A target tracking method, device, medium and product suitable for ultrasound contrast dual-mode dynamic images

Abstract

This application discloses a target tracking method, device, medium, and product applicable to dual-mode dynamic ultrasound imaging, relating to the field of target tracking technology. The method includes: configuring Kalman filters for both the anatomical and contrast-enhanced modalities; performing state prediction based on the fusion result of the previous frame; extracting dual-mode observation vectors and calculating innovation and residuals; simultaneously solving and correcting the noise covariance matrix using joint regression of innovation covariance and recursive least squares with matrix forgetting factors; performing two-stage measurement updates and cross-modal information interaction; and weighted fusion of the posterior state and closed-loop feedback to the next frame. This application can achieve deep dual-mode collaboration, adaptive estimation of noise covariance, and closed-loop correction, significantly improving the robustness and accuracy of target tracking in complex clinical scenarios.

Description

Technical Field

[0001] This application relates to the field of target tracking technology, and in particular to a target tracking method, device, medium and product suitable for dual-mode dynamic imaging of ultrasound contrast imaging. Background Technology

[0002] The existing dual-mode dynamic imaging target tracking in ultrasound contrast imaging suffers from the lack of cross-modal information interaction and collaboration mechanisms, and insufficient estimation accuracy of the covariance matrix (Q, R) of observation noise and process noise. This leads to technical challenges such as poor tracking robustness and difficulty in adapting to complex clinical scenarios, including respiratory motion, probe micro-movement, and lesion deformation. Summary of the Invention

[0003] The purpose of this application is to provide a target tracking method, device, medium and product applicable to dual-mode dynamic imaging of ultrasound contrast imaging, which can realize deep collaboration of dual-mode information, adaptive synchronous estimation of noise covariance and have closed-loop feedback capability, thereby adapting to the technical challenges of complex clinical scenarios such as respiratory motion, probe micro-motion and lesion deformation.

[0004] To achieve the above objectives, this application provides the following solution: In a first aspect, this application provides a target tracking method suitable for dual-mode dynamic ultrasound contrast imaging, comprising: Acquire dual-mode dynamic ultrasound contrast imaging; The ultrasound contrast imaging dual-mode dynamic images are segmented into anatomical modal images and contrast modal images, and the image coordinates are unified to establish a globally unified coordinate system. Independent Kalman filters are configured for the anatomical modality images and the contrast modality images respectively, to construct a dual-path Kalman filter; Based on the dual-path Kalman filter, state prediction is performed using the posterior state and posterior covariance obtained from the previous frame fusion to obtain the state prediction vector and covariance prediction matrix for the current frame. Based on the current frame state prediction vector and covariance prediction matrix, extract the dual-mode observation vector and calculate the innovation and state prediction residual; Based on the aforementioned information and state prediction residuals, the process noise covariance matrix and observation noise covariance matrix are obtained simultaneously by using the joint regression of information covariance and the recursive least squares method of matrix forgetting factor, and by constraint correction. Based on the process noise covariance matrix, the observation noise covariance matrix and the dual-mode observation vector, the dual-path Kalman filter is subjected to two measurement updates and cross-modal information interaction to obtain the second posterior state and the second posterior covariance. Based on the second posterior state and the second posterior covariance, state weighted fusion and closed-loop feedback are performed to obtain the globally optimal target state and the global fusion covariance, which are then used as the input for the state prediction of the next frame.

[0005] In a second aspect, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the target tracking method for dual-mode dynamic ultrasound imaging as described above.

[0006] Thirdly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the target tracking method for dual-mode dynamic ultrasound contrast imaging described above.

[0007] Fourthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the target tracking method for dual-mode dynamic ultrasound contrast imaging described above.

[0008] According to the specific embodiments provided in this application, this application has the following technical effects: This application provides a target tracking method, device, medium, and product applicable to dual-mode dynamic ultrasound imaging. This application improves the robustness and accuracy of dual-mode tracking by configuring independent Kalman filters for anatomical and contrast-enhanced images, constructing a dual-path Kalman filter architecture. Each filter can perform specialized modeling and state estimation based on the imaging characteristics of its respective modality (e.g., anatomical modality focuses on morphological contour stability, while contrast-enhanced modality focuses on dynamic changes in blood supply signals). This avoids tracking deviations caused by modal differences in traditional single-path filtering methods, significantly improving the robustness and position estimation accuracy of target tracking in complex clinical scenarios. This application also achieves high-precision adaptive estimation of the process noise and observation noise covariance matrices. It employs a novelty covariance joint regression and matrix forgetting factor recursive least squares method to simultaneously solve and constrain the process noise and observation noise covariance matrices, overcoming the shortcomings of existing technologies where fixed parameters or step-by-step estimation cannot match real-time dynamic noise changes. This method can capture real-time fluctuations in observation noise and uncertainties in target motion during ultrasound images, ensuring that the Kalman filter always maintains optimal gain and effectively suppressing the risk of filter divergence. This application enhances cross-modal information synergy and complementarity. During the dual-measurement update process, the dual-path Kalman filter achieves cross-modal interaction and deep synergy between anatomical structural information and contrast function information by exchanging the observation vector and the observation noise covariance matrix. This mechanism fully utilizes the complementary advantages of dual-mode images in target representation, compensating for information loss caused by artifacts, signal attenuation, or perfusion differences in a single mode, further improving the continuity and reliability of target state estimation. This application ensures the continuity and stability of long-term tracking. It optimally fuses the dual-path filtering results through a weighted fusion strategy based on the covariance matrix trace, and feeds the fused global optimal target state and the global fusion covariance in a closed loop to the next frame's state prediction stage, forming a closed-loop iterative tracking mechanism of "coordinate unification—prediction—estimation—update—fusion—feedback". This mechanism effectively eliminates inter-frame tracking jumps and error accumulation, ensuring smooth and stable target trajectory output under non-rigid interferences such as respiratory motion and probe micro-movements. This application improves clinical applicability and automation. The overall method requires no manual intervention for parameter adjustment. Through unified processing of image coordinates and online adaptive estimation of the noise covariance matrix, it can automatically adapt to differences in image characteristics under different patients and imaging conditions, reducing the operational complexity of target tracking in dual-mode dynamic ultrasound contrast imaging and improving the level of automation and universality in clinical applications. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments 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.

