Vte dynamic early warning method and system based on anatomical hemodynamics coupling

Abstract

The present application belongs to the technical field of anatomical hemodynamic data processing, and relates to a VTE dynamic early warning method and system based on anatomical hemodynamic coupling; the present application synchronously collects ultrasonic structure images and Doppler blood flow signals, extracts time-varying vorticity features and key anatomical constraint sets, generates an expected blood flow atlas by using a fluid dynamics simulation model, compares in real time to monitor continuous deviation events, and traces back to analyze to generate root cause candidate points; the system further constructs a dynamic risk field and a dynamic root cause field, which are updated in correlation to form a dynamic coupling field, realizing continuous and individualized tracking of the health status of blood vessels; finally, according to the dynamic coupling field situation, graded early warning data containing early warning state levels, risk visualization and traceability prompts are output; the present application improves the timeliness and diagnosis direction of early warning, and provides precise and dynamic risk assessment and decision support for clinical use.

Description

Technical Field

[0001] This invention belongs to the field of anatomical hemodynamic data processing technology, and relates to a dynamic early warning method and system for VTE based on anatomical hemodynamic coupling. Background Technology

[0002] In the field of medical diagnostics, particularly for the diagnosis of vascular diseases, there has long been a focus on obtaining internal information about the human body through non-invasive or minimally invasive methods to aid clinical decision-making. Ultrasound imaging, a widely used medical imaging technique, can visually display the anatomical structure of blood vessels, such as their diameter, wall thickness, and the presence of abnormal wall structures. Simultaneously, Doppler ultrasound technology can detect hemodynamic information such as blood flow velocity, direction, and flow patterns. These technologies play a crucial role in clinical practice, forming the cornerstone of vascular disease diagnosis. For diseases such as venous thromboembolism (VTE), ultrasound examination is a commonly used initial screening and diagnostic tool.

[0003] However, existing diagnostic methods often treat the vascular anatomy shown in ultrasound images and the hemodynamic parameters provided by Doppler flow signals as relatively independent pieces of information. Clinically, physicians typically need to interpret ultrasound images and Doppler spectrograms separately, relying on their personal experience and knowledge to perform correlation analysis to determine the presence and nature of lesions. This separate processing model usually involves judging fixed thresholds and manually interpreting static images or short-time-segment signals, lacking modeling and analysis of the dynamic, real-time, deep coupling relationship between vascular structure and function. For example, focusing solely on changes in vessel diameter or a decrease in Doppler flow velocity in ultrasound images often fails to comprehensively assess the complexity of vascular diseases.

[0004] Existing technologies have limitations in the following aspects: First, because anatomical structure and hemodynamic information are assessed separately, it is difficult to accurately reveal the dynamic causal chain of "how structural abnormalities affect function in real time" and "how functional abnormalities inversely indicate structural lesions." This leads to a lack of depth and completeness in the assessment results, and insufficient identification of early, subtle changes in vascular diseases. Second, current diagnoses rely heavily on single-point-in-time data snapshots or short-term observations, failing to effectively capture the continuous evolution of vascular status and the interactions and cascade effects between segments throughout the vascular network. This makes risk warnings typically passive, initiating diagnosis only when the lesion is relatively clear or symptoms have appeared, rather than proactively and prospectively predicting risk. Therefore, existing technologies are insufficient in providing early, accurate, dynamic, and personalized warning information, leading to sometimes delayed or imprecise clinical decisions. Summary of the Invention

[0005] In view of this, in order to solve the problems mentioned in the background technology, a dynamic early warning method and system for VTE based on anatomical hemodynamic coupling is proposed.

[0006] The objective of this invention can be achieved through the following technical solution: The first aspect of this invention provides a dynamic early warning method for VTE based on anatomical hemodynamic coupling, including: S1, acquiring a pre-stored sequence of ultrasound structural images and a sequence of Doppler blood flow signals synchronously acquired from a target vascular region.

[0007] S2. Extract time-varying vorticity features that characterize early signs of local flow instability from Doppler blood flow signal sequences, and extract key anatomical constraint sets from ultrasound structural image sequences. The key anatomical constraint sets include geometric contour data of the target blood vessel, dynamic parameters of vessel diameter change, venous valve motion trajectory, and physical property indicators of vessel wall.

[0008] S3. Based on the key anatomical constraint set, generate the expected blood flow map of each segment in the corresponding vascular region by running the fluid dynamics simulation model. The expected blood flow map includes the expected time-varying vortex range.

[0009] S4. The real-time extracted time-varying vorticity features are compared with the corresponding expected time-varying vorticity range in the expected blood flow map. When a continuous deviation event is detected, the source analysis of the key anatomical constraint set is triggered to generate at least one root cause candidate point.

[0010] S5. A dynamic risk field is generated based on the intensity and distribution of the continuous deviation events, and a dynamic root cause field is generated based on the results of the source analysis and the feedback verification of the fluid dynamics simulation model. The dynamic risk field and the dynamic root cause field are interconnected and dynamically updated, forming a dynamic coupling field.

[0011] S6. Based on the current state of the dynamic coupling field, output hierarchical early warning data including early warning status level, risk visualization information, and source tracing prompts.

[0012] A second aspect of the present invention provides a VTE dynamic early warning system based on anatomical hemodynamic coupling, comprising: an image and signal sequence acquisition module for acquiring a pre-stored ultrasound structural image sequence and Doppler blood flow signal sequence synchronously acquired from a target vascular region.

[0013] The feature and constraint set extraction module extracts time-varying vorticity features that characterize early signs of local flow instability from Doppler blood flow signal sequences, and extracts key anatomical constraint sets from ultrasound structural image sequences. The key anatomical constraint sets include geometric contour data of the target blood vessel, dynamic parameters of vessel diameter change, venous valve motion trajectory, and physical property indicators of the vessel wall.

[0014] The expected blood flow map generation module, based on the key anatomical constraint set, generates expected blood flow maps for each segment within the corresponding vascular region by running a fluid dynamics simulation model. The expected blood flow map includes the expected time-varying vorticity range.

[0015] The root cause candidate point generation module compares the real-time extracted time-varying vorticity features with the corresponding expected time-varying vorticity range in the expected blood flow map in real time. When a continuous deviation event is detected, it triggers the source analysis of the key anatomical constraint set and generates at least one root cause candidate point.

[0016] The dynamic coupling field module generates a dynamic risk field based on the intensity and distribution of continuous deviation events, and generates a dynamic root cause field based on the results of source analysis and feedback verification from the fluid dynamics simulation model. The dynamic risk field and the dynamic root cause field are interconnected and dynamically updated, forming a dynamic coupling field.

[0017] The graded early warning data output module outputs graded early warning data, including early warning status level, risk visualization information, and source tracing prompts, based on the current situation of the dynamic coupled field.

[0018] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: (1) The present invention realizes the transformation of vascular embolism risk warning from traditional passive response to active detection by introducing a bidirectional mutual driving coupling mechanism. This method can capture the weak early signals manifested by local flow turbulence before the blood flow velocity changes, which helps to detect the time window of potential embolism risk in advance. By closely linking real-time blood flow function abnormalities with individualized anatomical structural changes, the system no longer just reports the existence of abnormalities, but can directly point to the specific anatomical root causes of these abnormalities, improving the timeliness of warning and the directionality of diagnosis.

