 This is a schematic diagram of rigid registration comparison based on two Mark points; Figure 12 This is a structural block diagram of the coronary target vessel dynamic route map registration device according to an embodiment of the present invention; Figure 13 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0028] As an optional application scenario of this invention, such as Figure 1As shown, the coronary target vessel dynamic route map registration system may include at least one terminal device and at least one server. Figure 1 The system is illustrated in the example, which includes a computer 101, a mobile terminal 102, and a server 103, and the terminal devices such as the computer 101 and the mobile terminal 102 are connected to the server 103 through a network 110.

[0029] Specifically, the terminal device can be a smartphone, tablet, laptop, PDA, desktop computer, game console, smart TV, smart wearable device, in-vehicle terminal, VR (Virtual Reality) device, AR (Augmented Reality) device, etc. Server 103 can be a standalone physical server, a server cluster, a distributed system, or a cloud server providing cloud services. Network 110 can be a wired or wireless network, examples of which include, but are not limited to, the Internet, corporate intranet, local area network, wide area network, mobile communication network, and combinations thereof.

[0030] Existing dynamic road map image processing technologies include: (1) Philips Dynamic Coronary Roadmap (DCR): Currently a representative dynamic roadmap image processing solution. Its working principle is as follows: angiography sequence is obtained through a single coronary angiography; a roadmap is extracted using digital subtraction angiography; ECG gating is used to filter roadmap frames corresponding to cardiac phases; motion compensation is completed by combining image registration algorithms; and the roadmap is superimposed and output to a real-time fluoroscopic image. However, it has limitations: DCR uses ECG gating combined with image grayscale registration to achieve motion compensation, which has multiple technical constraints: it relies on an external ECG device to obtain phase information, and there is a time delay deviation between the ECG signal and the real cardiac motion; grayscale registration lacks stable reference features and is sensitive to respiratory motion; no mapping relationship between phase-mask-reference point is established, and the phase consistency between frames is difficult to guarantee, which easily leads to phase jumps; the registration accuracy of branch vessel regions is insufficient, which is difficult to meet the image processing needs of complex scenarios.

[0031] (2) Dynamic road map based on catheter tip tracking: Existing research proposes a dynamic road map processing scheme based on catheter tip tracking (Medical Image Analysis, 2020). This scheme identifies the position of the catheter tip through deep learning and uses the displacement of the catheter tip as the calculation parameter for road map transformation.

[0032] Limitations: The calculation is performed using only single-point features, which limits the degree of freedom for correction and makes rotation correction impossible. Introducing multi-point affine transformation can also lead to unnecessary scale distortion. The position of the catheter tip relative to the blood vessel changes during image acquisition, which can easily lead to error accumulation. The lack of Mark point spatiotemporal continuity constraints results in weak anti-interference ability and insufficient accuracy in branch vessel localization.

[0033] (3) Interventional device imaging mark point detection technology.

[0034] Existing technologies include interventional device imaging mark point detection and enhancement display related processing solutions. These solutions mainly complete the automatic detection of interventional device imaging mark points, image enhancement, and 3D reconstruction.

[0035] Limitations: This type of solution only applies Mark points to image localization and enhanced rendering of stents, and does not use Mark points as the registration reference for dynamic roadmaps; it also does not construct a mapping relationship between cardiac cycle phase, vascular mask, and Mark points, and cannot rely on the spatiotemporal continuity constraints of Mark points to determine cardiac phase, and cannot support roadmap generation and processing under angiography-free conditions.

[0036] In summary, existing dynamic roadmap image processing solutions lack a high-contrast, stable local registration benchmark, a phase determination method based on the spatiotemporal continuity constraints of the image's own Mark points, and a phase-vessel mask-Mark point mapping model. They rely heavily on external ECG equipment, which can easily lead to phase jumps. Some registration methods can introduce unnecessary scale distortions, resulting in poor anti-interference capabilities. They are not well adapted to non-angiography processing scenarios, have insufficient localization accuracy in key vascular regions, and require re-acquiring angiographic images for calibration when registration fails. Overall, registration stability and processing efficiency are limited.

[0037] According to an embodiment of the present invention, a method for dynamic registration of coronary target vessels is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0038] This embodiment provides a method for dynamic registration of coronary target vessel maps, which can be used on the aforementioned mobile terminals, such as mobile phones and tablets. Figure 2 This is a flowchart of the coronary target vessel dynamic route map registration method according to an embodiment of the present invention, such as... Figure 2 As shown, the process includes the following steps: Step S201: Obtain a reference angiographic image sequence of the target vascular region.

[0039] In this embodiment of the invention, the target vascular region refers to the local vascular area that needs to be closely observed and managed during coronary intervention surgery. This includes the target lesion segment, branch ostia, the expected stent deployment area, the balloon kissing area, the main branch-branch junction area, or the local vascular segment requiring precise navigation by the surgeon.

[0040] In this embodiment of the invention, the reference angiographic image sequence is a set of multiple consecutive images acquired through angiographic imaging of the target vascular region, which serves as the baseline reference data source for subsequent algorithm processing. This image sequence may include vascular filling images and Mark point image information, and the position sequence of the Mark points in consecutive frames is recorded synchronously.

[0041] Furthermore, the images are arranged sequentially in chronological order, covering at least one complete cardiac cycle.

[0042] In this embodiment of the invention, prior to the acquisition of angiographic images, pre-image preparation is completed by pre-positioning the interventional device with contrast-enhancing markers near the target vessel area of ​​interest, ensuring that the contrast-enhancing markers on the interventional device are within the imaging visibility range. The device itself does not alter the vessel morphology or obstruct image interpretation.

[0043] In one alternative implementation, pre-operative preparation may involve planning a double-stent procedure when the patient has severe stenosis at the bifurcation of the left anterior descending artery (LAD) and the diagonal branch.

[0044] First, balloon pre-positioning should be performed. Before the first angiography, an undilated balloon catheter with a contrast-enhancing mark should be inserted into the main branch of the LAD along the guidewire, with the balloon mark located in the main branch and close to the area of ​​interest where the LAD bifurcates with the diagonal branch. Without affecting subsequent operations and blood flow safety, the mark should be as close as possible to the branch opening, the target lesion segment, or the area where the stent is expected to be deployed.

[0045] The placement of the first stent should be based on local registration at the Mark point. After reaching the preset accuracy requirements under experimental calibration or phantom verification conditions, it is used to assist in the placement of the main stent.

[0046] Specifically, the insertion of the second stent should be based on the precise positioning of the branch opening using the branch mask after local registration with the main branch Mark point, which will assist in confirming the branch entrance and inserting the second stent.

[0047] Specifically, the method for implementing balloon kissing should detect the Mark points at both ends of the main branch balloon and the branch balloon respectively, calculate the axis of each balloon, the center of the balloon, the position of the effective expansion segment, and the relative distance and angle between the two balloons; when the positional deviation of the relative branch openings of the two balloons, the overlap length of the effective expansion segment, the angle of the axis and the spacing between Mark points are within the preset range, and the preset number of frames are maintained continuously and the detection confidence meets the requirements, an alignment prompt is given.

[0048] In the balloon kissing maneuver, the imaging markers of each of the two balloons or interventional instruments are detected to determine the axis, center position, and effective area of ​​action of each balloon or instrument. The relative distance, relative angle, and position relative to the target branch opening are displayed in real time. This provides a visual navigation reference, assisting the operator in determining whether the two balloons have reached the appropriate position for the kissing maneuver.

[0049] The preset alignment conditions include at least one of the following: the overlap length of the two effective action areas is within a preset range; the distance between the two centers or corresponding Mark points is within a preset range; the included angle between the two axes is within a preset angle range; the positioning deviation of each effective action area relative to the main branch or branch opening is less than a preset threshold; and the above state is maintained for a preset number of consecutive frames and the detection confidence is higher than a preset threshold; furthermore, when the preset alignment conditions are met, the corresponding electronic device issues an alignment confirmation signal.

[0050] In one alternative implementation, pre-implantation preparation can also be performed when a patient has a long, diffuse lesion requiring the tandem placement of multiple stents. This implementation can involve first obtaining a complete vascular mask of the lesion segment through full-cycle angiography. For the placement of the first stent, registration and navigation are performed using a first set of marker points. For subsequent stent placements, after the previous stent is released, registration is continued using the newly exposed marker points of the interventional device.

[0051] In cases involving long lesions or multiple stents, a new set of two imaging Mark points can be used as a new local rigid reference at different stages of the procedure to continue the two-point rigid registration. Two Mark points are still used in the same registration step, without using three-point affine or multi-point segmented affine.

[0052] Furthermore, angiographic image acquisition is performed to obtain a reference angiographic image sequence covering at least one complete cardiac cycle. Moreover, within this image sequence, each frame may contain imaging information of the target vessel after contrast agent filling, imaging information of contrast-enhancing markers on the interventional device, etc.

[0053] Furthermore, after acquisition, the reference contrast image can be corrected for bad spots, denoised, and normalized for brightness. If necessary, coarse motion alignment can be performed using a conduit or Mark points.

[0054] Step S202: Process the reference contrast image sequence and establish a target reference mapping table.

