A method, apparatus, device and storage medium for identifying a cardiac phase

Abstract

The present disclosure provides a cardiac phase recognition method, device, equipment and storage medium, to realize multi-frame automatic segmentation and accurate cardiac phase recognition. The cardiac phase recognition method comprises: selecting an initial segmentation frame in the ventricular contrast sequence image data; performing ventricular boundary recognition based on the feature points marked in the initial segmentation frame to determine the ventricular boundary contour; for the image frame adjacent to the initial segmentation frame, taking the ventricular boundary contour of the initial segmentation frame as the starting contour, emitting a ray in a narrow band range to search for a boundary point in a polar coordinate manner, connecting the boundary points and performing region-driven segmentation to obtain the ventricular boundary contour; taking the image frame as a new initial segmentation frame, repeating the above steps for the next image frame to obtain the full sequence ventricular boundary contour; determining the ventricular cavity area sequence corresponding to the ventricular contrast sequence image data based on the full sequence ventricular boundary contour, and dividing the cardiac phase according to the ventricular cavity area sequence.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and in particular to a method, apparatus, device and storage medium for recognizing the phase of heartbeats. Background Technology

[0002] Ventriculography is a crucial technique for assessing cardiac function. It requires segmenting ventricular images to obtain contour and volume parameters, and identifying cardiac phases such as diastole and systole. Currently, related techniques primarily rely on physicians manually segmenting multiple frames, or only partially segmenting frames and then manually determining the phases, making it impossible to achieve fully automated sequence tracking and segmentation, or automatic identification of key phases.

[0003] The aforementioned manual operations are time-consuming, labor-intensive, and inefficient, making it difficult to meet the needs of rapid clinical diagnosis. Moreover, human judgment discrepancies lead to instability in segmentation results and phase recognition, affecting the reliability of cardiac function assessment. With the development of precision medicine, clinical requirements for diagnostic efficiency and accuracy are constantly increasing, and traditional manual intervention methods can no longer meet the needs. There is an urgent need for a technology that can achieve automatic multi-frame segmentation and accurate cardiac phase recognition. Summary of the Invention

[0004] In view of this, the present disclosure provides a method, apparatus, device and storage medium for cardiac phase recognition, so as to realize automatic multi-frame segmentation and accurate cardiac phase recognition.

[0005] A first aspect provides a method for cardiac phase identification, comprising: acquiring ventriculography sequence image data; selecting an initial segmentation frame in the ventriculography sequence image data; performing ventricular boundary identification on the initial segmentation frame based on feature points marked in the initial segmentation frame to determine the ventricular boundary contour of the initial segmentation frame; for image frames in the ventriculography sequence image data adjacent to the initial segmentation frame, using the ventricular boundary contour of the initial segmentation frame as a starting contour, emitting rays in polar coordinates to search for boundary points within a narrow band, connecting the boundary points and performing region-driven segmentation to obtain the ventricular boundary contour of the image frame; the narrow band range is determined based on the intersection of the current ray and the starting contour; using the image frame as a new initial segmentation frame, repeating the steps of searching for boundary points and region-driven segmentation for the next image frame to obtain the full sequence ventricular boundary contour; determining the ventricular cavity area sequence corresponding to the ventriculography sequence image data based on the full sequence ventricular boundary contour, and performing cardiac phase segmentation according to the ventricular cavity area sequence.

[0006] In one embodiment, the feature points include the upper edge of the aortic root, the lower edge of the aortic root, and the apex of the heart; or, the feature points include the apex of the heart, the left hinge of the mitral valve annulus, and the right hinge of the mitral valve annulus.

[0007] In one embodiment, the ventricular boundary identification of the initial segmented frame may include: constructing a region of interest (ROI) based on the feature points; and performing ventricular boundary identification on the initial segmented frame within the ROI to determine the ventricular boundary contour of the initial segmented frame.

[0008] In one embodiment, constructing a Region of Interest (ROI) based on the feature points may include: determining a base position based on feature points other than the apex, and connecting the apex and the base position to form a major axis; or, performing coarse segmentation of the ventricular cavity of the initial segmented frame based on the feature points, performing principal component analysis on the coarse segmentation result to obtain the major axis direction; or, performing coarse segmentation of the ventricular cavity of the initial segmented frame based on the feature points, performing ellipse fitting on the coarse segmentation result to obtain the major axis direction of the fitted ellipse; constructing the ROI with the apex as the pole and the major axis direction as the reference, wherein the angle range of the ROI covers the left ventricular cavity, and the upper limit of the radius is a preset multiple of the length of the major axis.

[0009] In one embodiment, the step of identifying the ventricular boundary of the initial segmented frame includes: setting an initial seed in the initial segmented frame based on the feature points; using the initial seed as the initialization input of an energy model, performing iterative optimization of the model until convergence, and extracting the ventricular boundary contour of the initial segmented frame.

[0010] In one embodiment, setting an initial seed in the initial segmented frame based on the feature points includes: placing a hyperellipse that conforms to the geometry of the left ventricle along the major axis as an inner seed based on the feature points, and setting a thin ring as an outer seed around the periphery; the thin ring refers to a ring-shaped structure with a radial width smaller than a preset width.

[0011] In one embodiment, the feature point includes the apex of the heart. The emitting of rays in polar coordinates includes: using the apex marked in the initial segmentation frame as the pole, or extracting the centroid or geometric center of the initial contour as the pole, or adaptively determining the pole based on the shape features of the initial contour, and emitting multiple rays at fixed angular intervals from the pole. The searching for boundary points within a narrow band includes: extending a narrow band to both ends of each ray, centered on the intersection of each ray and the initial contour; searching for boundary points within the narrow band on each ray to obtain the ventricular boundary contour of the image frame.

[0012] In one embodiment, searching for boundary points within a narrow band includes: within the narrow band of each ray, using a gradient operator to calculate the brightness gradient value of each pixel position, and determining the position with the largest gradient value as the boundary point in that direction.

[0013] In one embodiment, connecting the boundary points and performing region-driven segmentation includes: connecting the boundary points in polar angle order to form a closed curve; performing smoothing and / or shape constraint processing on the closed curve to obtain a processed closed curve; and performing region-driven segmentation based on the processed closed curve to obtain the ventricular boundary contour of the image frame.

[0014] In one embodiment, performing region-driven segmentation based on the processed closed curve includes: initiating an active contour model based on the closed curve and obtaining the ventricular boundary contour of the image frame through energy minimization optimization; or, using the inner and outer regions of the closed curve as foreground and background initialization, initiating a graph cut model to perform region-driven segmentation to obtain the ventricular boundary contour of the image frame.

