Real-time image guidance methods and systems for coronary interventions

Abstract

The present application relates to the technical field of image analysis, in particular to a real-time image guiding method and system for coronary intervention treatment, which comprises the following steps: acquiring DSA image data and OCT cross-sectional image data of a patient's coronary artery; performing lesion block segmentation on static frames in the DSA image data to determine a lesion vessel segment and extract each target static frame of the lesion vessel segment in the OCT cross-sectional image data; determining a lipid calcification co-occurrence index and a lesion fracture confidence of each target static frame, and then determining a lesion fracture index of each target static frame; performing edge detection on each target static frame, and modifying a segmentation edge screening index in the edge detection process based on the lesion fracture index to obtain an edge detection result; and constructing a three-dimensional model of the coronary artery based on the edge detection result. The present application effectively avoids merging multiple discrete lesion regions into a pseudo-continuous lesion plate, and improves the modeling accuracy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image analysis, in particular to a real-time image guiding method and system for coronary intervention treatment. BACKGROUND

[0002] Coronary intervention treatment (PCI) is the cornerstone of treating coronary artery disease caused by myocardial ischemia, and its core goal is to reconstruct the coronary artery blood flow blocked by atherosclerotic plaques through minimally invasive catheter technology. In this highly delicate operation, real-time image guidance plays an indispensable "eye" and "navigator" role, and its accuracy is directly related to the success of the operation and the prognosis of the patient.

[0003] In the existing coronary intervention treatment, the PCI image guiding system relies on the fusion modeling of X-ray angiography and intraluminal images to achieve instrument navigation. However, in the real-time image guiding process, due to the multi-layer nested structure of coronary artery lesions, and the non-continuous distribution of lesion layers in three-dimensional space, the general reconstruction algorithm cannot accurately analyze the relationship between lesions, and is prone to merge multi-layer discrete lesion regions into pseudo-continuous lesion plates, resulting in inaccurate three-dimensional vascular models, such as significant deviations in lesion thickness and spatial orientation in the model, ultimately leading to failure of planning and guiding based on distorted models. SUMMARY

[0004] In order to solve the above technical problem that in the real-time image guiding process, due to the general reconstruction algorithm cannot accurately analyze the relationship between lesions, and is prone to merge multi-layer discrete lesion regions into pseudo-continuous lesion plates, resulting in inaccurate three-dimensional vascular models, the purpose of the present application is to provide a real-time image guiding method and system for coronary intervention treatment, and the technical solution adopted is as follows:

[0005] In the first aspect, the present application provides a real-time image guiding method for coronary intervention treatment, comprising the following steps:

[0006] Obtaining DSA image data and OCT cross-sectional image data of the patient's coronary artery;

[0007] Segmenting the lesion blocks in the static frames of the DSA image data, determining the lesion vessel segments, and extracting each target static frame of the lesion vessel segments in the OCT cross-sectional image data;

[0008] Based on the pixel distribution characteristics of the coronary lesion layer structure region in each target static frame, determining the lipid calcification co-occurrence index of each target static frame;

[0009] Based on the structure change characteristics of the coronary lesion layer structure in each target static frame, determining the lesion fracture confidence of each target static frame;

[0010] determine a lesion rupture index of each target static frame based on the lipid calcification co-occurrence index and the lesion rupture confidence;

[0011] perform edge detection on the target static frames, and modify a segmentation edge screening indicator in the edge detection process based on the lesion rupture index, to obtain an edge detection result, and construct a three-dimensional model of a coronary artery based on the edge detection result.

[0012] In some possible implementation manners, in combination with the first aspect, the lipid calcification co-occurrence index of each target static frame is determined by:

[0013] determine a lesion rupture index of each target static frame based on the lipid calcification co-occurrence index and the lesion rupture confidence;

[0014] determine a lesion rupture index of each target static frame based on the lipid calcification co-occurrence index and the lesion rupture confidence;

[0015] determine a lesion rupture index of each target static frame based on the lipid calcification co-occurrence index and the lesion rupture confidence;

[0016] In some possible implementation manners, in combination with the first aspect, the lesion rupture confidence of each target static frame is determined by:

[0017] determine a lesion rupture index of each target static frame based on the lipid calcification co-occurrence index and the lesion rupture confidence;

[0018] In some possible implementation manners, in combination with the first aspect, the determination of the coronary lesion structure region in each target static frame comprises:

[0019] determine a lesion rupture index of each target static frame based on the lipid calcification co-occurrence index and the lesion rupture confidence;

[0020] determine a lesion rupture index of each target static frame based on the lipid calcification co-occurrence index and the lesion rupture confidence;

[0021] determine a lesion rupture index of each target static frame based on the lipid calcification co-occurrence index and the lesion rupture confidence;

[0022] determine a lesion rupture index of each target static frame based on the lipid calcification co-occurrence index and the lesion rupture confidence;

[0023] In some possible implementation manners, in combination with the first aspect, the lesion rupture confidence of each target static frame is determined by:

[0024] determine a lipid pool arc variation rate corresponding to each target static frame based on a profile change of an intimal region in the layered structure of the coronary lesion in the target static frame;

[0025] determine a fibrous cap thickness corresponding to each target static frame based on a distance from a boundary point of the intimal region in the layered structure of the coronary lesion in the target static frame to a boundary of the blood vessel structure;

[0026] determine an adventitial disruption index corresponding to each target static frame based on a distribution of edge curves of an adventitial region in the layered structure of the coronary lesion in the target static frame;

[0027] determine a lesion rupture confidence of each target static frame based on a correlation between the lipid pool arc variation rate and a variation trend between the fibrous cap thickness and the adventitial disruption index.

