A method of vessel segmentation for a coronary angiography sequence and related apparatus

Abstract

Embodiments of the present application disclose a coronary angiography sequence blood vessel segmentation method and related equipment to solve the problem that in the prior art, due to the difficulty of a rectangular region of interest to fit the target blood vessel direction, the target blood vessel segmentation process is prone to branch segmentation. The method comprises: acquiring a coronary angiography image to be segmented, and acquiring blood vessel center line information for representing the target blood vessel direction in the coronary angiography image to be segmented; generating a strip-shaped polygon region of interest ROI extending along the target blood vessel direction according to the blood vessel center line information; wherein the strip-shaped boundary of the strip-shaped polygon region of interest ROI is determined according to the lateral offset positions at the plurality of center line positions in the blood vessel center line information; performing input constraint processing on the coronary angiography image to be segmented based on the strip-shaped polygon region of interest ROI, to generate a constraint segmentation input image; and performing blood vessel segmentation based on the constraint segmentation input image to obtain a target segmentation result of the coronary angiography image to be segmented.

Description

Technical Field

[0001] This application relates to the field of medical image processing technology, and in particular to a method and related equipment for vascular segmentation in coronary angiography sequences. Background Technology

[0002] Coronary angiography image sequences are continuous imaging results of the coronary arteries during the contrast agent filling and dissipation process. During coronary angiography image analysis, it is usually necessary to segment the target vessel in the angiography sequence frame by frame to obtain structural information of the target vessel in different image frames, thereby providing basic data for quantitative stenosis analysis, fractional flow reserve calculation, and subsequent vascular modeling.

[0003] In existing coronary angiography sequence segmentation methods, to reduce the interference of background regions and non-target vessels on the segmentation results, a region of interest (ROI) is typically defined, and the segmentation range is constrained based on this ROI. For example, a rectangular ROI can be defined near the target vessel, allowing the segmentation algorithm to search for vessels within this rectangular ROI region.

[0004] However, coronary arteries typically exhibit complex morphologies in angiographic images, such as tortuosity, bifurcation, overlap, and local proximity, making it difficult for rectangular regions of interest (ROIs) to accurately reflect the tortuous course of the target vessel. Secondly, to fully cover the target vessel, the rectangular ROI often needs to encompass a large image area, thus including adjacent branch vessels, overlapping vessels, and other background structures. Especially at vessel bifurcation points, where the target vessel and adjacent branches are spatially close, if the segmentation algorithm searches for vascular structures within a rectangular ROI containing numerous non-target regions, it is prone to missegmenting adjacent branches as the target vessel, resulting in cross-branch segmentation.

[0005] Furthermore, while some existing methods utilize regions of interest (ROIs) to filter candidate segmentation results, these ROIs are merely used as posterior filtering conditions after segmentation. They do not alter the input content of the current frame image before the segmentation algorithm runs, nor do they limit the search space of the segmentation algorithm from the input end. Therefore, when the segmentation algorithm has already generated erroneous segmentation results with structural continuity and complete morphology in non-target blood vessel regions, relying solely on posterior filtering is insufficient to fundamentally prevent the generation of erroneous segmentation results. Summary of the Invention

[0006] This application provides a method for segmenting blood vessels in a coronary angiography sequence to solve the problem in the prior art where the rectangular region of interest is difficult to fit the direction of the target blood vessel, which leads to the easy occurrence of branch segmentation during the segmentation of the target blood vessel.

[0007] This application also provides a vessel segmentation device for coronary angiography sequences, an electronic device, a computer-readable storage medium, and a computer program product.

[0008] The embodiments of this application adopt the following technical solutions: In a first aspect, embodiments of this application provide a method for vessel segmentation in a coronary angiography sequence, including: Acquire the coronary angiography image to be segmented, and obtain the vessel centerline information used to characterize the orientation of the target vessel in the coronary angiography image to be segmented; A region of interest (ROI) of polygonal shape extending along the direction of the target blood vessel to be segmented is generated based on the blood vessel centerline information; wherein, the strip-shaped boundary of the ROI of polygonal shape is determined based on the lateral offset position at multiple centerline positions in the blood vessel centerline information. The input constraint processing is performed on the coronary angiography image to be segmented based on the region of interest (ROI) of the strip polygon to generate a constrained segmentation input image. Based on constraint segmentation of the input image, blood vessel segmentation is performed to obtain the target segmentation result of the coronary angiography image to be segmented.

[0009] Optionally, obtain vessel centerline information to characterize the orientation of the target vessel in the coronary angiography image to be segmented, including: Determine the coronary angiography image sequence containing the coronary angiography images to be segmented, and determine the reference frame from the coronary angiography image sequence; Obtain the reference centerline of the reference blood vessel corresponding to the target blood vessel to be segmented in the reference frame; The reference centerline is used as the vessel centerline information to characterize the orientation of the target vessel in the coronary angiography image to be segmented.

[0010] Optionally, obtain vessel centerline information to characterize the orientation of the target vessel in the coronary angiography image to be segmented, including: Identify neighboring frames adjacent to the coronary angiography image to be segmented from the coronary angiography image sequence containing the coronary angiography image to be segmented; Determine whether the neighboring frame has completed the segmentation of the blood vessel corresponding to the target blood vessel and whether the segmentation result is valid; If the blood vessels in the neighboring frame have been segmented and the segmentation result is valid, then the center line of the blood vessels in the neighboring frame is obtained and the center line of the blood vessels in the neighboring frame is determined as the blood vessel center line information. If the segmentation of the blood vessels in the neighboring frame is not completed, or the segmentation result of the blood vessels in the neighboring frame is invalid, a reference frame is determined from the coronary angiography image sequence, and the reference center line of the reference blood vessel corresponding to the target blood vessel to be segmented in the reference frame is obtained. The reference center line is used as the information for the vascular center line.

[0011] Optionally, a region of interest (ROI) of the shape extending along the direction of the target blood vessel to be segmented is generated based on the vessel centerline information, including: The vascular centerline corresponding to the vascular centerline information is determined as the target centerline, and multiple centerline positions are selected along the target centerline; Determine the local extension direction at each centerline position based on the positions of adjacent centerlines; Determine the direction of the unit normal perpendicular to the direction of local extension based on the direction of local extension; Based on the position of each centerline and the corresponding unit normal direction, determine the first side lateral offset position and the second side lateral offset position on both sides of the target centerline. Multiple first-side lateral offset positions are connected in the order of extension of the target center line to form a first-side strip boundary, and multiple second-side lateral offset positions are connected in the order of extension of the target center line to form a second-side strip boundary. A closed polygonal region is formed based on the first and second side strip boundaries, and the closed polygonal region is determined as the region of interest (ROI) of the strip polygon.

[0012] Optionally, if the blood vessels in the adjacent frame have been segmented and the segmentation result is valid, a strip-shaped polygonal region of interest (ROI) extending along the direction of the target blood vessel to be segmented is generated based on the blood vessel centerline information, including: The center line of the blood vessel in the adjacent frame is determined as the target center line, and multiple center line positions are selected along the extension direction of the target center line; Determine the local extension direction at each centerline position based on the positions of adjacent centerlines; Determine the direction of the unit normal perpendicular to the direction of local extension based on the direction of local extension; Based on the position of each centerline and the corresponding unit normal direction, determine the first side lateral offset position and the second side lateral offset position on both sides of the target centerline. Multiple first-side lateral offset positions are connected in the order of extension of the target center line to form a first-side strip boundary, and multiple second-side lateral offset positions are connected in the order of extension of the target center line to form a second-side strip boundary. A closed polygonal region is formed based on the first and second side strip boundaries, and the closed polygonal region is determined as the region of interest (ROI) of the strip polygonal image to be segmented in the coronary angiography image.

[0013] Optionally, if the blood vessels in a neighboring frame are not segmented, or the segmentation result of the blood vessels in a neighboring frame is invalid, a strip-shaped polygonal region of interest (ROI) extending along the direction of the target blood vessel to be segmented is generated based on the blood vessel centerline information, including: Obtain the reference strip polygon region generated based on the reference centerline; Obtain the outer distance of the region corresponding to the reference strip polygon region; The reference strip polygon region is expanded based on the region expansion distance to obtain the expanded reference strip polygon region. Based on the image boundary of the coronary angiography image to be segmented, the boundary of the expanded reference strip polygon region is clipped. The cropped, extended reference polygonal region is determined as the region of interest (ROI) of the coronary angiography image to be segmented.

[0014] Optionally, the coronary angiography image to be segmented is subjected to input constraint processing based on the region of interest (ROI) of the strip polygon, generating a constrained segmentation input image, including: Based on the region of interest (ROI) of a polygonal strip, pixels within and outside the region in the coronary angiography image to be segmented are determined; where pixels within the region are those located within the ROI of the polygonal strip, and pixels outside the region are those located outside the ROI of the polygonal strip. The original image information of pixels within the region is preserved, and the gray values ​​of pixels outside the region are adjusted to the preset background gray values ​​to generate a constrained segmentation input image.

[0015] Optionally, before performing input constraint processing on the coronary angiography image to be segmented based on the region of interest (ROI) of the strip polygon to generate a constrained segmentation input image, the method further includes: Perform full-image initial vessel segmentation on the coronary angiography image to be segmented to obtain the initial segmentation result; Initial candidate vessels are determined based on the initial segmentation results, and the initial candidate vessel feature information is obtained. A reference frame is determined from the coronary angiography image sequence containing the coronary angiography image to be segmented. The reference frame is segmented to obtain the reference vessel corresponding to the target vessel to be segmented, and the reference vessel feature information of the reference vessel is obtained. Based on the initial candidate vessel feature information and the reference vessel feature information, the initial candidate vessels are preliminarily identified as standard vessels of the same branch. If the initial candidate vessel fails the preliminary judgment of the standard same branch vessel, the step of performing input constraint processing on the coronary angiography image to be segmented based on the region of interest (ROI) of the strip polygon is performed to generate a constrained segmentation input image.

[0016] Optionally, vessel segmentation is performed based on the constraint segmentation input image to obtain the target segmentation result of the coronary angiography image to be segmented, including: Perform blood vessel segmentation on the constrained segmentation input image to obtain the resegmentation result; Based on the resegmentation results, candidate vessels for resegmentation are determined, and the candidate vessel feature information is obtained. Based on the candidate vessel feature information and the reference vessel feature information, the determination of the same branch vessel is relaxed for the re-segmented candidate vessels. If the candidate vessel for resegmentation is determined by broadening the criteria for the same branch vessel, then the resegmentation result is determined as the target segmentation result for the coronary angiography image to be segmented. If the candidate vessel for resegmentation fails the broadened co-branch vessel determination, then a soft mask-enhanced fusion resegmentation based on a strip mask is performed to obtain the soft mask-enhanced fusion resegmentation result, and the target segmentation result of the coronary angiography image to be segmented is determined based on the soft mask-enhanced fusion resegmentation result.