[0015] In one embodiment, determining the ventricular cavity area sequence corresponding to the ventricular angiography sequence image data includes: based on the full sequence of ventricular boundary contours, counting the number of pixels inside each frame of the ventricular boundary contour; converting the number of pixels into actual area units according to the pixel spacing parameter, and arranging the ventricular cavity areas of each frame of the ventricular boundary contour in chronological order to form the ventricular cavity area sequence.

[0016] In one embodiment, the step of dividing cardiac phases according to the ventricular cavity area sequence includes: determining the maximum area value of each cardiac cycle in the ventricular cavity area sequence, and determining the end-diastolic frame and end-systolic frame based on the image frame corresponding to the maximum area value and the image frame corresponding to the minimum area value.

[0017] In one implementation, determining the end-diastolic frame and the end-systolic frame based on the image frame corresponding to the maximum area and the image frame corresponding to the minimum area includes: identifying the image frame corresponding to the maximum area as the end-diastolic frame and the image frame corresponding to the minimum area as the end-systolic frame; or, within a preset number of frames near the maximum area, calculating the movement speed of the ventricular boundary contour in each frame, and selecting the frame with the smallest boundary movement speed as the end-diastolic frame or the end-systolic frame.

[0018] Secondly, a cardiac phase recognition device is provided, comprising: The acquisition module is used to acquire ventricular angiography sequence image data and select an initial segmentation frame in the ventricular angiography sequence image data; The initial contour determination module is used to identify the ventricular boundary of the initial segmented frame based on the feature points marked in the initial segmented frame, and determine the ventricular boundary contour of the initial segmented frame. The sequence frame contour determination module is used to, for image frames adjacent to the initial segmentation frame in the ventriculography sequence image data, take the ventricular boundary contour of the initial segmentation frame as the starting contour, emit rays in polar coordinates to search for boundary points within a narrow band, connect the boundary points and perform region-driven segmentation to obtain the ventricular boundary contour of the image frame; the narrow band range is determined based on the intersection of the current ray and the starting contour; take the image frame as the new initial segmentation frame, and repeat the steps of searching for boundary points and region-driven segmentation for the next image frame to obtain the full sequence ventricular boundary contour; The cardiac phase segmentation module is used to determine the ventricular cavity area sequence corresponding to the ventricular angiography sequence image data based on the full sequence ventricular boundary contour, and to segment the cardiac phase according to the ventricular cavity area sequence.

[0019] Thirdly, a computer device is provided, the computer device including a processor and a memory, the memory storing computer-readable instructions, the processor being used to read the computer-readable instructions to execute the cardiac phase recognition method described in any of the above embodiments.

[0020] Fourthly, a computer-readable storage medium is provided, including instructions stored thereon, wherein, when the instructions are executed by a processor, the cardiac phase recognition method described in any of the above embodiments is executed.

[0021] This embodiment only requires manual marking of feature points in the initial segmentation frame. After determining the ventricular boundary contour of the initial segmentation frame based on these feature points, this ventricular boundary contour is used as the starting contour. Subsequent sequence frames automatically complete contour extraction through polar coordinate tracking (using polar coordinate rays to search for boundary points within a narrow band) and region-driven segmentation, eliminating the need for manual operation frame by frame, significantly reducing time consumption and meeting the needs of rapid clinical diagnosis. Furthermore, narrow-band constraints reduce noise interference, making full-sequence contour tracking more continuous and phase identification more reliable. In addition, automating full-sequence segmentation and cardiac phase division replaces the tedious steps of manually searching and determining diastolic / systolic phases, reducing operational complexity and the probability of human error. Attached Figure Description

[0022] Figure 1 A flowchart of a cardiac phase recognition method provided in this embodiment of the disclosure; Figure 2 A schematic diagram illustrating the setting of inner and outer seeds provided in an embodiment of this disclosure; Figure 3 A schematic diagram of a cardiac phase recognition device 300 provided in an embodiment of this disclosure; Figure 4 This is a schematic diagram of a computer device 400 provided in an embodiment of the present disclosure. Detailed Implementation

[0023] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0024] The terminology used in this disclosure is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The singular forms “a,” “the,” and “the” as used in this disclosure and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0025] It should be understood that although the terms first, second, third, etc., may be used in this disclosure to describe various information, such information should not be limited to these terms. These terms are used only to distinguish information of the same type from one another. For example, without departing from the scope of this disclosure, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0026] Furthermore, the symbol “ / ” in this disclosure indicates that there is an “or” relationship between the related objects before and after the symbol, or that the related objects have an exemplary relationship that can coexist.

[0027] like Figure 1 As shown, a cardiac phase recognition method provided in this disclosure includes: S101: Acquire ventricular angiography sequence image data, and select an initial segmentation frame from the ventricular angiography sequence image data.

[0028] Here, ventricular angiography sequence images are acquired for cardiac function assessment. One frame from this sequence is randomly selected as the reference frame for subsequent initial segmentation, i.e., the initial segmentation frame. The ventricular angiography sequence image data here can refer to digital subtraction angiography (DSA) image data.

[0029] In some embodiments, a physician may manually select any frame from the ventriculography sequence as the initial segmentation frame. Alternatively, the image frame with the highest image quality from the ventriculography sequence can be automatically selected as the initial segmentation frame based on image quality assessment metrics. For example, the image quality assessment metrics may include a sharpness metric and a contrast metric; that is, the sharpness metric and the contrast metric can be calculated for each frame in the ventriculography sequence, and the image frame with the highest combined score of the sharpness metric and the contrast metric can be selected as the initial segmentation frame.

[0030] S102: Based on the feature points marked in the initial segmented frame, perform ventricular boundary recognition on the initial segmented frame to determine the ventricular boundary contour of the initial segmented frame.

[0031] In practice, a small number of feature points can be manually marked on the selected initial segmentation frame. The region of interest can be constructed and seed points can be set with the help of the feature points. The ventricular boundary contour of the frame can be identified and determined by the active contour model with shape prior.

[0032] In some embodiments, taking the left ventricle as an example, the feature points may include (first type) the upper edge of the aortic root, the lower edge of the aortic root, and the apex (left ventricular apex); or (second type) the feature points may include the apex, the left hinge of the mitral annulus, and the right hinge of the mitral annulus.

[0033] The first type of feature points mentioned above have high image visibility. The aortic root is a large vascular structure, and its boundaries are clear after contrast filling in DSA images, with well-defined solid outlines at the upper and lower edges, making it highly distinguishable from the surrounding myocardium and blood vessels. Therefore, the operation threshold for doctors to mark the first type of feature points is low, and the marking process is intuitive and time-saving. Since the aortic root is the outflow tract of the left ventricle, and it is a different anatomical structure from the base of the left ventricle (mitral valve annulus region), the base constructed using these three points is an approximation. The region of interest (ROI) constructed based on the approximate base may have slight deviations in angle coverage, which can be refined through subsequent region-driven segmentation to ensure close alignment with the actual cavity of the left ventricle.