[0010] Figure 1 This is an application environment diagram of a target tracking method for dual-mode dynamic ultrasound imaging according to an embodiment of this application; Figure 2 A flowchart illustrating a target tracking method for dual-mode dynamic ultrasound imaging provided in an embodiment of this application; Figure 3 A schematic diagram of the overall process of a target tracking method for dual-mode dynamic imaging of ultrasound contrast imaging, provided in another embodiment of this application; Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0011] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0012] This application aims to address the technical challenges in existing ultrasound contrast imaging dual-mode dynamic image target tracking methods, such as the lack of cross-modal information interaction and collaboration mechanisms, insufficient estimation accuracy of the covariance matrices (Q, R) of observation noise and process noise, leading to poor tracking robustness and difficulty in adapting to complex clinical scenarios such as respiratory motion, probe micro-movement, and lesion deformation. It provides a dual-channel interactive adaptive Kalman filter-based ultrasound contrast imaging dual-mode dynamic image target tracking method. The core innovations of this application include two independent innovations: Innovation 1 is targeted modeling using dual-channel Kalman filters, where independent Kalman filters are configured for each mode, and the two Kalman filters achieve information collaboration by exchanging and updating the observation matrix, observation vector, and observation noise; Innovation 2 is that each Kalman filter employs a joint regression of the new covariance and a matrix forgetting factor recursive least squares method to simultaneously solve for its own process noise covariance matrix Q and observation noise covariance matrix R, replacing the traditional fixed Q, R or step-by-step estimation of Q, R methods. Meanwhile, this application incorporates a cross-modal information interaction and closed-loop feedback mechanism. Through dual measurement updates, state-weighted fusion, and fusion result feedback, it achieves deep collaboration of dual-mode information, ultimately achieving continuous, accurate, and stable tracking of the target in dual-mode dynamic ultrasound imaging.

[0013] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0014] The target tracking method for dual-mode dynamic ultrasound contrast imaging provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be set up independently, integrated into server 104, or placed in the cloud or on another server.

[0015] The terminal 102 can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, and smart in-vehicle devices. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted devices. The server 104 can be implemented using a standalone server or a server cluster composed of multiple servers, or it can be a cloud server.

[0016] In one exemplary embodiment, such as Figure 2 As shown, a target tracking method suitable for dual-mode dynamic ultrasound contrast imaging is provided. This method is executed by a computer device, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method is applied to... Figure 1 Taking server 104 as an example, the explanation includes the following steps S1 to S8. Wherein: S1. Acquire dual-mode dynamic images of ultrasound contrast imaging.

[0017] S2. The ultrasound contrast imaging dual-mode dynamic image is segmented into anatomical modal image and contrast modal image, and the image coordinates are unified to establish a global unified coordinate system.

[0018] S3. Configure independent Kalman filters for the anatomical modality image and the contrast modality image respectively, and construct a dual-path Kalman filter.

[0019] S4. Based on the dual-path Kalman filter, perform state prediction using the posterior state and posterior covariance obtained from the previous frame fusion to obtain the state prediction vector and covariance prediction matrix for the current frame.

[0020] S5. Based on the current frame state prediction vector and covariance prediction matrix, extract the dual-mode observation vector and calculate the innovation and state prediction residual.

[0021] S6. Based on the aforementioned information and state prediction residuals, the process noise covariance matrix and observation noise covariance matrix are obtained simultaneously by using the joint regression of information covariance and the recursive least squares method of matrix forgetting factor, and by constraint correction.

[0022] S7. Based on the process noise covariance matrix, the observation noise covariance matrix and the dual-mode observation vector, perform two measurement updates and cross-modal information interaction on the dual-path Kalman filter to obtain the second posterior state and the second posterior covariance.

[0023] S8. Based on the second posterior state and the second posterior covariance, perform state weighted fusion and closed-loop feedback to obtain the global optimal target state and the global fusion covariance, and use them as the input for the state prediction of the next frame.

[0024] This embodiment provides a target tracking method, device, medium, and product suitable for dual-mode dynamic ultrasound imaging. This embodiment improves the robustness and accuracy of dual-mode tracking by configuring independent Kalman filters for anatomical and contrast-enhanced images, constructing a dual-path Kalman filter architecture. Each filter can perform specialized modeling and state estimation based on the imaging characteristics of its respective modality (e.g., anatomical modality focuses on morphological contour stability, while contrast-enhanced modality focuses on dynamic changes in blood supply signals). This avoids tracking deviations caused by modal differences in traditional single-path filtering methods, significantly improving the robustness and position estimation accuracy of target tracking in complex clinical scenarios. This embodiment achieves high-precision adaptive estimation of the process noise and observation noise covariance matrices. It employs a novel covariance joint regression and matrix forgetting factor recursive least squares method to simultaneously solve and constrain the process noise and observation noise covariance matrices, overcoming the shortcomings of existing technologies where fixed parameters or step-by-step estimation cannot match real-time dynamic noise changes. This method can capture real-time fluctuations in observation noise and uncertainties in target motion during ultrasound images, ensuring that the Kalman filter always maintains optimal gain and effectively suppressing the risk of filter divergence. This embodiment enhances cross-modal information synergy and complementarity. During the dual-measurement update process, the dual-path Kalman filter achieves cross-modal interaction and deep synergy between anatomical structural information and contrast function information by exchanging the observation vector and the observation noise covariance matrix. This mechanism fully utilizes the complementary advantages of dual-mode images in target representation, compensating for information loss caused by artifacts, signal attenuation, or perfusion differences in a single mode, further improving the continuity and reliability of target state estimation. This embodiment ensures the continuity and stability of long-term tracking. It optimally fuses the dual-path filtering results through a weighted fusion strategy based on the covariance matrix trace, and feeds the fused globally optimal target state and the global fusion covariance in a closed loop to the next frame's state prediction stage, forming a closed-loop iterative tracking mechanism of "coordinate unification—prediction—estimation—update—fusion—feedback". This mechanism effectively eliminates inter-frame tracking jumps and error accumulation, ensuring smooth and stable target trajectory output under non-rigid interferences such as respiratory motion and probe micro-movements. This application improves clinical applicability and automation. The overall method requires no manual intervention for parameter adjustment. Through unified processing of image coordinates and online adaptive estimation of the noise covariance matrix, it can automatically adapt to differences in image characteristics under different patients and imaging conditions, reducing the operational complexity of target tracking in dual-mode dynamic ultrasound contrast imaging and improving the level of automation and universality in clinical applications.