[0017] Optionally, perform soft mask enhancement fusion resegmentation based on strip masks to obtain soft mask enhancement fusion resegmentation results, including: Obtain the reference center line of the reference vessel corresponding to the target vessel to be segmented in the reference frame, and the center line of the neighboring vessel of the adjacent frame that is adjacent to the coronary angiography image to be segmented and has been effectively segmented. The centerline points contained in the reference centerline are merged with the centerline points contained in the blood vessel centerlines of the adjacent frames to obtain a merged centerline point set; Using each centerline point in the merged centerline point set as the center, generate filled circular regions according to the preset mask radius; Multiple filled circular regions are combined to obtain a binary strip mask. Generating a soft mask fusion image based on a binary strip mask; Blood vessel segmentation is performed on the soft mask fusion image to obtain the soft mask enhanced fusion resegmentation result.

[0018] Secondly, embodiments of this application provide a vessel segmentation device for coronary angiography sequences, including an acquisition module, a generation module, a processing module, and a segmentation module, wherein: The acquisition module is used to acquire the coronary angiography image to be segmented and to acquire the vessel centerline information used to characterize the orientation of the target vessel to be segmented in the coronary angiography image to be segmented. The generation module is used to generate a strip-shaped polygonal region of interest (ROI) extending along the direction of the target blood vessel to be segmented based on the blood vessel centerline information; wherein, the strip-shaped boundary of the ROI is determined based on the lateral offset position at multiple centerline positions in the blood vessel centerline information. The processing module is used to perform input constraint processing on the coronary angiography image to be segmented based on the region of interest (ROI) of the strip polygon, and generate a constrained segmentation input image; The segmentation module is used to segment blood vessels from the input image based on constraints, and obtain the target segmentation result of the coronary angiography image to be segmented.

[0019] Thirdly, embodiments of this application provide an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the coronary angiography sequence vessel segmentation method as described above.

[0020] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the coronary angiography sequence vessel segmentation method as described above.

[0021] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the vessel segmentation method for coronary angiography sequences as described above.

[0022] The above-described technical solutions adopted in the embodiments of this application can achieve the following beneficial effects: The method provided in this application involves acquiring vessel centerline information to characterize the orientation of the target vessel to be segmented, and generating a strip-shaped polygonal Region of Interest (ROI) extending along the orientation of the target vessel based on the lateral offset positions at multiple centerline locations in the vessel centerline information. Then, input constraint processing is applied to the coronary angiography image to be segmented based on this strip-shaped polygonal ROI to generate a constrained segmentation input image, which is then used for vessel segmentation. This approach offers several advantages. First, the strip-shaped polygonal ROI extends along the curvature of the target vessel, reducing the inclusion of adjacent branch vessels, overlapping vessels, and background structures in the segmentation search compared to a rectangular ROI. Second, since the strip-shaped polygonal ROI is not only used for candidate result selection after segmentation but also participates in the input constraint processing of the coronary angiography image before vessel segmentation, it limits the search space of the segmentation algorithm from the segmentation input end, reducing the possibility of the segmentation algorithm forming structurally continuous and morphologically complete erroneous segmentation results in non-target vessel regions. This improves the segmentation accuracy of the target vessel in the coronary angiography sequence and reduces the risk of cross-branch segmentation between the target vessel and adjacent branch vessels. Attached Figure Description

[0023] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A schematic diagram illustrating the implementation process of a vessel segmentation method for a coronary angiography sequence provided in this application embodiment; Figure 2A schematic diagram of a region of interest (ROI) in the shape of a polygon provided in this application embodiment; Figure 3 A schematic diagram of a constrained segmentation input image provided in an embodiment of this application; Figure 4 A schematic diagram illustrating the segmentation result and initial segmentation result of an exemplary reference frame provided for embodiments of this application; Figure 5 A schematic diagram illustrating the target segmentation result of a re-segmented coronary angiography image to be segmented, provided in an embodiment of this application; Figure 6 This application provides a schematic diagram of the specific structure of a vessel segmentation device for coronary angiography sequences. Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0025] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0026] To address the problem in existing technologies where rectangular regions of interest are difficult to fit the target vessel's course, leading to the easy occurrence of branch segmentation during target vessel segmentation, this application provides a vessel segmentation method for coronary angiography sequences.

[0027] The execution subject of this method can be various types of computing devices, or it can be an application or app installed on the computing device. The computing device can be a user terminal such as a mobile phone, tablet computer, or smart wearable device, or it can be a server.

[0028] For ease of description, this application uses a server as the execution subject of the method in its embodiments to illustrate the method. Those skilled in the art will understand that this embodiment uses a server as an example to describe the method, which is merely an illustrative example and does not limit the scope of protection of the corresponding claims.

[0029] Specifically, the implementation flow of the method provided in this application embodiment is as follows: Figure 1 As shown, it includes the following steps: Step 102: Obtain the coronary angiography image to be segmented, and obtain the vessel centerline information used to characterize the orientation of the target vessel to be segmented in the coronary angiography image to be segmented.

[0030] The coronary angiography image to be segmented refers to the frame in the coronary angiography image sequence that currently requires vessel segmentation processing. Here, the coronary angiography image sequence refers to multiple consecutive coronary angiography images of the same coronary artery region acquired during the contrast agent filling, flow, or dissipation process.

[0031] The target vessel to be segmented refers to the main or branch coronary artery that needs to be continuously tracked and segmented in the coronary angiography image to be segmented. It can be specified by the user in the reference frame or determined by the system based on the segmentation results of the reference frame.

[0032] In some embodiments, when acquiring coronary angiography images to be segmented and obtaining vessel centerline information to characterize the orientation of the target vessel in the coronary angiography images, a coronary angiography image sequence can be acquired first. After acquiring the coronary angiography image sequence, a reference frame can be determined from the sequence. The reference frame refers to an image in the coronary angiography image sequence whose vessel visualization quality meets preset conditions and is suitable as a basis for vessel tracking in subsequent frames. For example, a frame with superior vessel visualization quality can be determined from the coronary angiography image sequence as the reference frame based on vessel visualization clarity, vessel length, vessel contrast, vessel continuity, or a pre-trained optimal frame selection model.

[0033] Subsequently, full-image vessel segmentation is performed on the reference frame to obtain reference vessels corresponding to the target vessel to be segmented, and reference centerlines are extracted from these reference vessels. Here, a reference vessel refers to the vessel in the reference frame that corresponds to the target vessel to be segmented; the reference centerline is a sequence of points located in the lumen or middle of the reference vessel's outline, used to characterize the extension path of the reference vessel. This reference centerline may include multiple centerline points arranged sequentially according to the vessel's extension direction, and each centerline point may have corresponding image coordinates.

[0034] Finally, the reference centerline, or at least one of the following information obtained from the reference centerline: the coordinates of the centerline points, the order of the centerline points, the start and end points of the centerline, and the direction of the centerline extension, can be determined as the vessel centerline information. This completes the acquisition of the vessel centerline information used to characterize the orientation of the target vessel in the coronary angiography image to be segmented.

[0035] It should be noted that in this embodiment, obtaining the vessel centerline information to characterize the orientation of the target vessel in the coronary angiography image is based on the consideration that coronary arteries in coronary angiography images typically exhibit morphological features such as bending, bifurcation, local proximity, and inter-frame displacement. If only a rectangular region of interest with a fixed shape is used to constrain the segmentation range, the rectangular region of interest is difficult to change with the bending orientation of the target vessel, and it is easy to include adjacent branch vessels, overlapping vessels, or background structures in the segmentation search range. Based on this, in this embodiment, by obtaining the vessel centerline information that can characterize the orientation of the target vessel, multiple centerline positions can be determined along the vessel centerline information, and the corresponding lateral offset direction can be determined according to the local extension direction at each centerline position, thereby forming a strip-shaped boundary extending along the orientation of the vessel on both sides of the target vessel. The resulting strip-shaped polygonal region of interest can fit the bending path of the target vessel with the target vessel centerline as the skeleton, so that when performing input constraint processing on the coronary angiography image to be segmented, the possibility of non-target branch vessels entering the segmentation search space can be reduced from the segmentation input end, thereby reducing the risk of cross-branch segmentation during the target vessel segmentation process.

[0036] In another optional implementation, considering that during the frame-by-frame vessel segmentation process of the coronary angiography image sequence, the target vessel to be segmented will be affected by the heartbeat, respiratory movement, and changes in contrast agent filling, resulting in positional shifts between different image frames, and that neighboring frames that are closer in time to the coronary angiography image to be segmented are generally better able to reflect the actual position and direction of the target vessel to be segmented at the current moment than reference frames, in this embodiment of the application, when obtaining the vessel centerline information used to characterize the direction of the target vessel to be segmented in the coronary angiography image to be segmented, neighboring frames adjacent to the coronary angiography image to be segmented can also be determined from the coronary angiography image sequence containing the coronary angiography image to be segmented.

[0037] In this context, a neighboring frame can be a frame that has been processed before the coronary angiography image to be segmented during frame-by-frame segmentation, or a frame that has been processed after the coronary angiography image to be segmented. For example, when segmenting along the forward time sequence of the coronary angiography image sequence, the neighboring frame can be the frame preceding the coronary angiography image to be segmented; when segmenting along the backward time sequence of the coronary angiography image sequence, the neighboring frame can be the frame following the coronary angiography image to be segmented.

[0038] After determining the neighboring frames, it can be determined whether the neighboring frames have completed the segmentation of the neighboring frame vessels corresponding to the target vessel to be segmented and whether the segmentation result is valid. Here, the neighboring frame vessels refer to the vessels in the neighboring frames that correspond to the target vessel to be segmented; the segmentation result is valid if the segmentation result of the neighboring frame vessels meets the preset validity conditions, such as the number of centerline points of the neighboring frame vessels meeting the preset number requirement, the length of the neighboring frame vessels meeting the preset length requirement, or the neighboring frame vessels have been identified through the same branch vessels between them and the reference vessel.

[0039] If the adjacent frame has already segmented the vessels and the segmentation result is valid, the centerline of the adjacent frame vessels can be obtained and determined as the vessel centerline information. Since the adjacent frame and the coronary angiography image to be segmented are acquired adjacently in time, the centerline of the adjacent frame vessels can more directly reflect the spatial position and extension direction of the target vessel near the current frame. Therefore, the vessel centerline information generated based on the centerline of the adjacent frame vessels can better adapt to the inter-frame displacement of the target vessel, thereby improving the fit between the subsequent region of interest (ROI) and the target vessel.