[0028] In some possible implementation manners, in combination with the first aspect, the determination of the lipid pool arc variation rate corresponding to each target static frame comprises:

[0029] determine a long axis inclination angle of a minimum circumscribed ellipse of the intimal region in the layered structure of the coronary lesion in the target static frame;

[0030] determine the lipid pool arc variation rate corresponding to each target static frame based on a difference between the long axis inclination angles corresponding to adjacent target static frames and a time interval between the acquisition times.

[0031] In some possible implementation manners, in combination with the first aspect, the determination of the adventitial disruption index corresponding to each target static frame comprises:

[0032] determine a slope of the edge curve of the adventitial region in the layered structure of the coronary lesion in the target static frame;

[0033] determine an average slope of the slopes of all the edge curves in the target static frame, and determine the average slope as the adventitial disruption index corresponding to the target static frame.

[0034] In some possible implementation manners, in combination with the first aspect, the determination of the lesion rupture confidence of each target static frame comprises:

[0035] determine a Pearson correlation coefficient between a first sequence formed by the fibrous cap thicknesses corresponding to adjacent target static frames of the target static frame and a second sequence formed by the adventitial disruption indexes;

[0036] determine the lesion rupture confidence of each target static frame based on a negative correlation mapping result of the lipid pool arc variation rate and the Pearson correlation coefficient.

[0037] In combination with the first aspect, in some possible implementation manners, edge detection is performed on the target static frames, and a segmentation edge screening index in the edge detection process is modified based on the lesion fracture index, so as to obtain an edge detection result, including:

[0038] The lesion fracture index is subjected to negative correlation normalization processing, so as to obtain an edge segmentation correction parameter;

[0039] The U-Net image segmentation method is used to perform edge segmentation on the target static frames, and in the edge segmentation process, the segmentation edge screening index is modified by using the edge segmentation correction parameter, and an edge segmentation result is obtained based on the adaptive segmentation edge screening index obtained after the modification.

[0040] In a second aspect, the present application further provides a real-time image guidance system for coronary intervention treatment, including a memory and a processor. The memory is used to store executable computer program code, and the processor is used to call and run the executable computer program code from the memory, so that the system executes the method in the first aspect or any one of the possible implementation manners of the first aspect.

[0041] In a third aspect, the present application further provides a real-time image guidance device for coronary intervention treatment, including:

[0042] An image data acquisition module is configured to acquire DSA image data and OCT cross-sectional image data of a patient's coronary artery;

[0043] A target frame acquisition module is configured to perform lesion block segmentation on a static frame in the DSA image data, determine a lesion vessel segment, and extract each target static frame of the lesion vessel segment in the OCT cross-sectional image data;

[0044] A co-occurrence index acquisition module is configured to determine a lipid calcification co-occurrence index of each target static frame based on pixel distribution characteristics of a coronary lesion layer structure region in the target static frame;

[0045] A confidence degree acquisition module is configured to determine a lesion fracture confidence degree of each target static frame based on structure change characteristics of a coronary lesion layer structure in the target static frame;

[0046] A fracture index acquisition module is configured to determine a lesion fracture index of each target static frame based on the lipid calcification co-occurrence index and the lesion fracture confidence degree;

[0047] The three-dimensional construction module is used for edge detection on the target static frames, and a segmentation edge screening index in the edge detection process is modified based on the lesion fracture index, so as to obtain an edge detection result. Based on the edge detection result, a three-dimensional model of the coronary artery is constructed.

[0048] In a fourth aspect, the present application further provides a computer program product, which comprises computer program codes, when the computer program codes are run on a computer, the computer is caused to execute the real-time image guidance method for coronary intervention treatment in the first aspect or any possible implementation manner of the first aspect.

[0049] In a fifth aspect, the present application further provides a computer readable storage medium, which stores computer program codes, when the computer program codes are run on a computer, the computer is caused to execute the real-time image guidance method for coronary intervention treatment in the first aspect or any possible implementation manner of the first aspect.

