Method and apparatus for guidewire tip tracking in dynamic fluorescence fluoroscopic imaging sequences

Abstract

This invention belongs to the field of computer vision technology and discloses a method and apparatus for guidewire tip tracking in dynamic fluorescence fluoroscopy image sequences. The method includes: acquiring a dynamic fluorescence fluoroscopy image sequence during PCI; wherein the dynamic fluorescence fluoroscopy image sequence during PCI includes several consecutive fluorescence fluoroscopy image frames; inputting the dynamic fluorescence fluoroscopy image sequence during PCI into a geometric constraint detection framework, performing cross-frame tracking and trajectory matching of guidewire markers, and obtaining the tracking result of the guidewire tip; this invention achieves stable and high-precision identification and cross-frame tracking of guidewire tip position by acquiring a dynamic fluorescence fluoroscopy image sequence during PCI and using a geometric constraint detection framework to perform cross-frame tracking and trajectory matching of guidewire markers. It can maintain good robustness even in poor fluorescence fluoroscopy image quality, especially in low-contrast scenes, and adapt to the dynamic changes in guidewire position during PCI.

Description

Technical Field

[0001] This invention belongs to the field of computer vision technology, and relates to medical image processing technology, and in particular to a method and device for tracking the tip of a guidewire in a dynamic fluorescence fluoroscopic image sequence. Background Technology

[0002] Percutaneous coronary intervention (PCI) is a mainstream minimally invasive procedure for treating coronary artery disease. It relies on the vascular puncture path to precisely deliver interventional devices such as balloons and stents to the stenotic or occluded lesions of the coronary arteries, restoring normal blood supply to the myocardium by opening the diseased lumen. The guidewire, as the core navigation device in PCI, guides various interventional devices to accurately locate the lesion site. Among them, the tracking accuracy and trajectory matching effectiveness of the guidewire tip directly determine the precision and safety of the surgical operation.

[0003] Currently, guidewire tip tracking technology is widely used in PCI clinical practice, interventional device development, and medical image processing. However, in practical applications, under low-dose X-ray imaging conditions, dynamic fluorescence fluoroscopy image sequences generally suffer from low contrast and strong noise and dryness, resulting in blurred guidewire tip contours and poor identification. Furthermore, the complex morphology of human coronary arteries and significant individual differences in vascular anatomy among patients, coupled with the dynamic movement of the guidewire during PCI, exacerbate the technical difficulty of guidewire tip tracking. Insufficient tracking accuracy can easily lead to increased contrast agent dosage and device guidance deviation, resulting in serious complications such as vascular injury and stent mispositioning. This not only prolongs the operation time but also reduces patient safety, failing to meet the clinical demands for low-contrast-agent, high-precision, and real-time diagnostic and treatment capabilities.

[0004] To address the aforementioned issues with guidewire tip tracking during PCI, current technologies primarily employ operator-manual tracking and traditional image tracking methods. However, these methods struggle to achieve stable, high-precision identification and cross-frame tracking of the guidewire tip position. This results in time-consuming and labor-intensive manual image registration, increasing surgical risks and operational complexity. Specifically, operator-manual tracking relies entirely on the surgeon's clinical experience and visual observation, which is time-consuming and labor-intensive. Furthermore, it is prone to tracking deviations due to surgeon fatigue and differences in experience, making it unsuitable for the high-speed dynamic movement of the guidewire in clinical settings. Traditional image tracking methods achieve automatic tracking through conventional target detection models combined with basic inter-frame matching logic. However, this method lacks algorithmic optimization for low-contrast imaging scenarios, easily misidentifying image noise as guidewire tip signals. Moreover, it lacks an inter-frame matching strategy adapted to the dynamic geometric characteristics of the guidewire's movement, leading to issues such as guidewire trajectory disorder and tip loss. Consequently, the tracking accuracy and stability fall far short of the stringent requirements of clinical surgery. Summary of the Invention

[0005] In view of the technical problems existing in the prior art, the present invention provides a method and apparatus for guide wire tip tracking in dynamic fluorescence fluoroscopic image sequences, so as to solve the technical problem that existing guide wire tip tracking methods are difficult to achieve stable, high-precision identification and cross-frame tracking of guide wire tip position.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: This invention provides a method for guidewire tip tracking in dynamic fluorescence fluoroscopy image sequences, comprising: Acquire a sequence of dynamic fluorescence fluoroscopy images during PCI; wherein, the sequence of dynamic fluorescence fluoroscopy images during PCI includes several consecutive fluorescence fluoroscopy image frames; The dynamic fluorescence fluoroscopic image sequence during PCI is input into the geometric constraint detection framework to perform cross-frame tracking and trajectory matching of guidewire markers and obtain the tracking results of the guidewire tip. The geometric constraint detection framework includes a target detection module, a geometric constraint tracking module, a filtering module, a comparison module, and a fitting module. The target detection module is used to acquire candidate points for guide wire markers in each fluorescence fluoroscopic image frame; The geometric constraint tracking module is used to perform cross-frame matching of guide wire marker candidate points for each fluorescence fluoroscopic image frame based on spatial proximity constraints and temporal consistency constraints, and generate several tracking sequences. The filtering module is used to filter several tracking sequences by comparing common points between tracking sequences, and obtain the filtered tracking sequences. The comparison module is used for spatial deviation analysis based on trajectory overlap segments to identify tip marker sequences from the filtered tracking sequences; The fitting module is used to interpolate and complete the sequence of identified tip markers to obtain the tracking results of the guidewire tip.

[0007] Furthermore, the process of obtaining candidate guidewire markers for each fluorescence fluoroscopic image frame is as follows: Using an improved YOLOv5 model, guidewire marker detection was performed on each frame of the dynamic fluoroscopic image sequence during PCI, obtaining candidate guidewire marker points for each frame. The improved YOLOv5 model is a YOLOv5 model that integrates normalized Wasserstein distance.

[0008] Furthermore, based on spatial proximity constraints and temporal consistency constraints, the process of cross-frame matching of guidewire marker candidate points for each fluorescence fluoroscopic image frame to generate several tracking sequences is as follows: For each guidewire marker candidate point in the first fluorescence fluoroscopic image frame, an independent tracking sequence is initialized. For guidewire marker candidate points in subsequent fluorescence fluoroscopic image frames, traverse existing tracking sequences and determine whether the guidewire marker candidate points simultaneously meet preset matching conditions. If there are guidewire marker candidate points that meet the preset matching conditions, add them to the corresponding existing tracking sequence to generate several tracking sequences. For guidewire marker candidate points that have not been added to any existing tracking sequence, initialize a new tracking sequence.