[0040] Alternatively, if the adjacent frame fails to segment the vessels in the adjacent frame, or the segmentation result of the adjacent frame is invalid, a reference frame can be determined from the coronary angiography image sequence, and the reference centerline of the reference vessel corresponding to the target vessel in the reference frame can be obtained. This reference centerline is then used as the vessel centerline information. Since the reference centerline originates from the reference vessel corresponding to the target vessel in the reference frame, it provides a stable orientation reference for the target vessel. Therefore, even when the adjacent frame cannot provide a reliable orientation reference, the availability and reliability of the vessel centerline information can still be maintained through the reference centerline, avoiding interruptions in subsequent segmentation processes due to adjacent frame segmentation failures, errors, or missing results.

[0041] Step 104: Generate a region of interest (ROI) that is a strip-shaped polygon extending along the direction of the target blood vessel to be segmented, based on the blood vessel centerline information.

[0042] The strip-shaped boundary of the region of interest (ROI) is determined based on the lateral offset position at multiple centerline locations in the vessel centerline information.

[0043] In some embodiments, if the vessel centerline information is the reference centerline information of the reference vessel corresponding to the target vessel to be segmented in the reference frame, then when generating a strip polygonal region of interest (ROI) extending along the direction of the target vessel to be segmented based on the vessel centerline information, the vessel centerline corresponding to the vessel centerline information can be determined as the target centerline first, and multiple centerline positions can be selected along the target centerline.

[0044] The target centerline is the centerline used to generate the region of interest (ROI) of the strip polygon. The multiple centerline positions can be multiple centerline points on the target centerline, or multiple position points obtained after interpolating or resampling the target centerline.

[0045] After determining multiple centerline positions, the local extension direction at each centerline position can be determined based on the adjacent centerline positions.

[0046] The local extension direction is used to characterize the extension trend of the target centerline near its location. For example, for a centerline location in the middle of the target centerline, the local extension direction can be determined based on the locations of adjacent centerlines before and after that centerline location; for a centerline location at the end of the target centerline, the local extension direction can be determined based on the location of that end centerline and its adjacent centerlines.

[0047] Subsequently, a unit normal direction perpendicular to the local extension direction can be determined based on the local extension direction. The unit normal direction is a direction perpendicular to the local extension direction and with normalized length, used to represent the lateral direction of the target centerline at the corresponding centerline position. Based on each centerline position and its corresponding unit normal direction, a first lateral offset position and a second lateral offset position can be determined on both sides of the target centerline. Specifically, the centerline position can be offset by a preset lateral offset distance along one side of the unit normal direction to obtain the first lateral offset position; and the centerline position can be offset by a preset lateral offset distance along the opposite side of the unit normal direction to obtain the second lateral offset position. The preset lateral offset distance can be determined based on the diameter of the target blood vessel to be segmented, the image resolution, the segmentation tolerance range, or the preset ROI width.

[0048] After obtaining multiple first-side lateral offset positions and multiple second-side lateral offset positions, the multiple first-side lateral offset positions can be connected in the order of extension of the target centerline to form a first-side strip boundary, and the multiple second-side lateral offset positions can be connected in the order of extension of the target centerline to form a second-side strip boundary.

[0049] Subsequently, a closed polygonal region can be formed based on the first and second side strip boundaries, and this closed polygonal region can be defined as the region of interest (ROI) of the strip polygon. For example, multiple first-side lateral offset positions can be arranged in the order of extension of the target centerline, and multiple second-side lateral offset positions can be arranged in the reverse order of extension of the target centerline, and the two can be connected end to end to form a closed polygonal region extending around the target centerline.

[0050] In some embodiments, the vertex coordinates of the closed polygon region can be cropped according to the image boundary of the coronary angiography image to be segmented, so that the region of interest (ROI) of the strip polygon is located within the image range.

[0051] In this way, since the first and second side strip boundaries of the polygonal region of interest (ROI) are determined based on the lateral offset positions at multiple centerline locations in the target centerline, the polygonal ROI can extend along the curvature of the target centerline. Compared with a rectangular ROI, it can reduce the inclusion range of non-target vessel regions and background regions, thereby providing spatial constraints that better fit the direction of the target vessel to be segmented for subsequent input constraint processing.

[0052] In other embodiments, if the vessel centerline information is the vessel centerline information of a neighboring frame, then when generating a strip-shaped polygonal region of interest (ROI) extending along the direction of the target vessel to be segmented based on the vessel centerline information, it can be divided into the following two cases: Scenario 1: The blood vessels in the neighboring frame have been segmented and the segmentation results are valid; In this embodiment, when the neighboring frame has completed the segmentation of the blood vessel corresponding to the target blood vessel and the segmentation result is valid, the center line of the neighboring frame blood vessel can be determined as the target center line, and multiple center line positions can be selected along the extension direction of the target center line. The subsequent process of generating a strip polygonal region of interest (ROI) can be the same as the process of generating a strip polygonal ROI when the blood vessel center line information is the reference center line information of the reference blood vessel corresponding to the target blood vessel in the reference frame. That is, the local extension direction at each center line position can be determined according to the adjacent center line positions, the unit normal direction perpendicular to the local extension direction can be determined according to the local extension direction, and based on each center line position and the corresponding unit normal direction, a first lateral offset position and a second lateral offset position can be determined on both sides of the target center line. Subsequently, multiple first-side lateral offset positions are connected in the order of extension of the target centerline to form a first-side strip boundary, and multiple second-side lateral offset positions are connected in the order of extension of the target centerline to form a second-side strip boundary. A closed polygonal region is formed based on the first-side strip boundary and the second-side strip boundary, and this closed polygonal region is determined as the strip polygonal region of interest (ROI) corresponding to the coronary angiography image to be segmented.

[0053] Scenario 2: The adjacent frame has not completed the segmentation of the adjacent frame's blood vessels, or the segmentation result of the adjacent frame's blood vessels is invalid; In this embodiment of the application, if the segmentation of blood vessels in a neighboring frame is incomplete or the segmentation result of blood vessels in a neighboring frame is invalid, a strip-shaped polygonal region of interest (ROI) can be generated based on the reference centerline. Specifically: In one optional implementation, the reference centerline of the reference vessel corresponding to the target vessel to be segmented in the reference frame can be determined as the target centerline, and multiple centerline positions can be selected along the target centerline. After determining the multiple centerline positions, the local extension direction at each centerline position can be determined based on the adjacent centerline positions. Subsequently, a unit normal direction perpendicular to the local extension direction can be determined based on the local extension direction, and based on each centerline position and the corresponding unit normal direction, a first lateral offset position and a second lateral offset position can be determined on both sides of the target centerline. After obtaining multiple first lateral offset positions and multiple second lateral offset positions, the multiple first lateral offset positions can be connected in the extension order of the target centerline to form a first lateral strip boundary, and the multiple second lateral offset positions can be connected in the extension order of the target centerline to form a second lateral strip boundary. Finally, a closed polygonal region can be formed based on the first lateral strip boundary and the second lateral strip boundary, and the closed polygonal region can be determined as the region of interest (ROI) of the strip polygon.

[0054] In another alternative implementation, a reference strip polygon region generated based on a reference centerline can also be obtained.

[0055] The reference polygonal region can be the region obtained by generating a polygon based on the reference centerline when the aforementioned vessel centerline information is the reference centerline information of the reference vessel corresponding to the target vessel to be segmented in the reference frame. In other words, the reference centerline can be used as the target centerline, and a reference polygonal region can be formed based on multiple centerline positions on the reference centerline and their corresponding lateral offset positions.

[0056] Subsequently, the region expansion distance corresponding to the reference strip polygon region can be obtained. This region expansion distance is used to increase the coverage area based on the reference strip polygon region to compensate for any potential positional shift of the target vessel relative to the reference frame in the coronary angiography image to be segmented.

[0057] After obtaining the region expansion distance, the reference strip polygon region can be expanded based on the region expansion distance to obtain the expanded reference strip polygon region.

[0058] In one implementation, the envelope of the reference strip polygon region can be determined first, and then the envelope range can be extended outward based on the region expansion distance to obtain the expanded reference strip polygon region.

[0059] Finally, based on the image boundary of the coronary angiography image to be segmented, the outer expanded reference strip polygon region can be cropped, and the cropped outer expanded reference strip polygon region can be determined as the strip polygon region of interest (ROI) corresponding to the coronary angiography image to be segmented.

[0060] For example, such as Figure 2 The diagram shown is a schematic representation of a strip-shaped polygonal region of interest (ROI) provided in an embodiment of this application. Figure 2 In the diagram, the white area represents the image range covered by the region of interest (ROI) of the polygonal strip, while the black background area represents the image range outside the ROI.

[0061] It should be noted that the above Figure 2 The schematic diagram of the region of interest (ROI) shown is merely an example in the embodiments of this application and does not limit the embodiments of this application in any way.

[0062] Step 106: Perform input constraint processing on the coronary angiography image to be segmented based on the region of interest (ROI) of the strip polygon to generate a constrained segmentation input image.

[0063] Input constraint processing refers to adjusting the pixel content of the coronary angiography image to be segmented based on the Region of Interest (ROI) of the polygonal strip before the blood vessel segmentation algorithm is executed. This ensures that the image input received by the segmentation algorithm is already constrained by the ROI range, rather than using the ROI for posterior filtering after the segmentation results are generated.

[0064] Constrained segmentation of the input image refers to an image that retains the original grayscale information of the coronary angiography image to be segmented within the region of interest (ROI) of the polygonal shape, while the image outside the ROI has been adjusted to the background grayscale.

[0065] like Figure 3 The diagram shown is an exemplary constrained segmentation input image provided in an embodiment of this application. Figure 3 In the diagram, the original image information of the coronary angiography image is preserved within the region of interest (ROI) of the polygonal shape, while the background grayscale outside the ROI is adjusted to 255, i.e. Figure 3 The white background area shown in the image.

[0066] It should be noted that, Figure 3 Adjusting the background grayscale outside the region of interest (ROI) of the polygonal strip to 255 is merely an exemplary illustration in this application. In practical applications, the background grayscale value can be set according to actual needs, and is not limited here.

[0067] In some embodiments, when generating a constrained segmentation input image, the pixels within and outside the region of interest in the coronary angiography image to be segmented can be determined first based on the region of interest (ROI) of the strip polygon. Pixels within the region are those located within the ROI, and pixels outside the region are those located outside the ROI. Then, the original image information of the pixels within the region is preserved, and the grayscale values ​​of the pixels outside the region are adjusted to a preset background grayscale value to generate the constrained segmentation input image.

[0068] In practical applications, after obtaining the region of interest (ROI) of a polygonal shape, the pixels within and outside the ROI in the coronary angiography image to be segmented can be determined based on the coverage area of ​​the ROI. After determining the pixels within and outside the ROI, the original image information of the pixels within the ROI can be preserved, and the grayscale values ​​of the pixels outside the ROI can be adjusted to the preset background grayscale values ​​to generate the constrained segmentation input image.