[0019] (2) This invention constructs an adaptive and self-learning early warning system through the design of a dynamic coupling field, realizing continuous and individualized tracking of vascular health status. The dynamic coupling field integrates the current manifestations of risk, potential root causes, and future evolution trends within a unified framework, and can be corrected and updated in real time according to actual changes in vascular status. This dynamism and personalization enable the system to provide tailored risk assessment and early warning information for each patient's unique vascular pathological progression, improving the precision and reliability of early warning information and surpassing the limitations of traditional static assessment methods in dealing with complex and dynamic disease development.

[0020] (3) This invention improves the objectivity of early warning data through a two-way driving and feedback verification mechanism of risk field and root cause field. Blood flow abnormalities can guide the system to deeply trace potential anatomical lesions, while the verification of anatomical lesions, in turn, optimizes the hemodynamic prediction model. This mutually verifying closed loop makes the system's judgment logic more rigorous, and the early warning data is based on a solid physical model and real-time data support. It avoids the misjudgment and uncertainty that may be caused by a single data source or linear analysis, and ensures that the final output graded early warning information is not only a simple numerical value of risk, but also a comprehensive situational judgment that includes risk intensity, location and cause, providing clinicians with highly operational decision support. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram of the method steps of the present invention.

[0023] Figure 2 This is a schematic diagram of the system structure connection of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] The "early warning status level" and "graded early warning data" output by the method and system of this invention are essentially physical situation characteristic parameters calculated based on the anatomy-flow field coupling model, reflecting the evolutionary state of the hemodynamic model. The above outputs serve only as intermediate reference data for computer system auxiliary processing and do not directly indicate a qualitative diagnosis of the disease, nor do they include clinical treatment recommendations.

[0026] Please see Figure 1 The first aspect of the present invention provides a dynamic early warning method for VTE based on anatomical hemodynamic coupling, comprising: S1, acquiring a pre-stored sequence of ultrasound structural images and a sequence of Doppler blood flow signals synchronously acquired from a target vascular region.

[0027] It should be noted that by using ultrasound dual-modal synchronous scanning technology, the anatomical morphology sequence and real-time hemodynamic signals of the target blood vessel are continuously acquired under a unified time reference and spatial dimension. This breaks the limitation of the separation of structure and function in traditional diagnosis, and constructs a spatiotemporally aligned "structure-flow field" dual-dimensional data source. This lays the physical foundation for subsequent blood flow simulation driven by anatomical constraints, capturing weak signs of flow instability, and ultimately achieving accurate dynamic coupling prediction of VTE risk.

[0028] S2. Extract time-varying vorticity features that characterize early signs of local flow instability from Doppler blood flow signal sequences, and extract key anatomical constraint sets from ultrasound structural image sequences. The key anatomical constraint sets include geometric contour data of the target blood vessel, dynamic parameters of vessel diameter change, venous valve motion trajectory, and physical property indicators of vessel wall.

[0029] In a specific embodiment of the present invention, extracting time-varying vorticity features characterizing early signs of local flow instability from a Doppler blood flow signal sequence includes: calculating the curl of the blood flow velocity vector in the local space from consecutive frames of color Doppler images.

[0030] By performing differential operations on the time series of curl, the time-varying vorticity characteristics are obtained.

[0031] In a specific embodiment of the present invention, the time-varying vortex characteristic specifically refers to the instantaneous rotation intensity of local blood flow in a blood vessel, i.e., curl. The rate of change over time not only reflects the existence of microscopic vortices within the blood vessel at a given moment, but also dynamically quantifies the continuous evolution of the generation, development, or decay of these vortices. Essentially, it is a dynamic quantitative indicator used to accurately characterize early, subtle signs of local blood flow instability, enabling the system to sensitively capture early signals of potential venous thromboembolism risk before macroscopic blood flow velocity changes.

[0032] Specifically, the engineering objective is to accurately capture and quantify the rate of change of local microscopic rotational intensity and direction of blood flow from continuously varying blood flow Doppler signals, thereby characterizing early signs of embolism risk, i.e., time-varying vorticity features. This process is achieved through frame-by-frame processing and time-domain analysis of color Doppler blood flow image sequences.

[0033] First, preprocessing is required for the continuous frames of color Doppler images acquired from the synchronous acquisition module to obtain an accurate blood flow velocity vector field. This includes image denoising, such as using wavelet transform-based denoising algorithms to filter out artifacts and random noise, thereby improving the accuracy of velocity vector extraction. Next, techniques such as cross-correlation or optical flow are applied to each frame to calculate the blood flow velocity vector at each pixel in the image. Specifically, for any point in the image... Its velocity vector It can be represented as ,in and These are the velocity components of the point in the horizontal and vertical directions, respectively. These are unit vectors in the horizontal and vertical directions, respectively. The velocity vector field extraction employs a sub-pixel level optical flow algorithm, and its computational accuracy must ensure that the mean square error of the velocity residual is less than a preset threshold. The extraction accuracy of this velocity vector field must reach the sub-pixel level to ensure the accuracy of subsequent vortex calculations. For example, for a frame rate of 100... For a color Doppler image, a valid velocity vector must be obtained on at least 90% of the pixels.

[0034] In a specific embodiment of the present invention, the subpixel-level optical flow algorithm specifically refers to breaking the physical limitation of the inherent pixel resolution of the ultrasound equipment between consecutive ultrasound image frames through mathematical interpolation technology, and performing minute displacement tracking and velocity vector calculation on blood flow acoustic scatter points or contrast agents with a precision smaller than the size of a single pixel (such as 0.1 pixels). This high-precision technique aims to accurately reconstruct the real two-dimensional or three-dimensional blood flow velocity distribution field, enabling the system to capture extremely weak flow anomalies or tiny eddy currents in early blood vessels, thereby providing high-fidelity underlying data support for subsequent accurate calculation of "time-varying vorticity characteristics" and early warning of VTE.

[0035] In a specific embodiment of the present invention, the preset threshold refers to the maximum allowable error limit set to filter out speckle noise in ultrasound images and algorithm calculation errors. Its typical value is usually set to 0.5 mm / s to 1.5 mm / s of absolute velocity error or equivalent to 0.01 to 0.05 pixels / frame of spatial displacement residual. Setting this accuracy standard is to ensure that the reconstructed blood flow velocity vector field has an extremely high signal-to-noise ratio, thereby ensuring that the system can accurately extract the real and extremely weak "fluid instability" signal caused by small lesions in the blood vessel wall or early microthrombi, avoiding misjudging the calculation jitter of the underlying image as physical eddy current, and providing the underlying data foundation for subsequent time-varying vortex calculation and reverse source tracing analysis.