[0055] In one alternative implementation, a target reference mapping table is used to characterize the mapping relationship between cardiac cycle phase, reference vascular mask, reference Mark point, and local geometry of Mark point.

[0056] In this embodiment of the invention, the target reference mapping table can be a structured reference dataset indexed by cardiac phase, used to store the reference morphology of the target blood vessel, the position of the reference marker, and the inherent local spatial association information between the marker and the blood vessel at different cardiac times.

[0057] Furthermore, this target reference mapping table also supports the rapid retrieval and acquisition of reference information on blood vessels and markers corresponding to the cardiac state based on the real-time cardiac phase.

[0058] In this embodiment of the invention, a reference angiography image sequence can be used as input, and various reference features related to the surgical target blood vessel can be extracted from the original angiography image through image processing methods. Then, the feature information can be linked according to the correspondence of cardiac phases to construct a structured reference dataset that can be called and queried in real time.

[0059] In one optional implementation, the target reference mapping table can adopt a dictionary index array data structure, where the target identifier is used as the index key of the dictionary, and the values ​​of the dictionary are arrays. Furthermore, the array stores all the reference data corresponding to the target, thereby achieving a one-to-many mapping between the target and multiple reference data.

[0060] Specifically, in this embodiment, the dictionary index array uses the normalized cardiac phase value as the index key for each record, and each index key stores a set of vascular contour points, a set of marker coordinates, and a marker-vascular relative spatial feature descriptor.

[0061] For example, the reference angiography image sequence is first processed by parsing the original angiography image frame by frame and extracting the vascular contour point set and Mark point coordinate set corresponding to each frame. At the same time, the normalized cardiac phase corresponding to each frame is calculated. Then, based on the position of the blood vessel and the Mark point in the image, the relative spatial feature descriptor of the marker point-blood vessel is calculated. This descriptor is used to record the spatial correlation information such as the distance and orientation of the Mark point relative to the target blood vessel.

[0062] Furthermore, a target reference mapping table in the form of a dictionary index array is established. Each normalized cardiac phase is used as the index key of the dictionary. The set of vascular contour points, the coordinate group of marker points, and the marker point-vascular relative spatial feature descriptor extracted under that phase are used as the storage value corresponding to the key. These are stored one by one into the dictionary index array to complete the construction of the target reference mapping table.

[0063] Step S203: Obtain an angiography-free real-time fluoroscopic image of the target vascular region and extract multiple sets of real-time Mark points from the angiography-free real-time fluoroscopic image.

[0064] In this embodiment of the invention, the real-time Mark point is the interventional device imaging mark detected in the non-contrast fluoroscopic image, and it belongs to the same device as the reference Mark point in the reference contrast image.

[0065] In one optional embodiment, a sequence of angiography-free real-time fluoroscopic images of the target vascular region is acquired. In this image sequence, no contrast agent is input during the imaging process, and the images retain imaging information of contrast markers carried by the interventional device pre-positioned near the target vascular region of interest.

[0066] Furthermore, for each frame of angiography-free real-time fluoroscopic image, the Mark point search area is first defined by combining the region of interest, the Mark point output position of the previous frame, and the instrument axis information. Image denoising and local contrast enhancement are then performed on the search area. Next, candidate connected components for Mark points can be output using one of the following methods: local grayscale extremum detection, speckle detection, template matching, morphological enhancement, or deep learning detection.

[0067] Furthermore, the gray-scale weighted centroid, area, major and minor axes, roundness, and local contrast are calculated for each candidate connected component, and initial screening is completed by combining the preset size of the Mark points and development features. The screened candidate points are then combined in pairs. Further, the combination matching score can be calculated based on the nominal physical spacing of the Mark points, the allowable range of the projection spacing, collinearity of the instrument axes, the proximal-to-farward order, and the inter-frame motion consistency. Candidate point combinations with scores exceeding the confidence threshold are selected as the real-time Mark point detection results for this frame, and finally, multiple sets of real-time Mark points, corresponding proximal-to-farward labels, and detection confidence scores are output.

[0068] Step S204: Determine the real-time cardiac cycle phase of each real-time Mark point in different frames based on the spatiotemporal continuity constraint relationship between multiple real-time Mark points.

[0069] In this embodiment of the invention, the spatiotemporal continuity constraint relationship refers to a constraint condition that simultaneously includes spatial and temporal dimensions. Specifically, it may include: in time, the motion velocity and displacement of Mark points in adjacent frames will not undergo abrupt changes; in space, the relative positions, spacing, and orientation of the connecting lines between Mark points will not exhibit abnormal jumps that violate the physical structure of the device.

[0070] For example, in terms of time, the distance the Mark points move between consecutive frames cannot exceed the reasonable speed of the instrument under X-ray imaging; in terms of space, the projection distance between two Mark points on the same balloon should be within a reasonable range of fluctuation and should not suddenly increase or decrease.

[0071] In this embodiment of the invention, the real-time cardiac cycle phase is used to characterize the relative position of the current image moment within the complete cardiac cycle, and is used to characterize the stage of the heartbeat.

[0072] For example, φ_cur=0 indicates a correspondence with the preset cycle start point of the reference template, while φ_cur close to 1 indicates that the next cycle is about to begin. This definition is only used to index the reference phase and does not require a one-to-one correspondence with the P wave, QRS wave, or T wave of the ECG.

[0073] Wherein, φ_cur is the normalized cardiac cycle position of the current real-time fluoroscopic frame, which is a dimensionless decimal used for table lookup indexing.

[0074] In one optional implementation, when determining the real-time cardiac cycle phase of each real-time Mark point in different frames, the phase is first predicted based on the phase of the previous frame and the estimated cycle: φ_pred = (φ_prev + Δt / T_cycle)mod1. Then, a real-time feature z_cur is constructed from the current and several previous Mark points. This real-time feature may include the midpoint projection after detrended modeling, the direction of the connecting line, the feature change rate, and the distance between two points. Here, φ_pred represents the predicted cardiac phase of the current frame; φ_prev represents the normalized cardiac cycle phase obtained in the previous frame; Δt represents the image time interval between the current frame and the previous frame; T_cycle represents the currently estimated cardiac cycle duration; and mod1 represents performing a modulo-1 operation on the calculation result to constrain the phase result within the normalized phase value range [0, 1).

[0075] Furthermore, the constructed real-time feature z_cur is compared one by one with the reference feature z_ref corresponding to each phase in the reference period template, and the feature matching cost corresponding to each candidate phase φi is calculated; at the same time, a phase continuity constraint is introduced to obtain the continuity constraint cost.

[0076] The phase continuity constraint is used to prevent phase jumps between adjacent frames. That is, the phase obtained in the current frame should be continuous with the output phase and the phase prediction value φ_pred of the previous frame, effectively suppressing erroneous cross-cycle phase jumps. Furthermore, the feature matching cost and the phase continuity constraint cost are fused to obtain the comprehensive matching cost for each candidate phase φi.

[0077] Furthermore, all candidate phases φi are traversed, and the candidate phase with the smallest comprehensive matching cost is selected as the final real-time cardiac cycle phase φ_cur of the current frame.

[0078] Where φi is the reference cardiac cycle phase obtained by pre-discrete sampling in the reference angiography image sequence, and its value also belongs to [0,1). Where i is the index, representing a discrete phase sample pre-stored in the reference mapping table.

[0079] The comprehensive matching cost can include feature amplitude difference, velocity direction difference, connection direction difference, two-point spacing error, and phase distance with φ_pred. Specifically, if the difference between the optimal cost and the second-best cost is insufficient, the detection confidence is too low, or an unreasonable phase jump occurs, the phase prediction value will be temporarily used, or the dynamic roadmap display will be reduced / paused.

[0080] In one optional implementation, the comprehensive matching cost can comprehensively measure the motion amplitude, motion velocity direction, the angle of the line connecting the Mark points, the distance between the two points, and the deviation from the estimated phase. If abnormal situations occur, such as poor matching result discrimination, unreliable Mark point detection, or unreasonable phase jumps, the phase selected by the matching is not used. The estimated phase calculated over time can be temporarily used to continue the calculation, or the reliability of the road map can be reduced or the road map display can be paused directly to prevent unreliable road maps from misleading the operation.

[0081] In one alternative implementation, the comprehensive matching cost is calculated as shown in the following formula:

[0082] in: Candidate phase The corresponding overall matching cost; The feature matching cost is composed of the real-time feature z_cur and the reference feature z_ref. The characteristic differences between them were calculated; As a cost of phase continuity constraints, characterizing candidate phases The degree of phase jump relative to the predicted phase φ_pred The greater the deviation from φ_pred, the higher the constraint cost. , These are the feature matching term weights and the phase continuity constraint term weights, respectively, both of which are preset coefficients greater than 0.

[0083] in, Control the proportion of image feature similarity in matching; The penalty intensity for phase continuity smoothing can be controlled; when the image signal-to-noise ratio is poor or the Mark point detection noise is high, it can be appropriately increased. It is used to strengthen phase continuity constraints and effectively suppress mismatches caused by phase jumps.