[0069] The original image information may include the original grayscale values ​​of the corresponding pixels in the coronary angiography image to be segmented; the preset background grayscale value may be a grayscale value used to characterize the background region. For example, in an 8-bit grayscale image, the preset background grayscale value may be set to 255, so that the outer region of the region of interest (ROI) of the strip polygon is presented as a white background.

[0070] In practice, the coronary angiography image to be segmented can be copied to obtain a copy of the image to be processed. Then, the position of each pixel in the copy is evaluated: if the pixel is within a region of interest (ROI), its original grayscale value is retained; if it is outside the ROI, its grayscale value is adjusted to a preset background grayscale value. The processed copy of the image becomes the constrained segmentation input image. In this way, the constrained segmentation input image still retains the original image details of the region containing the target blood vessel, while reducing or eliminating the influence of adjacent branch vessels, overlapping vessels, and background structures outside the ROI on the segmentation algorithm.

[0071] Using the above method, since the image content outside the region of interest (ROI) of the polygonal strip has been adjusted to the background grayscale before the segmentation algorithm runs, it is more difficult for the segmentation algorithm to form structurally continuous candidate blood vessels outside the ROI when performing blood vessel search. Therefore, the search space of the segmentation algorithm can be limited from the segmentation input end, thereby reducing the risk of neighboring branch blood vessels being mis-segmented as the target blood vessel to be segmented.

[0072] In some embodiments, before performing input constraint processing on the coronary angiography image to be segmented based on the region of interest (ROI) of the strip polygon to generate a constrained segmentation input image, the coronary angiography image to be segmented can first undergo full-image initial vessel segmentation, and the result of the full-image initial vessel segmentation can be used to determine whether further input constraint processing is needed.

[0073] Specifically, considering that not every frame in a coronary angiography image sequence will experience branch segmentation or target vessel localization errors, in some coronary angiography images to be segmented, direct full-image vessel segmentation can already obtain the vessel segmentation result corresponding to the target vessel to be segmented. If re-segmentation processing under the region of interest (ROI) constraint is directly performed on all coronary angiography images to be segmented, it may increase unnecessary image processing procedures and computational overhead. Therefore, in this embodiment, before performing input constraint processing on the coronary angiography images to be segmented, initial full-image vessel segmentation can be performed on the coronary angiography images to be segmented to obtain initial segmentation results.

[0074] Among them, the initial blood vessel segmentation of the whole image refers to performing blood vessel segmentation on the entire image range of the coronary angiography image to be segmented without applying pixel constraints based on the region of interest (ROI) of the strip polygon.

[0075] Optionally, the initial segmentation result may include vascular regions, vascular contours, vascular centerlines, or other vascular structure information output by the vascular segmentation algorithm.

[0076] Subsequently, initial candidate vessels can be determined based on the initial segmentation results, and the initial candidate vessel feature information can be obtained.

[0077] The initial candidate vessel refers to a vessel object identified from the initial segmentation results that may correspond to the target vessel to be segmented. The feature information of the initial candidate vessel may include at least one of the following: centerline information, vessel length information, and main direction information; in some embodiments, it may also include the start and end point positions, centroid position, average diameter, contour point information, etc.

[0078] Simultaneously, reference frames can be determined from a coronary angiography image sequence containing images of coronary angiography to be segmented, and blood vessel segmentation can be performed on the reference frames to obtain reference blood vessels corresponding to the target blood vessels to be segmented, and reference blood vessel feature information of the reference blood vessels can be obtained.

[0079] Optionally, the reference vessel feature information may include at least one of the reference centerline, vessel length information, and principal direction information of the reference vessel; in some embodiments, it may also include the start and end points, centroid position, average diameter, and contour point information of the reference vessel. The aforementioned reference vessel feature information is used as a comparison benchmark to determine whether the initial candidate vessel corresponds to the target vessel to be segmented.

[0080] After obtaining the initial candidate vessel feature information and the reference vessel feature information, the initial candidate vessel can be preliminarily identified as a standard branch vessel based on the initial candidate vessel feature information and the reference vessel feature information.

[0081] The preliminary determination of standard co-branch vessels refers to the process of determining whether the initial candidate vessels and the reference vessels belong to the same target vessel before performing the region of interest (ROI) constraint segmentation of the polygonal strip.

[0082] In one implementation, when initially determining the standard co-branch vessels for the initial candidate vessels, this can be achieved by comparing the centerline distance, vessel length difference, and principal direction angle difference between the initial candidate vessels and the reference vessels. Specifically, the centerline distance characterizes the spatial proximity between the initial candidate vessels and the reference vessels; the vessel length difference characterizes the consistency in their vessel extension scale; and the principal direction angle difference characterizes the consistency in their overall extension direction.

[0083] Specifically, based on the initial candidate vessel feature information and the reference vessel feature information, the average centerline distance, length ratio, and principal direction angle difference between the initial candidate vessel and the reference vessel can be extracted. The average centerline distance is used to characterize the proximity of the initial candidate vessel and the reference vessel in the image space; the length ratio is used to characterize the consistency of the initial candidate vessel and the reference vessel in terms of vessel extension scale; and the principal direction angle difference is used to characterize the consistency of the overall extension direction of the initial candidate vessel and the reference vessel.

[0084] The average centerline distance can be determined using bidirectional nearest point distances. Specifically, for each centerline point on the centerline of the initial candidate vessel, the nearest reference centerline point on the reference centerline of the reference vessel can be found, and the average of all nearest distances is calculated to obtain a first average distance from the initial candidate vessel to the reference vessel. Simultaneously, for each reference centerline point on the reference centerline, the nearest centerline point on the centerline of the initial candidate vessel can be found, and the average of all nearest distances is calculated to obtain a second average distance from the reference vessel to the initial candidate vessel. Then, the average centerline distance can be determined based on the first and second average distances; for example, the average of the first and second average distances can be used as the average centerline distance. This avoids the bias caused by calculating distances from only one direction, resulting in better symmetry in the average centerline distance.

[0085] The length ratio can be determined by comparing the total length of the initial candidate vessels with the total length of the reference vessels. For example, the length ratio can be defined as the ratio between the total length of the initial candidate vessels and the total length of the reference vessels. The total length of the initial candidate vessels can be obtained by summing the distances between adjacent centerline points on the centerline of the initial candidate vessels; the total length of the reference vessels can be obtained by summing the distances between adjacent reference centerline points on the reference centerline.

[0086] The principal direction angle difference can be determined based on the start and end points of the initial candidate vessel and the reference vessel. Specifically, the principal direction vector of the initial candidate vessel can be determined based on its start and end points, and the principal direction vector of the reference vessel can be determined based on its start and end points. Then, the corresponding principal direction angles are determined based on the two principal direction vectors, and the difference between the two principal direction angles is calculated. Considering that the start and end directions of the vessel centerline may have a 180° reverse equivalence relationship, the principal direction angle difference can be normalized to the range of 0° to 90°.

[0087] After obtaining the average centerline distance, length ratio, and principal direction angle difference, the features of these three dimensions can be mapped to corresponding scores. Specifically, the distance score can be determined based on the average centerline distance, the length score based on the length ratio, and the angle score based on the principal direction angle difference.

[0088] In some embodiments, the distance score, length score, and angle score can be determined using an exponential decay method. For example, the distance score can decay as the average distance to the centerline increases; the length score can decay as the length ratio deviates from 1 increases; and the angle score can decay as the principal direction angle difference increases. As an example, the distance score can be expressed as S_dist=exp(-d / (d_max / 3)), where d represents the average distance to the centerline, and d_max represents the maximum distance parameter corresponding to the distance dimension; the length score can be expressed as S_len=exp(-|ratio-1|×2), where ratio represents the length ratio, and when ratio exceeds a preset length ratio range, the length score can be directly set to 0; and the angle score can be expressed as S_angle=exp(-θ / (θ_max / 3)), where θ represents the principal direction angle difference, and θ_max represents the maximum angle parameter corresponding to the angle dimension. The above formulas are merely examples and should not be construed as limiting the scope of protection of this application.

[0089] After obtaining the distance score, length score, and angle score, these scores can be weighted and fused to obtain a comprehensive score of the initial candidate vessel relative to the reference vessel. For example, a weighted fusion can be performed with a weight of 0.4×S_dist + 0.2×S_len + 0.4×S_angle to obtain the comprehensive score.

[0090] In some embodiments, a hard upper limit check can be performed before initial determination of standard co-branch vessels. Specifically, it can be determined whether the average centerline distance exceeds the hard upper limit for distance, whether the length ratio exceeds the hard range for length ratio, and whether the angle difference in the principal direction exceeds the hard upper limit for angle. If any of the average centerline distance, length ratio, or angle difference in the principal direction does not meet the corresponding hard condition, the initial candidate vessel can be directly determined to have failed the initial determination of standard co-branch vessels. As an example, the hard upper limit for distance can be 1.5 times the configured distance threshold, the hard range for length ratio can be 0.6 to 1.3 times, and the hard upper limit for angle can be 1.5 times the configured angle threshold.

[0091] After the initial candidate vessel passes the hard upper limit check, a soft constraint violation count can be performed. Specifically, the average centerline distance, length ratio, and principal direction angle difference can be compared with their corresponding standard thresholds, and the number of dimensions that do not meet the standard thresholds can be counted to obtain the number of soft constraint violations. In some embodiments, the preliminary determination of the standard co-branch vessel may require that the number of soft constraint violations be 0, that is, the average centerline distance, length ratio, and principal direction angle difference all meet their respective standard thresholds, before the initial candidate vessel is determined to have passed the preliminary determination of the standard co-branch vessel; otherwise, the initial candidate vessel is determined to have failed the preliminary determination of the standard co-branch vessel.

[0092] If the initial candidate vessels are preliminarily determined by the standard co-branch vessels, it means that the initial segmentation result obtained by the initial vessel segmentation of the whole image can correspond well to the target vessel to be segmented. At this time, the initial segmentation result can be used as the segmentation result of the coronary angiography image to be segmented, or the subsequent input constraint processing based on the region of interest (ROI) of the strip polygon will no longer be triggered.

[0093] If the initial candidate vessel fails the preliminary judgment of the standard same branch vessel, it means that the initial segmentation result obtained by the initial vessel segmentation of the whole image may correspond to a non-target vessel, a nearby branch vessel, or have a significant deviation from the target vessel to be segmented in position, length, or direction. In this case, the step of input constraint processing based on the region of interest (ROI) of the strip polygon to be segmented coronary angiography image can be performed to generate a constrained segmentation input image so that the vessel segmentation can be re-performed within the limited input range.