[0025] This embodiment employs a complete technical solution including dual-mode collaborative processing, dual-path interactive adaptive Kalman filtering, cross-modal dual measurement updates, state-weighted fusion, and closed-loop feedback. It is implemented in modules around the core innovations, with specific technical details as follows: In another exemplary embodiment of this application, in step S2 above, the ultrasound contrast-enhanced dual-mode dynamic image is segmented into an anatomical modality image and a contrast-enhanced modality image, and image coordinate unification processing is performed to establish a globally unified coordinate system. Specifically, this includes: dividing the ultrasound contrast-enhanced dual-mode dynamic image of size W×H into two equal regions along the vertical midline, with the left region serving as the anatomical modality image and the right region serving as the contrast-enhanced modality image, each region having a size of (W / 2)×H; establishing a local coordinate system with the upper left corner of each region as the origin, the horizontal axis as the x-axis, and the vertical axis as the y-axis, so that the coordinate range of the left and right images is unified as x∈[0,W / 2) and y∈[0,H).

[0026] Specifically, this embodiment performs image coordinate unification processing: to ensure that the state vector coordinates of the two Kalman filters are completely consistent, eliminate positional offsets and coordinate ambiguities in the dual-mode dynamic images of ultrasound contrast imaging, and ensure coordinate consistency in all subsequent stages such as state prediction, observation vector extraction, noise covariance estimation, and measurement updates, the two Kalman filters achieve information collaboration by exchanging observation matrices, observation vectors, and observation noise and then updating them. Specifically, coordinate unification processing of the dual-mode dynamic images of ultrasound contrast imaging needs to be performed first, and the steps are as follows: 1. Vertically centered image segmentation: The ultrasound contrast-enhanced dual-mode dynamic image of size W×H is uniformly divided into two equal-sized regions along the vertical midline (where W is the width of the stitched image and H is the height of the stitched image). The left region is the anatomical modality image I1 (the anatomical modality is specifically the B-mode image), and the right region is the contrast-enhanced modality image I2 (the contrast-enhanced modality is specifically the contrast-enhanced image). The size of each region is (W / 2)×H, realizing the physical separation and correspondence of the ultrasound contrast-enhanced dual-mode dynamic image.

[0027] 2. Establishment of a unified global coordinate system: Both the left and right modal images adopt a unified local coordinate system, with the upper left corner of their respective regions as the origin (0, 0), the horizontal axis as the x-axis, and the vertical axis as the y-axis. The coordinate range of the ultrasound contrast dual-mode dynamic images is clearly defined as x∈[0,W / 2) and y∈[0,H), achieving complete alignment of the dual-mode coordinate system and eliminating coordinate ambiguity.

[0028] 3. Global coordinate normalization mapping: The detection center, predicted position, and region of interest of the target in the anatomical modal image I1 (ultrasound B-mode image) and the contrast modal image I2 (ultrasound contrast image) are directly represented by the above-mentioned unified coordinate system, without using the global offset coordinates of the stitched image. This ensures that the coordinates of the state vector, observation vector, predicted value, and updated value of the two Kalman filters are completely unified, providing a unified coordinate reference for subsequent dual-path filtering, noise estimation, and cross-modal interaction.

[0029] In another exemplary embodiment of this application, step S3 described above can be implemented by the following method: In this embodiment, the basic parameters of the state model and Kalman filter are set: after completing the unified processing of image coordinates, a unified multi-dimensional state vector is defined for target tracking of dual-mode ultrasound contrast imaging. X This vector encompasses the core feature parameters of the dual-mode target, including key information such as target position, velocity, and deformation degree, and can comprehensively describe the dynamic changes of the target. Considering the inherent characteristics of dual-mode dynamic ultrasound imaging, an independent Kalman filter is configured for each mode (two filters in total, corresponding to the anatomical modality image I1 (ultrasound B-mode image) and the contrast modality image I2 (ultrasound contrast image) respectively). Both filters use the same state transition matrix. A With observation matrix H This ensures consistency in dual-mode tracking, and also configures an initial process noise covariance matrix for each filter. Q 0 With the initial observation noise covariance matrix R 0 This serves as the initial input for subsequent adaptive updates. Unlike the traditional method of using a single Kalman filter to process dual modes, the two filters in this embodiment are independent of each other and can exchange information, which can not only give full play to the inherent advantages of each mode, but also provide support for cross-modal cooperative tracking.