[0084] Step S205: Based on the target reference mapping table, determine multiple reference Mark points and multiple reference vascular masks corresponding to multiple real-time cardiac cycle phases.

[0085] In this embodiment of the invention, the reference vascular mask is vascular morphology data extracted from a reference angiographic image sequence and used as a baseline to mark the location region of the target vascular vessel (main vessel, branch vessel). For example, the reference vascular mask can be in the form of a binary label map, a vascular contour point set, etc., and can be used to distinguish which pixels belong to the vascular vessel and which belong to the background.

[0086] In one alternative implementation, a pre-built target reference mapping table can be used as the base data source, and the calculated real-time cardiac cycle phase of each frame can be used as the query index to retrieve reference Mark point data and reference vascular mask data that match the phase from the mapping table.

[0087] Step S206: Based on multiple reference Mark points and multiple real-time Mark points, the multiple reference vascular masks are spatially mapped and superimposed onto the angiography-free real-time fluoroscopic image using a preset rigid registration transformation formula to obtain a dynamic roadmap of the target local blood vessels.

[0088] In this embodiment of the invention, the preset rigid registration transformation formula only performs spatial translation and rotation transformation rules, and does not use three or more Mark points to construct affine transformations, in order to align the reference coordinate system with the real-time perspective image coordinate system.

[0089] Furthermore, spatial mapping represents the process of converting each pixel / contour point on the reference vascular mask stored in the coordinate system of the reference angiographic image to the coordinate system of the real-time fluoroscopic image without angiography, based on the transformation parameters obtained by solving the rigid registration transformation relationship; overlay is the process of overlaying the mapped vascular mask onto the original fluoroscopic image without angiography, wherein the fluoroscopic image retains the real images of instruments, bones, etc., and the mask provides virtual vascular information.

[0090] In one specific implementation, a set of reference marker coordinates and a vascular reference template have been obtained by looking up a table; simultaneously, real-time X-ray images detect the real-time markers of the same group of instruments in the current image. Further, by inputting these two sets of markers into a preset rigid registration transformation formula, it can be calculated that the vascular template needs to be translated 8 pixels to the right and rotated 3° counterclockwise. Then, this translation and rotation operation is performed on the entire vascular reference template to complete the spatial mapping.

[0091] Furthermore, by overlaying the transformed vascular template layer onto the original angiography-free X-ray fluoroscopy image, a corresponding dynamic vascular path map of the target locality can be generated. The fluoroscopy image retains the true image of the skeleton and interventional instruments, while the overlaid vascular template is used to virtually draw the contours of the blood vessels.

[0092] In an optional instance, such as Figure 3 As shown, a specific processing procedure for a local dynamic roadmap registration method for a coronary target vessel is provided: using the Mark points near the region of interest as the local registration reference, the following steps are performed in sequence: balloon pre-positioning, extraction of vessel and Mark point data by the first angiography, construction of a reference mapping table, Mark point detection, phase determination, registration to generate a dynamic roadmap on the non-angiography fluoroscopic image, and registration reliability monitoring, and re-angiography calibration if necessary.

[0093] In an optional instance, such as Figure 4 As shown, a schematic diagram of balloon Mark point pre-positioning and initial angiography reference model establishment is provided: After placing the balloon with Mark points close to the area of ​​interest of the blood vessel in accordance with the device placement principle, the balloon is pre-positioned first and then the reference angiography image sequence is acquired. After processing the reference angiography sequence, the blood vessel mask, the set of reference Mark points, the motion period curve of Mark points and the phase-mask-Mark point reference mapping table are obtained, thus completing the construction of the reference model for subsequent real-time registration.

[0094] The coronary target vessel dynamic path map registration method provided in this embodiment constructs a target reference mapping table with phase index by acquiring a single segment of reference angiography sequence covering the complete cardiac cycle. Furthermore, during the operation, the double Mark points of the instrument are detected under non-angiography fluoroscopy conditions, and the real-time cardiac phase is solved by combining the spatiotemporal continuity constraints of the Mark points. The reference data of the corresponding phase is retrieved to complete the local rigid registration and generate a dynamic vessel path map.

[0095] This method eliminates the need for continuous contrast agent infusion, thus effectively reducing the risk of kidney damage in patients. Furthermore, it does not rely on ECG signals, performing phase calculations and local registration solely based on instrument imaging markers. Moreover, through multi-level detection, weighted cost constraints, and anomaly tolerance mechanisms, it suppresses mismatches and phase jumps caused by noise and occlusion, effectively eliminating roadmap misalignment and drift issues caused by cardiac pulsation. While meeting the intraoperative real-time computation frame rate requirements, it effectively improves the accuracy and reliability of navigation in the coronary target area, especially in bifurcation lesion regions.

[0096] This embodiment provides a method for dynamic registration of coronary target vessel maps, which can be used on the aforementioned mobile terminals, such as mobile phones and tablets. Figure 5 This is a second flowchart of the coronary target vessel dynamic route map registration method according to an embodiment of the present invention, as follows: Figure 5 As shown, the process includes the following steps: Step S501: Obtain a reference angiographic image sequence of the target vascular region.

[0097] Please see details Figure 2 Step S201 of the illustrated embodiment will not be described again here.

[0098] Step S502: Process the reference contrast image sequence and establish a target reference mapping table.

[0099] Specifically, step S502 includes: Step S5021: Based on the reference contrast image sequence, extract multiple reference Mark points and construct the Mark point motion period curve.

[0100] In this embodiment of the invention, the cardiac cycle curve of the Mark point is a curve representing the periodic reciprocating motion of the Mark point caused by the heartbeat. Furthermore, this cardiac cycle curve records the positional change of the Mark point over time; the horizontal axis can be the image acquisition time, and the vertical axis can be the spatial position of the Mark point. Further, this spatial position can be coordinates, displacement, midpoint offset, etc.

[0101] Furthermore, the periodic curve of the Mark point can be obtained from the coordinates of a single Mark point, a weighted combination of the coordinates of multiple Mark points, the trajectory of the midpoint between two Mark points, the change in the direction of the line connecting the Mark points, or a combination of the above features.

[0102] In this embodiment of the invention, the coordinates p1(ti) and p2(ti) of two Mark points are obtained in each frame of the reference imaging sequence. Further, the midpoint c(ti) = [p1(ti) + p2(ti)] / 2 and the direction of the line connecting the two points are calculated. θ(ti) and the distance between the two points d(ti), then interpolate or predict the trajectory of the missed frame and denoise and smooth the coordinate sequence.

[0103] Furthermore, using the projection of the main motion direction of the trajectory of the midpoint between the two points as the main motion feature, a one-dimensional or multi-dimensional motion sequence m(t) can be constructed by fusing the changes in the direction of the connecting line. This motion sequence m(t) is used to characterize the overall translational motion of the two Mark points under the influence of cardiac motion, as well as the dynamic changes in the rotation of the connecting line over time.

[0104] In one alternative implementation, the midpoint c(t) of the two Mark points and the angle θ(t) of the connecting line are calculated in each frame. Furthermore, principal component analysis is performed on the midpoint trajectory of a reference sequence to obtain the principal direction e of local cardiac motion.

[0105] Furthermore, let m1(t) = e^T[c(t) - c_mean] represent the displacement along the principal direction, and m2(t) = unwrap[ θ (t)- θ [_mean] represents the local rotation change. Construct m(t) = a·m1_norm(t) + b·m2_norm(t), where a and b are preset or determined weights through training.

[0106] In this case, when only the phase translation feature is needed, b=0 can be set, and only the midpoint projection curve is used; when the midpoint displacement amplitude is small and the direction of the connecting line changes significantly, the weight of the direction feature b can be increased. The resulting m(t) exhibits a repeating peak-valley pattern in continuous cardiac cycles, which can be used as a template for determining the cycle boundary and the current phase.

[0107] In one optional implementation, two Mark points are detected and tracked in each reference angiography frame to obtain the Mark points and detection confidence of the reference angiography frames. Then, the Mark point motion period curve can be constructed based on the midpoint trajectory, the connection direction, and the inter-frame motion. Furthermore, each frame can be assigned to a corresponding phase φi, and the coordinates of the concurrent Mask and Mark points, along with the branch features, are written into a reference mapping table.

[0108] In a preferred embodiment, the Mark points in the reference contrast image sequence used for processing should have high contrast and be clearly visible under X-ray fluoroscopy; at the same time, the image quality, pixel size and calibration conditions should also meet the requirements.

[0109] For example, at least the following conditions should be met: the local contrast or signal-to-noise ratio of the Mark point is higher than the preset detection threshold; the Mark point occupies a preset pixel range in the image where the centroid can be stably calculated and is not severely occluded or truncated; the DSA pixel spacing and magnification can be obtained or calibrated; and the image spacing between two Mark points and their nominal physical spacing are within the allowable error range.

[0110] Furthermore, when the above conditions are met, the centroid positioning error of the Mark point can be verified by phantom or playback data to be no more than a preset pixel error; preferably no more than 1 pixel.

[0111] Furthermore, under these conditions, Mark point detection is performed, and after extracting multiple reference Mark points, the corresponding Mark point motion cycle curve can be constructed.