[0094] Thus, since this embodiment does not directly perform re-segmentation under the ROI constraint of the strip polygon in every frame, but first identifies the reliability of the initial segmentation result through initial vessel segmentation of the entire image and preliminary determination of standard co-branch vessels, unnecessary re-segmentation processing can be reduced when the initial segmentation result is reliable, and input constraint processing can be triggered in a timely manner when the initial segmentation result is unreliable, thereby balancing processing efficiency and target vessel segmentation accuracy. Furthermore, since the preliminary determination of standard co-branch vessels is based on the difference between the initial candidate vessel feature information and the reference vessel feature information, the risk of cross-branch segmentation can be identified earlier when neighboring branches are incorrectly segmented as the target vessel to be segmented, and the search space can be limited from the segmentation input end through subsequent input constraint processing, reducing the possibility that the cross-branch segmentation result will continue to be accepted.

[0095] Step 108: Perform vessel segmentation based on the constraint segmentation input image to obtain the target segmentation result of the coronary angiography image to be segmented.

[0096] In some embodiments, after generating the constrained segmentation input image, the constrained segmentation input image can be input into a blood vessel segmentation algorithm to segment the blood vessel structures in the constrained segmentation input image and obtain a re-segmentation result. The re-segmentation result refers to the blood vessel region, blood vessel contour, blood vessel centerline, or other blood vessel structure information obtained after re-performing blood vessel segmentation based on the constrained segmentation input image.

[0097] The blood vessel segmentation algorithm can be any segmentation algorithm that can extract blood vessel structures from coronary angiography images. This application does not limit the specific type of blood vessel segmentation algorithm.

[0098] Because pixels within the Region of Interest (ROI) of the polygonal area in the constrained segmentation input image retain their original image information, while pixels outside the ROI are adjusted to a preset background grayscale value, the blood vessel segmentation algorithm primarily searches for vascular structures within the image region covered by the ROI when processing the constrained segmentation input image. This reduces the interference of adjacent branch vessels, overlapping vessels, or background structures on the blood vessel segmentation results.

[0099] After obtaining the resegmentation results, candidate vessels for resegmentation can be determined based on these results. These candidate vessels refer to vessel objects in the resegmentation results that may correspond to the target vessel to be segmented. For example, the vessel connectivity region, vessel centerline, and vessel contour can be extracted from the resegmentation results, and vessel objects located within the region of interest (ROI) of a polygonal strip and meeting preset length, connectivity, or morphological conditions can be identified as candidate vessels for resegmentation.

[0100] In some embodiments, the validity of the resegmented candidate vessels can be assessed to determine whether they can be used as the segmentation result corresponding to the target vessel to be segmented. For example, candidate vessel feature information can be obtained and compared with reference vessel feature information. The candidate vessel feature information may include at least one of the centerline information, vessel length information, and principal direction information of the resegmented candidate vessel, while the reference vessel feature information may include at least one of the reference centerline, vessel length information, and principal direction information of the reference vessel. If the resegmented candidate vessel and the reference vessel meet preset criteria for determining the same branch vessel in dimensions such as centerline distance, vessel length, and principal direction angle, then the resegmented candidate vessel can be determined to correspond to the target vessel to be segmented, and the resegmentation result can be determined as the target segmentation result of the coronary angiography image to be segmented.

[0101] The target segmentation result refers to the final output segmentation result used to characterize the target vessel in the coronary angiography image to be segmented. This target segmentation result may include the vessel region mask, vessel centerline, vessel contour, vessel origin and termination points, vessel length, or any combination of the above information.

[0102] Using the above method, the constrained segmentation input image is not merely used for posterior filtering of the segmentation results, but rather serves as the actual input to the vessel segmentation algorithm during the segmentation process. Thus, since the image region outside the region of interest (ROI) of the strip polygon has already been processed as background before segmentation, the search space of the vessel segmentation algorithm can be limited from the segmentation input end. This reduces the possibility of the algorithm forming continuous erroneous vessel structures in non-target vessel regions, thereby improving the segmentation accuracy of the target vessel in the coronary angiography image to be segmented.

[0103] In some embodiments, if the resegmented candidate vessels obtained based on the constraint segmentation input image still do not meet the preset criteria for determining the same branch vessel, the resegmentation result can be used as the target segmentation result instead of directly being used. Instead, a soft mask enhancement fusion resegmentation based on a strip mask can be further performed, and the target segmentation result of the coronary angiography image to be segmented can be determined based on the soft mask enhancement fusion resegmentation result. Thus, when simple strip polygonal region of interest (ROI) input constraint processing is insufficient to obtain reliable segmentation results, the vessel segmentation effect can be further improved by enhancement fusion.

[0104] Specifically, the reference center line of the reference vessel corresponding to the target vessel to be segmented in the reference frame can be obtained, as well as the center line of the neighboring vessel of the adjacent frame that is adjacent to the coronary angiography image to be segmented and has been effectively segmented.

[0105] In this embodiment, by simultaneously acquiring the reference centerline and the blood vessel centerline of the adjacent frame, the stability of the reference frame and the adaptability of the adjacent frame to changes in the position of the current frame can be taken into account.

[0106] After obtaining the reference centerline and the vessel centerlines of adjacent frames, the centerline points contained in the reference centerline and the vessel centerlines of adjacent frames can be merged to obtain a merged centerline point set. This merged centerline point set can be understood as a set of points composed of the centerline points in the reference centerline and the vessel centerlines of adjacent frames. This set of points is used to cover the possible spatial distribution range of the target vessel to be segmented in the reference frame and adjacent frames. This merging can be done by direct stitching, or by performing deduplication, sorting, or outlier removal after stitching.

[0107] Subsequently, using each centerline point in the merged centerline point set as the center, filled circular regions can be generated according to a preset mask radius. The preset mask radius controls the coverage area of ​​the binary strip mask around each centerline point, and can be determined based on the vessel diameter, image resolution, inter-frame vessel displacement amplitude, or a preset tolerance range. For example, the preset mask radius can be 35 pixels. Since each filled circular region is generated around a centerline point, multiple filled circular regions can form a coverage band along the distribution direction of the reference centerline and the vessel centerlines of adjacent frames.

[0108] After generating multiple filled circular regions, a union operation can be performed on these regions to obtain a binary stripe mask. The binary stripe mask refers to a binary image used to distinguish between the mask's foreground and background regions; the mask's foreground region corresponds to the union range of the multiple filled circular regions, and the mask's background region corresponds to the range outside the multiple filled circular regions. In some embodiments, morphological closing operations can also be performed on the binary stripe mask to fill any small gaps that may exist between adjacent filled circular regions and to smooth the boundaries of the binary stripe mask. For example, a 5×5 structuring element can be used to perform the closing operation on the binary stripe mask.

[0109] In some embodiments, before generating a soft mask fusion image based on the binary strip mask, the validity of the binary strip mask can be checked. For example, connected component analysis can be performed on the binary strip mask to count the number of foreground connected regions and the area of ​​the largest foreground connected region. If the binary strip mask has no foreground region, or if the area of ​​the largest foreground connected region is less than a preset area ratio threshold for the area of ​​the coronary angiography image to be segmented, the binary strip mask can be determined to be invalid, and soft mask enhancement fusion re-segmentation based on the binary strip mask can be stopped. This avoids fragmented or empty masks from entering the subsequent segmentation process due to abnormalities in the reference centerline or the vessel centerline of adjacent frames.

[0110] When the binary strip mask is effective, a soft mask fusion image can be generated based on the binary strip mask. Specifically, a distance transformation can be performed on the binary strip mask to obtain the distance values ​​from each pixel in the mask foreground region to the mask boundary. These distance values ​​are then normalized to obtain a soft mask with values ​​ranging from 0 to 1. This soft mask represents the fusion weights at different pixel positions; the soft mask values ​​are larger near the center line and gradually decrease near the boundary of the binary strip mask, thus forming a weight distribution that smoothly decays from the center region to the edge region.

[0111] Furthermore, image enhancement processing can be performed on the coronary angiography image to be segmented to obtain an enhanced image. For example, Gamma correction can be applied to the coronary angiography image to be segmented, and a gray-level shift can be superimposed to adjust the gray-level relationship between the vessel region and the background region. As an example, a Gamma correction with a Gamma value of 0.15 can be used, and a gray-level shift of 60 gray levels can be superimposed. Then, based on a soft mask, the enhanced image and the coronary angiography image to be segmented can be weighted and fused to obtain a soft-mask fused image. For example, pixel-level fusion can be performed as follows: the gray-level value of a pixel in the soft-mask fused image is equal to the product of the soft mask weight corresponding to that pixel and the gray-level value of the corresponding pixel in the enhanced image, plus 1 minus the soft mask weight, and then multiplied by the product of the gray-level value of the corresponding pixel in the coronary angiography image to be segmented.

[0112] Through the above fusion method, in regions where the soft mask weight is large, the fused image retains more grayscale information from the enhanced image; in regions where the soft mask weight is small, the fused image retains more original grayscale information from the coronary angiography image to be segmented; and in transitional regions where the soft mask weight changes from large to small, the grayscale of the fused image also transitions smoothly. This avoids the abrupt grayscale changes caused by directly using a hard mask, reducing the possibility of the segmentation algorithm misidentifying the mask boundary as the blood vessel boundary.

[0113] After obtaining the soft mask fusion image, vessel segmentation can be performed on the soft mask fusion image to obtain the soft mask enhanced fusion re-segmentation result. The soft mask enhanced fusion re-segmentation result can include at least one of the following vascular structural information: vessel region, vessel centerline, vessel contour, vessel origin and termination points, and vessel length. Subsequently, corresponding candidate vessels can be determined based on the soft mask enhanced fusion re-segmentation result, and further combined with reference vessel feature information or adjacent frame vessel feature information to determine if the candidate vessel corresponds to the target vessel to be segmented. If the candidate vessel is determined to correspond to the target vessel to be segmented, the soft mask enhanced fusion re-segmentation result can be determined as the target segmentation result of the coronary angiography image to be segmented.

[0114] In this way, since the binary strip mask is generated by merging the centerline points contained in the reference centerline and the adjacent frame's vessel centerline, rather than relying solely on a single reference centerline or a single adjacent frame's vessel centerline to generate segmentation constraints, the coverage of possible locations of the target vessel to be segmented can be expanded, improving tolerance to inter-frame displacement. Furthermore, since a soft mask fusion image is further generated based on the binary strip mask, a smooth transition is formed between the image enhancement region and the original image region. Therefore, segmentation artifacts caused by hard mask boundaries can be reduced, thereby further improving the segmentation success rate and segmentation stability of the target vessel when re-segmentation with polygonal ROI constraints still does not yield reliable results.