[0036] Secondly, for each frame of the obtained blood flow velocity vector field, the curl of the blood flow velocity vector in the local space is calculated to quantify the degree of rotation of the blood flow. Curl is a concept in vector calculus, which measures the "intensity of rotation" or "vortex" of a vector near a point in a vector field. In a two-dimensional plane, for a velocity vector field... Its curl is denoted as , can be represented as: ; in, Represents the component of velocity in the y-direction Follow Rate of change of direction Represents the component of velocity in the x-direction Follow The rate of change of direction. For engineering implementation, partial derivatives are usually approximated using the finite difference method. For example, the central difference method is used: , in The spatial step size is typically 1-3 pixels. This step yields the instantaneous curl field for each frame, reflecting the local vortex distribution of blood flow within the vessel at that moment.

[0037] Next, the time series of the instantaneous curl field is differentiated to obtain the time-varying vorticity characteristics, i.e., the rate of change of vorticity with time. The time-domain differentiation of vorticity aims to capture the dynamic changes in the local rotational state of the flow field, rather than just the instantaneous state. Let the curl field at time t be... , No. The curl field at time t is Then vorticity in time The change within is Furthermore, the time-varying vorticity characteristic is denoted as... By standardizing this change, such as by taking its absolute value or root mean square, a time-varying vorticity characteristic representing the degree of flow instability can be obtained. For example, for a Doppler signal with a sampling frequency of 100 Hz, the time step... Typically, the vorticity time is 0.01 seconds. This time-varying vorticity characteristic not only reflects the existence of vortices, but more importantly, it embodies the dynamic process of vortex generation, development, or decay. This is of great significance for identifying early, subtle embolism risks. For example, time-varying vorticity consistently above a preset threshold is identified as an early sign of abnormal flow instability. Through these refined processing steps, the time-varying vorticity characteristics crucial for early embolism warning can be extracted from the raw Doppler signal.

[0038] S3. Based on the key anatomical constraint set, generate the expected blood flow map of each segment in the corresponding vascular region by running the fluid dynamics simulation model. The expected blood flow map includes the expected time-varying vortex range.

[0039] In a specific embodiment of the present invention, based on a set of key anatomical constraints, the expected blood flow patterns of each segment within the corresponding vascular region are generated by running a fluid dynamics simulation model, including: inputting the set of key anatomical constraints as boundary conditions into the fluid dynamics simulation model.

[0040] Run a fluid dynamics simulation model to simulate the blood flow state within a vascular network under boundary conditions.

[0041] The expected range of hemodynamic parameters for each vascular segment is extracted from the simulation results and integrated to form the expected blood flow map.

[0042] In one specific embodiment of the present invention, the fluid dynamics simulation model adopts... The core governing equations are used, and spatial discretization is performed using the finite volume method. The inputs to the key anatomical constraint set specifically include: converting the extracted vessel contour into a 3D STL mesh model, and mapping the vessel wall elastic modulus to stiffness parameters of the wall boundary conditions. The simulation process employs a transient solver with a time step of [missing information]. Set to 0.001 seconds to capture high-frequency vortex fluctuations.

[0043] Specifically, the engineering objective is to transform the vascular geometry and physical state extracted from actual ultrasound images into boundary conditions and parameters that can be used for numerical calculations. This allows for the generation of the theoretically expected hemodynamic performance under these anatomical constraints—the anticipated blood flow map—within a simulation environment. This process is crucial for realizing the "anatomical-driven loop," providing a benchmark for subsequent comparisons and predictions of actual blood flow.

[0044] First, a set of key anatomical constraints is input as boundary conditions into a pre-defined fluid dynamics simulation model. This set of constraints comprises the vessel geometry, wall features, and physiological state extracted from ultrasound structural image sequences acquired by the synchronous acquisition module. For example, the vessel diameter is the area parameter of the inflow and outflow cross-sections; the presence of plaque or calcification on the vessel wall alters the boundary layer behavior of the fluid. This information is encoded into the geometric model data and wall characteristic parameters of the simulation model. Specifically, the vessel contour data identified in the ultrasound images is converted into a three-dimensional mesh model, and complex regions such as vessel bifurcation and tortuosity are finely meshed. For example, tetrahedral or hexahedral meshes are used to ensure sufficient mesh density to capture key fluid dynamic features, such as wall shear stress distribution. For valve opening and closing states, if incomplete closure or stenosis is observed, structural parameters impeding flow are introduced at the corresponding locations. Inlet boundary conditions are typically set as velocity or pressure curves representing physiological pulsating flow, while outlet conditions are set as constant pressure or no-backflow conditions.

[0045] Secondly, a pre-defined fluid dynamics simulation model is run to simulate the blood flow state within the vascular network under boundary conditions. This model is a numerical solver based on the fundamental equations of fluid mechanics. The solver employs numerical algorithms such as the finite element method, finite volume method, or lattice Boltzmann method to solve for the velocity and pressure fields of the blood flow on a discretized three-dimensional mesh model of the blood vessels. The simulation considers the non-Newtonian characteristics of blood flow, treating blood as a shear-thinned fluid and introducing physical parameters such as blood viscosity and density as needed. For example, a transient solver is used during the simulation, with the time step set to the millisecond level to capture the pulsation and turbulence details of the blood. The convergence criterion for each iteration is set to a residual reduction to below 10E⁻⁵ to ensure the stability of the calculation results.

[0046] Next, the expected range of hemodynamic parameters for each vascular segment is extracted from the simulation results and integrated to form an expected blood flow map. Once the fluid dynamics simulation model is running stably and yields results such as blood flow velocity and pressure at each node within the vascular network, parameters useful for the core "time-varying vorticity" calculation of this invention can be extracted. Specifically, the instantaneous curl of each vascular segment is calculated from the blood flow velocity vector field obtained from the simulation. Due to the variability of physiological processes, the "expected range" here refers to the statistical interval of the curl value output by the model under reasonable physiological fluctuations, rather than a fixed value. For example, by statistically analyzing the simulation results, the mean and standard deviation of the time-varying vorticity value of each vascular segment are obtained, and its expected range is defined as the mean plus or minus twice the standard deviation. The expected blood flow map is displayed visually on the 3D vascular model, and the expected time-varying vorticity range corresponding to each segment is marked, becoming a comparison benchmark for actual blood flow monitoring.

[0047] S4. The real-time extracted time-varying vorticity features are compared with the corresponding expected time-varying vorticity range in the expected blood flow map. When a continuous deviation event is detected, the source analysis of the key anatomical constraint set is triggered to generate at least one root cause candidate point.

[0048] In a specific embodiment of the present invention, the specific implementation logic for triggering the source analysis of the key anatomical constraint set is as follows: For the captured flow field anomaly features, a preset anatomy-flow field simulation model is invoked for inverse parameter iteration; by sequentially or randomly adjusting the boundary geometric parameters in the anatomical constraint set in the simulation model and repeatedly running the simulation calculation, the simulated flow field generated by each simulation is compared with the measured flow field anomaly features until the feature deviation between the two is less than a preset threshold, thereby inversely locating the root cause anatomy site leading to flow field instability. This process, through closed-loop verification of numerical calculation, achieves a precise mapping from the shift of flow field physical quantities to the anomaly of the anatomical structure.

[0049] In a specific embodiment of the present invention, the real-time extracted time-varying vorticity features are compared with the corresponding expected time-varying vorticity range in the expected blood flow map in real time. When a continuous deviation event is detected, the source analysis of the key anatomical constraint set is triggered to generate at least one root cause candidate point, including: continuously calculating the deviation between the measured time-varying vorticity value and the boundary value of the expected time-varying vorticity range.