[0050] The present application has the following beneficial effects: by simultaneously acquiring multi-dimensional image data of the coronary artery of the patient, i.e., DSA image data and OCT cross-sectional image data, and by performing lesion block segmentation on the static frames in the DSA image data to determine the lesion vessel segment, and then extracting each target static frame of the lesion vessel segment in the OCT cross-sectional image data; since the coronary artery exhibits a multi-layer structure, such as the intima, media and adventitia, based on the pixel distribution characteristics of each layer structure region of the coronary artery lesion in each target static frame, the pathological coupling strength of the lipid disorder degree and the calcification density is dynamically analyzed to determine the lipid calcification co-occurrence index of each target static frame; at the same time, since the rupture nature of the coronary artery lesion is the structural collapse caused by the mechanical instability of the lipid pool, based on the structural change characteristics of the layered structure of the coronary artery lesion in each target static frame, the lesion rupture confidence of each target static frame is determined to quantify the lesion block rupture possibility of the lesion vessel segment in each target static frame; the lipid calcification co-occurrence index and the lesion rupture confidence are fused to determine the lesion rupture index of each target static frame, when the lesion rupture index is larger, it indicates that the structure of the lesion region is more complex, and there is significant rupture or discontinuity, in this case, the degree of edge pixel point presentation should be limited to the detection algorithm of the rupture region, to realize the structural independence of the discrete lesion region and the clarity of the rupture boundary; then, edge detection is performed on each target static frame, and based on the lesion rupture index, the segmentation edge screening index in the edge detection process is modified, so that the accurate edge detection result is obtained, and finally based on the edge detection result, the three-dimensional model of the coronary artery is accurately constructed. The present application determines the lesion rupture index based on the lipid calcification co-occurrence index and the lesion rupture confidence, and adaptively modifies the segmentation edge screening index in the edge detection process of each target static frame, which can effectively preserve the real edge structure of the independent and discrete lesion blocks in the lesion block, and can realize accurate differentiation between the real lesion structure edge pixel and the cross-rupture region pseudo-boundary pixel, thereby improving the coronary artery modeling accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0051] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0052] Figure 1 A step flow chart of a real-time image guided method for coronary artery intervention treatment according to an embodiment of the present application;

[0053] Figure 2 A schematic diagram of a static frame in DSA image data according to an embodiment of the present application;

[0054] Figure 3 A static frame diagram in the OCT cross-sectional image data of an embodiment of the present application;

[0055] Figure 4 A structural diagram of a real-time image guiding device for coronary intervention treatment of an embodiment of the present application. DETAILED DESCRIPTION

[0056] In order to clearly illustrate the technical features of the scheme, specific implementation manners will be described below in combination with the drawings.

[0057] Embodiments of the present application will be described in more detail by referring to the drawings. Although certain embodiments of the present application are shown in the drawings, it is understood that the present application can be implemented in various forms and should not be interpreted as being limited to the embodiments set forth herein, but rather these embodiments are provided so as to more completely and thoroughly understand the present application. It is understood that the drawings and embodiments of the present application are for exemplary purposes only and are not intended to limit the scope of protection of the present application.

[0058] It should be understood that each of the steps recited in the method embodiments of the present application can be performed in different orders and / or in parallel. In addition, the method embodiments can include additional steps and / or omit the steps shown. The scope of the present application is not limited in this respect.

[0059] The term "comprising" and variations thereof as used herein are open-ended, that is, "comprising but not limited to". The term "based on" is "based, at least in part, on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Related definitions will be given in the description below.

[0060] It should be noted that the concepts of "first", "second", etc. mentioned in the present application are only used to distinguish different devices, modules or units, and are not intended to limit the functions performed by these devices, modules or units in order or interdependence.

[0061] In the embodiments of the present application, although the operations or steps are described in a specific order in the drawings, it should not be understood as requiring the operations or steps to be performed in the specific order or serial order shown, or requiring all of the operations or steps to be performed to obtain the desired results. In the embodiments of the present application, these operations or steps can be performed in series; they can also be performed in parallel; and a part of them can be performed.

[0062] Meanwhile, it can be understood that the data involved in the technical solutions of the present application (including but not limited to the data itself, the acquisition or use of the data) should comply with the requirements of the corresponding laws and regulations and relevant provisions. Unless otherwise defined, all technical and scientific terms used in the present application have the same meaning as understood by those skilled in the art to which the present application belongs, and all parameters or indicators in the formulas involved in the present application are normalized values after eliminating the dimension influence.

[0063] In order to solve the problem that in real-time image guidance process, due to the general reconstruction algorithm cannot accurately analyze the relationship between lesions, it is easy to merge multi-layer discrete lesion regions into pseudo-continuous lesion plates, thereby causing the constructed three-dimensional model of blood vessels to be not accurate enough, the embodiment of the present application provides a real-time image guidance method and system for coronary intervention treatment, which breaks through the technical bottleneck of distortion of pseudo-continuous lesion plates in traditional three-dimensional reconstruction, and first realizes multi-modal dynamic fusion modeling of pathological structure and mechanical instability characteristics, thereby effectively improving the accuracy of coronary artery model construction.

[0064] In the following, a real-time image guidance method and system for coronary intervention treatment provided by the embodiment of the present application will be described in detail with reference to the accompanying drawings.

[0065] Figure 1 A basic flowchart of a real-time image guidance method for coronary intervention treatment provided by the embodiment of the present application is shown, as shown in Figure 1 The method specifically includes the following steps:

[0066] Step S100: Obtain DSA image data and OCT cross-sectional image data of the patient's coronary artery.