[0115] The method provided in this application involves acquiring vessel centerline information to characterize the orientation of the target vessel to be segmented, and generating a strip-shaped polygonal Region of Interest (ROI) extending along the orientation of the target vessel based on the lateral offset positions at multiple centerline locations in the vessel centerline information. Then, input constraint processing is applied to the coronary angiography image to be segmented based on this strip-shaped polygonal ROI to generate a constrained segmentation input image, which is then used for vessel segmentation. This approach offers several advantages. First, the strip-shaped polygonal ROI extends along the curvature of the target vessel, reducing the inclusion of adjacent branch vessels, overlapping vessels, and background structures in the segmentation search compared to a rectangular ROI. Second, since the strip-shaped polygonal ROI is not only used for candidate result selection after segmentation but also participates in the input constraint processing of the coronary angiography image before vessel segmentation, it limits the search space of the segmentation algorithm from the segmentation input end, reducing the possibility of the segmentation algorithm forming structurally continuous and morphologically complete erroneous segmentation results in non-target vessel regions. This improves the segmentation accuracy of the target vessel in the coronary angiography sequence and reduces the risk of cross-branch segmentation between the target vessel and adjacent branch vessels.

[0116] The following describes how the methods provided in the embodiments of this application are applied in practice, taking into account real-world scenarios.

[0117] Please refer to the following content, which is an application flow of the method provided in the embodiments of this application in practice. The flow specifically includes the following steps: In a specific application scenario, in order to obtain continuous segmentation results of the target blood vessel in each frame of coronary angiography image, a reference frame can be determined from the coronary angiography image sequence.

[0118] For example, a reference frame can be selected from a coronary angiography image sequence based on the clarity, continuity, and contrast of the blood vessels. The reference frame is then segmented to obtain the reference blood vessel corresponding to the target vessel in the frame, and its reference centerline, length, principal direction angle, and start and end point positions are acquired.

[0119] After obtaining the reference centerline, each coronary angiography image to be segmented can be processed along the forward and reverse time directions of the coronary angiography image sequence, starting from the reference frame. For any coronary angiography image to be segmented, it can be first determined whether the neighboring frames adjacent to the image to be segmented have completed the segmentation of the neighboring frame vessels corresponding to the target vessel and whether the segmentation result is valid. If the neighboring frame has completed the segmentation of the neighboring frame vessels and the segmentation result is valid, then the centerline of the neighboring frame vessels is obtained and determined as the vessel centerline information; if the neighboring frame has not completed the segmentation of the neighboring frame vessels, or the segmentation result of the neighboring frame vessels is invalid, then the reference centerline is determined as the vessel centerline information. In this way, when the neighboring frame results are reliable, the neighboring frame vessel centerline can be used to adapt to the inter-frame position changes of the target vessel to be segmented, and when the neighboring frame results are unavailable, the reference centerline can be used to provide a stable basis.

[0120] After determining the vessel centerline information, a region of interest (ROI) of a strip-shaped polygon extending along the direction of the target vessel to be segmented can be generated based on the vessel centerline information.

[0121] Specifically, the vessel centerline corresponding to the vessel centerline information can be determined as the target centerline. Multiple centerline positions are selected along the target centerline, and the local extension direction at each centerline position is determined based on the adjacent centerline positions. Then, the unit normal direction perpendicular to the local extension direction is determined based on the local extension direction. Based on each centerline position and the corresponding unit normal direction, a first-side lateral offset position and a second-side lateral offset position are determined on both sides of the target centerline. Then, multiple first-side lateral offset positions are connected to form a first-side strip boundary, and multiple second-side lateral offset positions are connected to form a second-side strip boundary. A closed polygonal region is formed based on the first-side strip boundary and the second-side strip boundary, and this closed polygonal region is determined as the strip polygonal region of interest (ROI). If the segmentation results of adjacent vessel frames are invalid, a reference strip polygonal region generated based on the reference centerline can be obtained. The reference strip polygonal region is then expanded and its boundaries are clipped according to the region expansion distance. The clipped expanded reference strip polygonal region is determined as the strip polygonal region of interest (ROI) corresponding to the coronary angiography image to be segmented.

[0122] After generating the region of interest (ROI) in a polygonal shape, initial full-image vessel segmentation can be performed on the coronary angiography image to be segmented to obtain initial segmentation results. Initial candidate vessels are then identified based on these initial segmentation results, and their initial candidate vessel feature information is obtained. Subsequently, based on the initial candidate vessel feature information and reference vessel feature information, standard intra-branch vessel identification is performed on the initial candidate vessels.

[0123] For example, the average centerline distance, length ratio, and principal direction angle difference between the initial candidate vessel and the reference vessel can be compared. If the average centerline distance, length ratio, and principal direction angle difference all meet the corresponding standard thresholds, the initial candidate vessel can be preliminarily determined through the standard coronary artery, and the initial segmentation result can be used as the target segmentation result of the coronary angiography image to be segmented.

[0124] If the initial candidate vessel fails the preliminary judgment of the standard same branch vessel, it means that the initial candidate vessel obtained from the initial vessel segmentation of the whole image may deviate from the target vessel to be segmented, or may correspond to a neighboring branch vessel.

[0125] As a specific example, coronary angiography image sequences can include, for example, Figure 4 The coronary angiography image with frame number 28 shown in (a) is shown in the image. Figure 4 The coronary angiography image shown in (b) is frame number 18. When performing vessel segmentation in reverse chronological order along the coronary angiography image sequence, frame 28 can be designated as the reference frame, and frame 18 as the coronary angiography image to be segmented. Note that the frame numbers mentioned above are only used to distinguish different image frames in this example and do not constitute a limitation on the specific location of the reference frame or the coronary angiography image to be segmented within the coronary angiography image sequence.

[0126] like Figure 4 As shown in (a), the 28th frame of coronary angiography image can be segmented to obtain a reference vessel corresponding to the target vessel to be segmented, and the reference centerline can be extracted from the reference vessel. Figure 4 The reference centerline is shown as a green curve superimposed on the coronary angiography image of frame 28 in (a).

[0127] like Figure 4 As shown in (b), after performing full-image initial vessel segmentation on the 18th frame of coronary angiography image, the initial candidate vessels can be determined based on the initial segmentation results, and the center line of the initial candidate vessels can be extracted from the initial candidate vessels. Figure 4 (b) shows the initial candidate vessel centerline as a green curve superimposed on the 18th frame of the coronary angiography image. This initial candidate vessel centerline is then compared with... Figure 4 By comparing the reference centerline shown in (a) in the figure, it can be determined that the centerline of the initial candidate vessel extends along the adjacent branch in the bifurcation region of the vessel, and therefore the initial candidate vessel did not pass the preliminary judgment of the standard same branch vessel.

[0128] When the initial candidate vessel fails the preliminary judgment of the standard same branch vessel, the input constraint processing of the coronary angiography image to be segmented can be performed based on the region of interest (ROI) of the strip polygon to generate a constrained segmentation input image.

[0129] Specifically, the pixels within and outside the coronary angiography image to be segmented can be determined. The original image information of the pixels within the region is preserved, while the grayscale value of the pixels outside the region is adjusted to a preset background grayscale value, such as 255. Thus, the input image is constrained to retain the original vascular image information within the region of interest (ROI) and to present the background grayscale outside the ROI.

[0130] Subsequently, vessel segmentation can be performed on the constrained segmentation input image to obtain resegmentation results, and candidate vessels for resegmentation can be determined based on these results. Then, candidate vessel feature information is obtained, and based on the candidate vessel feature information and reference vessel feature information, broadened same-branch vessel determination is performed on the candidate vessels for resegmentation.

[0131] If the resegmentation candidate vessels are determined by relaxing the criteria for same-branch vessels, the resegmentation result can be used as the target segmentation result for the coronary angiography image to be segmented. In this way, when there is a risk of cross-branch involvement during the initial full-image segmentation, the vessel search space can be restricted from the input end by using a region of interest (ROI) of a polygonal shape, so that the segmentation algorithm mainly resegments around the region where the target vessel to be segmented is located.

[0132] For example, such as Figure 5 The diagram shown is a schematic representation of the target segmentation result of a re-segmented coronary angiography image to be segmented, provided in an embodiment of this application. Figure 5 The centerline of the target vessel to be segmented is marked by a green curve. The centerline is located inside the target vessel and extends continuously from the proximal end to the distal end, with its extension shape corresponding to the curvature of the target vessel.

[0133] If the candidate vessel for resegmentation fails the broadened co-branch vessel determination, then soft mask-enhanced fusion resegmentation based on strip masks can be further performed.

[0134] Specifically, a reference centerline and adjacent, effectively segmented vessel centerlines from neighboring frames of the coronary angiography image to be segmented can be obtained. The centerline points contained in the reference centerline are then merged with those contained in the adjacent frame vessel centerlines to obtain a merged centerline point set. Next, using each centerline point in the merged centerline point set as a center, filled circular regions are generated according to a preset mask radius. These filled circular regions are then combined to obtain a binary stripe mask. For example, the preset mask radius can be 35 pixels. After obtaining the binary stripe mask, morphological closing operations can be performed on it to smooth its boundaries and fill local gaps.

[0135] After obtaining the binary strip mask, connected component analysis can be performed on the binary strip mask to determine whether the binary strip mask is effective.

[0136] For example, the number of foreground connected regions and the area of ​​the largest foreground connected region in the binary strip mask can be counted. If the binary strip mask has no foreground region, or if the area of ​​the largest foreground connected region is less than a preset area ratio threshold for the area of ​​the coronary angiography image to be segmented, the binary strip mask can be determined to be invalid, and soft mask enhancement fusion re-segmentation based on the binary strip mask can be stopped. If the binary strip mask is valid, a soft mask fusion image can be generated based on the binary strip mask.

[0137] When generating the soft mask fused image, a distance transformation can be performed on the binary strip mask, and the result can be normalized to the range of 0 to 1 to obtain the soft mask. Subsequently, image enhancement processing can be performed on the coronary angiography image to be segmented to obtain the enhanced image. For example, Gamma correction can be performed on the coronary angiography image to be segmented, and a gray-level shift can be superimposed; in one example, the Gamma value can be 0.15, and the gray-level shift can be 60 gray levels. Then, pixel-level weighted fusion can be performed on the enhanced image and the coronary angiography image to be segmented based on the soft mask to obtain the soft mask fused image. Specifically, in areas with a large soft mask weight, the soft mask fused image uses more gray-level information from the enhanced image; in areas with a small soft mask weight, the soft mask fused image uses more original gray-level information from the coronary angiography image to be segmented; in transition regions, the gray-level value changes smoothly with the soft mask weight.