[0030] In another exemplary embodiment of this application, step S4 above can be implemented by the following method: In this embodiment, the state update process of the dual Kalman filter is as follows: each of the two Kalman filters independently executes the state update process, based on the posterior state output from the previous frame. With posterior covariance (in m =1,2, m =1 corresponds to the anatomical modality image I1 (ultrasound B-mode image) filter. m =2 corresponds to the contrast modal image I2 (ultrasound contrast image) filter), to complete the state prediction and covariance prediction of this frame. The specific calculation process is as follows.

[0031] 1. State prediction: ,in, m =1,2, m =1 corresponds to the I1 filter in the anatomical modality image. m =2 corresponds to the I2 filter in the contrast modality image. For this frame (the first) k The state prediction vector of the frame. A Here is the state transition matrix. For the previous frame (the first frame) k The prediction process uses the posterior state of the target (frame -1) as a basis to provide a preliminary estimate of the target state in the current frame, based on the dynamic change pattern of the target.

[0032] 2. Covariance prediction: ,in This is the covariance prediction vector for the current frame (the k-th frame). For the previous frame (the first frame) k -1 frame) posterior covariance matrix, For the previous frame (the first frame) k The process noise covariance matrix obtained by adaptive estimation (-1 frame) is used to quantify the uncertainty of state prediction.

[0033] During the state update process described above, the two filters perform the calculations independently to ensure that each mode can complete the prediction for the current frame based on its tracking results from the previous frame.

[0034] In another exemplary embodiment of this application, step S5 is specifically implemented by the following steps: S51. Within the state prediction region of each Kalman filter, a feature extraction algorithm is used to extract the observation vectors to obtain dual-mode observation vectors; the dual-mode observation vectors include: anatomical mode observation vectors and contrast mode observation vectors.

[0035] S52, Using the formula Calculate the new information; among which, Represents the state prediction vector. Let H represent the observation vector, and let H represent the observation matrix.

[0036] S53. Use formula ( m =1,2) Calculate the state prediction residuals; where, A Here is the state transition matrix. This is the posterior state of the previous frame.

[0037] In another exemplary embodiment of this application, step S6 is specifically implemented by the following steps: S61. Set the observation noise forgetting matrix, process noise forgetting matrix, and separate least squares forgetting factor for each Kalman filter.

[0038] S62. Based on the new information and the observation noise forgetting matrix, recursively estimate the new information covariance matrix. The state prediction residual covariance matrix is ​​recursively estimated based on the state prediction residual and the process noise forgetting matrix. The recursive formula is: ; ;in, , These are the information covariance and the state prediction residual covariance obtained from the recursion of the previous frame, respectively. To observe the noise forgetting matrix; For new information; This is the process noise forgetting matrix; This is the residual for state prediction.

[0039] S63. Construct a covariance matrix including process noise. Covariance matrix of observation noise Joint vector to be estimated And construct a linear observation equation; where, vec (·) represents the matrix column vectorization operation.

[0040] S64. Using the least squares forgetting factor, the weighting matrix and covariance matrix are updated synchronously according to the recursive least squares rule. The process noise covariance matrix and observation noise covariance matrix of this frame are extracted from the updated weighting matrix.

[0041] S65. Apply positive semidefinite constraint correction to the process noise covariance matrix and the observation noise covariance matrix to ensure their positive semidefiniteness.

[0042] In another exemplary embodiment of this application, steps S5-S6 are specifically implemented by the following steps of observation vector extraction and Q / R synchronous adaptive estimation: After the two Kalman filters complete the state update, the observation vectors of the dual-mode dynamic ultrasound contrast imaging are extracted based on a unified coordinate system. Each filter independently performs synchronous estimation of the process noise covariance Q and the observation noise covariance R. The specific algorithm flow is as follows: 1. Observation Vector Extraction. For dual-mode dynamic ultrasound imaging (anatomical modality I1 (ultrasound B-mode image) and contrast-enhanced modality I2 (ultrasound contrast-enhanced image)), lightweight feature extraction algorithms, such as deep convolutional neural networks, are used to extract the observation vectors of the two modalities within the state prediction region of each Kalman filter. (anatomical modal observation vector) (Animation modal observation vector); The observation vector extracted by the algorithm can reflect the target state, and since the coordinates have been unified, the coordinate references of the two observation vectors are completely consistent, which meets the input requirements of the linear observation model.

[0043] 2. Residual calculation. Based on the state prediction vector of each path. Observation vector Given the observation matrix H, calculate two types of residuals: one is the innovation. The residual reflects the deviation between the observed and predicted values, indicating the degree of interference from observation noise; secondly, the state prediction residual... (m=1,2), the residual reflects the accuracy of the state prediction and the degree of interference from process noise; by calculating the two types of residuals, the interference of noise on the tracking process is captured, providing input for the subsequent estimation of Q and R matrices.

[0044] 3. Forgetting Factor Configuration. Each Kalman filter has an independently configured matrix forgetting factor with different dimensions to adjust the update weights of components in different dimensions, balancing tracking response speed and stability; among which, the observation noise forgetting matrix... With process noise forgetting matrix All are diagonal matrices, and the elements on the diagonal are... , All values ​​are taken in (0,1]. Elements of different dimensions can be set with different values ​​according to modal characteristics to achieve asynchronous adaptive updates. At the same time, to further optimize the dynamic adaptation capability of least squares estimation, a separate least squares forgetting factor is introduced. This forgetting factor is independent of and Similarly, let's set it as a diagonal matrix, with its diagonal elements... The value is set to (0,1], used to individually adjust the update rate of the least squares estimation process to adapt to the dynamic changes in the estimation of the Q and R covariance matrices; the global joint forgetting matrix... ,in The Kronecker product, the global forgetting matrix, is used for collaborative updates of the dual-path Kalman filter, while the least-squares forgetting factor... It can be applied independently to the least squares estimation process to achieve independent adaptive adjustment of filter update and least squares estimation, ensuring the stability and adaptability of the estimation process.