[0067] The existing coronary image acquisition system is used to acquire multi-source image data of the patient's coronary artery, and the DSA image data and the OCT cross-sectional image data of the patient's coronary artery are obtained.

[0068] In a specific example, after the guide catheter is positioned to the coronary artery of the patient, first, the pulse time-space encoding contrast agent injection is used to ensure the visualization capture of the coronary blood vessels. Then, the double-axis rotating DSA system performs a 200° rotating scan at an angular velocity of 40° / s, and the scan parameters are configured as: tube voltage 90kV, tube current 120mA, layer thickness 0.28mm, frame rate 30fps, to obtain the coronary angiography sequence image, that is, the DSA image data of the patient's coronary artery is obtained. Figure 2 A static frame schematic diagram in the DSA image data is shown.

[0069] The three-mode coaxial catheter is driven by a stepper motor to perform axial retraction at a constant speed of 0.5 mm / s, and the total retraction stroke is 100 mm. During this process: the middle layer OCT fiber adopts a dual-wavelength alternating scanning mode of 1310 nm and 1550 nm, and outputs a continuous sequence of coronary OCT images in the scanning period, thereby obtaining the OCT cross-sectional image data of the patient's coronary artery. Figure 3 A static frame diagram in the OCT cross-sectional image data is shown.

[0070] Step S200: Perform lesion block segmentation on the static frames in the DSA image data, determine the lesion vessel segment, and extract each target static frame of the lesion vessel segment in the OCT cross-sectional image data.

[0071] The static frames in the DSA image data are segmented into coronary artery lesion blocks, thereby determining the lesion vessel segment in the static frames in the DSA image data.

[0072] In a specific example, first, a pre-trained U-Net image segmentation method is used to pre-segment the coronary artery lesion blocks in each static frame in all DSA image data, generate a coronary artery lesion mask, and determine the vessel segment in the corresponding static frame as the lesion vessel segment.

[0073] Next, for all coronary artery lesion masks, the corresponding lesion vessel segment in the OTC cross-sectional image in the OCT cross-sectional image data is extracted synchronously, and these OTC cross-sectional images are recorded as each target static frame of the lesion vessel segment in the OCT cross-sectional image data. Each target static frame refers to the continuous frame OTC cross-sectional image corresponding to the lesion vessel segment in the OCT cross-sectional image data.

[0074] Step S300: Based on the pixel distribution characteristics of the coronary lesion structure region in each target static frame, determine the lipid calcification co-occurrence index of each target static frame.

[0075] The coronary artery exhibits a three-layered structure, including the intima-lipid metabolism zone, the media-calcification zone, and the adventitia-feeding vascular network. During the three-dimensional reconstruction of the coronary artery, the lipid and calcification regions within rapidly calcified lesions are highly complex and intertwined. The 3D reconstruction process requires refining the selection scale of edge pixels in the final two-dimensional image of this region and removing false boundary pixels across fracture areas to ensure accurate identification of the true boundaries of discrete lesion regions. Therefore, this embodiment first identifies the structural regions (intima, media, and adventitia) of the coronary lesion in the target static frame and determines the lipid-calcification co-occurrence index of the target static frame based on the pixel distribution characteristics of each structural region. This lipid-calcification co-occurrence index reflects the degree of synchronous activity of lipid inflammation and calcification deposition in the diseased vascular segment. When the lipid-calcification co-occurrence index significantly increases, it indicates that lipid inflammation and calcification deposition are synchronously active in the lesion, forming a "high-fluidity lipid-high-rigidity calcification" interlayer. Under blood flow shear force, this induces interlayer attention concentration, increasing the complexity of cross-layer coupling in the lesion.

[0076] Due to differences in their composition, the various structural regions of coronary lesions (intima, media, and adventitia) exhibit significant pixel differences across different layers. For example, the intima (lipid pool), media (calcification deposition), and adventitia (vascular integrity) layers exhibit low, high, and medium grayscale, respectively. Therefore, by analyzing the pixel distribution characteristics of the structural regions of coronary lesions in each target static frame, the structural regions of coronary lesions in the target static frame can be identified first.

[0077] Furthermore, the process of determining the layer structure regions of the coronary lesions in each target static frame in step S300 above includes: determining the distance and pixel difference value between any two pixels in the target static frame; determining the distance metric improvement value between the two pixels based on the distance and pixel difference value; clustering all pixels in the target static frame based on the distance metric improvement value to obtain each cluster; and determining the specific type of the layer structure region corresponding to each cluster based on the pixel mean value of the pixels in each cluster.

[0078] In a specific example, firstly, for each target static frame of the diseased blood vessel segment in the OCT cross-sectional image data, the pixel value of each pixel in the target static frame is counted. And calculate the distance between any two pixels. and pixel difference values that distance The pixel difference value is the Euclidean distance between any two pixels. It represents the absolute value of the difference between the grayscale values ​​of any two pixels.