[0009] Furthermore, the preset matching conditions include dynamic space constraints, proximity constraints, and maximum distance constraints; The dynamic spatial constraint is as follows: the candidate points of the guide wire markers in the current fluorescence fluoroscopic image frame are located within a dynamic rectangular region centered on the last valid point of the target tracking sequence; The proximity constraint condition is as follows: the Euclidean distance between the candidate point of the guide wire marker in the current fluorescence fluoroscopic image frame and the last valid point in the target tracking sequence satisfies the preset Euclidean distance threshold. The maximum distance constraint is as follows: the distance between the candidate point of the guide wire marker in the current fluorescence fluoroscopic image frame and the last valid point in the target tracking sequence in both the horizontal and vertical directions does not exceed the preset maximum allowable distance threshold.

[0010] Furthermore, by comparing common points among the tracking sequences, several tracking sequences are filtered to obtain the filtered tracking sequences, as follows: Calculate the number of common points between any two tracking sequences and compare the number of common points between any two tracking sequences with a preset common point threshold; If the number of common points between two tracking sequences exceeds a preset common point threshold, the number of valid points in the two tracking sequences is compared, and the tracking sequence with fewer valid points is removed to obtain the filtered tracking sequences.

[0011] Furthermore, based on spatial deviation analysis of the trajectory overlap segments, the process of identifying the tip marker sequence from the filtered tracking sequences is as follows: From the filtered tracking sequences, the tracking sequence with the longest total length is selected as the reference sequence; The spatial deviation of the overlapping segment between each filtered tracking sequence and the reference sequence is calculated and compared with a preset deviation threshold to obtain the tip marker sequence.

[0012] The present invention also provides a guidewire tip tracking system for dynamic fluorescence fluoroscopy image sequences, for implementing the aforementioned guidewire tip tracking method for dynamic fluorescence fluoroscopy image sequences, comprising: The image sequence acquisition module is used to acquire dynamic fluorescence fluoroscopy image sequences during PCI procedures; wherein, the dynamic fluorescence fluoroscopy image sequence during PCI procedures includes several consecutive fluorescence fluoroscopy image frames; The trajectory tracking sequence inputs the dynamic fluorescence fluoroscopic image sequence during PCI into the geometric constraint detection framework to perform cross-frame tracking and trajectory matching of guidewire markers, and obtains the tracking results of the guidewire tip; The geometric constraint detection framework includes a target detection module, a geometric constraint tracking module, a filtering module, a comparison module, and a fitting module. The target detection module is used to acquire candidate points for guide wire markers in each fluorescence fluoroscopic image frame; The geometric constraint tracking module is used to perform cross-frame matching of guide wire marker candidate points for each fluorescence fluoroscopic image frame based on spatial proximity constraints and temporal consistency constraints, and generate several tracking sequences. The filtering module is used to filter several tracking sequences by comparing common points between tracking sequences, and obtain the filtered tracking sequences. The comparison module is used for spatial deviation analysis based on trajectory overlap segments to identify tip marker sequences from the filtered tracking sequences; The fitting module is used to interpolate and complete the sequence of identified tip markers to obtain the tracking results of the guidewire tip.

[0013] The present invention also provides an electronic device, comprising: A processor is used to execute computer programs; A computer-readable storage medium storing a computer program that, when executed by the processor, performs the described guidewire tip tracking method for dynamic fluorescence fluoroscopic image sequences.

[0014] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the described method for guidewire tip tracking for dynamic fluorescence fluoroscopic image sequences.

[0015] The present invention also provides a computer program product comprising a computer program that, when executed by a processor, implements the aforementioned guidewire tip tracking method for dynamic fluorescence fluoroscopic image sequences.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention provides a guidewire tip tracking method for dynamic fluorescence fluoroscopy image sequences. By acquiring dynamic fluorescence fluoroscopy image sequences during PCI surgery and utilizing a geometric constraint detection framework for cross-frame tracking and trajectory matching of guidewire markers, it achieves stable, high-precision identification and cross-frame tracking of the guidewire tip position. This method maintains good robustness even in low-contrast scenes with poor fluorescence fluoroscopy image quality, adapting to dynamic changes in guidewire position during PCI surgery. Specifically, within the geometric constraint detection framework, a target detection module accurately captures candidate points of guidewire markers in each frame, effectively solving the problems of blurred guidewire tip contours and poor recognition in high-contrast, high-noise images. Furthermore, the geometric constraint tracking module relies on spatial proximity constraints and temporal consistency... This invention employs a constraint mechanism to achieve precise cross-frame matching of guidewire marker candidate points, effectively adapting to individual differences in guidewire dynamic movement characteristics and human coronary vascular anatomy, thus avoiding the inter-frame matching disorder problem of traditional methods. Next, a filtering module screens effective tracking sequences by comparing common points between tracking sequences, while a comparison module identifies tip marker sequences based on spatial deviation analysis of trajectory overlap segments. This dual screening eliminates noise and misidentified signals, reducing tracking deviation. Finally, a fitting module completes the tip marker sequence through interpolation, effectively improving the continuity and accuracy of tracking results. This invention significantly reduces the time and workload of manual image registration, lowers the dependence on additional contrast agents, and thus effectively improves the safety and clinical application value of real-time guidance during PCI.

[0017] The guidewire tip tracking system, electronic device, computer-readable storage medium, and computer program product for dynamic fluorescence fluoroscopy image sequences provided by this invention possess all the advantages of the aforementioned guidewire tip tracking method for dynamic fluorescence fluoroscopy image sequences. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart of the guidewire tip tracking method for dynamic fluorescence fluoroscopic image sequences provided in Example 1; Figure 2 This is a structural block diagram of the geometric constraint detection framework in Example 1; Figure 3 This is a schematic diagram illustrating the working principle of the geometric constraint detection framework in Example 1. Figure 4This is a qualitative detection and tracking result diagram of the coronary angiography image in Example 1; Figure 5 This is a schematic diagram illustrating the clinical application of the guidewire tip tracking method described in Example 1; Figure 6 This is a graph showing the quantitative performance evaluation results of the guidewire tip tracking method described in Example 1; Figure 7 This is a structural block diagram of the guidewire tip tracking system for dynamic fluorescence fluoroscopic image sequences provided in Example 2; Figure 8 This is a structural block diagram of the electronic device provided in Example 3. Detailed Implementation

[0020] To make the technical problems, technical solutions, and beneficial effects solved by this application clearer, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0021] Before describing the specific embodiments of this application, some of the technical terms involved in the embodiments of this application are explained as follows: The YOLOv5 model is a high-efficiency object detection model. It adopts a single-stage detector design and can process video frames in real time, making it suitable for deployment in scenarios with high real-time requirements, such as video surveillance and autonomous driving.