[0050] When the deviation exceeds a preset threshold and the duration reaches a preset duration, a continuous deviation event is determined to have occurred.

[0051] Based on the location of the vascular segment where the persistent deviation event occurred, combined with ultrasound structural image sequences, the upstream or local anatomical structures were analyzed in reverse to identify at least one root cause candidate point.

[0052] Specifically, the engineering objective is to detect discrepancies between actual hemodynamic characteristics and theoretical states predicted based on anatomical structures in real time, and to initiate the tracing of potential anatomical causes when discrepancies persist, in order to identify functional lesions or structural changes leading to abnormal blood flow.

[0053] First, the deviation between the measured time-varying vorticity value and the expected time-varying vorticity range boundary value is continuously calculated. During real-time monitoring, the system continuously calculates the real-time time-varying vorticity characteristics. This is applied to each segment of the vascular network. At the current point in time Real-time time-varying vorticity characteristic value This will correspond to the expected time-varying vorticity range in the generated expected blood flow map. A comparison is made. The expected range of time-varying vorticity is defined by an upper limit. and lower limit value Composition, that is . Deviation It can be defined as, if Within the expected range, 0. If If the deviation exceeds the range, it is quantified as the degree to which it exceeds the boundary. For example, when the result... > hour, ;when < hour, Here and Adaptive adjustments can be made using historical simulations or population data to reflect individual physiological fluctuations. For example, a deviation exceeding 10% can be considered an initial deviation.

[0054] Secondly, when the deviation exceeds a preset threshold and lasts for a preset duration, a persistent deviation event is determined to have occurred. This step aims to filter out transient, non-pathological deviations caused by physiological fluctuations or measurement noise, thereby focusing on identifying truly persistent abnormalities that require attention. (Preset threshold) Typically, the parameters are set based on clinical experience and statistical analysis; for example, for the degree of deviation. , It can be set between 15% and 25%. Preset duration. The second method is determined based on the type of blood vessel being monitored and its physiological rhythm; for example, for venous blood flow, 2 can be set to a continuous monitoring cycle of 30 seconds to 1 minute. When in a vascular segment... Above, continuously sampling points ( )of All greater than At that time, the system determined that a persistent deviation event had occurred in that vascular segment. This judgment mechanism ensures the clinical significance and pathological relevance of the event, avoiding false alarms.

[0055] Secondly, based on the location of the vascular segment where the persistent deviation event occurs, and combined with the ultrasound structural image sequence, the system performs reverse analysis of the upstream or local anatomical structures to identify at least one root cause candidate point. Once the persistent deviation event is determined, the system immediately initiates an automated source tracing algorithm centered on the abnormal segment. This algorithm first uses a pre-defined vascular topology map to search for anatomical constraint points that may lead to hemodynamic abnormalities in the segment, either downstream to upstream or within the local region, along the retrograde direction of blood flow. For example, if persistent deviation occurs in the mid-segment of the femoral vein, the system will check upstream for extrinsic compression of the iliac vein or for valvular insufficiency in the femoral vein. During this process, the system will re-recall and analyze the mentioned ultrasound structural image sequence, and through image processing and pattern recognition techniques, based on a deep learning-based vascular lesion detection model, automatically identify points where key anatomical constraints have changed or may be abnormal. These identified points, such as new external compression points, thickened vessel wall areas, and abnormal valve structures, are marked as root cause candidate points. For example, by comparing current ultrasound images with historical images, newly identified stenosis with a reduction in vessel diameter of more than 20% is identified as a candidate root cause. This step aims to establish a direct link between abnormal blood flow function and specific anatomical changes, providing targets for subsequent interventions.

[0056] Based on the location of the vascular segment where the persistent deviation event occurred, combined with ultrasound structural image sequences, the upstream or local anatomical structures were analyzed in reverse to identify at least one root cause candidate point.

[0057] In a specific embodiment of the present invention, based on the location of the vascular segment where the continuous deviation event occurs, and combined with the ultrasound structural image sequence, the upstream or local anatomical structure is analyzed in reverse to identify at least one root cause candidate point. The method also includes: adjusting the constraint parameters corresponding to the key anatomical constraint set based on at least one root cause candidate point.

[0058] The fluid dynamics simulation model was run again using the adjusted set of key anatomical constraints to obtain the corrected expected blood flow pattern.

[0059] The matching degree between the corrected expected blood flow pattern and the real-time extracted time-varying vorticity features is calculated. If the matching degree is improved beyond the preset level, at least one root cause candidate point is confirmed as a high-probability root cause.

[0060] Specifically, the engineering objective is to validate the identified root cause candidates, that is, to confirm which anatomical changes(s) are most likely to have caused the observed persistent deviations in blood flow. This validation is conducted through a simulation model, enabling the "blood flow-driven loop" to provide feedback and correction to the "anatomical-driven loop," thereby enhancing the reliability of root cause determination.

[0061] First, the constraint parameters corresponding to the key anatomical constraint set are adjusted based on at least one root cause candidate point. Root cause candidate points identified through inverse analysis, such as local vascular compression, valvular insufficiency, or vessel wall thickening, each correspond to one or a set of specific parameters in the key anatomical constraint set. For example, if a root cause candidate point indicates mild external vascular compression, the vessel diameter parameter of the corresponding segment in the key anatomical constraint set is reduced by a preset amount, or an external pressure boundary condition is added outside the vessel. If the root cause candidate point is valvular insufficiency, the degree of leaflet closure in the valve geometry model is modified. This adjustment process quantitatively models the hypothetical root cause, and the parameter adjustment range is based on clinical experience or a preset pathological model; for example, increasing the valve regurgitation volume ratio by 5%-10%, or reducing the vessel diameter by 10%-20%.

[0062] Next, the pre-set hydrodynamic simulation model is run again using the adjusted set of key anatomical constraints to obtain a revised expected blood flow map. This step repeats the simulation process, but the input is no longer the original anatomical constraints, but a revised set of constraints that considers the anatomical changes caused by root cause candidate points. The simulation model recalculates the blood flow state within the vascular network under these revised boundary conditions, including the blood flow velocity and pressure field of each vascular segment. The final output will be a new expected blood flow map, reflecting the theoretically expected hemodynamic state under the assumption of a root cause. For example, if the assumed root cause is iliac vein stenosis, the revised expected blood flow map will show expected changes such as slower blood flow velocity and increased vorticity at the femoral vein.

[0063] Next, the matching degree between the corrected expected blood flow map and the real-time extracted time-varying vorticity features is calculated. If the matching degree improves beyond a preset level, at least one root cause candidate point is confirmed as a high-probability root cause. This step is the core of the verification. The system compares the newly generated corrected expected blood flow map with the real-time extracted time-varying vorticity features. The matching degree index can be the mean squared error, correlation coefficient, or structural similarity index. For example, by calculating the mean squared error between the corrected expected time-varying vorticity field and the real-time time-varying vorticity field, let the error be... Meanwhile, we possess the mean square error of the original expected blood flow profile and the real-time time-varying vorticity characteristics. .if ,in If the target level is set, for example, 20%-30%, it indicates that the corrected anatomical constraints better explain the measured blood flow abnormalities. This means that the introduced root cause candidate successfully brings the simulation prediction closer to the actual observation, and therefore the root cause candidate is identified as a high-probability root cause. This verification mechanism avoids the one-sidedness of judging the root cause based solely on image features or a single indicator, ensuring the physical rationality and high reliability of the root cause determination.