[0034] The image visibility of the second type of feature points is relatively low. The mitral valve annulus is a delicate ring-shaped fibrous structure with low contrast in DSA images and no clear solid boundary. It can only be indirectly judged through the transition area between the left ventricle and the left atrium, and the left and right hinge points are not visually marked. Therefore, for the second type of feature points, it depends on the doctor's accurate grasp of the cardiac anatomy and the experience required to judge the location of the left and right hinge points of the mitral valve annulus. However, since the mitral valve annulus is the direct connection between the left ventricle and the left atrium and is the true anatomical boundary of the left ventricular base, the long axis (from the apex to the midpoint of the left and right hinge points of the mitral valve annulus) determined based on these three points perfectly matches the anatomical long axis of the left ventricle without approximation error. ROI construction can be completed in one step (angle coverage directly matches the left ventricular cavity defined by the mitral valve annulus), the placement of the inner seed (ellipse / hyperellipse) in the initial frame is more precise, and the shape prior constraint is more effective, resulting in higher accuracy for initial segmentation and thus improving the accuracy of subsequent sequence frame segmentation.

[0035] In practical implementation, after manually marking feature points in the initial segmented frame, an automatic feature point detection algorithm can be used for precise correction. This algorithm corrects the subjective bias of manual marking, ensuring that the feature point positions accurately align with the boundaries of the actual anatomical structures in the DSA image. For example, feature points can be verified based on the intensity gradient features of the DSA image (such as the transition zone between the aortic root and myocardium, and the intensity difference between the left ventricle and left atrium on both sides of the mitral valve annulus) and prior constraints of cardiac anatomy (such as the aortic root being a tubular structure with symmetrically distributed boundaries; the mitral valve annulus being a ring structure, requiring specific spacing and angular relationships between the left and right hinge points). For instance, a small search area (such as a 10×10 pixel window) is defined centered on the manually marked point; within this search area, the brightness gradient is calculated along the long axis of the aortic root (perpendicular to the aortic direction), and the location corresponding to the maximum gradient value (i.e., the actual aortic root boundary) is found; combined with the symmetry constraints of the aortic root boundary, it is ensured that the line connecting the upper and lower edge points is perpendicular to the aortic direction, and the distance between the two points conforms to the normal anatomical range (avoiding abnormal spacing after correction). For example, using manually marked left and right hinge points as initial references, a continuous annular boundary (the anatomical morphology of the mitral valve annulus) is searched within the ROI. The curvature of this annular boundary is calculated, and the two endpoints corresponding to the maximum curvature are the true left and right hinge points (the anatomical feature of the mitral valve annulus is that the left and right hinge points are the curvature extreme points of the annular structure). Simultaneously, it is ensured that the constraints are met: the line connecting the left and right hinge points must be perpendicular to the long axis of the left ventricle, and the distance between the two points conforms to the anatomical range of a normal mitral valve annulus. Another example is the apex, which is the most distal point of the left ventricle, corresponding to the endpoint of the ventricular contour plus the center of symmetry of the brightness gradient. Using the manually marked point as the center, the contour endpoint (i.e., the termination position of the ventricular boundary) is searched along the long axis direction (apex → base). Combined with the left and right boundary symmetry constraints, it is ensured that the apex is located on the axis of symmetry of the left ventricular contour and is the most distal point of the long axis.

[0036] Accordingly, the embodiments of this disclosure are also applicable to the right ventricle. For the right ventricle, the first type of feature points can be replaced by the upper edge of the pulmonary valve annulus, the lower edge of the pulmonary valve annulus, and the apex of the right ventricle; the second type of feature points can be replaced by the apex of the right ventricle, the left hinge of the tricuspid valve annulus, and the right hinge of the tricuspid valve annulus.

[0037] In some embodiments, ventricular boundary identification of the initial segmented frame may include: constructing a region of interest (ROI) based on the feature points; and performing ventricular boundary identification of the initial segmented frame within the ROI to determine the ventricular boundary contour of the initial segmented frame.

[0038] Here, ventricular-related geometric benchmarks are established based on feature points. By delineating the Region of Interest (ROI) that adapts to the ventricular region, the target is focused and irrelevant interference is eliminated. Then, within the ROI, a segmentation strategy that adapts to the ventricular morphology and image features is adopted to accurately identify and extract the ventricular boundary contour of the initial segmented frame.

[0039] In practice, the long axis direction of the ventricle can be determined first based on manually marked feature points; then, with the apex of the heart as the pole, combined with the long axis direction, a sector-shaped ROI can be defined, covering the ventricular cavity at an angle, thereby narrowing the search range and eliminating interference from irrelevant areas such as bones and other blood vessels.

[0040] This disclosure provides several specific methods for constructing the aforementioned ROI, including: determining the base position based on feature points other than the apex of the heart, and connecting the apex and the base position to form a major axis; or, performing coarse segmentation of the ventricular cavity of the initial segmented frame based on the feature points, and performing principal component analysis on the coarse segmentation result to obtain the major axis direction; or, performing coarse segmentation of the ventricular cavity of the initial segmented frame based on the feature points, and performing ellipse fitting on the coarse segmentation result to obtain the major axis direction of the fitted ellipse; constructing the ROI with the apex as the pole and the major axis direction as the reference, wherein the angle range of the ROI covers the left ventricular cavity, and the upper limit of the radius is a preset multiple of the length of the major axis.

[0041] The above methods include: Feature point derivation: First, the base position (midpoint of the line connecting the feature points) is determined using feature points other than the apex (such as the upper and lower edges of the aortic root, and the left and right hinge points of the mitral valve annulus). The apex and the base position are then connected to form the major axis. Next, using the apex as the pole and the major axis direction as the reference, a Region of Interest (ROI) is defined with an angle covering the left ventricular cavity and a radius that is a preset multiple of the major axis (e.g., 1.2 times). Other methods include: Coarse segmentation + Principal Component Analysis (PCA): Based on the feature points, the ventricular cavity of the initial segmented frame is coarsely segmented (roughly delineating the cavity area). PCA is performed on the coarse segmentation result to obtain the major axis direction of the ventricle. Subsequently, an ROI is constructed with the apex as the pole, the angle covering the cavity, and the radius being a preset multiple of the major axis. Still other methods include: Coarse segmentation + Ellipse fitting: First, the ventricular cavity is coarsely segmented using feature points. An ellipse is fitted to the coarse segmentation result, and the major axis of the fitted ellipse is used as the major axis direction of the ventricle. Then, an ROI is constructed following the same rules as the previous two methods.