[0022] Normalized Wasserstein Distance (NWD) is an improved similarity metric, particularly suitable for small object detection. It measures the similarity between predicted and ground truth boxes by modeling bounding boxes as two-dimensional Gaussian distributions and calculating the Wasserstein distance between the two distributions.

[0023] This invention provides a method for guidewire tip tracking in dynamic fluorescence fluoroscopic image sequences, comprising the following steps: Step 100: Obtain the dynamic fluorescence fluoroscopy image sequence during PCI; wherein, the dynamic fluorescence fluoroscopy image sequence during PCI includes several consecutive fluorescence fluoroscopy image frames.

[0024] Step 200: Input the dynamic fluorescence fluoroscopic image sequence during PCI into the geometric constraint detection framework, perform cross-frame tracking and trajectory matching of the guidewire markers, and obtain the tracking results of the guidewire tip.

[0025] Specifically, the geometric constraint detection framework includes a target detection module, a geometric constraint tracking module, a filtering module, a comparison module, and a fitting module. The target detection module is used to acquire candidate guidewire markers for each fluorescence fluoroscopy image frame. The geometric constraint tracking module is used to perform cross-frame matching of candidate guidewire markers for each fluorescence fluoroscopy image frame based on spatial proximity constraints and temporal consistency constraints, generating several tracking sequences. The filtering module is used to filter several tracking sequences by comparing common points between tracking sequences, obtaining the filtered tracking sequences. The comparison module is used to identify tip marker sequences from the filtered tracking sequences based on spatial deviation analysis of trajectory overlap segments. The fitting module is used to interpolate and complete the identified tip marker sequences to obtain the tracking result of the guidewire tip.

[0026] In the above embodiments, by acquiring dynamic fluorescence fluoroscopy image sequences during PCI, a complete dynamic data foundation is provided for guidewire tip tracking. A geometric constraint detection framework is used for cross-frame tracking and trajectory matching of guidewire markers, achieving stable and high-precision tracking of the guidewire tip in low-contrast, high-noise dynamic fluorescence fluoroscopy image sequences. Within the geometric constraint detection framework, a target detection module first acquires candidate guidewire marker points for each fluorescence fluoroscopy image frame. Then, a geometric constraint tracking module applies spatial proximity constraints and temporal consistency constraints for cross-frame matching, effectively filtering out noise interference and establishing a continuous tracking sequence. Subsequently, through common point comparison filtering and spatial deviation analysis of trajectory overlap segments, the tip marker sequence is accurately identified, and finally, complete tracking results are obtained through interpolation.

[0027] This invention eliminates the need for manual operation by the surgeon, enabling stable and high-precision identification and cross-frame tracking of the guidewire tip position. This effectively reduces the complexity of surgical procedures and the workload of the surgeon, meeting the clinical needs of PCI surgery for low-contrast agent, high-precision, and real-time diagnosis and treatment, and effectively reducing surgical risks and operational complexity.

[0028] The following specific embodiments further explain the guidewire tip tracking method for dynamic fluorescence fluoroscopic image sequences provided by the present invention: Example 1 As attached Figure 1 As shown, this embodiment 1 provides a guidewire tip tracking method for dynamic fluorescence fluoroscopic image sequences, including the following steps: Step 1: Obtain the dynamic fluorescence fluoroscopy image sequence during PCI; wherein, the dynamic fluorescence fluoroscopy image sequence during PCI includes several consecutive fluorescence fluoroscopy image frames.

[0029] Step 2: Input the dynamic fluorescence fluoroscopic image sequence obtained in Step 1 into the geometric constraint detection frame, perform cross-frame tracking and trajectory matching of the guidewire markers, and obtain the tracking results of the guidewire tip.

[0030] Specifically, the geometric constraint detection framework comprises an object detection module, a geometric constraint tracking module, a filtering module, a comparison module, and a fitting module, as shown in the appendix. Figure 2 As shown.

[0031] In this embodiment 1, the geometric constraint detection framework achieves cross-frame tracking of markers by fusing spatial and temporal constraints, aiming to ensure consistency in the dynamic marker tracking process; its working process is shown in the attached figure. Figure 3 As shown; specifically, firstly, the improved YOLOv5 model is used to detect candidate points of guidewire markers in each fluorescence fluoroscopy image frame, thus achieving candidate point detection; then, a tracking strategy based on geometric constraints is used to initialize a new tracking sequence for each detected candidate point; subsequently, the corresponding target point is matched in subsequent fluorescence fluoroscopy image frames according to spatial geometric constraints to complete trajectory matching; next, the trajectory is optimized by distinguishing between candidate points at the proximal and tip of the guidewire; finally, interpolation is performed to complete missing candidate points, thus achieving robust cross-frame tracking of the guidewire tip.

[0032] It should be noted that in the first frame, the improved YOLOv5 model is used to detect candidate points and initialize an independent trajectory sequence for each candidate point. For each subsequent frame, the candidate points are matched with the existing tracking sequence based on spatial geometric constraints, combined with spatial proximity and temporal consistency criteria. Secondly, calibration is achieved through dynamically adjusted spatial boundaries, and the trajectories are compared and filtered based on common points. At the same time, the trajectory similarity index is used to distinguish between proximal and distal candidate points of the guidewire, ensuring stable identification of the target guidewire trajectory. Finally, missing points in the tracking sequence are interpolated and filled in, and the detection accuracy is evaluated by comparing with the results of manual annotation.

[0033] This geometric constraint detection framework integrates trajectory comparison, distance measurement, and dynamic alignment mechanisms, resulting in high accuracy in dynamic marker tracking tasks. It effectively addresses temporal misalignment issues, achieves robust marker type classification, and dynamically adapts to changes in marker motion states, providing an efficient and reliable marker tracking solution for medical imaging scenarios. Specifically, when markers shift or disappear between frames, the tracking sequence can be adaptively adjusted to maintain inter-frame consistency, while also improving resistance to partial occlusion. Furthermore, as a robust algorithm suitable for dynamic fluorescence fluoroscopy sequences, the geometric constraint detection framework enables cross-frame tracking and trajectory matching of guidewire tips. It incorporates an improved YOLOv5 model and NWD, significantly enhancing the reliability of detection and tracking tasks through optimized IoU calculation logic. Even in complex scenarios common in PCI procedures such as partial occlusion, marker rotation, rapid displacement, and low contrast, it can stably capture the guidewire tip position and maintain tracking continuity. Clinical angiography images have validated that this framework can operate in real-time in dynamic and uncontrollable clinical environments while maintaining stable tracking performance.