[0064] S5. A dynamic risk field is generated based on the intensity and distribution of the continuous deviation events, and a dynamic root cause field is generated based on the results of the source analysis and the feedback verification of the fluid dynamics simulation model. The dynamic risk field and the dynamic root cause field are interconnected and dynamically updated, forming a dynamic coupling field.

[0065] In a specific embodiment of the present invention, a dynamic risk field is generated based on the intensity and distribution of persistent deviation events, including mapping the vascular segments where persistent deviation events occur to a three-dimensional vascular tree model.

[0066] A dynamic risk value is assigned to each vascular segment based on the magnitude and duration of the deviation.

[0067] Based on the dynamic risk values ​​of all vascular segments, a dynamic risk field is generated by rendering on a three-dimensional vascular tree model.

[0068] In a specific embodiment of the present invention, the three-dimensional vascular tree model specifically refers to a high-precision digital twin geometric skeleton of the venous system of the examined site reconstructed based on multi-angle ultrasound scanning data. As a spatial carrier of dynamic risk data, it performs real-time spatial topological mapping and visual rendering of the "time-varying vorticity" and "dynamic risk value" dispersed in each vascular segment, thereby transforming abstract hemodynamic data into an intuitive panoramic risk heat map with anatomical positioning significance. This allows clinicians to clearly identify the specific origin, diffusion trend, and anatomical relationship of the risk in the entire venous network, providing a decision-making basis with three-dimensional spatial guidance for the precise intervention and early intervention of VTE.

[0069] Specifically, the engineering goal is to transform hemodynamic abnormalities, i.e., persistent deviation events, into intuitive and understandable risk quantification indicators, and dynamically present them on a three-dimensional spatial model of blood vessels, forming a "dynamic risk field" that facilitates clinicians to quickly interpret and locate risk areas.

[0070] First, the vascular segments experiencing persistent deviation events are mapped onto a 3D vascular tree model. When a vascular segment... When a persistent deviation event is detected, the system retrieves the precise geometric location and topological connectivity of the segment in three-dimensional space from the key anatomical constraint set. This information is used to match the segment with a pre-established three-dimensional vascular tree model representing the patient's vascular network. For example, by identifying anatomical landmarks such as vascular branch points and anastomoses, it ensures that each persistent deviation event is accurately located to the corresponding segment on the vascular tree model. This mapping is typically achieved through pre-processed vascular centerline extraction algorithms and vascular segmentation techniques, ensuring spatial consistency.

[0071] Secondly, a dynamic risk value is assigned to each vascular segment based on the magnitude and duration of the deviation. Deviation This describes the degree of deviation between the real-time time-varying vorticity characteristics and the expected time-varying vorticity range. Duration This refers to the cumulative time since the deviation event in that segment occurred. Dynamic Risk Value The calculation takes both factors into account to reflect the severity and trend of risk. For example, dynamic risk value... The following formula can be used for calculation: ; in, and These are weighting coefficients, calibrated based on clinical experience, for example... Set to 0.7, Set it to 0.3. This represents the natural exponential function. It is a time constant used to adjust the degree to which the duration affects the risk value, for example... Setting it to 15 seconds means that the risk weighting factor increases exponentially for every 15-second increase in duration. According to this formula, the segment with the greater the deviation and the longer the duration, the higher its dynamic risk value. This value changes continuously, reflecting the real-time risk dynamics of each vascular segment.

[0072] Next, based on the dynamic risk values ​​of all vascular segments, a dynamic risk field is generated and rendered on the 3D vascular tree model. The system uses the geometric data of the 3D vascular tree model as the visualization basis, and displays the calculated... This is visualized and mapped onto corresponding vascular segments. For example, the risk level of a vascular segment can be represented by changing its color or brightness. Low-risk values ​​might be displayed as green or blue, medium-risk values ​​as yellow, and high-risk values ​​as red. Color gradients can be used to clearly show subtle differences in risk levels. Furthermore, flashing or vibrating effects can be overlaid on high-risk areas to attract the attention of physicians. The dynamic risk field is a real-time, dynamically updated image that visually displays which areas in the entire vascular network face a higher risk of embolism, as well as the geographical distribution and intensity changes of these risk areas. For example, if a segment of a blood vessel... If the threshold for orange is exceeded, the blood vessel segment will be rendered as orange in real time on the model.

[0073] In one specific embodiment of the present invention, the preset orange threshold refers to the quantitative critical point that triggers a "high-risk" state. Its typical value is usually set between 0.7 and 0.8, based on a 0-1 normalized risk index or corresponding to a time-varying vorticity deviation of 3 to 5 times the normal baseline value. This threshold setting serves to accurately classify the risk level of a certain blood vessel segment. Exceeding this threshold immediately signifies that the hemodynamic characteristics at that location have undergone changes sufficient to lead to thrombus formation. The system will then issue a "high warning" signal to clinicians through color switching in visual rendering, allowing intervention to be taken during the "window period" before thrombus formation is visible to the naked eye.

[0074] In a specific embodiment of the present invention, a dynamic root cause field is generated based on the results of the source tracing analysis and the feedback verification of the fluid dynamics simulation model, including: mapping at least one root cause candidate point to a three-dimensional vascular tree model.

[0075] Each root cause candidate is assigned a root cause confidence value based on the probability that at least one root cause candidate is identified as a high-probability root cause and its contribution to the persistent deviation event.

[0076] A dynamic root cause field is generated by rendering on a 3D vascular tree model based on the root cause confidence values ​​of all root cause candidate points.

[0077] Specifically, the project aims to establish a clear causal link between the functional manifestations of abnormal blood flow and potential anatomical changes. By verifying and quantifying root cause candidate sites, it visualizes the fundamental structural problems leading to risk on a three-dimensional vascular model, forming a "dynamic root cause field." This complements the dynamic risk field, one indicating "where the risk is" and the other explaining "why the risk is."

[0078] First, at least one root cause candidate point is mapped to a 3D vascular tree model. During source analysis and feedback validation, the system identifies at least one root cause candidate point leading to persistent deviations. These points represent specific anatomical changes in vessel wall morphology, valve structure, or external compression. The system uses a pre-built 3D model of the patient's vascular network to precisely locate the geometric positions of these root cause candidate points within the model. For example, if a root cause candidate point is identified as thickening of the femoral vein valve, the corresponding valve location will be marked on the 3D vascular tree model. This mapping ensures consistency between root cause information and actual anatomical locations, facilitating intuitive visualization of lesions by physicians.