[0079] Secondly, based on the distance between any two pixels and pixel difference value , the distance metric improvement value between any two pixel points . Wherein, represents the distance between any two pixel points represents the pixel difference value between any two pixel points represents the max-min value normalization function, which is used to standardize and

[0080] Next, the elbow method is used to determine the (corresponding to intima / media / adventitia) in the k-means clustering algorithm, and all pixel points in the target static frame are clustered based on the traditional k-means clustering algorithm and the distance metric improvement value Through iteration (traditional clustering iteration), the clustering results are constantly updated until all clustering results clusters no longer change, thereby obtaining three clustering clusters, each of which represents the same type of structure pixel points in the diseased vessel segment.

[0081] Finally, the pixel mean value of the pixel points in each clustering cluster is calculated, and the three clustering clusters are divided into low gray, high gray and medium gray pixel point clusters according to the of the three clustering clusters, and the regions formed by the three types of pixel point clusters correspond to the intima (lipid pool), media (calcification deposition) and adventitia (nourishing blood vessel integrity) layer structures of the diseased vessel segment. Thus, the identification of the coronary artery lesion layer structure region (intima, media and adventitia) in the target static frame can be obtained. It should be understood that other ways in the prior art can also be used to identify the coronary artery lesion layer structure region in the target static frame, which is not limited here.

[0082] For each target static frame of the diseased vessel segment in the OCT cross-sectional image data, the pixel distribution in the intima region is disordered, indicating that the pathological activity of the lipid pool in the current target static frame is high, indicating that the cell debris / cholesterol crystals in the lipid pool of the lesion are increasing, meaning that the lesion inflammation is active. At the same time, as the lesion worsens, the calcification deposition in the calcification area (adjacent to the media layer) becomes more significant, showing a significant increase in the regional pixel mean value, so the lipid calcification co-occurrence index of each target static frame can be determined by analyzing the pixel distribution characteristics of the coronary artery lesion layer structure region in each target static frame.

[0083] ​​​​Further, the step S300 of determining the lipid calcification co-occurrence index of each target static frame comprises: determining a plaque lipid heterogeneity index based on pixel distribution difference in the intima region in the coronary lesion layer structure region in each target static frame; determining a plaque calcification density coefficient based on pixel distribution level in the intima region in the coronary lesion layer structure region in each target static frame; and determining the lipid calcification co-occurrence index of each target static frame based on the plaque lipid heterogeneity index and the plaque calcification density coefficient.

[0084] In a specific example, first, for the coronary lesion layer structure region in each target static frame of the lesion vessel segment in the OCT cross-sectional image data, for the lipid pool pixel set i.e. the low gray pixel point cluster corresponding to the intima region, the pixel variance of all pixel points in the set is calculated. The pixel variance is taken as the plaque lipid heterogeneity index to quantify the pathological activity intensity of the lipid pool in the current target static frame.

[0085] Secondly, for the calcification region pixel set i.e. the high gray pixel point cluster corresponding to the media, the pixel mean of all pixel points in the set is calculated. The pixel mean is taken as the plaque calcification density coefficient to quantify the calcification deposition degree in the current target static frame. The plaque calcification density coefficient has a larger value, indicating that the calcification deposition is more significant and the brittleness increases.

[0086] Finally, the plaque lipid heterogeneity index and the plaque calcification density coefficient are combined to measure the lipid calcification co-occurrence index of the lesion vessel segment in the current target static frame, wherein, represents the lipid calcification co-occurrence index of the target static frame n.

[0087] In the step S300, by accurately decoupling the coronary lesion layer structure in each target static frame and analyzing the pixel distribution characteristics in each layer structure region, the plaque lipid heterogeneity index and the plaque calcification density coefficient are determined to quantify the lipid pool pathological activity intensity and the calcification deposition degree of the lesion vessel segment in each target static frame, so as to accurately determine the lipid calcification co-occurrence index of each target static frame.

[0088] Step S400: determining a lesion rupture confidence of each target static frame based on the structural change characteristics of the coronary lesion layer structure in each target static frame.

[0089] In the process of coronary artery three-dimensional reconstruction, for the plaque area with multi-layer nested structure and the discrete (non-continuous) distribution of the lesion structure in space, the edge detection algorithm in the traditional segmentation model cannot accurately distinguish the true lesion structure edge pixels and the pseudo boundary pixels across the broken area (both have higher edge pixel gradient compared with normal pixels) only relying on the edge point pixel gradient, resulting in the blurring of the final broken area boundary, and the multiple discrete lesion layers are miscombined as pseudo continuous lesion area.

[0090] Since the broken nature of coronary artery lesions is the structural collapse caused by the mechanical instability of lipid pools, specifically manifested as the entanglement of bending mutation and calcification and lipid area in the image, by analyzing the target static frame of the lesion vessel segment in the OCT cross-sectional image data, the true broken condition of the lesion can be accurately identified, thereby constraining the edge point identification in the three-dimensional reconstruction process, and finally improving the accuracy of the lesion area segmentation.