[0034] In this embodiment 1, the target detection module is used to obtain the guidewire marker candidate points for each fluorescence fluoroscopy image frame. Specifically, the improved YOLOv5 model is used to detect guidewire markers for each fluorescence fluoroscopy image frame in the dynamic fluorescence fluoroscopy image sequence during PCI, and to obtain the guidewire marker candidate points for each fluorescence fluoroscopy image frame. The improved YOLOv5 model is a YOLOv5 model that integrates normalized Wasserstein distance.

[0035] Explained, in the improved YOLOv5 model, integrating the normalized Wasserstein distance into the loss function to optimize the Intersection over Union (IoU) calculation method effectively improves the accuracy of small target detection and enhances model robustness. Specifically, the insensitivity of the normalized Wasserstein distance to the positional deviation of small targets provides a more stable similarity metric, helping the model learn the features of small targets better during training and detect them more accurately during inference. Secondly, due to the scale invariance and sensitivity to non-overlapping targets, the normalized Wasserstein distance maintains stable performance across scenarios of varying scales and complexities, contributing to better robustness in various practical applications. Furthermore, the smooth gradient provided by the normalized Wasserstein distance helps the model converge faster during training and avoid getting trapped in local optima, effectively shortening training time and improving the final performance of the model.

[0036] In this embodiment 1, the geometric constraint tracking module is used to perform cross-frame matching of guidewire marker candidate points for each fluorescence fluoroscopy image frame based on spatial proximity constraints and temporal consistency constraints, generating several tracking sequences. Specifically, firstly, independent tracking sequences are initialized for each guidewire marker candidate point in the first fluorescence fluoroscopy image frame; then, for guidewire marker candidate points in subsequent fluorescence fluoroscopy image frames, the existing tracking sequences are traversed to determine whether the guidewire marker candidate points simultaneously meet the preset matching conditions; if there are guidewire marker candidate points that meet the preset matching conditions, they are added to the corresponding existing tracking sequences to generate several tracking sequences.

[0037] The preset matching conditions include dynamic spatial constraints, proximity constraints, and maximum distance constraints. The dynamic spatial constraints are as follows: the candidate point of the guide wire marker in the current fluorescence fluoroscopy image frame is located within a dynamic rectangular region centered on the last valid point of the target tracking sequence. The proximity constraints are as follows: the Euclidean distance between the candidate point of the guide wire marker in the current fluorescence fluoroscopy image frame and the last valid point of the target tracking sequence meets a preset Euclidean distance threshold. The maximum distance constraints are as follows: the distance between the candidate point of the guide wire marker in the current fluorescence fluoroscopy image frame and the last valid point of the target tracking sequence in both the horizontal and vertical directions does not exceed a preset maximum allowable distance threshold.

[0038] Example explanation: Assume the first fluorescence perspective image frame (denoted as...) Candidate points for guidewire markers are denoted as follows: ;in, The first fluorescence transillumination image frame detected the first One guidewire marker candidate point , The total number of guidewire marker candidate points detected in the first fluorescence fluoroscopic image frame.

[0039] First, candidate points for guidewire markers. Each is initialized with a unique tracking sequence, denoted as . ;in, Candidate points for guidewire markers Initialize the independent tracking sequence.

[0040] Next, in processing subsequent fluorescence perspective image frames (when...) When ), the existing tracking sequences are traversed, and candidate point matching is completed by combining the spatial proximity criterion and the temporal consistency constraint; where, for a given target tracking sequence Its last valid point is denoted as , will determine the first Candidate points in a fluorescence perspective image frame Does it meet the following matching criteria: (1) The candidate points of the guide wire markers in the current fluorescence fluoroscopic image frame are located within a dynamic rectangular region centered on the last valid point of the target tracking sequence; that is, the candidate points It must be located at a valid point Within the dynamic rectangular region centered on the candidate point; wherein, the boundary of the dynamic rectangular region is determined by the candidate point and effective points The horizontal and vertical boundaries define the area as follows:

[0041] in, For valid points The x-coordinate; For valid points The ordinate; The time interval between the current fluorescence perspective image frame and the target tracking sequence. ; Candidate points The x-coordinate; Candidate points The ordinate.

[0042] It should be noted that this is achieved by introducing the time interval between the current fluorescence perspective image frame and the target tracking sequence. It enables the spatial search area to dynamically expand or shrink over time intervals; by utilizing the adaptive spatial adjustment mechanism of the dynamic rectangular area, it can effectively adapt to the variable speed movement and displacement of markers caused by patient movement, system noise or instrument rotation, thereby improving tracking stability in complex scenarios.

[0043] (2) The Euclidean distance between the candidate point of the guide wire marker in the current fluorescence fluoroscopic image frame and the last valid point in the target tracking sequence satisfies the preset Euclidean distance threshold; wherein, by introducing the Euclidean distance metric, the spatial proximity between the candidate point and the last valid point in the target tracking sequence is guaranteed, and the far end of different markers is not mistakenly associated with the current tracking sequence.

[0044] (3) The distance between the candidate point of the guide wire marker in the current fluorescence fluoroscopic image frame and the last valid point of the target tracking sequence in both the horizontal and vertical directions does not exceed the preset maximum allowable distance threshold; wherein, by applying the maximum spatial distance bidirectional constraint, it is ensured that the candidate point and the last valid point of the target tracking sequence do not exceed a reasonable range in both the horizontal and vertical directions: preferably, the maximum allowable distance threshold is set to 120 units, so as to eliminate potential mismatched candidate points that are too far away.

[0045] If candidate point If all conditions (1)-(3) above are met, then the candidate point will be... Add it to the corresponding target tracking sequence.

[0046] Specifically, if no matching candidate point is found in a certain tracking sequence, a None value will be filled at the corresponding frame position of the sequence, indicating that the corresponding guidewire marker in the fluorescence fluoroscopy image frame has not been detected. Furthermore, for guidewire marker candidate points that have not been added to any existing tracking sequence, a new tracking sequence is initialized. Specifically, for candidate points that do not match any existing tracking sequence, a new tracking sequence is initialized for them. At this time, each new detection point... Each generates an independent tracking sequence This enables effective tracking of newly appearing markers in subsequent frames.

[0047] It is worth noting that the geometric constraint tracking module executes iteratively across all image frames, ensuring both the temporal continuity and consistency of guide wire marker tracking and the flexibility to adapt to the inter-frame appearance and disappearance behavior of markers. In particular, the dynamic updating of the tracking sequence based on spatial local proximity and temporal consistency significantly improves robustness to partial occlusion and appearance changes. Partial occlusion includes temporary occlusion between markers, and appearance changes include shape / size variations caused by motion, rotation, or changes in shooting angle.