[0112] In one optional implementation, the extraction of multiple reference Mark points can be based on local gray-level extremum detection, candidate region screening, and dual Mark point spatial relationship verification, outputting a Mark point coordinate set P_mark={p1, p2}, corresponding near / far labels, and detection confidence. Specifically, during detection, a search region is first set according to the region of interest, the predicted position of the previous frame, or the instrument axis, and the search region is denoised and its local contrast enhanced. Then, candidate connected components are obtained using local gray-level extremum detection, LoG / DoG speckle response, template matching, morphological enhancement, or deep learning detection.

[0113] Furthermore, for each candidate region, the gray-scale weighted centroid, area, major and minor axes, roundness, and local contrast are calculated, and initial screening is performed according to the expected size, development intensity, and shape of the Mark points. Subsequently, the candidate points are combined in pairs, and the combination score is calculated based on the nominal distance between the two Mark points, the allowable projection distance range, collinearity with the instrument axis, the proximal / far-end order, and the consistency with the position of the previous frame. The group with the highest score that exceeds the confidence threshold is selected as the detection result.

[0114] In one alternative implementation, the motion trajectory of the Mark points is tracked in consecutive frames to establish inter-frame correspondences between the Mark points, while maintaining consistency in identity between near-end and far-end Mark points. Specifically, based on the position and velocity of the previous frame or several previous frames, a uniform velocity model, Kalman filtering, or optical flow can be used to predict the search threshold for the current frame.

[0115] Furthermore, within this search threshold, a matching cost is constructed based on positional distance, direction of movement, changes in the distance between two points, changes in the direction of the connecting line, and detection confidence, to establish a one-to-one correspondence between the current candidate point and the existing trajectory. Furthermore, the proximal / distal identity can be determined based on the order of their projection along the instrument axis.

[0116] Furthermore, when a Mark point is briefly occluded by a device or skeleton, the predicted trajectory is preserved and its possible location is estimated using another Mark point and the nominal spacing; when more than a preset number of frames are lost consecutively or the geometric constraints are not met, the expanded area is re-detected and the roadmap display is reduced or paused.

[0117] Furthermore, this implementation method can be used to obtain the motion trajectory of continuous Mark points based on a reference angiographic image sequence to construct a cardiac cycle curve, and it can also be used for real-time fluoroscopic images without angiography, providing a spatiotemporal continuity constraint relationship for determining the phase of the real-time cardiac cycle.

[0118] In an optional instance, such as Figure 6 The diagram shows the detection and location modeling of Mark points for interventional devices. First, Mark points are identified by grayscale extreme values, spots, or templates. Then, spatial relationship verification is completed using the distance between two points and the direction of the line. Phase-Mark point mapping is obtained through continuous frame tracking modeling. A set of Mark points is output, and a rigid reference is constructed with two Mark points. Only two-point rigid registration with translation and rotation is used.

[0119] Step S5022: Based on the reference angiography image sequence, determine multiple reference vascular masks and the local geometric relationship between multiple reference Mark points and the multiple reference vascular masks.

[0120] In this embodiment of the invention, by taking a reference angiography image sequence as input, on the one hand, the reference vascular mask corresponding to each cardiac phase is obtained, and the mask is subjected to skeletonization and topological analysis to extract vascular structural features such as centerline, branch openings, and branch angles; on the other hand, the reference Mark point of the corresponding phase is obtained, and the inherent local geometric constraint relationship of the reference Mark point relative to the vascular structure under the same phase is determined.

[0121] In one optional implementation, a coronary reference angiography sequence covering the entire cardiac cycle is input, wherein the sequence contains 25 angiography images, each corresponding to a cardiac phase. For each image, a vessel probability map is first generated, and then adaptive threshold binarization, connected component filtering, and morphological artifact removal are performed to obtain 25 binary reference vessel masks corresponding to each phase.

[0122] Furthermore, the generated reference vessel mask is processed into a skeleton, the vessel centerline is extracted, and a vessel topology map is constructed based on the obtained vessel centerline. Next, bifurcation nodes on the skeleton are identified from this vessel topology map, and these identified bifurcation nodes are used as candidate branch openings. Then, by combining vessel diameter, centerline continuity, connectivity with the catheter inlet direction, and the location of the region of interest, the main branch vessel and branch vessels are distinguished from the candidate branch openings.

[0123] Furthermore, for the determined bifurcation nodes, the tangent vectors of the center lines of each segment at the bifurcation node are calculated, the branch angles are solved based on the tangent vectors, and the branch opening coordinates, branch direction, and boundary of the region of interest are recorded. Then, the different nodes and related node information and the boundary of the region of interest are marked in the reference vascular mask to generate multiple final reference vascular masks.

[0124] Furthermore, the coordinate set P_mark_ref of reference Mark points and the target vessel mask are obtained synchronously in the first angiographic reference frame, and the relative spatial relationship between the Mark points and the reference vessel mask near the region of interest is recorded. This relative spatial relationship is not absolutely constant at all times, but is established separately for each reference cardiac phase.

[0125] Specifically, when the balloon or instrument is near the same phase and does not advance, retract, twist, or slip relative to the target vessel, the relationship between the Mark point and the local reference vessel mask is approximated as a local rigid relationship. When the position of the instrument relative to the vessel changes, the imaging gantry angle or magnification changes, or the Mark point spacing / local image consistency exceeds the preset error, the original reference relationship is determined to be invalid, and the system prompts to re-establish the reference relationship or re-calibrate the angiography.

[0126] Specifically, step S5022 above includes: Step a1: Extract the contrast agent flow region from the reference contrast image sequence and generate multiple vessel probability maps.

[0127] In this embodiment of the invention, the blood vessel probability map is used to characterize the probability that each pixel in the image belongs to a blood vessel region. Specifically, the larger the pixel value, the more likely that the location is a blood vessel.

[0128] In one optional implementation, a frame or time median image where the contrast agent has not yet fully entered the target blood vessel is selected as background B. The difference D_i(x)=|I_i(x)-B(x)| is then performed on each frame. The difference results are then used to enhance the blood vessel structure. Weights w_i are set according to the filling degree of the contrast agent, local contrast, or frame quality. Weighted maximum, weighted average, or time accumulation is then performed to obtain a blood vessel probability map.

[0129] For enhancing vascular structures, multi-scale vascular enhancement, directional filtering, morphological top-hat / bottom-hat processing, convolutional neural network segmentation, or combinations thereof can be used. Specifically, for sequences with obvious cardiac motion, coarse inter-frame registration can be performed first using Mark points or image features, followed by temporal accumulation to reduce ghosting at vascular edges.

[0130] Step a2: Perform adaptive threshold segmentation on multiple blood vessel probability maps and generate multiple reference blood vessel masks.

[0131] In this embodiment of the invention, an adaptive threshold can be determined based on the local background grayscale, the probability map histogram, or the pixel classification probability obtained through training.

[0132] Furthermore, based on an adaptive threshold, the vascular probability map is binarized, and then a reference vascular mask is obtained by connecting component filtering, hole filling, morphological opening and closing operations, and removal of small artifacts. The continuity of the vascular system is maintained by centerline constraints or boundary smoothing.

[0133] Step S5023: Separate the cardiac motion component and respiratory motion component in the heart rate cycle curve of Mark and identify multiple cardiac cycle phases.

[0134] In this embodiment of the invention, the cardiac motion component is the high-frequency periodic motion of the Mark point caused by cardiac contraction and relaxation, which is an effective signal for identifying the phase of the cardiac cycle. The respiratory motion component is the low-frequency interference signal caused by the patient's breathing, resulting in the slow drift of the entire heart. This respiratory motion component is superimposed on the movement trajectory of the Mark point and needs to be separated and removed.

[0135] In this embodiment of the invention, the temporal motion signal of the heartbeat cycle curve of Mark is decomposed, and the slow drift motion caused by breathing and the fast periodic motion caused by heartbeat are separated. Only the motion component corresponding to the heartbeat is retained, and on this basis, each heartbeat cycle is identified and divided, and the normalized heartbeat cycle phase corresponding to each frame is calculated.

[0136] In one alternative implementation, the missed detection points are first interpolated, and high-frequency noise is removed by moving average, Savitzky-Golay filtering, or low-pass filtering; then, slow respiratory / positional drift and faster heart rate cycle components are separated by sliding baseline, trend term fitting, low-frequency filtering, or bandpass filtering.

[0137] Furthermore, autocorrelation function processing or spectral analysis is performed on the heartbeat component to obtain candidate periods T_cycle, and stable repeating peaks, valleys, or feature zero-crossing points are searched within a reasonable period range. Further, a period is defined between adjacent feature points of the same type, and the phase of time t within the k-th period is defined as φ(t) = (t - t_k) / (t_{k+1} - t_k), φ ∈ [0, 1); alternatively, the range can be multiplied by 2π to convert to radian phase, in which case the range corresponds to [0, 2π).

[0138] Furthermore, each reference frame can be assigned a value according to this phase, and the features of the Mask and Mark points of the same period can be saved.

[0139] Step S5024: Establish a target reference mapping table based on multiple cardiac cycle phases, multiple reference vascular masks, local geometric relationships of Mark points, and multiple reference Mark points.