[0045] 4. Covariance recursive estimation. Each Kalman filter calculates its own innovation. With state prediction residual Recursively estimate the new information covariance matrix With state prediction residual covariance matrix The specific recursive formula is as follows: ; ,in , These are the information covariance and the state prediction residual covariance obtained from the previous frame, respectively. Through this recursive method, the covariance matrix is ​​updated in real time to adapt to the real-time changes in noise.

[0046] 5. Joint Q / R Solution. A linear observation model with joint Q and R estimation is constructed, and the vector to be estimated is solved based on the least squares criterion. The specific process is as follows: and By column vectorization, the joint observation vector is obtained. (where vec(·) is a matrix column vectorization operation); constructing linear observation equations ,in Let be the vector of parameters to be estimated. The observation matrix (by) , and covariance matrix , Construction, specifically , , They are respectively with , (identity matrix of the same dimension) The error of the recursive least squares solution follows a Gaussian distribution with a mean of 0.

[0047] 6. Recursive Least Squares Synchronous Update: Following the recursive least squares rule based on the matrix forgetting factor, combined with the separately introduced least squares forgetting factor. Simultaneously with the Kalman filter, the weighting matrix and covariance matrix are updated synchronously. The specific update formula is as follows: (1) Calculate the recursive least squares gain matrix: ,in for (Right now k The state error covariance matrix (the parameter vector estimated at time -1), for (Right now k The observation matrix of the parameter vector to be estimated at time step (time step). The least squares forgetting factor is introduced separately. Its inverse matrix is ​​used to individually adjust the update rate of recursive least squares, adapting to the dynamic changes in the estimation of Q and R covariance matrices.

[0048] (2) Update the weighted matrix: ,in The parameter vector estimated at time k-1. for k The least-squares gain matrix at time step 1. Let be the vector to be estimated.

[0049] (3) Update the covariance matrix: Where I is the same as Identity matrices of the same dimension, using the least squares forgetting factor Its dynamic adjustment function completes the dynamic recursion of the covariance matrix and adapts to the real-time changes in noise.

[0050] (4) Extract Q and R estimates: The updated weighted matrix Split by column to obtain the following: ( (first half of the column vector) ( The latter half of the column vector), where This is the inverse operation of matrix column vectorization, which completes the synchronous estimation of Q and R at the same time as the Kalman filter.

[0051] 7. Constraint Correction: Adjustments to the simultaneous estimation results... and To apply semi-positive definite constraints, the eigenvalue decomposition method is used, and the specific correction formula is as follows: , (where U and V are orthogonal matrices,) , (As a non-negative diagonal matrix), ensuring , (Positive semidefinite) to avoid Kalman filter divergence caused by the non-positive semidefinite noise covariance matrix.

[0052] In another exemplary embodiment of this application, step S7 described above can be implemented by the following method: First measurement update: Each Kalman filter independently performs a measurement update based on its own current frame state prediction vector, covariance prediction matrix, observation vector, observation noise covariance matrix, and process noise covariance matrix to obtain the first posterior state and the first posterior covariance. Second measurement update: The two Kalman filters exchange observation vectors and observation noise covariance matrices. The first Kalman filter calls the observation vectors and observation noise covariance matrices of the second Kalman filter, and the second Kalman filter calls the observation vectors and observation noise covariance matrices of the first Kalman filter. Each of them performs a second measurement update based on the first posterior state and the first posterior covariance to obtain the second posterior state and the second posterior covariance.

[0053] In another exemplary embodiment of this application, step S7 is specifically implemented by the following steps of two-time measurement updates and cross-modal information interaction: To address the issues of lack of cross-modal information interaction and collaboration mechanisms and low information utilization in existing dual-mode tracking technologies, this embodiment adopts a dual measurement update strategy, combining independent updates for its own mode with cross-modal collaborative updates. The specific execution steps are as follows.

[0054] 1. Initial measurement update (independent update of its own mode). The two Kalman filters are based on their own estimated observation noise covariance. Covariance of process noise Combined with its own observation vector With state prediction vector It independently performs measurement updates and calculates its own posterior state. With posterior covariance The specific formula is as follows: Kalman gain: ; Posterior state: ; Posterior covariance: ;in (m=1,2) represents the values ​​obtained through the initial measurement update. k Kalman gain at time step.

[0055] This step is used by each mode to complete the initial state correction using its own observation information.

[0056] 2. Secondary measurement update (cross-modal collaborative update). After completing the initial self-modal update, the two filters exchange observation information, i.e., the first filter ( m =1, corresponding to the anatomical modality B mode image) calls the observation vector of the second path. Covariance of observation noise The second filter ( m =2, corresponding to the angiography modality image) calls the first path observation vector. Covariance of observation noise At the same time, retain their respective post-update states. With posterior covariance Perform a secondary measurement update, the specific formula is as follows: Kalman gain: (in m (Indicates another mode); Posterior state: ; Posterior covariance: ;in, For another filter k The observation noise matrix at time t, Updated after secondary measurement k Kalman gain at time step For another filter k The observation vector at time t.

[0057] Secondary cross-modal updates are used for dual-mode information collaboration to compensate for the limitations of single-mode observation.

[0058] In another exemplary embodiment of this application, step S8 described above can be implemented by the following method: S81, Secondary Posterior Covariance Based on Two-Way Kalman Filter and According to the formula ( m =1,2) Calculate the fusion weights and , where tr(·) is the matrix trace operation.

[0059] S82. Based on the fusion weights, the second posterior state of the two Kalman filters... and By performing weighted fusion, the globally optimal objective state is obtained. And calculate the global fusion covariance. ;in, ; In the formula, This represents the posterior state of the first Kalman filter. This represents the posterior state of the second Kalman filter. Let be the posterior covariance of the first Kalman filter. Let be the posterior covariance of the second Kalman filter.

[0060] S83. Feed the global optimal target state and the global fusion covariance back to the two Kalman filters as input for the state prediction of the next frame.