[0048] In this embodiment 1, the filtering module is used to filter several tracking sequences by comparing common points between tracking sequences to obtain the filtered tracking sequences. Specifically, the process of filtering several tracking sequences by comparing common points between tracking sequences to obtain the filtered tracking sequences is as follows: Calculate the number of common points between any two tracking sequences and compare it with a preset common point threshold. If the number of common points between two tracking sequences exceeds the preset common point threshold, compare the number of valid points in the two tracking sequences and remove the tracking sequence with fewer valid points to obtain the filtered tracking sequences.

[0049] Example explanation: definition , indicating the first Tracking sequences With the Tracking sequences The number of common points; if If the threshold for common points is exceeded, further comparison is performed. Tracking sequences With the Tracking sequences The number of valid points (not None values) in the sequence is used to determine the reliability of the sequence. Sequences with more valid points are considered to have higher reliability, while sequences with lower reliability are discarded to obtain the filtered tracking sequence.

[0050] Preferably, the preset common point threshold is 10; in this case, the filtering criteria can be expressed as follows:

[0051] in, For the first Tracking sequences With the Tracking sequences The number of common points.

[0052] It should be noted that pairwise alignment is performed on all tracking sequences. By comparing the number of shared common points between sequences, erroneous tracking sequences and sequences with low accuracy are eliminated, ensuring that only the most correlated sequences are retained for subsequent analysis. In complex and dynamic environments where markers are temporarily occluded or partially lost, the necessity of this filtering module is particularly prominent. Through a prudent filtering mechanism, it can effectively avoid interference from incorrect matches and abnormal sequences to subsequent analysis, ensuring the reliability of tracking results.

[0053] In this embodiment 1, the comparison module is used to identify the tip marker sequence from the screened tracking sequences based on the spatial deviation analysis of the trajectory overlap segment; specifically, the tracking sequence with the longest total length is selected as the reference sequence from the screened tracking sequences; the spatial deviation of the overlap segment between each screened tracking sequence and the reference sequence is calculated and compared with the preset deviation threshold to obtain the tip marker sequence.

[0054] Specifically, in angiography imaging scenarios, guidewire markers typically correspond to the proximal end and the tip (or distal end) of the guidewire, i.e., they include proximal markers and tip markers. In this embodiment 1, the inherent similarity of the spatial distribution of guidewire markers in a specific frame is utilized to enhance the classification and recognition effect.

[0055] More specifically, proximal markers mostly originate from the near end of the observed object, while apical markers originate from the farthest end. The motion trajectories of the two types of markers are both unique and share inherent common features, providing an important basis for achieving robust identification and classification. To further optimize matching performance, the degree of overlap between trajectory segments of sequence pairs is compared. The comparison process is based on a distance metric that can quantify the similarity of the motion paths of the two sequences.

[0056] Example explanation: First, define For the set of all filtered tracking sequences; where each All are spatial point sequences arranged in chronological order; next, the total length of the filtered tracking sequences is calculated, and the tracking sequence with the longest total length is selected as the reference sequence. The total length of the filtered tracking sequence is used to characterize the cumulative distance the tracked marker has traveled over time; for the tracking sequence Its total length is obtained by accumulating the Euclidean distance between adjacent points in the sequence; this distance metric is not only the core basis for analyzing the trajectory features of markers, but also provides a quantitative basis for subsequent comparisons of different tracking sequences.

[0057] Then, each tracking sequence is calculated. With reference sequence Trajectory differences To achieve the goal of using trajectory differences Quantifying the spatial deviation between two tracking sequences: This involves calculating the spatial deviation of each tracking sequence. With reference sequence Trajectory differences The process is as follows:

[0058] in, and Representing trajectories Compared with reference trajectory The time index corresponding to the overlapping frames; The summation term represents the number of overlapping frames between the two sequences; the summation term is used to calculate the average absolute spatial deviation of corresponding spatial points within the overlapping frames, thereby quantifying the difference between the trajectories.

[0059] If the trajectory differs Less than the preset deviation threshold Then determine the reference trajectory. For the proximal marker sequence, and simultaneously the trajectory The sequence is divided into remote marker sequences, and the trajectory is output. As the final target sequence; if the trajectory difference Exceeding the preset deviation threshold Then the reference trajectory is preserved. As the final target sequence; output the final target sequence to obtain the tip marker sequence; preferably, a preset deviation threshold. The value is 62.

[0060] Explained, this embodiment 1 demonstrates significant superiority in complex scenarios where markers exhibit temporal misalignment, such as due to changes in motion speed or nonlinear motion patterns, or appearance variations such as occlusion. It improves the accuracy of classifying proximal and distal marker sequences by dynamically adjusting the comparison process based on trajectory overlap and spatial alignment. The comparison module, as a core component of the framework, ensures the accuracy and consistency of dynamic marker tracking. Specifically, it calculates the spatial overlap of the tracking sequences and compares their topological attributes to optimize the marker association process. Furthermore, the comparison module is highly practical in distinguishing between proximal and distal marker ends, particularly crucial in scenarios such as angiography. This is because, in angiography scenarios, markers may be randomly distributed within any position within a frame, and accurately distinguishing between proximal and distal ends is a prerequisite for ensuring tracking effectiveness.

[0061] In this embodiment 1, the fitting module is used to interpolate and complete the identified tip marker sequence to obtain the tracking result of the guidewire tip. Specifically, the identified tip marker sequence is interpolated and completed according to a preset interpolation and completion strategy to obtain the tracking result of the guidewire tip. The preset interpolation and completion strategy includes strategy A and strategy B. Strategy A is that if there is a gap between two non-empty points in the identified tip marker sequence, the gap is filled by linear interpolation based on the two non-empty points. Strategy B is that if there is a gap after the last non-empty point in the identified tip marker sequence or if the starting point is missing a preceding point, linear extrapolation or linear interpolation is used for completion.

[0062] It should be noted that the pseudocode of the geometric constraint detection framework in this embodiment 1 is shown in Table 1 below.

[0063] Table 1. Pseudocode of the geometric constraint detection framework

[0064] Verification experiments and results: (1) Dataset In this embodiment 1, 70 clinical cases of coronary angiography were used. Each case included several consecutive fluorescence fluoroscopic image frames, with a single frame resolution of 512×512 pixels. For each case, key markers in each frame were manually annotated in advance, and the results of the manual annotation were established as the true standard for the image registration task. All cases were derived from real clinical patients, covering typical imaging features of cardiovascular diseases, which can effectively reflect the actual imaging manifestations of cardiovascular lesions in clinical scenarios and have good clinical representativeness.