[0091] Since the coronary plaque rupture usually shows structural characteristic changes of coronary lesion layered structure, for example, in the coronary plaque structure, the lipid pool is relatively weak and is easily affected by blood flow pressure and surrounding stress, and breaks when exceeding the bearing limit, resulting in sharp changes in the edge, reflecting the plaque structure rupture, therefore, by identifying the structural change characteristics of the coronary lesion layered structure in each target static frame, the lesion rupture confidence of each target static frame can be determined to quantify the plaque rupture possibility of the lesion vessel segment in each target static frame in the OCT cross-sectional image data.

[0092] Further, the step S400 of determining the lesion rupture confidence of each target static frame comprises: determining the lipid pool curvature change rate corresponding to each target static frame based on the contour change of the intimal region in the coronary lesion layered structure in each target static frame; determining the fiber cap thickness corresponding to each target static frame based on the distance from the boundary point of the intimal region in the coronary lesion layered structure in each target static frame to the blood vessel structure boundary; determining the adventitial rupture index corresponding to each target static frame based on the edge curve distribution of the adventitial region in the coronary lesion layered structure in each target static frame; and determining the lesion rupture confidence of each target static frame based on the correlation between the change trend of the lipid pool curvature change rate and the fiber cap thickness and the adventitial rupture index.

[0093] In a specific example, first, since the coronary plaque rupture usually shows the curvature mutation of the lipid pool region, by analyzing the contour change of the intimal region in the coronary lesion layered structure in each target static frame, the lipid pool curvature change rate corresponding to each target static frame can be determined.

[0094] Specifically, the major axis tilt angle of the minimum circumscribed ellipse of the intima region within the coronary lesion layered structure in each target static frame is determined. That is, for the lipid pool region (i.e., the intima region composed of low-grayscale pixel clusters) within the coronary lesion layered structure, the minimum circumscribed ellipse of the lipid pool region is determined, and the major axis tilt angle of this ellipse is extracted. The major axis tilt angle This refers to the angle between the major axis of the ellipse and the horizontal line, using the inclination angle of the major axis. This reflects the curvature of the circumscribed ellipse. Then, based on the difference in the major axis tilt angles between adjacent target static frames and the acquisition time interval, the rate of change of lipid pool curvature for each target static frame is determined. For target static frame n, the rate of change is determined based on the major axis tilt angles corresponding to target static frame n and its next target static frame n+1. and And combined with the acquisition time interval between adjacent target static frames Determine the rate of change of lipid pool curvature corresponding to the target static frame n. The greater the rate of change in the curvature of the lipid pool, the more drastic the morphological changes in the lipid pool region of the lesion in the corresponding lesion vessel segment in adjacent static frames n and n+1, suggesting a possible break in the lesion structure distribution.

[0095] Secondly, to eliminate false-positive curvature noise caused by catheter displacement artifacts and calcification shadows, mechanical instability verification was performed using the negative correlation coupling mechanism between the effective thickness of the fibrous cap and the adventitia rupture index within the axial sequence. Therefore, the fibrous cap thickness corresponding to each target static frame was determined based on the distance from the boundary point of the intima region in the coronary lesion layered structure to the boundary of the vascular structure in each target static frame; simultaneously, the adventitia rupture index corresponding to each target static frame was determined based on the edge curve distribution of the adventitia region in the coronary lesion layered structure in each target static frame.

[0096] Specifically, for each target static frame, a discrete point set is extracted from the boundary of the intima region within the layered structure of the coronary lesion. The distance from each point in the discrete point set to the boundary of the vascular structure (the outermost edge of the Canny edge detection) is calculated, and the minimum value among all distances is taken. The thickness of the fibrous cap is used to represent the current location of the coronary lesion structure. A smaller fibrous cap thickness indicates a weakened intimal pressure-bearing capacity of the lesion structure at that location. Simultaneously, the slope of the edge curves of the adventitia region within the layered coronary lesion structure in each target static frame is determined. For the adventitia region within the layered coronary lesion structure, i.e., the region composed of mid-grayscale pixel clusters, all complete edge curves in the adventitia region are extracted, and the curvature of each edge curve is calculated. Then, the average slope of the slopes of all edge curves in each target static frame is determined, and the average slope is determined as the adventitia rupture index corresponding to each target static frame. That is, for the adventitia region in the layered structure of the coronary lesion in each target static frame, the average curvature of all edge curves in the adventitia region is calculated , and the average curvature is taken as the adventitia rupture index of the coronary plaque in the target static frame. The greater the value of the adventitia rupture index, the greater the loss of the integrity of the adventitia structure, i.e., the folding / breaking of the adventitia collagen fiber network.

[0097] Finally, the lipid pool arc rate of the target static frame is corrected using the correlation between the fiber cap thickness and the adventitia rupture index corresponding to each adjacent target static frame of the target static frame, so as to determine the lesion rupture confidence of each target static frame.