[0079] Secondly, a root cause confidence value is assigned to each root cause candidate based on the probability that at least one root cause candidate will be identified as a high-probability root cause and its contribution to the persistent deviation event. We identify high-probability root causes by improving the matching degree of the simulation model; this process generates the probability of identification. (Contribution) Individual root cause candidates were evaluated. For observed persistent deviation events, the time-varying vorticity deviation of a certain segment The relative impact. For example, if multiple root cause candidates exist, sensitivity analysis can be performed to remove or adjust the parameters of each candidate in the simulation model one by one, and observe their impact on the matching degree between the simulation results (i.e., the expected time-varying vorticity field) and the measured time-varying vorticity field. The greater the impact, the higher the contribution. Root cause confidence value It can be defined as: ; in, Indicates root cause candidate points The probability of being identified as a high-probability root cause, and the degree of improvement in the matching degree, can be normalized to a value between 0 and 1. This indicates the root cause candidate point. The contribution of the event to the overall sustained deviation. This confidence value. This reflects the likelihood and importance of the anatomical change being the true root cause of blood flow abnormalities. For example, if the contribution of a matching improvement at an extravascular compression point is 0.6 and its matching probability is 0.9, then the root cause confidence value for that compression point is 0.54.

[0080] Next, based on the root cause confidence values ​​of all root cause candidate points, a dynamic root cause field is generated and rendered on a 3D vascular tree model. The system uses the 3D vascular tree model as a carrier to visually display the confidence values ​​of each root cause candidate point. Similar to the risk field, color coding or symbol size can be used to represent the intensity of the root cause confidence value. For example, high-confidence root cause points may be displayed in darker or brighter colors, possibly accompanied by specific icons to emphasize their importance, while lower-confidence root cause points may be represented by lighter colors or smaller icons. If multiple root cause points exist, their distribution and relative importance in 3D space can be displayed through color gradients, transparency, or overlay effects. The dynamic root cause field is a dynamically changing, spatially distributed root cause map that clearly reveals which specific anatomical changes in the patient's vascular network are causing hemodynamic risks. This visualization information has direct guiding significance for clinical intervention decisions.

[0081] In a specific embodiment of the present invention, the dynamic risk field and the dynamic root cause field are interconnected and dynamically updated to form a dynamic coupling field, including: establishing an association mapping relationship between high-risk regions in the dynamic risk field and high-confidence root cause points in the dynamic root cause field.

[0082] When the dynamic risk field is updated, the update direction of the dynamic root cause field is guided according to the correlation mapping relationship.

[0083] When the dynamic root cause field is updated through feedback verification, the fluid dynamics simulation model is corrected in reverse based on the updated root cause information, thereby updating the expected blood flow spectrum to drive the next round of evolution of the dynamic risk field.

[0084] Specifically, the engineering goal is to tightly integrate the "dynamic risk field" representing blood flow abnormalities with the "dynamic root cause field" explaining the causes of abnormalities, forming a "dynamic coupling field" that provides mutual feedback and co-evolution. This integration not only provides a dual perspective of "risk" and "root cause," but also enables the system to continuously learn and optimize its predictive and tracing capabilities from dynamic interactions.

[0085] First, a mapping relationship is established between high-risk areas in the dynamic risk field and high-confidence root cause points in the dynamic root cause field. When the generated dynamic risk field shows a high-risk vascular segment, the system queries the generated dynamic root cause field to find high-confidence root cause points that are spatially adjacent to or causally related to the high-risk segment through blood flow pathways. For example, if the risk value of the mid-segment of the femoral vein... If the risk level remains consistently above the preset high-risk threshold, the system will check for root cause confidence values ​​in the upstream or same segment. Root causes exceeding a preset high-confidence threshold. This association mapping can be implemented using search algorithms based on distance and topological connectivity. For example, the association strength can be determined by calculating the shortest path length between high-risk areas and root causes, and their dependence on blood flow direction. This step clarifies the physical connection between risk manifestations and potential lesions, providing a basis for subsequent guidance and correction.

[0086] Secondly, when the dynamic risk field is updated, the update direction of the dynamic root cause field is guided by the correlation mapping relationship. The dynamic risk field changes in real time; when the risk value of a certain segment changes... The risk changed, rising rapidly from 0.5 to 0.8, and through correlation mapping, this risk was linked to a specific high-confidence root cause. When there is a strong association, the system will improve its understanding of the root cause. The system may prioritize attention to root causes and their corresponding regions. For example, the system might increase attention to root causes. Analysis of the frequency or depth of ultrasound structural image sequences in the surrounding area to identify root causes. More subtle changes. Simultaneously, if a high-risk region emerges or expands within the dynamic risk field, but the initial dynamic root cause field fails to explain this region, it triggers a broader source analysis. This guiding mechanism makes root cause analysis directional, enabling computational resources to be focused on the areas most likely to lead to new risks, improving the efficiency and real-time nature of root cause identification.

[0087] Furthermore, when the dynamic root cause field is updated through feedback verification, the preset fluid dynamics simulation model is corrected in reverse based on the updated root cause information, thereby updating the expected blood flow pattern to drive the next round of evolution of the dynamic risk field. This demonstrates the system's adaptive learning and optimization capabilities. When a certain root cause point is confirmed... As the system further validates the data or its parameters are estimated more accurately, adjustments to the key anatomical constraint set and corrections to the fluid dynamics simulation model will be solidified or updated. Specifically, the system will update the relevant parameters in the preset fluid dynamics simulation model based on the latest and most accurate root cause information. This updated model will be used to generate the expected blood flow map for the next round. For example, if the simulation results are closer to real-world observations, the relevant parameters in the model will be considered more accurate. This updated expected blood flow map will serve as a new benchmark for real-time comparison, directly impacting the subsequent dynamic risk field. The calculations form a closed loop from risk to root cause, and then from root cause to prediction, enabling the system to dynamically track and adaptively learn the evolution of the patient's vascular pathophysiology.

[0088] S6. Based on the current state of the dynamic coupling field, output hierarchical early warning data including early warning status level, risk visualization information, and source tracing prompts.

[0089] In a specific embodiment of the present invention, based on the current state of the dynamic coupling field, hierarchical early warning data including early warning status level, risk visualization information and source tracing prompts is output, including: analyzing whether there is a strongly correlated focus in the dynamic coupling field where both the dynamic risk field and the dynamic root cause field point to the same anatomical location.

[0090] The warning status level is determined by comprehensively considering the risk intensity of strongly correlated focal points, the confidence level of the dynamic root cause field, and the risk diffusion trend.

[0091] The visualization rendering of the dynamic coupled field, the warning status level, and the location information of strongly correlated focal points are integrated into a hierarchical warning result for output.

[0092] In a specific embodiment of this invention, the dynamic coupling field specifically refers to a comprehensive data field that deeply integrates hemodynamic characteristics and vascular pathological evolution trends in four-dimensional spacetime and three-dimensional space + time axis. It is no longer an isolated physical parameter, but rather a process that real-time correlates and interacts the microscopic physical variations of the underlying "blood flow field" with the macroscopic quantitative logic of the top-level "risk field." Through this coupling field, the system can characterize the overall trend of risk evolution and diffusion in the vascular network over time, providing a core analytical foundation for accurately determining the warning status level and achieving reverse tracing from "abnormal results" to "physical causes."