[0065] (2) Evaluation indicators In this embodiment 1, the criteria for successful registration are set as follows: if the spatial distance between the marker position identified by the automatic registration algorithm and the real standard is less than or equal to 15 pixels, the marker is considered to be successfully registered; for each case, the marker registration success rate is used as the core evaluation index. If the marker registration success rate of a single case exceeds 90%, the registration result of the case is determined to meet the preset accuracy threshold; wherein, the registration success rate is calculated by the following formula (4).

[0066]

[0067] Using the above formula (4), the registration performance differences of the proposed algorithm in different cases can be systematically evaluated, providing reliable empirical support and theoretical basis for the subsequent optimization and iteration and performance improvement of the registration algorithm.

[0068] (3) Experimental results To verify the performance of the geometric constraint detection framework proposed in Embodiment 1, the attached... Figure 4 The paper presents qualitative analysis results of different frames from a coronary angiography sequence of the same patient; Appendix Figure 4 In the figure, Figures a, b, c, and d show the qualitative analysis results of different frames of images. The first column is the original fluorescence fluoroscopic image, and the second column is the detection and tracking results of the geometric constraint detection framework. The red dots mark the positions of the detected markers (guidewire tips), and the blue bounding boxes identify the candidate points. From the dynamic process of the guidewire gradually retracting in the sequence frames, it can be seen that the geometric constraint detection framework can achieve robust and accurate cross-frame tracking of the marker positions and can effectively adapt to the dynamic evolution of the imaging scene. The results fully demonstrate that in complex clinical angiography scenarios, the geometric constraint detection framework can exhibit stable and reliable performance in marker recognition and dynamic tracking tasks.

[0069] Secondly, appendix Figure 5 The video screenshots showcasing the clinical deployment of the algorithm provide a schematic diagram of the clinical application of the guidewire tip tracking method, demonstrating the product's interactive interface with OCT images. This interface, built on an angiography-OCT fusion imaging platform, can intuitively present the real-time detection and tracking results of interventional devices. Clinical deployment verification results further demonstrate that even under the complex imaging conditions brought about by OCT fusion data, the algorithm can maintain stable performance in actual clinical procedures, achieving accurate identification and tracking of target markers.

[0070] Furthermore, in this embodiment 1, the performance of the method was quantitatively evaluated based on clinical data from 50 patients. The registration accuracy results for each patient are shown in the attached figure. Figure 6As shown; it is noteworthy that 24 patients (48%) achieved 100% registration accuracy, verifying the robustness of the method described in Example 1 under ideal clinical conditions; simultaneously, 40 patients achieved registration accuracy exceeding 90%, indicating its high performance across diverse clinical scenarios; the overall average registration accuracy of 50 patients reached 92.7%, fully reflecting the clinical reliability of the method in various complex registration tasks during PCI surgery; from the appendix Figure 6 As can be seen from the bar chart, the registration success rate and mean accuracy (mAP) of different cases are fully presented. Among them, 24 patients (about half) achieved a 100% registration success rate, and 40 patients had a success rate of over 90%, which confirms the robustness of the method. The mAP reached 92.72%, which further verifies the reliability of the method in various registration tasks.

[0071] As shown in Table 2, the performance comparison results of different guidewire tip tracking methods in PCI fluorescence fluoroscopy sequences are presented. Among them, the tracking accuracies of the baseline models YOLOv5+GeoModule and YOLO11+GeoModule are 79.53% and 83.03%, respectively. However, the geometric constraint detection framework provided in Example 1 performs better, with a tracking accuracy of 92.72%, which is significantly better than the two baseline models. The core of the performance improvement comes from the optimized YOLOv5 backbone network, the IoU calculation method improved by NWD, and the geometric constraint tracking strategy in the geometric constraint detection framework, which can specifically solve the core challenge of low contrast in fluorescence fluoroscopy imaging during PCI. The results verify that the geometric constraint detection framework can provide robust guidewire tip tracking capability, which has important clinical significance for reducing intraoperative contrast agent dependence and improving the safety of PCI interventional surgery.

[0072] Table 2 Performance Comparison of Different Guidewire Tip Tracking Methods

[0073] As shown in Table 3 below, Table 3 presents the ablation experiment results of key modules in the geometric constraint detection framework; among them, the geometric constraint tracking module (GeoModule) is the core basic module. When this module is deployed alone, the tracking accuracy is 79.53%; after combining the filtering process defined by formula (2), i.e. the filtering module, with GeoModule, the tracking accuracy is improved to 82.15%; among them, the filtering module is used to remove erroneous and low-precision tracking sequences, and only retains highly correlated sequences for subsequent analysis; then, the comparison module that distinguishes between the near end and the far end is integrated into the architecture of GeoModule + filtering module, and the accuracy is further improved to 89.36%; finally, after fully integrating the fitting module with GeoModule + filtering module + comparison module, the tracking accuracy reaches 92.72%; the above results show that the step-by-step integration of each module can significantly improve the overall tracking and analysis accuracy of the algorithm, and verify the effectiveness of the design of each module.

[0074] Table 3 Ablation Experiment Results of Key Modules

[0075] As shown in Table 4 below, Table 4 quantifies the impact of four different parameter combinations on the algorithm's tracking accuracy. Scope and The range is used to define the spatial search area for the marker points; as shown in Table 4, comparing the first and third experimental groups, The range is fixed as [ 29,25]、 The range is fixed as [ [27,25], when only other parameter values ​​are adjusted, the accuracy shows a significant difference: Secondly, comparing the first and third groups of experiments, when the threshold of the number of common points is reached... When the threshold for the number of common points was increased from 10 to 15, the accuracy dropped from 92.72% to 83.69%, indicating that setting the threshold too high weakens the algorithm's ability to identify valid trajectories, leading to a decrease in accuracy. Secondly, comparing the first and fourth groups of experiments... Fixed at 10, threshold for comparing only trajectory differences As the number increased from 62 to 70, the accuracy decreased from 92.72% to 85.27%; the results show that... The higher the value, the lower the tracking accuracy of the algorithm; that is, an excessively high trajectory difference threshold can cause the algorithm to misclassify near-end and far-end marker sequences, thereby reducing the overall tracking accuracy; when Scope and The spatial search range is adjusted to [ At [25,25], the algorithm accuracy dropped to 91.60%; the shrinkage of the spatial search range changed the effective search domain of the marker points, resulting in some candidate points being missed or having matching deviations. This shows that unreasonable adjustments to the spatial search range can also have an adverse effect on the algorithm accuracy, verifying the sensitivity of this parameter setting.