[0140] In this embodiment of the invention, specifically, a cardiac cycle is normalized to [0, 1) or [0, 2π), and discrete phases φi are set according to the reference sequence frame rate. Further, for each phase φi, a data item R_i={M(φi), p1_ref(φi), p2_ref(φi), c_ref(φi), θ_ref(φi), d_ref(φi), B(φi)} is stored, where R_i is a single data record in the target reference mapping table, M(φi) is the reference vascular mask for phase φi, and p1_ref(φi) is the pixel coordinate of the first Mark point of the reference frame. p2_ref(φi) is the pixel coordinate of the second Mark point in the reference frame. c_ref(φi) is the midpoint of the reference Mark point, θ_ref(φi) is the angle of the line connecting the two reference points, d_ref(φi) is the Euclidean distance between the two reference Mark points, and B(φi) represents local vascular features such as branch openings and centerline direction.

[0141] When multiple cardiac cycles are collected, the coordinates of Mark points, Mask contours, or distance fields of the same phase can be robustly averaged to generate a cycle template after removing missed detections and abnormal frames. During runtime, the corresponding data item can be queried using the current phase index, or interpolation can be performed between adjacent phase data items.

[0142] In one optional implementation, the obtained reference angiography sequence I={I1, I2, ..., In}, along with optional pixel spacing, imaging geometry, frame rate, physical spacing of Mark points, and region of interest information, are taken as input. Through the processing of the blood vessel mask, branch features, and Mark point reference model construction algorithm, the following can be output: reference blood vessel mask set M_ref={M(φi)}, Mark point reference coordinate set P_mark_ref={P_mark_ref(φi)}, local blood vessel features such as branch opening position and branch angle, and phase-mask-Mark point reference mapping table.

[0143] Where I1, I2, ..., In are single-frame angiographic images in the sequence; subscripts 1 to n represent the images arranged in chronological order. M_ref is the set of reference vascular masks; P_mark_ref is the set of reference mark point coordinates.

[0144] Step S503: Obtain an angiography-free real-time fluoroscopic image of the target vascular region and extract multiple sets of real-time Mark points from the angiography-free real-time fluoroscopic image.

[0145] Please see details Figure 2 Step S203 of the illustrated embodiment will not be described again here.

[0146] Step S504: Determine the real-time cardiac cycle phase of each real-time Mark point in different frames based on the spatiotemporal continuity constraint relationship between multiple real-time Mark points.

[0147] Please see details Figure 2 Step S204 of the illustrated embodiment will not be described again here.

[0148] Step S505: Based on the target reference mapping table, determine multiple reference Mark points and multiple reference vascular masks corresponding to multiple real-time cardiac cycle phases.

[0149] Please see details Figure 2 Step S205 of the illustrated embodiment will not be described again here.

[0150] Step S506. Based on multiple reference Mark points and multiple real-time Mark points, using a preset rigid registration transformation formula, spatially map multiple reference vascular masks and overlay them onto an angiography-free real-time fluoroscopic image to obtain a dynamic roadmap of the target local blood vessels.

[0151] Please see details Figure 2 Step S206 of the illustrated embodiment will not be described again here.

[0152] The coronary artery target vessel dynamic route map registration method provided in this embodiment requires only one reference angiography and eliminates the need for continuous contrast agent application during the procedure, effectively reducing the risk of complications for patients. Furthermore, it eliminates the need for ECG acquisition hardware, completing phase resolution solely based on the multidimensional motion characteristics of dual Mark points. Further, through Mark point multidimensional motion modeling, fine segmentation of bifurcation vessels, phase-separated local geometric constraints, separation of heartbeat and respiratory signals, and storage of a phase index mapping table, it effectively suppresses errors caused by occlusion, noise, respiratory drift, and cardiac pulsation. This overcomes the shortcomings of traditional static route maps, which cannot adapt to cardiac pulsation and suffer from poor accuracy, effectively improving the accuracy, stability, and operational safety of navigation.

[0153] This embodiment provides a method for dynamic registration of coronary target vessel maps, which can be used on the aforementioned mobile terminals, such as mobile phones and tablets. Figure 7 This is a third flowchart of the coronary target vessel dynamic route map registration method according to an embodiment of the present invention, such as... Figure 7 As shown, the process includes the following steps: Step S701: Obtain a reference angiographic image sequence of the target vascular region.

[0154] Please see details Figure 2Step S201 of the illustrated embodiment will not be described again here.

[0155] Step S702: Process the reference contrast image sequence and establish a target reference mapping table.

[0156] Please see details Figure 2 Step S202 of the illustrated embodiment will not be described again here.

[0157] Step S703: Obtain an angiography-free real-time fluoroscopic image of the target vascular region and extract multiple sets of real-time Mark points from the angiography-free real-time fluoroscopic image.

[0158] Please see details Figure 2 Step S203 of the illustrated embodiment will not be described again here.

[0159] Step S704: Determine the real-time cardiac cycle phase of each real-time Mark point in different frames based on the spatiotemporal continuity constraint relationship between multiple real-time Mark points.

[0160] Please see details Figure 2 Step S204 of the illustrated embodiment will not be described again here.

[0161] Step S705: Based on the target reference mapping table, determine multiple reference Mark points and multiple reference vascular masks corresponding to multiple real-time cardiac cycle phases.

[0162] Specifically, step S705 above includes: Step S7051: Determine whether there is a corresponding real-time cardiac cycle phase in the target reference mapping table.

[0163] In this embodiment of the invention, the currently calculated real-time cardiac cycle phase is compared with each discrete phase pre-stored in the target reference mapping table to determine whether there is a phase entry in the table that matches the real-time cardiac cycle phase.

[0164] Step S7052: When a corresponding real-time cardiac cycle phase exists in the target reference mapping table, the real-time cardiac cycle phase is used as an index to search the target reference mapping table and determine the corresponding reference Mark point and reference vascular mask.

[0165] In this embodiment of the invention, if a phase that matches the real-time cardiac cycle phase exists in the target reference mapping table, the real-time cardiac cycle phase is used as a search keyword to search the target reference mapping table and read the coordinates of the reference Mark point and the reference vascular mask data corresponding to the phase entry.

[0166] In one optional implementation, the discrete phases pre-stored in the target reference mapping table can be 0, 0.2, 0.4, 0.6, or 0.8. Further, assuming the currently calculated real-time cardiac cycle phase is 0.4, and the table happens to contain an index for phase 0.4, then 0.4 is used as the search keyword to retrieve the mapping table, directly extracting the reference Mark point coordinates p1_ref and p2_ref under that entry, along with the corresponding reference vascular mask data.

[0167] Step S7053: When the real-time cardiac cycle phase is between two adjacent discrete phases in the target reference mapping table, linear interpolation is performed on the coordinates of the reference Mark points and the reference vascular mask of the two adjacent discrete phases respectively, and a reference Mark point and a reference vascular mask matching the current real-time phase are generated.

[0168] In one alternative implementation, if the current cycle position does not perfectly match the reference phase, interpolation compensation is performed on the neighboring phase Mark points and Mask.

[0169] For example, assuming φ_cur is located between adjacent reference phases φi and φi+1, the interpolation coefficient α = (φ_cur - φi) / (φi+1 - φi). Furthermore, the positions of the two Mark points are linearly interpolated according to p_k(φ_cur) = (1 - α)p_k(φi) + αp_k(φi+1), while the connecting angle is either linearly interpolated after angle expansion or circularly interpolated to avoid jumps when crossing ±π.

[0170] Here, φi+1 is the next discrete reference phase, which can be the next neighboring phase index stored in the mapping table that is larger than φ_cur. p_k(φi) is the two-dimensional pixel coordinate of the k-th reference Mark point under phase φi; p_k(φ_cur) is the Mark point coordinate corresponding to the current phase obtained after interpolation.

[0171] Furthermore, the vessel mask is preferably interpolated using contour, centerline, or signed distance field. This can be achieved by calculating the signed distance fields D_i and D_i+1 of adjacent masks separately, obtaining D_cur = (1-α)D_i + αD_i+1, and generating the interpolation mask using the zero level set of D_cur. Alternatively, adjacent masks can be registered to the interpolation mark point coordinate system before weighted fusion and morphological correction. A circular interpolation method is used between the end-phase and the beginning-phase of the cycle.

[0172] Where D_i is the signed distance field of the vascular mask corresponding to the reference phase φi; D_i+1 is the signed distance field of the vascular mask corresponding to the next adjacent reference phase φi+1; and D_cur is the signed distance field corresponding to the current real-time phase φ_cur obtained after interpolation.

[0173] Step S706: Based on multiple reference Mark points and multiple real-time Mark points, the multiple reference vascular masks are spatially mapped and superimposed onto the angiography-free real-time fluoroscopic image using a preset rigid registration transformation formula to obtain a dynamic roadmap of the target local blood vessels.

[0174] Please see details Figure 2 Step S206 of the illustrated embodiment will not be described again here.