[0061] In another exemplary embodiment of this application, step S8 is specifically implemented by the following state fusion and closed-loop feedback steps: To achieve global optimization in dual-mode tracking, this embodiment employs a state covariance weighted fusion strategy, combined with a closed-loop feedback mechanism, to ensure the continuity and stability of the tracking process. The specific implementation process is as follows: 1. Weight Calculation: Based on the posterior covariance after the second update of the two-channel Kalman filter. , Calculate the fusion weights of the two modes. , The specific calculation formula is as follows: ( m =1,2), where tr (·) represents the trace operation of the matrix; the smaller the trace of the covariance matrix, the more accurate the state estimation, and the larger the corresponding weight, which can ensure that the fusion result is tilted towards a more accurate mode.

[0062] 2. State Fusion: Based on the calculated fusion weights, the posterior state of the two filters is updated twice. , By performing weighted fusion, the globally optimal objective state is obtained. This fusion state integrates the advantages of both modes, avoiding the deviations of single-mode tracking and improving the accuracy of target tracking.

[0063] 3. Closed-loop feedback: The fused globally optimal state Global fusion covariance The data is synchronously fed back to the two Kalman filters, serving as the initial state for the next frame's state update of both filters. with initial covariance ( m =1,2, m =1 corresponds to the anatomical modality B mode image. m =2 corresponds to the contrast imaging modality), forming a closed-loop tracking mechanism of "coordinate unification → prediction → estimation → update → fusion → feedback", ensuring the consistency and continuity of dynamic image tracking of dual-mode ultrasound contrast imaging, avoiding tracking jumps, and improving tracking stability and accuracy.

[0064] Finally, the iterative tracing process is executed: The entire process of "image coordinate unification → dual-path Kalman state update → observation vector extraction and Q / R synchronous estimation → dual measurement update and cross-modal interaction → state fusion and closed-loop feedback" is executed sequentially frame by frame to form a continuous iterative tracking process. At the same time, the target tracking deviation is monitored in real time for each frame. Based on the dynamic relationship between the information covariance matrix and the state prediction residual covariance matrix, the value of the matrix forgetting factor is dynamically adjusted to optimize the accuracy of Q and R synchronous estimation, correct the tracking deviation, and ensure that the entire tracking process is stable and reliable, adapting to the needs of dual-mode dynamic image tracking in complex clinical application scenarios.

[0065] In another exemplary embodiment of this application, such as Figure 3 As shown, to further illustrate the implementation and feasibility of this application, a specific embodiment is provided for liver lesion tracking in a clinical ultrasound contrast-enhanced imaging dual-mode dynamic imaging scenario (left anatomical modal B-mode image + right contrast-enhanced modal image, a dual-mode ultrasound contrast-enhanced image sequence with a resolution of 1024×768), detailing the implementation process and parameter settings of this application. The left anatomical modal B-mode image is used to capture the location and morphological contour of the lesion, while the right contrast-enhanced modal image is used to present the blood supply characteristics of the lesion. The specific implementation steps for lesion tracking based on the technical solution of this application are as follows.

[0066] Step 1. Image coordinate processing: To ensure that the state vector coordinates of the two Kalman filters are completely consistent and to eliminate positional offsets and coordinate ambiguities in the dual-mode dynamic images of ultrasound contrast imaging, this embodiment performs centered segmentation and global coordinate mapping on the dual-mode dynamic images of ultrasound contrast imaging. The specific steps are as follows: 1) Vertically centered image segmentation: The contrast-enhanced ultrasound dual-mode dynamic stitched image adopts the clinically commonly used resolution, specifically set to W=1024 pixels and H=768 pixels (i.e., resolution 1024×768). This 1024×768 contrast-enhanced ultrasound dual-mode dynamic stitched image is evenly divided into two equal-sized regions along the vertical midline. The left region is the anatomical modality image I1 (anatomical modality is B mode image), and the right region is the contrast-enhanced modality image I2 (contrast-enhanced modality is contrast image). The size of each region is (512×768) pixels, realizing the physical separation and correspondence of the contrast-enhanced ultrasound dual-mode dynamic image.

[0067] 2) Establishment of a unified global coordinate system: Both the left and right images adopt a unified local coordinate system, with the upper left corner of their respective regions as the origin (0,0), the horizontal axis as the x-axis, and the vertical axis as the y-axis; the coordinate range of the anatomical modality I1 (B-mode image) and the contrast modality I2 (contrast image) is unified as: x∈[0,512), y∈[0,768), achieving complete alignment of the dual-mode coordinate system.

[0068] 3) Global coordinate normalization mapping: The detection center, predicted position, and ROI region of the target in the anatomical modality (B mode image) and the contrast modality (contrast image) are directly represented by the above unified coordinate system, without using the global offset coordinate of the stitched image, to ensure that the coordinates of the state vector, observation vector, predicted value, and update value are completely unified.

[0069] Step 2. Parameter initialization: 1) Definition of State Vector: Define a 4-dimensional state vector. Where x and y are the pixel coordinates of the lesion center in the dual-mode dynamic ultrasound imaging. , The velocity of the lesion in the x and y directions (unit: pixels / frame) can comprehensively describe the location and motion state of the lesion.

[0070] 2) Matrix parameter configuration: State transition matrix Where T is the ultrasound image frame interval (set according to the frame rate of the clinical ultrasound equipment; in this embodiment, T = 1 / 30 frames / second); observation matrix It only observes the coordinates of the lesion center, which aligns with the actual clinical tracking needs.

[0071] 3) Initial covariance setting: Initial process noise covariance matrix Initial observation noise covariance matrix Adapt to the initial noise level.