[0098] Specifically, the Pearson correlation coefficient between the first sequence composed of the fiber cap thickness corresponding to each adjacent target static frame of each target static frame and the second sequence composed of the adventitia rupture index is determined. That is, for any target static frame, the nearest several target static frames (such as 5 frames) are taken as the adjacent target static frames, and the fiber cap thickness and the adventitia rupture index corresponding to these adjacent target static frames are arranged in the order of the arrangement of the adjacent target static frames, so as to obtain the first sequence and the second sequence . The Pearson correlation coefficient of the first sequence and the second sequence is calculated, and the value of the correlation coefficient ranges from -1 to 1. The closer the value is to -1, the more synchronous the thinning of the fiber cap and the tearing of the adventitia in the lesion segment, which indicates that the lesion structure in the lesion segment is significantly mechanically unstable, and indicates that the possibility of the existence of a rupture point in the lesion structure at the position corresponding to the current OCT target static frame is greater. Further, based on the negative correlation mapping result of the lipid pool arc rate and the Pearson correlation coefficient, the lesion rupture confidence of each target static frame is determined. That is, for each target static frame distributed along the axial direction of the blood vessel, the lesion rupture confidence of the position corresponding to each target static frame is determined in combination with the lipid pool arc rate and the Pearson correlation coefficient of the fiber cap thickness and the adventitia rupture index of the corresponding coronary plaque structure. For the target static frame n, based on the lipid pool arc rate corresponding to the target static frame n and the Pearson correlation coefficient of the first sequence and the second sequence , the corresponding lesion rupture confidence The value of the lesion rupture confidence is larger, indicating that the lipid pool of the current position of the lesion structure has a high degree of deformation, the fiber cap is thinned, and the synchronization of the outer membrane rupture is enhanced, meaning that the corresponding target static frame corresponds to a coronary lesion in a physically unstable state, indicating that there is a significant breaking point or breaking area at this position.

[0099] In the above step S400, by determining the lipid pool arc rate of change, the fiber cap thickness and the outer membrane rupture index corresponding to each target static frame, and finally based on the change trend correlation between the lipid pool arc rate of change and the fiber cap thickness and the outer membrane rupture index, the lesion rupture confidence of each target static frame is accurately determined.

[0100] Step S500: Based on the lipid calcification co-occurrence index and the lesion rupture confidence, determine the lesion rupture index of each target static frame.

[0101] The lipid calcification co-occurrence index and the lesion rupture confidence of each target static frame determined above are fused to obtain the lesion rupture index of each target static frame. The lesion rupture index is a comprehensive quantitative index of the lesion blood vessel segment at the position corresponding to the target static frame, and its value directly reflects the decision basis for the final edge pixel point retention degree of the rupture region in the region segmentation in the three-dimensional reconstruction process. A larger value means that the structure of the lesion region is more complex, and there is a significant rupture or discontinuity. In this case, the presentation retention degree of the edge pixel point detected by the rupture region algorithm should be limited to achieve the structural independence of the discrete lesion region and the clarity of the rupture boundary, and to ensure the accuracy of the subsequent three-dimensional modeling result.

[0102] In a specific example, for any target static frame n, based on its corresponding lipid calcification co-occurrence index and the lesion rupture confidence , its corresponding lesion rupture index is determined. Wherein, the logarithmic function aims to compress the lipid calcification co-occurrence index to a reasonable range.

[0103] Step S600: Perform edge detection on the target static frames, and based on the lesion rupture index, modify the segmentation edge screening index in the edge detection process to obtain the edge detection result. Based on the edge detection result, a three-dimensional model of the coronary artery is constructed.

[0104] In the U-Net image segmentation method, first, the input image is feature extracted by a convolutional neural network (CNN). The U-Net network structure has an encoder and a decoder, the encoder gradually compresses the features of the input image, and the decoder restores the compressed features to the size of the original image. In this process, the U-Net network combines the shallow features in the encoder with the deep features in the decoder through a skip connection to retain more detailed information.

[0105] Based on this, the U-Net image segmentation method can effectively extract the detailed information in the image by fusing features of different scales and accurately predict the pixel-level segmentation result at the boundary. In the segmentation process, all edge pixels are sorted according to the edge point pixel gradient to obtain an edge point gradient sequence, and then when obtaining the edge curve of the final image, the edge pixel points ranked in the top proportion (the proportion is also called a segmentation edge screening index) are selected as the final segmentation edge pixel points of the image (segmentation method in the prior art). In order to further optimize the segmentation result, the embodiment of the present application further constrains and screens the final edge pixel points of the lesion vessel region of each target static frame during algorithm segmentation by using the lesion fracture index, so as to effectively retain more detailed information and reduce information loss caused by edge overlap or small edge.

[0106] Further, the step S600 detects the edge of each target static frame, and modifies the segmentation edge screening index in the edge detection process based on the lesion fracture index, so as to obtain an edge detection result, including: performing negative correlation normalization processing on the lesion fracture index to obtain an edge segmentation correction parameter; using the U-Net image segmentation method to perform edge segmentation on each target static frame, in the edge segmentation process, using the edge segmentation correction parameter to correct the segmentation edge screening index, and based on the adaptive segmentation edge screening index obtained after correction, obtaining an edge segmentation result.