[0175] The coronary target vessel dynamic route map registration method provided in this embodiment can effectively adapt to the continuously changing cardiac state by performing a lookup table for cardiac phase or interpolation processing of adjacent phases. It can complete the dynamic registration of coronary vessels under real-time fluoroscopic images without angiography, effectively eliminate the jump tomography phenomenon caused by discrete phase sampling, and achieve smooth output of vessel route maps, ensuring registration accuracy and image continuity, and improving the stability and clinical usability of interventional surgery navigation.

[0176] This embodiment provides a method for dynamic registration of coronary target vessel maps, which can be used on the aforementioned mobile terminals, such as mobile phones and tablets. Figure 8 This is a fourth flowchart of the coronary target vessel dynamic route map registration method according to an embodiment of the present invention, such as... Figure 8 As shown, the process includes the following steps: Step S801: Obtain a reference angiographic image sequence of the target vascular region.

[0177] Please see details Figure 2 Step S201 of the illustrated embodiment will not be described again here.

[0178] Step S802: Process the reference contrast image sequence and establish a target reference mapping table.

[0179] Please see details Figure 2 Step S202 of the illustrated embodiment will not be described again here.

[0180] Step S803: Obtain an angiography-free real-time fluoroscopic image of the target vascular region and extract multiple sets of real-time Mark points from the angiography-free real-time fluoroscopic image.

[0181] Please see details Figure 2 Step S203 of the illustrated embodiment will not be described again here.

[0182] Step S804: Determine the real-time cardiac cycle phase of each real-time Mark point in different frames based on the spatiotemporal continuity constraint relationship between multiple real-time Mark points.

[0183] Please see details Figure 2Step S204 of the illustrated embodiment will not be described again here.

[0184] Step S805: Based on the target reference mapping table, determine multiple reference Mark points and multiple reference vascular masks corresponding to multiple real-time cardiac cycle phases.

[0185] Please see details Figure 2 Step S205 of the illustrated embodiment will not be described again here.

[0186] Step S806: Based on multiple reference Mark points and multiple real-time Mark points, a preset rigid registration transformation formula is used to spatially map multiple reference vascular masks and superimpose them onto an angiography-free real-time fluoroscopic image to obtain a dynamic roadmap of the target local blood vessels.

[0187] Specifically, step S806 above includes: Step S8061: Using multiple reference Mark points and multiple real-time Mark points, solve the preset rigid registration transformation formula to obtain the rigid registration transformation matrix.

[0188] In this embodiment of the invention, the rigid registration transformation matrix is ​​the result matrix obtained by operating on a preset rigid registration transformation formula. This matrix includes rotation and translation parameters and can map and transform all the blood vessel mask and bifurcation key points on the reference side to the pixel coordinate system of the current real-time image.

[0189] In an alternative implementation, the translation and rotation can be calculated from the midpoints of the two mark points and the difference in parallel directions, thus obtaining the rigid registration transformation T, as shown in the following formula: T(x) = R(Δθ)x + t.

[0190] Here, let p1 and p2 be the two Mark points in the reference phase, and q1 and q2 be the corresponding Mark points in the current frame; the reference midpoint is c_ref=(p1+p2) / 2, the current midpoint is c_cur=(q1+q2) / 2; the reference connection angle is θ_ref=atan2(p2_y-p1_y, p2_x-p1_x), the current connection angle is θ_cur=atan2(q2_y-q1_y, q2_x-q1_x), and the rotation angle is Δθ=θ_cur-θ_ref.

[0191] Furthermore, a two-dimensional rotation matrix R = [[cosΔθ, -sinΔθ], [sinΔθ, cosΔθ]] is constructed from Δθ, and the translation vector t = c_cur - R·c_ref. Here, for any point x on the reference mask, its current frame position is x_cur = R·x + t. Accordingly, this transformation only includes translation and rotation, without introducing scale changes.

[0192] Where p1_x, p1_y, p2_x, and p2_y are the x and y pixel coordinate components of reference Mark points p1 and p2, respectively; q1_x, q1_y, q2_x, and q2_y are the x-direction pixel components and y-direction pixel components of q1 and q2, respectively; and x_cur is the position of reference Mask point x after it is mapped to the real-time image coordinate system.

[0193] In one optional implementation, r_d=|‖q2-q1‖-‖p2-p1‖| can be used as the residual between the two points, and the reprojection residual between the transformed reference Mark point and the real-time Mark point can be calculated; when the residual exceeds a preset threshold, the transformation is not used or the display reliability is reduced.

[0194] In an optional implementation, based on the above closed-form solution, short-time smoothing can be performed using the detection confidence of two Mark points as weights, and the rigid registration can be optimized. Furthermore, reliability verification can be performed by the consistency of the two-point spacing, the rate of change of the rotation angle, and the reprojection residual. Thus, reliability verification can effectively avoid road map jumps caused by false detections in a single frame.

[0195] In step S8062, a rigid registration transformation matrix is ​​used to spatially map multiple reference vascular masks and superimpose them onto a non-contrast real-time fluoroscopic image to obtain a dynamic roadmap of the target local blood vessels.

[0196] Specifically, step S8062 includes: Step b1: Using a rigid registration transformation matrix, perform two-dimensional spatial translation and rotation mapping on multiple reference vascular masks to obtain multiple mapped vascular masks.

[0197] In this embodiment of the invention, the obtained rigid registration transformation matrix is ​​used to perform translation and rotation coordinate transformations on the reference blood vessel mask obtained from the target reference mapping table. This can transform the blood vessel mask in the reference coordinate system to the real-time perspective image coordinate system, thereby obtaining multiple mapped blood vessel masks.

[0198] Furthermore, the pixel coordinates ((120,200)) are substituted into the rigid registration transformation matrix operation. First, the coordinate rotation is completed, and then the translation is performed. After the operation, the new pixel coordinates (132,191) are obtained.

[0199] In this process, all the pixels of the blood vessels in the reference blood vessel mask are rotated and translated according to the same set of matrices. After all the transformations, the mapped blood vessel mask is obtained, and the blood vessel contours of the mask are aligned with the spatial position of the real-time perspective image.

[0200] Step b2: Using the local geometric relationship of Mark points in the target reference mapping table as a constraint verification condition, the relative spatial position deviation between multiple mapped blood vessel masks and multiple real-time Mark points is verified.

[0201] In this embodiment of the invention, the local geometric relationship of the Mark points stored in the target reference mapping table is used as a verification standard to detect the actual relative spatial position between the mapped blood vessel mask and the real-time Mark points, and the position deviation is calculated by comparing it with the inherent geometric relationship of the reference, thereby determining whether the rigid registration result is reliable.

[0202] In one optional implementation, the bifurcation opening is recorded in the reference mapping table as being 30 pixels to the right and 10 pixels down from the midpoint of the Mark point. After the rigid matrix T transformation, the bifurcation opening of the mapped blood vessel mask is offset relative to the real-time Mark midpoint by 29 pixels to the right and 11 pixels down. If this deviation is small, the verification passes. If the actual calculation shows that the bifurcation opening is 80 pixels to the right of the real-time Mark midpoint, this relative deviation is too large, indicating that the registration has failed and the verification fails.

[0203] Step b3: The verified mapped vascular mask is superimposed onto the angiography-free real-time perspective image to distinguish the main vessels and branch vessels using a differentiated display style, thus obtaining an initial local vascular dynamic roadmap.

[0204] In this embodiment of the invention, for the qualified mapped vascular mask, based on the pre-stored vascular topology information, the main blood vessels and branch blood vessels are superimposed and drawn on the real-time perspective image without angiography according to different display styles to generate an initial local vascular dynamic roadmap.

[0205] In one optional implementation, by setting differentiated display rules, the main blood vessel is outlined with a thick red line, 3 pixels wide, while branch blood vessels are outlined with thin yellow lines, 1 pixel wide. Furthermore, the red outline of the main blood vessel and the yellow outline of the branch blood vessels are drawn and overlaid on the X-ray grayscale image. At this point, the outlines of the main and branch blood vessels are clearly visible in the image.

[0206] Step b4: Mark the branch opening location, branch angle, stent expected release area or balloon kissing area in the initial local vascular dynamic roadmap, and generate the target local vascular dynamic roadmap.

[0207] In this embodiment of the invention, based on the initial local vascular dynamic roadmap with the main and branch vessel outlines superimposed, key markers such as branch opening positions, branch angle values, stent expected release areas, and balloon kissing areas are further drawn. After all annotations are completed, the target local vascular dynamic roadmap is obtained.

[0208] In one embodiment, such as Figure 9The image shown is a dynamic overlay display of the road map, used to demonstrate the visualization results of overlaying vascular masks, Mark points, and branch opening markers in the non-angiography stage.

[0209] In one optional implementation, in the reference imaging mask, the center lines of the main branch and the branch are skeletonized, and the bifurcation node connecting the two is determined as the branch opening point b_ref; a center line segment of a preset length is selected on both sides of the bifurcation node and a straight line or curve is fitted to obtain the main branch tangent vector v_main_ref and the branch tangent vector v_branch_ref, and the reference branch angle is calculated accordingly.