[0138] After obtaining the soft mask fusion image, vessel segmentation can be performed on the soft mask fusion image to obtain the soft mask enhanced fusion re-segmentation result. Subsequently, the corresponding candidate vessels can be determined based on the soft mask enhanced fusion re-segmentation result, and the candidate vessels can be judged as being from the same branch. If the candidate vessel corresponds to the target vessel to be segmented, the soft mask enhanced fusion re-segmentation result is determined as the target segmentation result of the coronary angiography image to be segmented. Through the above processing, when re-segmentation constrained by polygonal ROI still cannot obtain reliable results, the coverage of the possible location of the target vessel to be segmented can be improved by using a binary strip mask jointly generated by the reference center line and the vessel center lines of adjacent frames, and the segmentation artifacts caused by hard mask boundaries can be reduced by using the soft mask fusion image.

[0139] After processing the current coronary angiography image to be segmented, the target segmentation result of the current coronary angiography image to be segmented can be used as the segmentation result of the neighboring frame of the subsequent image frame, and the above process can be repeated for the next coronary angiography image to be segmented in the coronary angiography image sequence.

[0140] Ultimately, a set of target segmentation results for the target vessel to be segmented across multiple frames in the coronary angiography image sequence can be obtained. This set of target segmentation results can include information such as the vessel region, vessel centerline, vessel contour, vessel length, and tracking status of the target vessel in each frame, and can be used for subsequent quantitative stenosis analysis, fractional flow reserve calculation, or 3D vessel modeling.

[0141] To address the problem in existing technologies where rectangular regions of interest (ROIs) are difficult to align with the target vessel's course, leading to frequent branch segmentation during target vessel segmentation, this application provides a vessel segmentation device for coronary angiography sequences. A schematic diagram of the device's specific structure is shown below. Figure 6 As shown, it includes an acquisition module 61, a generation module 62, a processing module 63, and a segmentation module 64. The functions of each module are as follows: The acquisition module 61 is used to acquire the coronary angiography image to be segmented and to acquire the vessel centerline information used to characterize the orientation of the target vessel to be segmented in the coronary angiography image to be segmented. The generation module 62 is used to generate a strip-shaped polygonal region of interest (ROI) extending along the direction of the target blood vessel to be segmented based on the blood vessel centerline information; wherein, the strip-shaped boundary of the ROI is determined based on the lateral offset position at multiple centerline positions in the blood vessel centerline information. Processing module 63 is used to perform input constraint processing on the coronary angiography image to be segmented based on the region of interest (ROI) of the strip polygon, and generate a constrained segmentation input image; The segmentation module 64 is used to segment blood vessels based on the constraint segmentation input image to obtain the target segmentation result of the coronary angiography image to be segmented.

[0142] Optionally, module 61 is used for: Determine the coronary angiography image sequence containing the coronary angiography images to be segmented, and determine the reference frame from the coronary angiography image sequence; Obtain the reference centerline of the reference blood vessel corresponding to the target blood vessel to be segmented in the reference frame; The reference centerline is used as the vessel centerline information to characterize the orientation of the target vessel in the coronary angiography image to be segmented.

[0143] Optionally, module 61 is used for: Identify neighboring frames adjacent to the coronary angiography image to be segmented from the coronary angiography image sequence containing the coronary angiography image to be segmented; Determine whether the neighboring frame has completed the segmentation of the blood vessel corresponding to the target blood vessel and whether the segmentation result is valid; If the blood vessels in the neighboring frame have been segmented and the segmentation result is valid, then the center line of the blood vessels in the neighboring frame is obtained and the center line of the blood vessels in the neighboring frame is determined as the blood vessel center line information. If the segmentation of the blood vessels in the neighboring frame is not completed, or the segmentation result of the blood vessels in the neighboring frame is invalid, a reference frame is determined from the coronary angiography image sequence, and the reference center line of the reference blood vessel corresponding to the target blood vessel to be segmented in the reference frame is obtained. The reference center line is used as the information for the vascular center line.

[0144] Optionally, module 62 is used for: The vascular centerline corresponding to the vascular centerline information is determined as the target centerline, and multiple centerline positions are selected along the target centerline; Determine the local extension direction at each centerline position based on the positions of adjacent centerlines; Determine the direction of the unit normal perpendicular to the direction of local extension based on the direction of local extension; Based on the position of each centerline and the corresponding unit normal direction, determine the first side lateral offset position and the second side lateral offset position on both sides of the target centerline. Multiple first-side lateral offset positions are connected in the order of extension of the target center line to form a first-side strip boundary, and multiple second-side lateral offset positions are connected in the order of extension of the target center line to form a second-side strip boundary. A closed polygonal region is formed based on the first and second side strip boundaries, and the closed polygonal region is determined as the region of interest (ROI) of the strip polygon.

[0145] Optionally, module 62 is used for: If the blood vessels in the neighboring frame have been segmented and the segmentation result is valid, the center line of the blood vessels in the neighboring frame is determined as the target center line, and multiple center line positions are selected along the extension direction of the target center line. Determine the local extension direction at each centerline position based on the positions of adjacent centerlines; Determine the direction of the unit normal perpendicular to the direction of local extension based on the direction of local extension; Based on the position of each centerline and the corresponding unit normal direction, determine the first side lateral offset position and the second side lateral offset position on both sides of the target centerline. Multiple first-side lateral offset positions are connected in the order of extension of the target center line to form a first-side strip boundary, and multiple second-side lateral offset positions are connected in the order of extension of the target center line to form a second-side strip boundary. A closed polygonal region is formed based on the first and second side strip boundaries, and the closed polygonal region is determined as the region of interest (ROI) of the strip polygonal image to be segmented in the coronary angiography image.

[0146] Optionally, module 62 is used for: If the blood vessels in the neighboring frame are not segmented, or the segmentation result of the blood vessels in the neighboring frame is invalid, then the reference strip polygon region generated based on the reference center line is obtained. Obtain the outer distance of the region corresponding to the reference strip polygon region; The reference strip polygon region is expanded based on the region expansion distance to obtain the expanded reference strip polygon region. Based on the image boundary of the coronary angiography image to be segmented, the boundary of the expanded reference strip polygon region is clipped. The cropped, extended reference polygonal region is determined as the region of interest (ROI) of the coronary angiography image to be segmented.

[0147] Optionally, processing module 63 is used for: Based on the region of interest (ROI) of a polygonal strip, pixels within and outside the region in the coronary angiography image to be segmented are determined; where pixels within the region are those located within the ROI of the polygonal strip, and pixels outside the region are those located outside the ROI of the polygonal strip. The original image information of pixels within the region is preserved, and the gray values ​​of pixels outside the region are adjusted to the preset background gray values ​​to generate a constrained segmentation input image.

[0148] Optionally, the device is also used for: Perform full-image initial vessel segmentation on the coronary angiography image to be segmented to obtain the initial segmentation result; Initial candidate vessels are determined based on the initial segmentation results, and the initial candidate vessel feature information is obtained. A reference frame is determined from the coronary angiography image sequence containing the coronary angiography image to be segmented. The reference frame is segmented to obtain the reference vessel corresponding to the target vessel to be segmented, and the reference vessel feature information of the reference vessel is obtained. Based on the initial candidate vessel feature information and the reference vessel feature information, the initial candidate vessels are preliminarily identified as standard vessels of the same branch. If the initial candidate vessel fails the preliminary judgment of the standard same branch vessel, the step of performing input constraint processing on the coronary angiography image to be segmented based on the region of interest (ROI) of the strip polygon is performed to generate a constrained segmentation input image.

[0149] Optionally, segmentation module 64 includes: The segmentation unit is used to perform blood vessel segmentation on the constrained segmentation input image to obtain the resegmentation result; The first determining unit is used to determine the candidate vessels for resegmentation based on the resegmentation results and to obtain the candidate vessel feature information of the candidate vessels for resegmentation. The determination unit is used to broaden the determination of the same branch vessels of the re-segmented candidate vessels based on the candidate vessel feature information and the reference vessel feature information. The second determining unit is used to determine the resegmentation result as the target segmentation result of the coronary angiography image to be segmented if the candidate vessel for resegmentation is determined by relaxing the same branch vessel. The resegmentation unit is used to perform soft mask enhancement fusion resegmentation based on strip mask if the candidate vessel for resegmentation fails to pass the broadened same branch vessel determination, and to obtain the soft mask enhancement fusion resegmentation result, and determine the target segmentation result of the coronary angiography image to be segmented based on the soft mask enhancement fusion resegmentation result.

[0150] Optional, resegmentation unit, used for: Obtain the reference center line of the reference vessel corresponding to the target vessel to be segmented in the reference frame, and the center line of the neighboring vessel of the adjacent frame that is adjacent to the coronary angiography image to be segmented and has been effectively segmented. The centerline points contained in the reference centerline are merged with the centerline points contained in the blood vessel centerlines of the adjacent frames to obtain a merged centerline point set; Using each centerline point in the merged centerline point set as the center, generate filled circular regions according to the preset mask radius; Multiple filled circular regions are combined to obtain a binary strip mask. Generating a soft mask fusion image based on a binary strip mask; Blood vessel segmentation is performed on the soft mask fusion image to obtain the soft mask enhanced fusion resegmentation result.

[0151] Using the apparatus provided in this application embodiment, the device acquires vessel centerline information to characterize the orientation of the target vessel to be segmented, and generates a strip-shaped polygonal Region of Interest (ROI) extending along the orientation of the target vessel based on the lateral offset positions at multiple centerline locations in the vessel centerline information. Then, input constraint processing is applied to the coronary angiography image to be segmented based on the strip-shaped polygonal ROI to generate a constrained segmentation input image, and vessel segmentation is performed based on this constrained segmentation input image. In this way, on the one hand, the strip-shaped polygonal ROI can extend along the curvature of the target vessel to be segmented, reducing the inclusion of adjacent branch vessels, overlapping vessels, and background structures in the segmentation search range compared to a rectangular ROI. On the other hand, since the strip-shaped polygonal ROI is not only used for candidate result selection after segmentation but also participates in the input constraint processing of the coronary angiography image to be segmented before vessel segmentation, it can limit the search space of the segmentation algorithm from the segmentation input end, reducing the possibility of the segmentation algorithm forming structurally continuous and morphologically complete erroneous segmentation results in non-target vessel regions. This can improve the segmentation accuracy of the target vessel in the coronary angiography sequence and reduce the risk of cross-branch segmentation between the target vessel and adjacent branch vessels.

[0152] Figure 7 To illustrate the hardware structure of an electronic device according to various embodiments of this application, the electronic device may include a processor 701 and a memory 702 storing computer program instructions. Specifically, the processor 701 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of this application.

[0153] Memory 702 may include mass storage for data or instructions. For example, and not limitingly, memory 702 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 702 may include removable or non-removable (or fixed) media. Where appropriate, memory 702 may be internal or external to an electronic device. In a particular embodiment, memory 702 may be a non-volatile solid-state memory.

[0154] In one embodiment, memory 702 may be read-only memory (ROM). In one embodiment, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM), or flash memory, or a combination of two or more of these.

[0155] The processor 701 reads and executes computer program instructions stored in the memory 702 to implement any of the coronary angiography sequence vessel segmentation methods in the above embodiments.