[0076] Table 4. Algorithm performance metrics under different parameter combinations

[0077] In this embodiment 1, the geometric constraint detection framework, as a novel method for robust guidewire tip tracking and trajectory matching in dynamic fluorescence fluoroscopy sequences during PCI, has good potential for integration with existing angiography imaging systems. It can provide technical support for automated monitoring of PCI procedures. Its core advantage lies in providing interventional cardiologists with more accurate visual feedback while reducing reliance on contrast agents. Clinical angiography image validation results show that the guidewire tip tracking accuracy of the geometric constraint detection framework reaches 92.7%, and it has real-time operation capability. It is seamlessly compatible with existing clinical workflows, demonstrating its good clinical applicability.

[0078] Example 2 This embodiment 2 provides a guidewire tip tracking system for dynamic fluorescence fluoroscopy image sequences, which is used in the guidewire tip tracking method for dynamic fluorescence fluoroscopy image sequences described in embodiment 1 above; as shown in the attached... Figure 7 As shown, the guidewire tip tracking system includes an image sequence acquisition module and a trajectory tracking sequence.

[0079] The image sequence acquisition module is used to acquire dynamic fluorescence fluoroscopy image sequences during PCI procedures; wherein, the dynamic fluorescence fluoroscopy image sequences during PCI procedures include several consecutive fluorescence fluoroscopy image frames.

[0080] The trajectory tracking sequence inputs the dynamic fluorescence fluoroscopic image sequence during PCI into the geometric constraint detection framework to perform cross-frame tracking and trajectory matching of guidewire markers, thereby obtaining the tracking results of the guidewire tip.

[0081] In this embodiment 2, the geometric constraint detection framework includes a target detection module, a geometric constraint tracking module, a filtering module, a comparison module, and a fitting module. Specifically, the target detection module is used to acquire candidate guidewire markers for each fluorescence fluoroscopy image frame; the geometric constraint tracking module is used to perform cross-frame matching of candidate guidewire markers for each fluorescence fluoroscopy image frame based on spatial proximity constraints and temporal consistency constraints to generate several tracking sequences; the filtering module is used to filter several tracking sequences by comparing common points between tracking sequences to obtain filtered tracking sequences; the comparison module is used to identify tip marker sequences from the filtered tracking sequences based on spatial deviation analysis of trajectory overlap segments; and the fitting module is used to interpolate and complete the identified tip marker sequences to obtain the tracking result of the guidewire tip.

[0082] Example 3 As attached Figure 8 As shown, this embodiment 3 provides an electronic device, including: a memory for storing a computer program; a processor for executing the computer program to implement the steps of a guidewire tip tracking method for dynamic fluorescence fluoroscopy image sequences; or, the processor for executing the computer program to implement the functions of each module in the above-mentioned guidewire tip tracking system for dynamic fluorescence fluoroscopy image sequences.

[0083] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a preset function, the instruction segments describing the execution process of the computer program in the electronic device.

[0084] The electronic device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above are examples of electronic devices and do not constitute a limitation on the electronic device. It may include more components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0085] The processor can be a central processing unit, or other general-purpose processors, digital signal processors, application-specific integrated circuits, off-the-shelf programmable gate arrays or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor, or any conventional processor, etc. The processor is the control center of the electronic device, connecting various parts of the entire electronic device through various interfaces and lines.

[0086] The memory can be used to store the computer program and / or module. The processor implements various functions of the electronic device by running or executing the computer program and / or module stored in the memory and by calling the data stored in the memory.

[0087] The memory may primarily include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a given function (such as sound playback, image playback, etc.). The data storage area may store data created based on the use of the mobile phone (such as audio data, phonebook, etc.). Furthermore, the memory may include high-speed random access memory and non-volatile memory, such as hard disks, RAM, plug-in hard disks, smart memory cards, secure digital cards, flash memory cards, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0088] Example 4 This embodiment 4 also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the guidewire tip tracking method for dynamic fluorescence fluoroscopic image sequences.

[0089] If the integrated module / unit of the guidewire tip tracking system for dynamic fluorescence fluoroscopic image sequences is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium.

[0090] Based on this understanding, the present invention can implement all or part of the process in the above-described method for tracking the guidewire tip in dynamic fluorescence fluoroscopy image sequences. This can also be accomplished by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of the above-described method for tracking the guidewire tip in dynamic fluorescence fluoroscopy image sequences. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or a preset intermediate form, etc.

[0091] The computer-readable storage medium may include any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory, random access memory, electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0092] Example 5 This embodiment 5 provides a computer product, which includes a computer program stored in a computer-readable storage medium. The processor of the electronic device reads the computer program from the computer-readable storage medium and executes the computer program, so that the electronic device can perform the guide wire tip tracking method for dynamic fluorescence fluoroscopic image sequence described in embodiment 1, which will not be described again here.

[0093] It should be noted that those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above methods.

[0094] The guidewire tip tracking method provided by this invention, based on existing image registration and motion tracking technologies, significantly improves the accuracy and efficiency of guidewire tip tracking and trajectory matching by integrating advanced deep learning and image processing algorithms. Specifically, by integrating a YOLOv5 model at the single-frame image level to locate candidate regions for the guidewire tip and introducing NWD-optimized cross-union ratio calculation logic, the reliability of target detection results under low-contrast imaging conditions is effectively improved. For inter-frame tracking tasks, a geometric constraint tracking strategy is designed to initialize an independent trajectory sequence for each candidate point detected in each frame and achieve accurate inter-frame target point matching based on spatial geometric constraint criteria. Simultaneously, a comparison module is embedded to effectively distinguish proximal and distal feature points of the guidewire, ultimately achieving stable identification of the target guidewire trajectory. In this invention, the geometric constraint detection framework maintains good robustness even in low-contrast scenarios with poor fluorescence fluoroscopy image quality, while adapting to dynamic changes in guidewire position during PCI. The geometric constraint detection framework automates the process of guidewire tip tracking and trajectory matching, effectively reducing the time and workload of manual image registration, lowering the dependence on additional contrast agents, and thus improving the safety and clinical application value of real-time guidance during PCI.

[0095] In this invention, the geometric constraint detection framework uses a tracking strategy based on geometric constraints to initialize a new trajectory sequence for each detected candidate point and match the corresponding target point in subsequent frames according to spatial geometric constraints. Simultaneously, a comparison module distinguishes between proximal and distal candidate points of the guidewire, achieving stable identification of the target guidewire trajectory. This geometric constraint detection framework provides clinicians with clear visual guidance of the guidewire tip without the need for additional contrast agent injection, significantly optimizing surgical planning and clinical decision-making processes during PCI. This results in more precise interventional guidance, effectively reducing the risk of kidney damage caused by excessive contrast agent use, and comprehensively improving the overall safety and clinical efficiency of PCI procedures.