[0210] Furthermore, in each real-time frame, the current rigid transformation T maps b_ref, the main branch centerline, the branch centerline, and their tangent vectors to the current perspective coordinate system, thus displaying the branch opening markers, branch direction, and included angle. The included angle can be calculated using the inverse cosine of the two tangent vectors or the directed included angle; where, when the registration confidence is insufficient, the markers are switched to a low confidence style, frozen at the previous confidence position, or temporarily not displayed.

[0211] The branch entry location is derived from the mapping of the vessel mask containing the branch in the first angiography after local registration with the main branch mark point, and no mark point needs to be set in the branch vessel.

[0212] In one alternative implementation, the Mark point detection error, Mask overlay error, and road map confidence threshold can be set based on the DSA system's pixel size, field of view, magnification, geometric calibration results, Mark point imaging quality, frame rate, and algorithm processing capabilities. Furthermore, these parameters can be validated through phantom experiments, in vitro experiments, clinical playback data, or prospective clinical data.

[0213] In one embodiment, such as Figure 10 The diagram illustrates precise branch inlet localization based on main branch Mark point registration. Mark points are placed only on the main branch vessel. By recording the local geometric relationships between the main and branch vessel masks and the branch inlet relative to the main branch Mark points, and by tracking the main branch Mark points in real time to complete local registration, the branch inlet can be accurately located and the branch angle obtained. Furthermore, no Mark points need to be placed inside the branch vessel, effectively improving the accuracy of navigation in the bifurcation area.

[0214] The coronary artery target vessel dynamic roadmap registration method provided in this embodiment constructs a target reference mapping table based on the phase, vessel mask, and local geometric relationships of Mark points from angiographic images. It further combines this with Mark points obtained from real-time angiographic-free image detection to solve the rigid registration transformation matrix, sequentially completing vessel mask mapping, geometric relationship verification, rendering of differentiated contours, and marking of key regions. Furthermore, by adapting to continuous cardiac motion changes through table lookup or phase interpolation, dynamic registration and navigation of coronary vessels can be achieved without continuous contrast agent injection, effectively suppressing roadmap jumps and improving the accuracy, reliability, and clinical applicability of localization in scenarios such as bifurcation lesions.

[0215] In an optional implementation, the method further includes: calculating a registration residual value based on the target local vascular dynamic roadmap; when the registration residual value is greater than a preset residual threshold, returning to the step of obtaining a reference angiographic image sequence of the target vascular region, iterating repeatedly until the registration residual value is less than or equal to the preset residual threshold, thereby obtaining the target local vascular dynamic roadmap; or, when the registration residual value is greater than the preset residual threshold, reducing the display reliability of the target local vascular dynamic roadmap.

[0216] In one alternative embodiment, the preset residual threshold can be calibrated based on the imaging pixels. For example, it can be set to 0.5 mm.

[0217] Specifically, after the registration is completed, the calculated registration residual value (1.2mm) is compared with the preset threshold (0.5mm). If the registration residual value is greater than the preset threshold, it indicates that the relative geometric deviation between the mapped vascular mask, branch opening marker and real-time Mark point after the rigid transformation is too large, and the registration is unreliable.

[0218] Furthermore, the process returns to the step of acquiring the reference angiographic image sequence of the target vascular region, re-acquires the reference angiographic sequence, extracts the reference Mark points again, constructs the cardiac cycle curve, establishes the target reference mapping table, and then performs the entire process of phase solving, table lookup / interpolation, rigid registration, geometric verification, and generating the roadmap. Iterative calculations are performed until the registration residual decreases to 0.5 mm or less, and a reliable target local vascular dynamic roadmap is output.

[0219] In one optional implementation, the registration residual calculation result is 1.1mm, exceeding the preset threshold of 0.5mm. In this case, re-angiography acquisition is not triggered. Instead of removing the roadmap outline, the dynamic roadmap of the target local blood vessel is switched to a semi-transparent display, and the branch openings, angles, and stent release area markers are changed to a gray low-confidence style to indicate to the physician that the current navigation result is not reliable and is for reference only.

[0220] In an alternative implementation, two imaging Mark points on the same balloon or the same interventional device can be used as a set of local rigid registration references; wherein, the two Mark points can be used to constrain the local translation and rotation of the target blood vessel, and the detection and verification can be performed by using the known Mark point spacing.

[0221] In one instance, such as Figure 11 As shown, a schematic diagram of rigid registration comparison based on two Mark points is provided. Furthermore, this... Figure 11 The paper compares global feature registration, single-point tracking of the catheter tip, and the rigid registration scheme of two Mark points in the fourteen rational methods of this invention. Figure 11 It is known that global registration is prone to local target area shift, and single-point tracking of the catheter cannot correct local rotation. However, this invention sets two Mark points near the region of interest, and relies on the two points to achieve local rigid registration including translation and rotation, without using affine transformation.

[0222] In one alternative embodiment, the components and specifications used in this invention may be as shown in Table 1.

[0223] Table 1

[0224] The X-ray imaging system uses a DSA device with a preferred real-time processing frame rate of no less than 15fps for acquiring contrast and fluoroscopic images. The image processing workstation relies on GPU acceleration and has no less than 16GB of memory, and undertakes all algorithm calculation tasks. The display device uses a high-definition medical monitor with a resolution of no less than 1920×1080 to output intraoperative navigation visualization. The software module is deployed on the image processing workstation, integrating image acquisition, mask extraction, mark point detection and tracking, registration, and road map display functions.

[0225] This embodiment also provides a coronary target vessel dynamic roadmap registration device, which is used to implement the above embodiments and preferred embodiments, and will not be repeated as described thereon. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0226] This embodiment provides a coronary target vessel dynamic path mapping registration device, such as... Figure 12 As shown, it includes: The acquisition module 1201 is used to acquire a reference angiographic image sequence of the target vascular region, wherein the reference angiographic image sequence includes target vascular filling information and reference Mark point information.

[0227] The module 1202 is used to process the reference angiographic image sequence and establish a target reference mapping table, wherein the target reference mapping table is used to characterize the mapping relationship between cardiac cycle phase, reference vascular mask, reference Mark point and local geometric relationship of Mark point.

[0228] The extraction module 1203 is used to acquire an angiography-free real-time fluoroscopic image of the target vascular region and extract multiple sets of real-time Mark points from the angiography-free real-time fluoroscopic image.

[0229] The cardiac cycle phase determination module 1204 is used to determine the real-time cardiac cycle phase of each real-time Mark point in different frames based on the spatiotemporal continuity constraint relationship between multiple real-time Mark points.

[0230] The reference vessel mask determination module 1205 is used to determine multiple reference Mark points and multiple reference vessel masks corresponding to multiple real-time cardiac cycle phases based on a target reference mapping table.

[0231] The mapping module 1206 is used to spatially map multiple reference vascular masks based on multiple reference Mark points and multiple real-time Mark points using a preset rigid registration transformation formula and superimpose them onto a non-contrast real-time fluoroscopic image to obtain a dynamic roadmap of the target local blood vessels. The preset rigid registration transformation formula only performs two-dimensional rigid transformations of translation and rotation, and does not use three or more Mark points to construct affine transformations.

[0232] In some alternative implementations, the establishment module 1202 includes: The extraction unit is used to extract multiple reference Mark points and construct the Mark point motion period curve based on the reference contrast image sequence.

[0233] The determination unit is used to determine multiple reference vascular masks and the local geometric relationships between multiple reference Mark points and multiple reference vascular masks based on a reference angiographic image sequence.

[0234] In some alternative implementations, the determining unit includes: The extraction subunit is used to extract the contrast agent flow region in the reference contrast image sequence and generate multiple vessel probability maps.

[0235] The segmentation subunit is used to perform adaptive threshold segmentation on multiple blood vessel probability maps and generate multiple reference blood vessel masks.

[0236] The identification unit is used to separate the heartbeat and respiratory components in the heart rate cycle curve of Mark points and to identify multiple heart rate cycle phases.

[0237] A unit is established to create a target reference mapping table based on multiple cardiac cycle phases, multiple reference vascular masks, local geometric relationships of Mark points, and multiple reference Mark points.

[0238] In some alternative implementations, the reference vascular mask determination module 1205 includes: The judgment unit is used to determine whether a corresponding real-time cardiac cycle phase exists in the target reference mapping table.

[0239] The retrieval unit is used to search the target reference map table and determine the corresponding reference mark point and reference vascular mask when the corresponding real-time cardiac cycle phase exists in the target reference map table, using the real-time cardiac cycle phase as the index.

[0240] The interpolation unit is used to linearly interpolate the coordinates of the reference Mark points and the reference vascular mask of the two adjacent discrete phases when the real-time cardiac cycle phase is between two adjacent discrete phases in the target reference mapping table, and generate reference Mark points and reference vascular masks that match the current real-time phase.

[0241] In some alternative implementations, the mapping module 1206 includes: The solving unit is used to solve the preset rigid registration transformation formula using multiple reference Mark points and multiple real-time Mark points to obtain the rigid registration transformation matrix.

[0242] The mapping unit is used to spatially map multiple reference vascular masks using a rigid registration transformation matrix and superimpose them onto a non-contrast real-time fluoroscopic image to obtain a dynamic roadmap of the target local blood vessels.