[0042] For the aforementioned feature point derivation method, for example, when the feature points include the upper edge of the aortic root, the lower edge of the aortic root, and the apex of the heart, the midpoint of the line connecting the aortic root is determined based on the upper edge of the aortic root and the lower edge of the aortic root, and the midpoint of the line connecting the apex of the heart and the aortic root is connected to form a long axis; when the feature points include the apex of the heart, the left hinge point of the mitral valve annulus, and the right hinge point of the mitral valve annulus, the base position is determined based on the left hinge point of the mitral valve annulus and the right hinge point of the mitral valve annulus, and the base position is connected to form a long axis.

[0043] Here, since the mitral annulus is the connection between the left ventricle and the left atrium, its left and right hinge points are direct markers of the lateral extent of the left ventricular base (bottom), forming the core geometric axis of the left ventricular apex-base together with the apex. Therefore, by using the apex as the pole and directly covering the lateral extent of the left ventricular cavity defined by the left and right hinge points of the mitral annulus, the ROI can be obtained. However, since the two points at the aortic root cannot directly reflect the boundary of the left ventricular base, the missing geometric information can be compensated for through refinement. For example, a temporary base can be constructed using the apex and aortic root to preliminarily estimate the approximate long axis and rough angle range of the left ventricle. At this point, the angle coverage of the ROI may have deviations (it cannot completely fit the left ventricular cavity). Then, edge / curvature analysis can be used to automatically identify the annulus line (i.e., the true contour of the mitral annulus), completing the precise lateral boundary of the left ventricular base. Finally, based on the midpoint of the annulus line, the long axis direction is corrected and the angle range is adjusted to determine the ROI. Thus, relying on the algorithm's subsequent automatic refinement, the difficulty for doctors to mark the mitral valve annulus hinge point can be reduced (only two easily identifiable points at the aortic root need to be marked).

[0044] In some embodiments, the ventricular boundary identification of the initial segmented frame may include: setting an initial seed in the initial segmented frame based on the feature points; using the initial seed as the initial input of an energy model (such as an active contour model, graph cut model, geodesic active contour model, level set model, or other models with optimization objectives), performing iterative optimization of the model until convergence, and extracting the ventricular boundary contour of the initial segmented frame.

[0045] Here, based on the ventricular geometric reference provided by feature points, an initial seed adapted to the ventricular morphology is configured in the initial segmentation frame to limit the segmentation range. This initial seed is used as the initial input of an energy model with optimization objectives (such as an active contour model with shape constraints and region / edge feature awareness). The model is iteratively optimized to convergence, and the ventricular boundary contour of the initial segmentation frame is accurately extracted.

[0046] In some embodiments, setting an initial seed in the initial segmentation frame based on the feature points may include: placing a hyperellipse (including an ellipse) that conforms to the geometry of the left ventricle along the major axis as an inner seed based on the feature points, and setting a thin ring (a narrow ring structure with a radial width less than a preset width) as an outer seed on the periphery.

[0047] Here, as Figure 2 As shown, based on the major axis established by feature points and the ventricular geometric reference, a closed contour adapted to the ventricular morphology is configured as an inner seed in the initial segmentation frame, and a narrow ring structure is set around it as an outer seed, thus completing the construction of the initial seed. Using the inner and outer seeds as the initialization inputs of the energy model, iterative optimization is performed to extract the aforementioned ventricular boundary contour.

[0048] As described in the above embodiments, the energy model can be a model with an optimization objective, such as an active contour model, a graphical cut model, a geodesic active contour model, or a level set function. Depending on the energy model, the process of extracting the ventricular boundary contour will also be different.

[0049] For example, when the energy model adopts the active contour model, the ventricular boundary identification of the initial segmented frame may include: starting the active contour model with shape prior and edge potential, using the inner and outer seeds as initial inputs, using regional energy constraint terms (Chan-Vese regional terms) to ensure the statistical stability of intracavitary and extracavitary intensity and gradually convergence, and determining the ventricular boundary contour of the initial segmented frame.

[0050] Here, the hyperelliptical shape of the inner seed is used as a guide to limit the contour deformation during the model convergence process, ensuring that it fits the anatomical geometry of the left ventricle. A Chan-Vese regional energy constraint term is used to perform statistical analysis based on the intensity distribution of the inner seed (cavity) and the outer seed (myocardium) to ensure the stability of the intensity statistics inside and outside the cavity. In addition, edge information from the image can be combined to enhance the model's sensitivity to the boundary of brightness abrupt changes between the cavity and myocardium. In this way, through the synergistic effect of multiple constraint terms, the active contour model gradually converges iteratively, and finally accurately locks the ventricular boundary contour of the initial segmented frame.

[0051] For example, when the energy model adopts a graph cut model, the ventricular boundary identification of the initial segmented frame may include: starting the graph cut model, constructing a graph structure with pixels as nodes and neighborhood relationships as edges; setting source and sink points based on the initial seed, calculating the minimum cut to segment the graph structure, and determining the ventricular boundary contour of the initial segmented frame.

[0052] For example, when the energy model adopts the Geodesic Active Contour model, the ventricular boundary identification of the initial segmented frame may include: calculating the edge stopping function based on the image gradient, defining the geodesic energy function to shrink the contour towards the edge, iteratively optimizing the contour position to minimize the geodesic energy function, and determining the ventricular boundary contour of the initial segmented frame.

[0053] For example, when the energy model adopts the level set model, the contour is implicitly represented as the zero level set of the level set function. An energy function containing region terms, edge terms, and shape regularization terms is defined. The level set function is iteratively updated to minimize the energy function, and the ventricular boundary contour of the initial segmented frame is determined.

[0054] S103: For the image frames adjacent to the initial segmentation frame in the ventriculography sequence image data, the ventricular boundary contour of the initial segmentation frame is used as the starting contour. A ray is emitted in polar coordinates to search for boundary points within a narrow band. The boundary points are connected and region-driven segmentation is performed to obtain the ventricular boundary contour of the image frame. The narrow band range is determined based on the intersection of the current ray and the starting contour (for example, the target length is extended from the intersection of the current ray and the starting contour to both ends of the ray to form a narrow band).