[0096] The above embodiments are merely one of the implementation methods for achieving the technical solution of the present invention. The scope of protection claimed by the present invention is not limited to this embodiment, but also includes any variations, substitutions and other implementation methods that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention.

Claims

1. A method for tracking the tip of a guidewire in a dynamic fluorescence fluoroscopic image sequence, characterized in that, include: Acquire a sequence of dynamic fluorescence fluoroscopy images during PCI; wherein, the sequence of dynamic fluorescence fluoroscopy images during PCI includes several consecutive fluorescence fluoroscopy image frames; The dynamic fluorescence fluoroscopic image sequence during PCI is input into the geometric constraint detection framework to perform cross-frame tracking and trajectory matching of guidewire markers and obtain the tracking results of the guidewire tip. The geometric constraint detection framework includes a target detection module, a geometric constraint tracking module, a filtering module, a comparison module, and a fitting module. The target detection module is used to acquire candidate points for guide wire markers in each fluorescence fluoroscopic image frame; The geometric constraint tracking module is used to perform cross-frame matching of guide wire marker candidate points for each fluorescence fluoroscopic image frame based on spatial proximity constraints and temporal consistency constraints, and generate several tracking sequences. The filtering module is used to filter several tracking sequences by comparing common points between tracking sequences, and obtain the filtered tracking sequences. The comparison module is used for spatial deviation analysis based on trajectory overlap segments to identify tip marker sequences from the filtered tracking sequences; The fitting module is used to interpolate and complete the sequence of identified tip markers to obtain the tracking results of the guidewire tip.

2. The method for guidewire tip tracking for dynamic fluorescence fluoroscopic image sequences according to claim 1, characterized in that, The process of obtaining candidate guidewire markers for each fluorescence fluoroscopic image frame is as follows: Using an improved YOLOv5 model, guidewire marker detection was performed on each frame of the dynamic fluoroscopic image sequence during PCI, obtaining candidate guidewire marker points for each frame. The improved YOLOv5 model is a YOLOv5 model that integrates normalized Wasserstein distance.

3. The method for guidewire tip tracking for dynamic fluorescence fluoroscopic image sequences according to claim 1, characterized in that, Based on spatial proximity constraints and temporal consistency constraints, the process of cross-frame matching of guidewire marker candidate points for each fluorescence fluoroscopic image frame to generate several tracking sequences is as follows: For each guidewire marker candidate point in the first fluorescence fluoroscopic image frame, an independent tracking sequence is initialized. For guidewire marker candidate points in subsequent fluorescence fluoroscopic image frames, traverse existing tracking sequences and determine whether the guidewire marker candidate points simultaneously meet preset matching conditions. If there are guidewire marker candidate points that meet the preset matching conditions, add them to the corresponding existing tracking sequence to generate several tracking sequences. For guidewire marker candidate points that have not been added to any existing tracking sequence, initialize a new tracking sequence.

4. The method for guidewire tip tracking for dynamic fluorescence fluoroscopic image sequences according to claim 3, characterized in that, The preset matching conditions include dynamic space constraints, proximity constraints, and maximum distance constraints; The dynamic spatial constraint is as follows: the candidate points of the guide wire markers in the current fluorescence fluoroscopic image frame are located within a dynamic rectangular region centered on the last valid point of the target tracking sequence; The proximity constraint condition is as follows: the Euclidean distance between the candidate point of the guide wire marker in the current fluorescence fluoroscopic image frame and the last valid point in the target tracking sequence satisfies the preset Euclidean distance threshold. The maximum distance constraint is as follows: the distance between the candidate point of the guide wire marker in the current fluorescence fluoroscopic image frame and the last valid point in the target tracking sequence in both the horizontal and vertical directions does not exceed the preset maximum allowable distance threshold.

5. The method for guidewire tip tracking for dynamic fluorescence fluoroscopic image sequences according to claim 1, characterized in that, The process of filtering several tracking sequences by comparing common points among them to obtain the filtered tracking sequences is as follows: Calculate the number of common points between any two tracking sequences and compare the number of common points between any two tracking sequences with a preset common point threshold; If the number of common points between two tracking sequences exceeds a preset common point threshold, the number of valid points in the two tracking sequences is compared, and the tracking sequence with fewer valid points is removed to obtain the filtered tracking sequences.

6. The method for guidewire tip tracking for dynamic fluorescence fluoroscopic image sequences according to claim 1, characterized in that, The process of identifying tip marker sequences from the filtered tracking sequences based on spatial deviation analysis of overlapping trajectory segments is as follows: From the filtered tracking sequences, the tracking sequence with the longest total length is selected as the reference sequence; The spatial deviation of the overlapping segment between each filtered tracking sequence and the reference sequence is calculated and compared with a preset deviation threshold to obtain the tip marker sequence.

7. A guidewire tip tracking system for dynamic fluorescence fluoroscopic image sequences, characterized in that, A method for tracking the guidewire tip in a dynamic fluorescence fluoroscopic image sequence as described in any one of claims 1-6, comprising: The image sequence acquisition module is used to acquire dynamic fluorescence fluoroscopy image sequences during PCI procedures; wherein, the dynamic fluorescence fluoroscopy image sequence during PCI procedures includes several consecutive fluorescence fluoroscopy image frames; The trajectory tracking sequence inputs the dynamic fluorescence fluoroscopic image sequence during PCI into the geometric constraint detection framework to perform cross-frame tracking and trajectory matching of guidewire markers, and obtains the tracking results of the guidewire tip; The geometric constraint detection framework includes a target detection module, a geometric constraint tracking module, a filtering module, a comparison module, and a fitting module. The target detection module is used to acquire candidate points for guide wire markers in each fluorescence fluoroscopic image frame; The geometric constraint tracking module is used to perform cross-frame matching of guide wire marker candidate points for each fluorescence fluoroscopic image frame based on spatial proximity constraints and temporal consistency constraints, and generate several tracking sequences. The filtering module is used to filter several tracking sequences by comparing common points between tracking sequences, and obtain the filtered tracking sequences. The comparison module is used for spatial deviation analysis based on trajectory overlap segments to identify tip marker sequences from the filtered tracking sequences; The fitting module is used to interpolate and complete the sequence of identified tip markers to obtain the tracking results of the guidewire tip.

8. An electronic device, characterized in that, include: A processor is used to execute computer programs; A computer-readable storage medium storing a computer program that, when executed by the processor, performs the guidewire tip tracking method for dynamic fluorescence fluoroscopic image sequences as described in any one of claims 1-6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the guidewire tip tracking method for dynamic fluorescence fluoroscopic image sequences as described in any one of claims 1-6.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the guidewire tip tracking method for dynamic fluorescence fluoroscopic image sequences as described in any one of claims 1-6.

Images