[0156] In one example, the electronic device may also include a communication interface 707 and a bus 710. Wherein, as... Figure 7 As shown, the processor 701, memory 702, and communication interface 707 are connected through bus 710 and complete communication with each other.

[0157] The communication interface 707 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0158] Bus 710 includes hardware, software, or both, that couples components of an electronic device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 710 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, this application contemplates any suitable bus or interconnect.

[0159] Furthermore, in conjunction with the vessel segmentation method for coronary angiography sequences in the above embodiments, this application embodiment can provide a computer-readable storage medium for implementation. This computer-readable storage medium stores computer program instructions; when executed by a processor, these computer program instructions implement any of the vessel segmentation methods for coronary angiography sequences in the above embodiments.

[0160] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0161] The above description is merely a specific implementation example of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0162] Secondly, those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0163] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0164] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0165] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0166] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0167] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0168] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0169] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0170] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for vessel segmentation in a coronary angiography sequence, characterized in that, include: Acquire a coronary angiography image to be segmented, and acquire vessel centerline information to characterize the orientation of the target vessel to be segmented in the coronary angiography image to be segmented; A region of interest (ROI) of the shape, which extends along the direction of the target blood vessel to be segmented, is generated based on the blood vessel centerline information; wherein, the strip-shaped boundary of the ROI is determined based on the lateral offset position at multiple centerline positions in the blood vessel centerline information. The coronary angiography image to be segmented is subjected to input constraint processing based on the region of interest (ROI) of the polygonal strip to generate a constrained segmentation input image. Based on the constraint segmentation input image, blood vessel segmentation is performed to obtain the target segmentation result of the coronary angiography image to be segmented.

2. The method as described in claim 1, characterized in that, The step of obtaining the vessel centerline information used to characterize the orientation of the target vessel in the coronary angiography image to be segmented includes: Determine a coronary angiography image sequence containing the coronary angiography image to be segmented, and determine a reference frame from the coronary angiography image sequence; Obtain the reference centerline of the reference blood vessel corresponding to the target blood vessel to be segmented in the reference frame; The reference centerline is used as the vessel centerline information to characterize the orientation of the target vessel in the coronary angiography image to be segmented.

3. The method as described in claim 1, characterized in that, The step of obtaining the vessel centerline information used to characterize the orientation of the target vessel in the coronary angiography image to be segmented includes: Determine neighboring frames adjacent to the coronary angiography image to be segmented from the coronary angiography image sequence containing the coronary angiography image to be segmented; Determine whether the neighboring frame has completed the segmentation of the neighboring frame blood vessel corresponding to the target blood vessel to be segmented and whether the segmentation result is valid; If the neighboring frame has completed the segmentation of the blood vessels in the neighboring frame and the segmentation result is valid, then the center line of the blood vessels in the neighboring frame is obtained, and the center line of the blood vessels in the neighboring frame is determined as the center line information of the blood vessels. If the neighboring frame has not completed the segmentation of the neighboring frame's blood vessels, or the segmentation result of the neighboring frame's blood vessels is invalid, then a reference frame is determined from the coronary angiography image sequence, and the reference centerline of the reference blood vessel corresponding to the target blood vessel to be segmented in the reference frame is obtained. The reference centerline is determined as the blood vessel centerline information.

4. The method as described in claim 2, characterized in that, The step of generating a strip-shaped polygonal region of interest (ROI) extending along the direction of the target blood vessel to be segmented based on the blood vessel centerline information includes: The vascular centerline corresponding to the vascular centerline information is determined as the target centerline, and multiple centerline positions are selected along the target centerline; The local extension direction at each of the centerline positions is determined based on the adjacent centerline positions. The unit normal direction perpendicular to the local extension direction is determined based on the local extension direction; Based on each centerline position and the corresponding unit normal direction, a first lateral offset position and a second lateral offset position are determined on both sides of the target centerline. Multiple first-side lateral offset positions are connected in the order of extension of the target center line to form a first-side strip boundary, and multiple second-side lateral offset positions are connected in the order of extension of the target center line to form a second-side strip boundary; A closed polygonal region is formed based on the first side strip boundary and the second side strip boundary, and the closed polygonal region is determined as the region of interest (ROI) of the strip polygon.

5. The method as described in claim 3, characterized in that, If the neighboring frame has already completed the segmentation of the blood vessels in that neighboring frame and the segmentation result is valid, then the step of generating a strip-shaped polygonal region of interest (ROI) extending along the direction of the target blood vessel to be segmented based on the blood vessel centerline information includes: The adjacent frame vessel centerline is determined as the target centerline, and multiple centerline positions are selected along the extension direction of the target centerline; The local extension direction at each of the centerline positions is determined based on the adjacent centerline positions. The unit normal direction perpendicular to the local extension direction is determined based on the local extension direction; Based on each centerline position and the corresponding unit normal direction, a first lateral offset position and a second lateral offset position are determined on both sides of the target centerline. Multiple first-side lateral offset positions are connected in the order of extension of the target center line to form a first-side strip boundary, and multiple second-side lateral offset positions are connected in the order of extension of the target center line to form a second-side strip boundary; A closed polygonal region is formed based on the first side band boundary and the second side band boundary, and the closed polygonal region is determined as the region of interest (ROI) of the banded polygonal image to be segmented in the coronary angiography image.

6. The method as described in claim 3, characterized in that, If the neighboring frame has not completed the segmentation of the blood vessels in the neighboring frame, or the segmentation result of the blood vessels in the neighboring frame is invalid, then the step of generating a strip-shaped polygonal region of interest (ROI) extending along the direction of the target blood vessel to be segmented based on the blood vessel centerline information includes: Obtain a reference strip polygonal region generated based on the reference centerline; Obtain the outer expansion distance of the region corresponding to the reference strip polygon region; The reference strip polygon region is expanded based on the expansion distance of the region to obtain the expanded reference strip polygon region. Based on the image boundary of the coronary angiography image to be segmented, the boundary of the expanded reference strip polygon region is clipped. The cropped, extended reference polygonal region is determined as the region of interest (ROI) of the coronary angiography image to be segmented.

7. The method as described in claim 1, characterized in that, The step of performing input constraint processing on the coronary angiography image to be segmented based on the region of interest (ROI) of the polygonal shape to generate a constrained segmentation input image includes: Based on the region of interest (ROI) of the strip polygon, pixels within the region and pixels outside the region in the coronary angiography image to be segmented are determined; wherein, pixels within the region are pixels located within the region of interest of the strip polygon, and pixels outside the region are pixels located outside the region of interest of the strip polygon. The original image information of the pixels within the region is retained, and the grayscale values ​​of the pixels outside the region are adjusted to the preset background grayscale values ​​to generate the constrained segmentation input image.

8. The method as described in claim 1, characterized in that, Before performing input constraint processing on the coronary angiography image to be segmented based on the strip-shaped polygonal region of interest (ROI) to generate a constrained segmentation input image, the method further includes: Perform full-image initial vessel segmentation on the coronary angiography image to be segmented to obtain initial segmentation results; Initial candidate vessels are determined based on the initial segmentation results, and initial candidate vessel feature information is obtained. A reference frame is determined from the coronary angiography image sequence containing the coronary angiography image to be segmented, and the reference frame is segmented to obtain a reference vessel corresponding to the target vessel to be segmented, and the reference vessel feature information of the reference vessel is obtained. Based on the initial candidate vessel feature information and the reference vessel feature information, the initial candidate vessels are preliminarily determined to be standard same-branch vessels. If the initial candidate vessel fails the preliminary determination by the standard coronary artery, then the step of performing input constraint processing on the coronary angiography image to be segmented based on the region of interest (ROI) of the strip polygon is performed to generate the constrained segmentation input image.

9. The method as described in claim 8, characterized in that, The process of segmenting blood vessels based on the constrained segmentation input image to obtain the target segmentation result of the coronary angiography image to be segmented includes: The constrained segmentation input image is subjected to blood vessel segmentation to obtain a resegmentation result; Based on the resegmentation results, candidate vessels for resegmentation are determined, and candidate vessel feature information of the candidate vessels for resegmentation is obtained. Based on the candidate vessel feature information and the reference vessel feature information, the re-segmented candidate vessels are subjected to broadened same-branch vessel determination; If the candidate vessel for resegmentation passes the determination by the broadened coronary artery, then the resegmentation result is determined as the target segmentation result of the coronary angiography image to be segmented. If the candidate vessel for resegmentation fails the determination of the broadened co-branch vessel, then a soft mask-enhanced fusion resegmentation based on a strip mask is performed to obtain the soft mask-enhanced fusion resegmentation result, and the target segmentation result of the coronary angiography image to be segmented is determined based on the soft mask-enhanced fusion resegmentation result.

10. The method as described in claim 9, characterized in that, The process of performing soft mask enhancement fusion resegmentation based on strip masks to obtain soft mask enhancement fusion resegmentation results includes: Obtain the reference center line of the reference vessel corresponding to the target vessel to be segmented in the reference frame, and the center line of the neighboring vessel of the neighboring frame that is adjacent to the coronary angiography image to be segmented and has been effectively segmented. The centerline points contained in the reference centerline are merged with the centerline points contained in the adjacent frame blood vessel centerline to obtain a merged centerline point set. Using each centerline point in the merged centerline point set as the center, fill circular regions are generated according to the preset mask radius; The multiple filled circular regions are combined to obtain a binary strip mask. A soft mask fusion image is generated based on the binary strip mask; The soft mask fused image is segmented into blood vessels to obtain the soft mask enhanced fusion resegmentation result.

11. A vessel segmentation device for coronary angiography sequences, characterized in that, It includes an acquisition module, a generation module, a processing module, and a segmentation module, among which: The acquisition module is used to acquire the coronary angiography image to be segmented and to acquire the vessel centerline information used to characterize the orientation of the target vessel to be segmented in the coronary angiography image to be segmented. The generation module is used to generate a strip-shaped polygonal region of interest (ROI) extending along the direction of the target blood vessel to be segmented, based on the blood vessel centerline information; wherein, the strip-shaped boundary of the ROI is determined based on the lateral offset position at multiple centerline positions in the blood vessel centerline information. The processing module is used to perform input constraint processing on the coronary angiography image to be segmented based on the region of interest (ROI) of the strip polygon, and generate a constrained segmentation input image; The segmentation module is used to segment blood vessels based on the constraint segmentation input image to obtain the target segmentation result of the coronary angiography image to be segmented.

12. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the vascular segmentation method for a coronary angiography sequence as described in any one of claims 1 to 10.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the vascular segmentation method for coronary angiography sequences as described in any one of claims 1 to 10.

14. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the steps of the vascular segmentation method for coronary angiography sequences as described in any one of claims 1 to 10.

Images