[0055] Here, using the ventricular boundary contour of the initial segmented frame as the initial reference, a polar coordinate ray search method is employed for adjacent image frames to search for true boundary points within a focal range defined by the intersection of the ray and the initial reference contour. After connecting the boundary points, region-driven segmentation optimization is performed to obtain the ventricular boundary contour of that frame. The extension length of the aforementioned narrow band can be preset, for example, extending inward by 5 to 15 pixels and outward by 5 to 15 pixels to form a narrow band. Alternatively, for each ray, the intersection position of the ray and the initial contour is calculated, and the width of the narrow band is adaptively adjusted based on the degree of change in the ventricular boundary between the previous frame and the current frame; the narrow band is increased when the boundary change is large and decreased when the boundary change is small. Alternatively, for each ray, the intersection position of the ray and the initial contour is calculated, and the boundary position of the current frame is predicted based on the movement speed of the ventricular boundary in that direction in the previous frame, with the predicted position as the center to determine the narrow band.

[0056] In some embodiments, the feature points include apical points, and emitting rays in polar coordinates includes: using the apical point marked in the initial segmentation frame as the pole, or extracting the centroid or geometric center of the initial contour as the pole, or adaptively determining the pole based on the shape features of the initial contour, and emitting multiple rays at fixed angular intervals from the pole; searching for boundary points within a narrow band includes: using the intersection of each ray and the initial contour as the center, extending to both ends of the ray to form a narrow band; searching for boundary points within the narrow band on each ray to obtain the ventricular boundary contour of the image frame.

[0057] The above embodiments provide three methods for determining poles (the apex of the initial segmentation frame, the centroid / geometric center of the initial contour, and adaptive determination based on the shape features of the initial contour (dynamically calculating the pole that best fits the ventricular morphology of the current frame by analyzing the inherent shape features of the ventricular boundary contour of the previous frame)). Multiple rays are emitted from the poles at fixed angular intervals. A narrow band is defined extending towards both ends of each ray, centered on the intersection of each ray and the initial contour. Boundary points are searched within the narrow bands on each ray, ultimately forming the ventricular boundary contour of the current image frame. In specific implementations, the location with the most significant brightness abrupt change (the transition from the contrast-filled cavity to the myocardium) is searched within the narrow band, and this location is designated as the new boundary point in that direction.

[0058] In some embodiments, searching for boundary points within a narrow band may include: calculating the brightness gradient value of each pixel location using a gradient operator within the narrow band of each ray, and determining the location with the largest gradient value as the boundary point in that direction.

[0059] Here, gradient operators (such as the Sobel operator, Canny operator, or difference of Gaussian operator) are used to calculate the brightness gradient value of each pixel within the narrow band. The magnitude of the gradient value reflects the drastic change in pixel brightness. In DSA images, the ventricular boundary corresponds to the transition zone from the contrast-filled cavity (high brightness) to the myocardial tissue (low brightness), where the brightness abrupt change is most significant, and the corresponding gradient value is also the largest. Therefore, the pixel position with the largest gradient value within the narrow band can be determined as the ventricular boundary point in the current ray direction; if there is no obvious gradient peak within the narrow band (i.e., no clear brightness abrupt change), the boundary point position in that direction of the previous frame can be used to ensure boundary continuity.

[0060] In other embodiments, a comprehensive boundary confidence score can be calculated based on brightness gradient, edge response, and regional statistical characteristics within the narrow band of each ray, and the position with the highest confidence score can be determined as the boundary point in that direction. Alternatively, a dynamic programming method can be used to search for the optimal boundary point path within the narrow band of each ray (not judging the boundary point of a single ray in isolation, but considering the continuity of the boundary points of adjacent rays to find the optimal boundary path that is coherent in all directions), and the position on the optimal boundary point path can be determined as the boundary point.

[0061] In some embodiments, connecting the boundary points and performing region-driven segmentation may include: connecting the boundary points in polar angle order to form a closed curve, smoothing and / or shape-constraining the closed curve to obtain a processed closed curve; and performing region-driven segmentation based on the processed closed curve to obtain the ventricular boundary contour of the image frame.

[0062] Here, boundary points in each direction obtained by polar coordinate search are connected in polar angle order to form a closed initial contour; contour burrs are eliminated by smoothing or shape constraint processing to ensure that the contour fits the ventricular anatomy, thus completing the quality optimization of the initial contour; based on the optimized closed contour, region-driven segmentation is performed to further refine the boundary, and finally a ventricular boundary contour that accurately fits the image intensity boundary is obtained.

[0063] The smoothing process described above can employ methods such as moving average filtering, Gaussian filtering, and B-spline curve fitting. Shape constraint processing forces the closed contour to conform to the inherent anatomical shape of the ventricle. For example, using a hyperellipse that matches the geometry of the left ventricle as a reference, the connected closed contour is fitted with a hyperellipse. The deviation between the fitted shape and the initial closed contour is calculated, and boundary points on the contour that deviate too far from the fitted curve (such as points exceeding a preset deviation threshold) are adjusted to ensure that the overall contour maintains the core shape of the hyperellipse, conforming to the anatomical characteristics of the ventricle, which is approximately symmetrical and has one pointed end (apex) and one wide end (base).

[0064] In some embodiments, region-driven segmentation is performed based on the processed closed curve. The processed closed curve can be used as the initial input, and region segmentation can be achieved through energy optimization or iterative evolution of various energy models (including active contour, graph cut, level set, geodesic active contour, etc.), and finally the ventricular boundary contour of the image frame can be accurately extracted.

[0065] For example, an active contour model can be initiated based on the closed curve, and the ventricular boundary contour of the image frame can be obtained through energy minimization optimization; or, the internal and external regions of the closed curve can be initialized as foreground and background, and a graph cut model can be initiated to perform region-driven segmentation to obtain the ventricular boundary contour of the image frame. Another example is to construct a signed distance function based on the closed curve, initiate a level set model, and iteratively evolve the signed distance function until convergence to obtain the ventricular boundary contour of the image frame. Alternatively, a geodesic energy function can be defined based on the closed curve, an active geodesic contour model can be initiated, and the contour position can be iteratively optimized to minimize the geodesic energy function to obtain the ventricular boundary contour of the image frame.

[0066] S104: Using the image frame as a new initial segmentation frame, repeat the steps of searching for boundary points and region-driven segmentation for the next image frame to obtain the full sequence ventricular boundary contour.

[0067] Here, the next image frame is either the frame preceding or following the current image frame (the image frame from which the ventricular boundary contours are extracted). The current image frame is used again as the initial segmentation frame, and step S103 is repeated to extract the ventricular boundary contours of each image frame corresponding to the entire ventricular angiography sequence image data. Specifically, starting from the initial segmentation frame, the steps of searching for boundary points and region-driven segmentation can be performed bidirectionally forward and backward until the repetition termination condition is met (e.g., all image frames in the ventricular angiography sequence image data have been processed), thus obtaining the full sequence ventricular boundary contours.