[0243] Specifically, the mapping unit includes: The mapping subunit is used to perform two-dimensional spatial translation and rotation mapping on multiple reference vascular masks using a rigid registration transformation matrix, so as to obtain multiple mapped vascular masks.

[0244] The verification subunit is used to verify the relative spatial position deviation between multiple mapped blood vessel masks and multiple real-time Mark points, using the local geometric relationship of Mark points in the target reference mapping table as the constraint verification condition.

[0245] The overlay subunit is used to overlay the verified mapped vascular mask with differentiated display styles to distinguish between main vessels and branch vessels onto the angiography-free real-time perspective image to obtain an initial local vascular dynamic roadmap.

[0246] The generation sub-unit is used to mark the branch opening location, branch angle, stent expected release area or balloon kissing area in the initial local vascular dynamic roadmap, and generate the target local vascular dynamic roadmap.

[0247] The coronary target vessel dynamic route map registration device provided in this embodiment of the invention can execute the coronary target vessel dynamic route map registration method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method. Further functional descriptions of the above modules and units are the same as in the corresponding embodiments described above, and will not be repeated here.

[0248] Figure 13 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0249] The following is a detailed reference. Figure 13 This diagram illustrates a structural schematic suitable for implementing an electronic device according to embodiments of the present invention. The electronic device may include a processor (e.g., a central processing unit, a graphics processing unit, etc.) 1301, which can perform various appropriate actions and processes according to a program stored in ROM 1302 or a program loaded from memory 1308 into RAM 1303. RAM 1303 also stores various programs and data required for the operation of the electronic device. The processor 1301, ROM 1302, and RAM 1303 are interconnected via bus 1304. An input / output (I / O) interface 1305 is also connected to bus 1304; wherein ROM is a read-only memory and RAM is a random access memory.

[0250] Typically, the following devices can be connected to I / O interface 1305: input devices 1306 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 1307 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 1308 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1309. Communication device 1309 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 13 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0251] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 1309, or installed from a memory 1308, or installed from a ROM 1302. When the computer program is executed by the processor 1301, it performs the functions defined in the coronary target vessel dynamic path mapping registration method of the embodiments of the present invention.

[0252] Figure 13 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0253] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that the computer, processor, microprocessor controller, or programmable hardware includes storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the coronary target vessel dynamic path mapping registration method shown in the above embodiments is implemented.

[0254] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0255] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and all such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A method for registering a dynamic path map of a coronary target vessel, characterized in that, The method includes: Obtain a reference angiographic image sequence of the target vascular region, wherein the reference angiographic image sequence includes target vascular filling information and reference Mark point information; The reference angiography image sequence is processed, and a target reference mapping table is established, wherein the target reference mapping table is used to characterize the mapping relationship between cardiac cycle phase, reference vascular mask, reference Mark point, and local geometric relationship of Mark point; Acquire an angiography-free real-time fluoroscopic image of the target vascular region and extract multiple sets of real-time Mark points from the angiography-free real-time fluoroscopic image; Based on the spatiotemporal continuity constraint relationship between the multiple real-time Mark points, the real-time cardiac cycle phase of each real-time Mark point in different frames is determined; Based on the target reference mapping table, multiple reference Mark points and multiple reference vascular masks corresponding to the multiple real-time cardiac cycle phases are determined; Based on the multiple reference Mark points and the multiple real-time Mark points, the multiple reference vascular masks are spatially mapped and superimposed onto the angiography-free real-time fluoroscopic image using a preset rigid registration transformation formula to obtain a dynamic roadmap of the target local blood vessels. The preset rigid registration transformation formula only performs two-dimensional rigid transformations of translation and rotation, and does not use three or more Mark points to construct affine transformations.

2. The method according to claim 1, characterized in that, The reference contrast image sequence is processed, and a target reference mapping table is established, including: Based on the reference contrast image sequence, multiple reference Mark points are extracted and a Mark point motion period curve is constructed; Based on the reference angiography image sequence, determine multiple reference vascular masks and the local geometric relationships between multiple reference Mark points and the multiple reference vascular masks. The cardiac motion component and respiratory motion component in the cardiac cycle curve of the Mark point are separated, and multiple cardiac cycle phases are identified. The target reference mapping table is established based on the multiple cardiac cycle phases, the multiple reference vascular masks, the local geometric relationships of the Mark points, and the multiple reference Mark points.

3. The method according to claim 2, characterized in that, Based on the reference angiographic image sequence, multiple reference vascular masks are determined, including: Extract the contrast agent flow region from the reference contrast image sequence and generate multiple vessel probability maps; Adaptive threshold segmentation is performed on the multiple blood vessel probability maps to generate multiple reference blood vessel masks, wherein the reference blood vessel masks are marked with the branch vessels and branch opening positions in the segmented blood vessel masks.

4. The method according to claim 1, characterized in that, Based on the target reference mapping table, multiple reference Mark points and multiple reference vessel masks corresponding to the multiple real-time cardiac cycle phases are determined, including: Determine whether a corresponding real-time cardiac cycle phase exists in the target reference mapping table; When a corresponding real-time cardiac cycle phase exists in the target reference mapping table, the real-time cardiac cycle phase is used as an index to search the target reference mapping table and determine the corresponding reference Mark point and reference vascular mask. When the real-time cardiac cycle phase is between two adjacent discrete phases in the target reference mapping table, linear interpolation is performed on the coordinates of the reference Mark points and the reference vascular mask of the two adjacent discrete phases, and a reference Mark point and a reference vascular mask matching the current real-time phase are generated.

5. The method according to claim 1, characterized in that, Based on the multiple reference Mark points and the multiple real-time Mark points, using a preset rigid registration transformation formula, the multiple reference vascular masks are spatially mapped and superimposed onto the angiography-free real-time fluoroscopic image to obtain a dynamic path map of the target local blood vessels, including: Using the multiple reference Mark points and the multiple real-time Mark points, the preset rigid registration transformation formula is solved to obtain the rigid registration transformation matrix; Using the rigid registration transformation matrix, the multiple reference vascular masks are spatially mapped and superimposed onto the angiography-free real-time fluoroscopic image to obtain the target local vascular dynamic roadmap.

6. The method according to claim 5, characterized in that, Using the rigid registration transformation matrix, the multiple reference vascular masks are spatially mapped and superimposed onto the angiography-free real-time fluoroscopic image to obtain the target local vascular dynamic path map, including: Using the rigid registration transformation matrix, the multiple reference blood vessel masks are translated and rotated in two dimensions to obtain multiple mapped blood vessel masks. Using the local geometric relationship of Mark points in the target reference mapping table as a constraint verification condition, the relative spatial position deviation between the multiple mapped blood vessel masks and the multiple real-time Mark points is verified. The verified mapped vascular mask is superimposed onto the angiography-free real-time fluoroscopic image to obtain an initial local vascular dynamic roadmap. In the initial local vascular dynamic roadmap, the branch opening location, branch angle, stent expected release area or balloon kissing area are marked, and the target local vascular dynamic roadmap is generated.

7. The method according to claim 1, characterized in that, The method further includes: Based on the target local vascular dynamic path map, the registration residual value is calculated; When the registration residual value is greater than a preset residual threshold, the step of obtaining the reference angiographic image sequence of the target vascular region is returned, and the process is repeated until the registration residual value is less than or equal to the preset residual threshold, thus obtaining the target local vascular dynamic roadmap; or, when the registration residual value is greater than the preset residual threshold, the display reliability of the target local vascular dynamic roadmap is reduced.

8. A coronary artery target vessel dynamic route map registration device, characterized in that, The device includes: The acquisition module is used to acquire a reference angiographic image sequence of the target vascular region; A module is established to process the reference angiographic image sequence and establish a target reference mapping table, wherein the target reference mapping table is used to characterize the mapping relationship between cardiac cycle phase, reference vascular mask, reference Mark point and local geometric relationship of Mark point; The extraction module is used to acquire an angiography-free real-time fluoroscopic image of the target vascular region and extract multiple sets of real-time Mark points from the angiography-free real-time fluoroscopic image. The cardiac cycle phase determination module is used to determine the real-time cardiac cycle phase of each real-time Mark point in different frames based on the spatiotemporal continuity constraint relationship between the multiple real-time Mark points. The reference vessel mask determination module is used to determine multiple reference Mark points and multiple reference vessel masks corresponding to the multiple real-time cardiac cycle phases based on the target reference mapping table. The mapping module is used to spatially map the multiple reference vascular masks based on the multiple reference Mark points and the multiple real-time Mark points using a preset rigid registration transformation formula and superimpose them onto the angiography-free real-time fluoroscopic image to obtain a target local vascular dynamic roadmap. The preset rigid registration transformation formula only performs two-dimensional rigid transformations of translation and rotation, and does not use three or more Mark points to construct affine transformations.

9. An electronic device, characterized in that, include: The system includes a memory and a processor, which are interconnected. The memory stores computer instructions, and the processor executes the computer instructions to perform the coronary target vessel dynamic path mapping registration method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the coronary target vessel dynamic path registration method according to any one of claims 1 to 7.

Images