[0072] 4) Forgetting factor configuration: Observation noise forgetting matrix Process noise forgetting matrix Global Joint Forgetting Matrix Simultaneously, a separate least-squares forgetting factor is introduced. (A 20-dimensional diagonal matrix, and a joint vector to be estimated) θ (Dimensional consistency), used to individually adjust the update rate of recursive least squares.

[0073] 5) Initialization of recursive least squares parameters: weighting matrix W 0 A 20×1 zero matrix ( Q It is a 4×4 matrix, which becomes 16-dimensional after column vectorization; R It is a 2×2 matrix, which becomes 4-dimensional after column vectorization, and the vector to be estimated is combined. θ (20-dimensional), covariance matrix ( (It is a 20×20 identity matrix) to meet the initial estimation requirements.

[0074] Step 3. Dual-path Kalman modeling and state update: 1) Dual-path filter modeling: m =1 corresponds to the left anatomical modality B mode image filter, which focuses on suppressing image noise and capturing the contour position of lesions; m =2 corresponds to the right-side angiography modal image filter, which focuses on adapting to contrast agent perfusion signal fluctuations and tracking the lesion's blood supply area; both filters use the same state transition matrix. A With observation matrix H To maintain tracking consistency, initial covariance parameters are configured independently to adapt to the noise characteristics of each mode.

[0075] 2) State update execution: The two filters are based on the fusion state fed back from the previous frame. Global fusion covariance Execution status update: m =1 (Anatomy Modal B Mode Image): State Prediction Covariance prediction ; m =2 (contrast modal images): State prediction Covariance prediction ; in , This is the process noise covariance matrix obtained from the synchronous estimation of the previous frame.

[0076] Step 4. Observation vector extraction and Q / R synchronous adaptive estimation: 1) Observation Vector Extraction: Within the state prediction region of the two filters, a lightweight deep convolutional neural network is used to extract the observation vectors of the left anatomical modality B-mode image. With the observation vector of the right-side contrast modality contrast image .

[0077] 2) Residual calculation: When m When =1,2, calculate the new information respectively. With state prediction residual .

[0078] 3) Covariance Recursion: Recursively estimating the new information covariance State prediction residual covariance .

[0079] 4) Simultaneous Q / R solution and recursive least squares update: Constructing a joint vector to be estimated Construct the observation matrix Simultaneously with the Kalman filter, a recursive least squares update is performed to obtain... and After eigenvalue decomposition correction, ensure , .

[0080] 5) Output Results: m When =1, It is a 4×4 positive semi-definite matrix, adapted to the noise in the process of anatomical modality B-mode images; It is a 2×2 positive semi-definite matrix, adapted to the observation noise of anatomical modality B mode images; m When =2, It is a 4×4 positive semi-definite matrix, which is adapted to the noise in the contrast imaging process of contrast modal images; It is a 2×2 positive semi-definite matrix, adapted to the observation noise of contrast modal contrast images.

[0081] Step 5. Two-time measurement update and cross-modal interaction: 1) Initial Measurement Update: The two filters utilize their own... , and Perform the first measurement update and obtain , (Predicted values ​​corresponding to anatomical modality B image) and , (Predicted value corresponding to the contrast modality image).

[0082] 2) Cross-modal information interaction: m =1 (Anatomy Modality B Mode Image) Call and , m =2 (contrast modality contrast image) call and Perform a second measurement update to obtain , and , .

[0083] Step 6. State fusion and closed-loop feedback: 1) Weight calculation: , ; 2) State fusion: ; 3) Covariance fusion: ; 4) Closed-loop feedback: and These serve as the initial states for the next frame of the two filters, respectively. with initial covariance ( m =1 corresponds to the anatomical modality B mode image. m =2 corresponds to the angiography modality image), and then proceed to the next frame tracking process.

[0084] 7. Iterative tracing: Steps 2-6 above are executed in a loop, with real-time updates for each frame. , With fusion state This enables the tracking of lesion areas in dual-mode dynamic ultrasound imaging.

[0085] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 4 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and databases. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a target tracking method suitable for dual-mode dynamic ultrasound imaging.

[0086] Those skilled in the art will understand that Figure 4The structures shown are merely block diagrams of some structures related to the present application and do not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0087] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0088] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0089] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Moreover, the collection, use and processing of the relevant data are carried out in compliance with the relevant data protection laws and policies of the country where the location is located, and with the authorization granted by the owner of the corresponding device.

[0090] 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 computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0091] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0092] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0093] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A target tracking method suitable for dual-mode dynamic ultrasound contrast imaging, characterized in that, include: Acquire dual-mode dynamic ultrasound contrast imaging; The ultrasound contrast imaging dual-mode dynamic images are segmented into anatomical modal images and contrast modal images, and the image coordinates are unified to establish a globally unified coordinate system. Independent Kalman filters are configured for the anatomical modality images and the contrast modality images respectively, to construct a dual-path Kalman filter; Based on the dual-path Kalman filter, state prediction is performed using the posterior state and posterior covariance obtained from the previous frame fusion to obtain the state prediction vector and covariance prediction matrix for the current frame. Based on the current frame state prediction vector and covariance prediction matrix, extract the dual-mode observation vector and calculate the innovation and state prediction residual; Based on the aforementioned information and state prediction residuals, the process noise covariance matrix and observation noise covariance matrix are obtained simultaneously by using the joint regression of information covariance and the recursive least squares method of matrix forgetting factor, and by constraint correction. Based on the process noise covariance matrix, the observation noise covariance matrix and the dual-mode observation vector, the dual-path Kalman filter is subjected to two measurement updates and cross-modal information interaction to obtain the second posterior state and the second posterior covariance. Based on the second posterior state and the second posterior covariance, state weighted fusion and closed-loop feedback are performed to obtain the globally optimal target state and the global fusion covariance, which are then used as the input for the state prediction of the next frame.