[0068] In specific implementation, the condition for repetitive execution to stop can be that all image frames in the ventriculography sequence image data have been processed, or the validity of the boundary contour can be detected after each iteration. When the boundary contour is found to be invalid, the iteration is stopped (for example, when the smoothness index, the area change rate, or the similarity exceeds the preset threshold range, it is determined that the boundary contour is invalid and the iteration is stopped) and the user is prompted, thus obtaining the full sequence of ventricular boundary contours.

[0069] S105: Based on the full sequence of ventricular boundary contours, determine the ventricular cavity area sequence corresponding to the ventricular angiography sequence image data, and divide the cardiac phases according to the ventricular cavity area sequence.

[0070] Here, based on the full sequence of ventricular boundary contours, a corresponding ventricular cavity area sequence is generated, and the cardiac phase division is automatically completed based on the changing characteristics of this area sequence.

[0071] In some embodiments, determining the ventricular cavity area sequence corresponding to the ventricular angiography sequence image data may include: based on the full sequence of ventricular boundary contours, counting the number of pixels inside each frame of the ventricular boundary contour; converting the number of pixels into actual area units according to the pixel spacing parameter, and arranging the ventricular cavity area of ​​each frame of the ventricular boundary contour in chronological order to form a ventricular cavity area sequence.

[0072] Specifically, for example, the actual area unit is obtained by multiplying the square of the pixel spacing by the number of pixels. Subsequently, the ventricular cavity area sequence can be smoothed by applying median filtering (with a set filter window length) or moving average filtering to remove isolated spikes, resulting in a smoothed ventricular cavity area sequence.

[0073] In another implementation, based on the full sequence of ventricular boundary contours, the area value enclosed by the ventricular boundary contour of each frame can be calculated using Green's formula (a mathematical tool that connects the line integral of a closed plane curve with the double integral of the region enclosed by the curve), and arranged in chronological order to form the ventricular cavity area sequence.

[0074] In some embodiments, the step of dividing cardiac phases according to the ventricular cavity area sequence may include: determining the maximum area value of each cardiac cycle in the ventricular cavity area sequence, and determining the end-diastolic frame and end-systolic frame based on the image frame corresponding to the maximum area value and the image frame corresponding to the minimum area value.

[0075] Here, the cardiac cycle interval can be determined based on a preset heart rate range (which can be represented by the number of image frames). In one implementation, a minimum interval can be set to periodically identify the maximum and minimum area values, avoiding misjudging false peaks / troughs within the same cardiac cycle as true peaks / troughs in the next cycle due to minor fluctuations in the area curve (such as image noise or residual respiratory interference). For example, a preset percentage of the statistical average number of cardiac cycles (e.g., 24 frames) can be used as the minimum interval, such as 15 frames (ensuring no missed judgments across cycles while filtering out invalid fluctuations within the same cycle).

[0076] In some embodiments, determining the end-diastolic frame and the end-systolic frame based on the image frame corresponding to the maximum area and the image frame corresponding to the minimum area may include: The image frame corresponding to the maximum area is identified as the end-diastolic frame, and the image frame corresponding to the minimum area is identified as the end-systolic frame; or, within a preset number of frames near the maximum area, the moving speed of the ventricular boundary contour of each frame is calculated, and the frame with the smallest boundary moving speed is selected as the end-diastolic frame or the end-systolic frame.

[0077] In the above embodiments, one approach is to directly determine the image frame corresponding to the maximum area in the area curve as the end-diastolic frame and the image frame corresponding to the minimum area as the end-systolic frame. Another approach is to calculate the movement velocity of the ventricular boundary contour in each frame within a preset frame range (e.g., 3-5 frames before and after) near the extreme point when the extreme point may be affected by noise (e.g., image jitter, residual breathing). The frame with the smallest movement velocity is then selected as the final end-diastolic or end-systolic frame. Finally, the end-diastolic frame number, end-systolic frame number, corresponding time, and corresponding area value can be output.

[0078] The cardiac phase recognition method provided in this embodiment only requires the physician to manually mark a preset number of feature points on any frame of the ventriculography sequence. Automatic segmentation techniques, such as feature point matching (extracting ventricular boundary feature points from manually marked ventricular feature points in a single frame) and polar coordinate tracking (using the previous frame's ventricular contour as a starting reference, emitting rays at a fixed angle centered on the pole (e.g., the apex of the heart), searching for brightness abrupt change boundary points within the corresponding narrow band of the ray, and achieving continuous tracking of the current frame's ventricular contour after continuous refinement), can complete the ventricular contour segmentation of the entire sequence. This eliminates the need for physicians to manually segment multiple frames one by one, significantly reducing manual operation time and effectively meeting the actual needs of rapid clinical diagnosis. Furthermore, based on a unified feature point matching algorithm, ventricular area calculation rules, and cardiac phase recognition logic, automatic segmentation of the entire sequence and automatic determination of diastolic and systolic phases can be achieved. This avoids fluctuations in results caused by human judgment differences, improving the accuracy and stability of ventricular contour segmentation results and key cardiac phase recognition, providing reliable data support for subsequent cardiac function assessment.

[0079] like Figure 3 The diagram shown is a schematic of a cardiac phase recognition device 300 provided in an embodiment of this disclosure, comprising: The acquisition module 31 is used to acquire ventricular angiography sequence image data and select an initial segmentation frame in the ventricular angiography sequence image data; The initial contour determination module 32 is used to identify the ventricular boundary of the initial segmented frame based on the feature points marked in the initial segmented frame, and determine the ventricular boundary contour of the initial segmented frame. The sequence frame contour determination module 33 is used to, for image frames adjacent to the initial segmentation frame in the ventriculography sequence image data, take the ventricular boundary contour of the initial segmentation frame as the starting contour, emit rays in polar coordinates to search for boundary points within a narrow band, connect the boundary points and perform region-driven segmentation to obtain the ventricular boundary contour of the image frame; the narrow band range is determined based on the intersection of the current ray and the starting contour; take the image frame as the new initial segmentation frame, and repeat the steps of searching for boundary points and region-driven segmentation for the next image frame to obtain the full sequence ventricular boundary contour; The cardiac phase segmentation module 34 is used to determine the ventricular cavity area sequence corresponding to the ventricular angiography sequence image data based on the full sequence ventricular boundary contour, and to perform cardiac phase segmentation based on the ventricular cavity area sequence.