[0093] Specifically, the project aims to transform complex blood flow anomalies and structural root cause analysis results into clear and hierarchical early warning information that clinicians can quickly understand and use to make informed decisions. This tiered early warning data is the final output of the entire system, emphasizing the urgency of the risk and the clarity of its potential causes.

[0094] First, the system analyzes whether there exists a strongly correlated focus in the dynamic coupling field where both the dynamic risk field and the dynamic root cause field point to the same anatomical location. The dynamic risk field and the dynamic root cause field are interconnected and dynamically updated; the system continuously monitors their spatial overlap and the degree of coupling intensity. A strongly correlated focus refers to a local region within the vascular network where the dynamic risk value... The risk threshold of 0.7 or higher is reached, and one or more high-confidence root causes exist at the same or adjacent anatomical locations. Values ​​above 0.8 strongly suggest that these root causes can effectively explain the observed high risk. For example, if the risk of blood flow disturbance in the mid-segment of the superficial femoral vein is persistently high, and a root cause of time-varying vorticity abnormalities due to local valvular dysfunction is precisely identified in that segment, then a strong correlation focus is identified there. The identification of strong correlation focuses is achieved through weighted aggregation calculations combining the spatial distribution of risk values ​​with the location and confidence level of the root causes, highlighting the most urgent and definitive issues.

[0095] Secondly, the warning level is comprehensively determined based on the risk intensity of strongly correlated focal points, the confidence level of the dynamic root cause field, and the risk diffusion trend. Determining the warning level is a multi-factor decision-making process, the core of which is assessing the severity and development potential of strongly correlated focal points. Risk intensity refers to the risk value within the area of ​​strongly correlated focal points. The peak value and its duration. The confidence level of the dynamic root cause field is the root cause confidence value within the strongly correlated focal region. The average or maximum value. Risk diffusion trend assesses whether the current risk is spreading from one segment to surrounding or downstream vessels, for example, by comparing the rate of area growth of high-risk areas within adjacent time windows. These parameters are input into a pre-set decision matrix or machine learning model for comprehensive evaluation. For example, if the average risk value of a strongly correlated focal region... Higher than 0.9 and root cause confidence If the risk value is higher than 0.95, and the risk diffusion rate exceeds 0.05 cm² / min, the system may determine a Level 1 warning state. This Level 1 warning state indicates a strong instability and anomaly in the physical model of the target vascular flow field. This value serves as an intermediate parameter recorded by the system to prompt it to enter high-frequency monitoring mode. If the risk value is moderate, the root cause confidence is moderate, and the risk diffusion is not significant, it may be determined as a Level 2 warning state. This decision-making process is a real-time, adaptive assessment to adapt to the individualized pathological progression of the patient.

[0096] Furthermore, the final warning information is presented to clinicians through the user interface. The visualization rendering of the dynamic coupling field is a superposition of the dynamic risk field and the dynamic root cause field, intuitively showing the distribution of risks and root causes and their interrelationships throughout the vascular network. Warning status levels are displayed using striking colors or icons, such as red for high risk, yellow for intermediate risk, and blue for low risk. The location information of strongly correlated focal points clearly indicates the specific anatomical location of the risk and its root cause, such as "high-risk thrombosis due to proximal valve dysfunction of the right deep femoral vein." The graded warning data may also include suggested clinical treatment measures, such as "It is recommended to immediately conduct further ultrasound examination for diagnosis and evaluate anticoagulation therapy," or "It is recommended to have a follow-up examination within 24 hours and continuously monitor the dynamic risk." This integrated information output aims to provide comprehensive, intuitive, and timely data support for clinical decision-making, realizing a shift from passive diagnosis to proactive, precise, and dynamic warning.

[0097] In this invention, the source tracing analysis technically encompasses the verification logic of rerunning the simulation model. Essentially, source tracing analysis is a reverse inversion process based on a physical model. Through multiple iterative cycles of 'simulation-comparison-correction,' it reduces the dimensionality of anomalous data in the flow field space to the geometric features of the anatomical space, thereby locating the source of risk. This reverse localization method based on a closed-loop numerical simulation constitutes the core technical foundation for this invention's ability to accurately reveal the evolution mechanism of VTE (Vacuum-Activated Turbine Entrance).

[0098] Reference Figure 2 The second aspect of the present invention provides a VTE dynamic early warning system based on anatomical hemodynamic coupling, comprising: an image and signal sequence acquisition module, a feature and constraint set extraction module, a expected blood flow map generation module, a root cause candidate point generation module, a dynamic coupling field construction module, and a graded early warning data output module.

[0099] The image and signal sequence acquisition module is connected to the feature and constraint set extraction module, which is connected to the expected blood flow map generation module. Both the feature and constraint set extraction module and the expected blood flow map generation module are connected to the root cause candidate point generation module. Both the expected blood flow map generation module and the root cause candidate point generation module are connected to the dynamic coupling field construction module, which is connected to the hierarchical early warning data output module.

[0100] The image and signal sequence acquisition module acquires pre-stored ultrasound structural image sequences and Doppler blood flow signal sequences of the target vascular region simultaneously.

[0101] The feature and constraint set extraction module extracts time-varying vorticity features that characterize early signs of local flow instability from Doppler blood flow signal sequences, and extracts key anatomical constraint sets from ultrasound structural image sequences. The key anatomical constraint sets include geometric contour data of the target blood vessel, dynamic parameters of vessel diameter change, venous valve motion trajectory, and physical property indicators of the vessel wall.

[0102] The expected blood flow map generation module, based on the key anatomical constraint set, generates expected blood flow maps for each segment within the corresponding vascular region by running a fluid dynamics simulation model. The expected blood flow map includes the expected time-varying vorticity range.

[0103] The root cause candidate point generation module compares the real-time extracted time-varying vorticity features with the corresponding expected time-varying vorticity range in the expected blood flow map in real time. When a continuous deviation event is detected, it triggers the source analysis of the key anatomical constraint set and generates at least one root cause candidate point.

[0104] The dynamic coupling field module generates a dynamic risk field based on the intensity and distribution of continuous deviation events, and generates a dynamic root cause field based on the results of source analysis and feedback verification from the fluid dynamics simulation model. The dynamic risk field and the dynamic root cause field are interconnected and dynamically updated, forming a dynamic coupling field.

[0105] The graded early warning data output module outputs graded early warning data, including early warning status level, risk visualization information, and source tracing prompts, based on the current situation of the dynamic coupled field.

[0106] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, and all such modifications and additions should fall within the protection scope of the present invention.