2. The target tracking method for dual-mode dynamic ultrasound contrast imaging according to claim 1, characterized in that, The ultrasound contrast imaging dual-mode dynamic images are segmented into anatomical modal images and contrast modal images, and image coordinate unification processing is performed to establish a globally unified coordinate system, specifically including: The dynamic ultrasound contrast imaging dual-mode image with a size of W×H was evenly divided into two equal regions along the vertical midline. The left region was used as the anatomical modality image I1, and the right region was used as the contrast imaging modality image I2. The size of each region was (W / 2)×H. Establish a local coordinate system with the upper left corner of each region as the origin, the horizontal axis as the x-axis, and the vertical axis as the y-axis, so that the coordinate range of the left and right images is unified as x∈[0,W / 2) and y∈[0,H).

3. The target tracking method for dual-mode dynamic ultrasound contrast imaging according to claim 1, characterized in that, The formula for calculating state prediction is: ； in, m =1,2, m =1 corresponds to the I1 filter in the anatomical modality image. m =2 corresponds to the I2 filter in the contrast modality image. This is the state prediction vector for this frame. A Here is the state transition matrix. This represents the posterior state of the previous frame; ； in, This is the covariance prediction matrix for this frame. The posterior covariance of the previous frame. This is the process noise covariance matrix obtained from the adaptive estimation of the previous frame.

4. The target tracking method for dual-mode dynamic ultrasound contrast imaging according to claim 1, characterized in that, Based on the current frame state prediction vector and covariance prediction matrix, the dual-mode observation vector is extracted and the innovation and state prediction residual are calculated, specifically including: Within the state prediction region of each Kalman filter, a feature extraction algorithm is used to extract the observation vectors to obtain dual-mode observation vectors; the dual-mode observation vectors include: anatomical mode observation vectors and contrast mode observation vectors; Using formula Calculate the new information; among which, Represents the state prediction vector. Represents the observation vector. H Represents the observation matrix; Using formula ( m= 1 , 2) Calculate the state prediction residuals; where, A Here is the state transition matrix. This is the posterior state of the previous frame.

5. The target tracking method for dual-mode dynamic ultrasound contrast imaging according to claim 1, characterized in that, Based on the aforementioned innovation and state prediction residuals, a joint regression of innovation covariance and a recursive least squares method using the matrix forgetting factor are employed to simultaneously solve for and constrain the semi-definite process noise covariance matrix and observation noise covariance matrix, specifically including: For each Kalman filter, an independent observation noise forgetting matrix, a process noise forgetting matrix, and a separate least-squares forgetting factor are set. Based on the new information and the observation noise forgetting matrix, the new information covariance matrix is ​​recursively estimated. The state prediction residual covariance matrix is ​​recursively estimated based on the state prediction residual and the process noise forgetting matrix. The recursive formula is: ; ;in, , These are the information covariance and the state prediction residual covariance obtained from the recursion of the previous frame, respectively. To observe the noise forgetting matrix; For new information; This is the process noise forgetting matrix; For state prediction residuals; Construct a covariance matrix including process noise Covariance matrix of observation noise Joint vector to be estimated And construct a linear observation equation; where, vec (·) represents the matrix column vectorization operation; Using the least squares forgetting factor, the weighting matrix and covariance matrix are updated synchronously according to the recursive least squares rule. The process noise covariance matrix and observation noise covariance matrix of this frame are extracted from the updated weighting matrix. The process noise covariance matrix and the observation noise covariance matrix are corrected by a positive semidefinite constraint to ensure their positive semidefiniteness.

6. The target tracking method for dual-mode dynamic ultrasound contrast imaging according to claim 1, characterized in that, Based on the process noise covariance matrix, the observation noise covariance matrix, and the dual-mode observation vector, a dual-path Kalman filter is subjected to two-stage measurement updates and cross-modal information interaction to obtain the second-order posterior state and the second-order posterior covariance, specifically including: Initial measurement update: Each Kalman filter independently performs measurement update based on its own current frame state prediction vector, covariance prediction matrix, observation vector, observation noise covariance matrix, and process noise covariance matrix to obtain the first posterior state and the first posterior covariance. Secondary measurement update: The two Kalman filters exchange the observation vector and the observation noise covariance matrix. The first Kalman filter calls the observation vector and the observation noise covariance matrix of the second Kalman filter, and the second Kalman filter calls the observation vector and the observation noise covariance matrix of the first Kalman filter. Each of them performs secondary measurement update based on the first posterior state and the first posterior covariance to obtain the second posterior state and the second posterior covariance.

7. The target tracking method for dual-mode dynamic ultrasound contrast imaging according to claim 1, characterized in that, Based on the aforementioned second-order posterior state and second-order posterior covariance, state-weighted fusion and closed-loop feedback are performed to obtain the globally optimal target state and the global fusion covariance, which are then used as inputs for the state prediction of the next frame. Specifically, this includes: The second-order posterior covariance based on a two-channel Kalman filter and According to the formula ( m =1,2) Calculate the fusion weights and ,in tr (·) represents the matrix trace operation; Based on the fusion weights, the second-order posterior states of the two Kalman filters are... and By performing weighted fusion, the globally optimal objective state is obtained. And calculate the global fusion covariance. ;in, ; In the formula, This represents the posterior state of the first Kalman filter. This represents the posterior state of the second Kalman filter. Let be the posterior covariance of the first Kalman filter. Let be the posterior covariance of the second Kalman filter; The global optimal target state and the global fusion covariance are fed back to two Kalman filters as inputs for the state prediction of the next frame.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that the processor executes the computer program to implement the target tracking method for dual-mode dynamic imaging of ultrasound contrast imaging as described in any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the target tracking method for dual-mode dynamic imaging of ultrasound contrast imaging as described in any one of claims 1-7.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the target tracking method for dual-mode dynamic imaging of ultrasound contrast imaging as described in any one of claims 1-7.

Images