[0080] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0081] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0082] Based on the same technical concept, this disclosure also provides a computer device 400, referring to... Figure 4 The diagram shown is a schematic representation of the structure of a computer device according to an exemplary embodiment of this disclosure, comprising: The processor 410, memory 420, and bus 430 are included. The memory 420 is used to store execution instructions and includes main memory 421 and external memory 422. The main memory 421, also known as internal memory, is used to temporarily store the operation data in the processor 410 and the data exchanged with external memory 422 such as hard disk. The processor 410 exchanges data with external memory 422 through main memory 421.

[0083] In this embodiment, the memory 420 is specifically used to store application code that executes the scheme of this disclosure, and its execution is controlled by the processor 410. That is, when the computer device 400 is running, the processor 410 communicates with the memory 420 through the bus 430, or the processor 410 communicates with the memory 420 through other means, so that the processor 410 executes the application code stored in the memory 420, and then performs the steps of the cardiac phase recognition method described in any of the foregoing embodiments.

[0084] The memory 420 may be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.

[0085] Processor 410 may be an integrated circuit chip with signal processing capabilities. The aforementioned processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor.

[0086] It is understood that the structures illustrated in the embodiments of this disclosure do not constitute a specific limitation on the computer device 400. In other embodiments of this disclosure, the computer device 400 may include more or fewer components than illustrated, or combine some components, or split some components, or have different component arrangements. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

[0087] This disclosure also provides a computer-readable storage medium including instructions stored thereon, wherein, when executed by a processor, the cardiac phase recognition method described in any of the preceding claims is executed. The storage medium may be a volatile or non-volatile computer-readable storage medium.

[0088] This disclosure also provides a computer program product, which stores a computer program. When the computer program is run by a processor, it executes the steps of the cardiac phase recognition method provided in any of the above embodiments of this disclosure. For details, please refer to the above method embodiments, which will not be repeated here.

[0089] The aforementioned computer program product can be implemented through hardware, software, or a combination thereof. In one optional embodiment, the computer program product is specifically embodied in a computer storage medium, which can be a volatile or non-volatile computer-readable storage medium. In another optional embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0090] Furthermore, embodiments of the subject matter and functional operation described in this specification can be implemented in the following ways: digital electronic circuits, tangibly embodied computer software or firmware, computer hardware including the structures disclosed in this specification and their structural equivalents, or combinations thereof. Embodiments of the subject matter described in this specification can be implemented as one or more computer programs, i.e., one or more modules of computer program instructions encoded on a tangible, non-transitory program carrier for execution by a data processing apparatus or for controlling the operation of a data processing apparatus. Alternatively or additionally, program instructions may be encoded on artificially generated propagation signals, such as machine-generated electrical, optical, or electromagnetic signals, which are generated to encode information and transmit it to a suitable receiving device for execution by the data processing apparatus. The computer storage medium may be a machine-readable storage device, a machine-readable storage substrate, a random or serial access memory device, or combinations thereof.

[0091] The processing and logic flows described in this specification can be executed by one or more programmable computers that execute one or more computer programs to perform corresponding functions by operating on input data and generating output. The processing and logic flows can also be executed by dedicated logic circuits—such as field-programmable gate arrays (FPGAs) or application-specific integrated circuits (ASICs), and the device can also be implemented as dedicated logic circuits.

[0092] Suitable computers for executing computer programs include, for example, general-purpose and / or special-purpose microprocessors, or any other type of central processing unit. Typically, the central processing unit receives instructions and data from read-only memory and / or random access memory. The basic components of a computer include a central processing unit for implementing or executing instructions and one or more memory devices for storing instructions and data. Typically, a computer will also include one or more mass storage devices for storing data, such as disks, magneto-optical disks, or optical disks, or the computer will be operatively coupled to such mass storage devices to receive data from or transfer data to them, or both. However, a computer is not required to have such devices. Furthermore, a computer can be embedded in another device, such as a mobile phone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a Global Positioning System (GPS) receiver, or a portable storage device such as a Universal Serial Bus (USB) flash drive, to name a few.

[0093] Computer-readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media, and memory devices, such as semiconductor memory devices (e.g., EPROM, EEPROM, and flash memory devices), magnetic disks (e.g., internal hard disks or removable disks), magneto-optical disks, and CD-ROM and DVD-ROM disks. Processors and memory may be supplemented by or incorporated into dedicated logic circuitry.

[0094] While this specification contains numerous specific implementation details, these should not be construed as limiting the scope of any invention or the scope of the claims, but rather are primarily intended to describe features of specific embodiments of a particular invention. Certain features described in the various embodiments herein may also be implemented in combination in a single embodiment. Conversely, various features described in a single embodiment may also be implemented separately in various embodiments or in any suitable sub-combination. Furthermore, while features may function in certain combinations as described above and even initially claimed in this way, one or more features from a claimed combination may be removed from that combination in some cases, and a claimed combination may refer to a sub-combination or a variation thereof.

[0095] Similarly, although the operations are depicted in a specific order in the accompanying drawings, this should not be construed as requiring these operations to be performed in the specific order shown or sequentially, or requiring all illustrated operations to be performed to achieve the desired result. In some cases, multitasking and parallel processing may be advantageous. Furthermore, the separation of various system modules and components in the above embodiments should not be construed as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.

[0096] Thus, specific embodiments of the subject matter have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions recited in the claims may be performed in a different order and still achieve the desired result. Furthermore, the processes depicted in the drawings are not necessarily shown in a specific order or sequence to achieve the desired result. In some implementations, multitasking and parallel processing may be advantageous.

[0097] The above description is merely a preferred embodiment of this disclosure and is not intended to limit this disclosure. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for identifying the phase of heartbeat, characterized in that, include: Acquire ventricular angiography sequence image data, and select an initial segmentation frame from the ventricular angiography sequence image data; Based on the feature points marked in the initial segmented frame, ventricular boundary recognition is performed on the initial segmented frame to determine the ventricular boundary contour of the initial segmented frame. For the image frames adjacent to the initial segmentation frame in the ventriculography sequence image data, the ventricular boundary contour of the initial segmentation frame is used as the starting contour. A ray is emitted in polar coordinates to search for boundary points within a narrow band. The boundary points are connected and region-driven segmentation is performed to obtain the ventricular boundary contour of the image frame. The narrow band range is determined based on the intersection of the current ray and the starting contour. Using the image frame as the new initial segmentation frame, the steps of searching for boundary points and region-driven segmentation are repeated for the next image frame to obtain the full sequence of ventricular boundary contours. Based on the full sequence of ventricular boundary contours, the ventricular cavity area sequence corresponding to the ventricular angiography sequence image data is determined, and cardiac phases are divided according to the ventricular cavity area sequence.