Claims

1. A dynamic early warning method for VTE based on anatomical hemodynamic coupling, characterized in that, include: S1. Acquire the pre-stored sequence of ultrasound structural images and Doppler blood flow signals synchronously acquired from the target vascular region; S2. Extract time-varying vorticity features that characterize early signs of local flow instability from Doppler blood flow signal sequences, and extract key anatomical constraint sets from ultrasound structural image sequences. The key anatomical constraint sets include geometric contour data of the target blood vessel, dynamic change parameters of the vessel diameter, venous valve motion trajectory, and physical property indicators of the vessel wall. S3. Based on the key anatomical constraint set, the expected blood flow map of each segment in the corresponding vascular region is generated by running the fluid dynamics simulation model. The expected blood flow map includes the expected time-varying vorticity range. S4. The real-time extracted time-varying vorticity features are compared with the corresponding expected time-varying vorticity range in the expected blood flow map in real time. When a continuous deviation event is detected, the source analysis of the key anatomical constraint set is triggered to generate at least one root cause candidate point. S5. Generate a dynamic risk field based on the intensity and distribution of continuous deviation events, and generate a dynamic root cause field based on the results of source analysis and feedback verification of the fluid dynamics simulation model. The dynamic risk field and the dynamic root cause field are interconnected and dynamically updated, forming a dynamic coupling field. S6. Based on the current state of the dynamic coupling field, output hierarchical early warning data including early warning status level, risk visualization information, and source tracing prompts.

2. The VTE dynamic early warning method based on anatomical hemodynamic coupling according to claim 1, characterized in that, The extraction of time-varying vorticity features characterizing early signs of local flow instability from Doppler blood flow signal sequences includes: Calculate the curl of the blood flow velocity vector in local space from consecutive frames of color Doppler images; By performing differential operations on the time series of curl, the time-varying vorticity characteristics are obtained.

3. The VTE dynamic early warning method based on anatomical hemodynamic coupling according to claim 1, characterized in that, The process of generating expected blood flow maps for each segment within a corresponding vascular region based on a set of key anatomical constraints and by running a fluid dynamics simulation model includes: The key anatomical constraint set is input as boundary conditions into the fluid dynamics simulation model; Run a fluid dynamics simulation model to simulate the blood flow state within a vascular network under boundary conditions; The expected range of hemodynamic parameters for each vascular segment is extracted from the simulation results and integrated to form the expected blood flow map.

4. The VTE dynamic early warning method based on anatomical hemodynamic coupling according to claim 1, characterized in that, The process involves real-time comparison of the extracted time-varying vorticity features with the corresponding expected time-varying vorticity range in the expected blood flow map. When a continuous deviation event is detected, a source tracing analysis of the key anatomical constraint set is triggered, generating at least one root cause candidate point, including: Continuously calculate the deviation between the measured time-varying vorticity value and the expected time-varying vorticity range boundary value; When the deviation exceeds a preset threshold and the duration reaches a preset duration, a continuous deviation event is determined to have occurred. Based on the location of the vascular segment where the persistent deviation event occurred, combined with ultrasound structural image sequences, the upstream or local anatomical structures were analyzed in reverse to identify at least one root cause candidate point.

5. The VTE dynamic early warning method based on anatomical hemodynamic coupling according to claim 4, characterized in that, The method of identifying at least one root cause candidate point by retrospectively analyzing the upstream or local anatomical structures based on the location of the vascular segment where the persistent deviation event occurs, combined with ultrasound structural image sequences, further includes: Adjust the constraint parameters corresponding to the key anatomical constraint set based on at least one root cause candidate point. The fluid dynamics simulation model was run again using the adjusted set of key anatomical constraints to obtain the corrected expected blood flow pattern. The matching degree between the corrected expected blood flow pattern and the real-time extracted time-varying vorticity features is calculated. If the matching degree is improved beyond the preset level, at least one root cause candidate point is confirmed as a high-probability root cause.

6. The VTE dynamic early warning method based on anatomical hemodynamic coupling according to claim 1, characterized in that, The generation of a dynamic risk field based on the intensity and distribution of persistent deviation events includes: Map the vascular segments that experience persistent deviation events to a 3D vascular tree model; A dynamic risk value is assigned to each vascular segment based on the magnitude and duration of the deviation. Based on the dynamic risk values ​​of all vascular segments, a dynamic risk field is generated by rendering on a three-dimensional vascular tree model.

7. The VTE dynamic early warning method based on anatomical hemodynamic coupling according to claim 6, characterized in that, The generation of the dynamic root cause field based on the results of the source analysis and feedback verification of the fluid dynamics simulation model includes: Map at least one root cause candidate point to a 3D vascular tree model; Each root cause candidate is assigned a root cause confidence value based on the probability that at least one root cause candidate is identified as a high-probability root cause and its contribution to the persistent deviation event. A dynamic root cause field is generated by rendering on a 3D vascular tree model based on the root cause confidence values ​​of all root cause candidate points.

8. The VTE dynamic early warning method based on anatomical hemodynamic coupling according to claim 7, characterized in that, The dynamic risk field and the dynamic root cause field are interconnected and dynamically updated, forming a dynamically coupled field, including: Establish a correlation mapping relationship between high-risk areas in the dynamic risk field and high-confidence root points in the dynamic root cause field; When the dynamic risk field is updated, the update direction of the dynamic root cause field is guided according to the correlation mapping relationship; When the dynamic root cause field is updated through feedback verification, the fluid dynamics simulation model is corrected in reverse based on the updated root cause information, thereby updating the expected blood flow spectrum to drive the next round of evolution of the dynamic risk field.

9. The VTE dynamic early warning method based on anatomical hemodynamic coupling according to claim 1, characterized in that, Based on the current state of the dynamic coupling field, the system outputs tiered early warning data, including early warning status levels, risk visualization information, and source tracing prompts, comprising: Analyze whether there exists a strongly correlated focus in the dynamic coupling field where both the dynamic risk field and the dynamic root cause field point to the same anatomical location; The warning status level is determined by comprehensively considering the risk intensity of strongly correlated focal points, the confidence level of the dynamic root cause field, and the risk diffusion trend. The visualization rendering of the dynamic coupled field, the warning status level, and the location information of strongly correlated focal points are integrated into a hierarchical warning result for output.

10. A VTE dynamic early warning system based on anatomical hemodynamic coupling, characterized in that, include: The image and signal sequence acquisition module acquires pre-stored ultrasound structural image sequences and Doppler blood flow signal sequences of the target vascular region simultaneously acquired; The feature and constraint set extraction module extracts time-varying vorticity features that characterize early signs of local flow instability from Doppler blood flow signal sequences, and extracts key anatomical constraint sets from ultrasound structural image sequences. The key anatomical constraint sets include geometric contour data of the target blood vessel, dynamic parameters of vessel diameter change, venous valve motion trajectory, and physical property indicators of vessel wall. The expected blood flow map generation module, based on the key anatomical constraint set, generates expected blood flow maps for each segment within the corresponding vascular region by running a fluid dynamics simulation model. The expected blood flow map includes the expected time-varying vorticity range. The root cause candidate point generation module compares the real-time extracted time-varying vorticity features with the corresponding expected time-varying vorticity range in the expected blood flow map in real time. When a continuous deviation event is detected, it triggers the source analysis of the key anatomical constraint set and generates at least one root cause candidate point. The dynamic coupling field module generates a dynamic risk field based on the intensity and distribution of continuous deviation events, and generates a dynamic root cause field based on the results of source analysis and feedback verification from the fluid dynamics simulation model. The dynamic risk field and the dynamic root cause field are interconnected and dynamically updated, forming a dynamic coupling field. The graded early warning data output module outputs graded early warning data, including early warning status level, risk visualization information, and source tracing prompts, based on the current situation of the dynamic coupled field.

Images