2. The cardiac phase recognition method according to claim 1, characterized in that, The feature points include the upper edge of the aortic root, the lower edge of the aortic root, and the apex of the heart; or... The feature points include the apex of the heart, the left hinge of the mitral annulus, and the right hinge of the mitral annulus.

3. The cardiac phase recognition method according to claim 1, characterized in that, The step of identifying ventricular boundaries in the initial segmented frame includes: Construct a region of interest (ROI) based on the feature points; Ventricular boundary identification is performed on the initial segmented frame within the ROI to determine the ventricular boundary contour of the initial segmented frame.

4. The cardiac phase recognition method according to claim 3, characterized in that, The construction of the ROI based on the feature points includes: The basal position is determined based on the feature points other than the apex, and the apex and the basal position are connected to form the major axis; or, the ventricular cavity of the initial segmented frame is coarsely segmented based on the feature points, and principal component analysis is performed on the coarse segmentation result to obtain the major axis direction; or, the ventricular cavity of the initial segmented frame is coarsely segmented based on the feature points, and ellipse fitting is performed on the coarse segmentation result to obtain the major axis direction of the fitted ellipse. The ROI is constructed with the apex of the heart as the pole and the long axis direction as the reference. The angle range of the ROI covers the left ventricular cavity, and the upper limit of the radius is a preset multiple of the long axis length.

5. The cardiac phase recognition method according to claim 1, characterized in that, The step of identifying ventricular boundaries in the initial segmented frame includes: An initial seed is set in the initial segmentation frame based on the feature points; The initial seed is used as the initial input for the energy model, and the model is iteratively optimized until convergence is achieved, thereby extracting the ventricular boundary contour of the initial segmented frame.

6. The cardiac phase recognition method according to claim 5, characterized in that, Setting an initial seed in the initial segmentation frame based on the feature points includes: Based on the aforementioned feature points, a hyperellipse conforming to the geometry of the left ventricle is placed along the long axis as an inner seed, and a thin ring is set around the periphery as an outer seed; the thin ring refers to a ring-shaped structure with a radial width smaller than a preset width.

7. The cardiac phase recognition method according to claim 1, characterized in that, The feature points include the apex point. The method of emitting rays in polar coordinates includes: taking the apex point marked in the initial segmentation frame as the pole, or extracting the centroid or geometric center of the starting contour as the pole, or adaptively determining the pole based on the shape features of the starting contour, and emitting multiple rays from the pole at fixed angle intervals. The step of searching for boundary points within a narrow band includes: extending a narrow band to both ends of each ray, centered on the intersection of each ray and the initial contour; and searching for boundary points within the narrow band on each ray to obtain the ventricular boundary contour of the image frame.

8. The cardiac phase recognition method according to claim 1, characterized in that, The search for boundary points within a narrow band includes: Within the narrow band of each ray, the brightness gradient value of each pixel position is calculated using a gradient operator, and the position with the largest gradient value is determined as the boundary point in that direction.

9. The cardiac phase recognition method according to claim 1, characterized in that, The process of connecting the boundary points and performing region-driven segmentation includes: Connect the boundary points in polar angle order to form a closed curve, and then perform smoothing and / or shape constraint processing on the closed curve to obtain the processed closed curve. Based on the processed closed curve, region-driven segmentation is performed to obtain the ventricular boundary contour of the image frame.

10. The cardiac phase recognition method according to claim 9, characterized in that, The process of performing region-driven segmentation based on the processed closed curve includes: The active contour model is initiated based on the closed curve, and the ventricular boundary contour of the image frame is obtained through energy minimization optimization; or, Using the inner and outer regions of the closed curve as foreground and background initialization, the graph cut model is started to perform region-driven segmentation to obtain the ventricular boundary contour of the image frame.

11. The cardiac phase recognition method according to claim 1, characterized in that, The step of determining the ventricular cavity area sequence corresponding to the ventricular angiography sequence image data includes: Based on the full sequence of ventricular boundary contours, the number of pixels inside the ventricular boundary contours of each frame is counted. The number of pixels is converted into actual area units based on the pixel spacing parameter, and the ventricular cavity area of ​​each frame of ventricular boundary contour is arranged in chronological order to form the ventricular cavity area sequence.

12. The cardiac phase recognition method according to claim 1, characterized in that, The step of dividing cardiac phases based on the ventricular cavity area sequence includes: In the ventricular cavity area sequence, the maximum area value for each cardiac cycle is determined. Based on the image frame corresponding to the maximum area value and the image frame corresponding to the minimum area value, the end-diastolic frame and the end-systolic frame are determined.

13. The cardiac phase recognition method according to claim 12, characterized in that, Based on the image frames corresponding to the maximum and minimum area values, the end-diastolic frames and end-systolic frames are determined, including: The image frame corresponding to the maximum area is identified as the end-diastolic frame, and the image frame corresponding to the minimum area is identified as the end-systolic frame; or... Within a preset number of frames near the positive value of the surface area, the moving speed of the ventricular boundary contour in each frame is calculated, and the frame with the smallest boundary moving speed is selected as the end-diastolic frame or the end-systolic frame.

14. A cardiac phase recognition device, characterized in that, include: The acquisition module is used to acquire ventricular angiography sequence image data and select an initial segmentation frame in the ventricular angiography sequence image data; The initial contour determination module is used to identify the ventricular boundary of the initial segmented frame based on the feature points marked in the initial segmented frame, and determine the ventricular boundary contour of the initial segmented frame. The sequence frame contour determination module is used to, for image frames adjacent to the initial segmentation frame in the ventriculography sequence image data, take the ventricular boundary contour of the initial segmentation frame as the starting contour, emit rays in polar coordinates to search for boundary points within a narrow band, connect the boundary points and perform region-driven segmentation to obtain the ventricular boundary contour of the image frame; the narrow band range is determined based on the intersection of the current ray and the starting contour; take the image frame as the new initial segmentation frame, and repeat the steps of searching for boundary points and region-driven segmentation for the next image frame to obtain the full sequence ventricular boundary contour; The cardiac phase segmentation module is used to determine the ventricular cavity area sequence corresponding to the ventricular angiography sequence image data based on the full sequence ventricular boundary contour, and to segment the cardiac phase according to the ventricular cavity area sequence.

15. A computer device, characterized in that, The computer device includes a processor and a memory, the memory storing computer-readable instructions, and the processor being configured to read the computer-readable instructions to execute the cardiac phase recognition method according to any one of claims 1 to 13.

16. A computer-readable storage medium, characterized in that, Includes instructions stored thereon, wherein, when the instructions are executed by a processor, the cardiac phase recognition method as described in any one of claims 1-13 is executed.

Images