[0107] In a specific example, the U-Net image segmentation method is used to perform edge segmentation on each target static frame, and in the segmentation process, for any target static frame n, based on its corresponding lesion fracture index , the segmentation edge screening index in the U-Net image segmentation method is corrected . Wherein, represents the initial segmentation edge screening index of the target static frame n using the U-Net image segmentation method, that is, the above-mentioned proportion; represents the adaptive segmentation edge screening index of the target static frame n after correction, represents a linear normalization function, which is used to normalize the lesion fracture index The value of the lipid calcification co-occurrence index is normalized to the range of (0, 1). Thus, based on the adaptive segmentation edge screening index obtained after the correction of each target static frame, the edge of each target static frame is segmented by using a U-Net image segmentation method to obtain a final edge pixel set of each target static frame, and then the real edge curve of each target static frame is obtained, so that the edge segmentation result is obtained, and finally the precise segmentation of the discrete lesion area is realized.

[0108] Based on the real edge curve of each target static frame, i.e. the edge segmentation result, a three-dimensional reconstruction algorithm is used to perform three-dimensional reconstruction of the blood vessels, so that a three-dimensional model of the coronary artery blood vessels is obtained. The three-dimensional model of the blood vessels includes two types of structures: a continuous lesion area (a smooth over-surface) and a discrete lesion block (an independent geometric body separated by a broken boundary), thereby effectively eliminating the modeling error of the "pseudo-continuous lesion plate" in the traditional method, and helping the doctor to perform subsequent interventional treatment based on the three-dimensional model of the coronary artery.

[0109] In the real-time image guidance method for coronary artery interventional treatment provided in the embodiments of the present application, first, the three-layer lesion structure is accurately decoupled based on the clustering algorithm for OCT cross-section influence, a lipid calcification co-occurrence index quantification model is created, the pathological coupling strength of lipid disorder degree and calcification density is dynamically analyzed, and the problem of mis-merging of discrete lesion structures in the nested lesion layer is solved; secondly, a lesion fracture confidence double-factor verification mechanism is proposed, the correlation between the lipid pool arc mutation detection and the mechanical instability of the fibrous cap outer membrane is analyzed, and multi-dimensional cross verification of the fracture boundary is realized; finally, a lesion connectivity index dynamic constraint engine is creatively constructed, the pathological complexity and the mechanical instability signal are converted into spatial edge pixel screening rules, i.e. the value of the lesion fracture index determined based on the lipid calcification co-occurrence index and the lesion fracture confidence is used to correct the edge pixel screening scale in the original segmentation model, the number of edge points output by the algorithm is reduced for the high lesion fracture index area, so as to realize the preservation of the real edge structure of the independent discrete lesion block in the lesion block, provide millimeter-level precision fracture boundary navigation capability for the coronary artery interventional surgery, and effectively weaken the problem that the real lesion structure edge pixel and the pseudo-boundary pixel point across the fracture area cannot be accurately distinguished in the traditional method, so that the fuzzy and indistinguishable fracture area boundary is obtained, and the modeling distortion of the vulnerable plaque is further caused.

[0110] Based on the same inventive concept, as shown in Figure 4 The present application also provides a real-time image guidance device for coronary artery interventional treatment, which comprises:

[0111] An image data acquisition module is configured to acquire DSA image data and OCT cross-sectional image data of the coronary artery of a patient.

[0112] a target frame acquisition module configured to perform lesion block segmentation on static frames in the DSA image data, determine a lesion vessel segment, and extract each target static frame of the lesion vessel segment in the OCT cross-sectional image data;

[0113] a co-occurrence index acquisition module configured to determine a lipid calcification co-occurrence index of each target static frame based on pixel distribution characteristics of a coronary artery lesion layer structure region in the target static frame;

[0114] a confidence degree acquisition module configured to determine a lesion rupture confidence degree of each target static frame based on structure change characteristics of a coronary artery lesion layer structure in the target static frame;

[0115] a rupture index acquisition module configured to determine a lesion rupture index of each target static frame based on the lipid calcification co-occurrence index and the lesion rupture confidence degree;

[0116] a three-dimensional construction module configured to perform edge detection on the target static frames, modify a segmentation edge screening index in the edge detection process based on the lesion rupture index, obtain an edge detection result, and construct a three-dimensional model of a coronary artery based on the edge detection result.

[0117] It should be noted that the apparatus provided in the above embodiments is only used as an example for the division of the above functional modules, and in actual applications, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the computer device is divided into different functional modules to complete all or part of the above-described functions.

[0118] Based on the same inventive concept, the embodiments of the present application also provide a real-time image guidance system for coronary intervention treatment, which comprises a memory, a processor and computer program code stored in the memory and running on the processor, wherein when the processor executes the computer program code, the system can execute any one of the real-time image guidance methods for coronary intervention treatment introduced above.

[0119] The embodiments of the present application can divide the system into functional modules according to the above method examples, for example, each functional module can be corresponding, or two or more functions can be integrated in one processing module, and the integrated module can be realized in the form of hardware. It should be noted that the division of the modules in the present embodiment is illustrative, and is only a logical function division, and another division mode can be used in actual implementation.

[0120] Based on the same inventive concept, the embodiments of the present application further provide a computer program product, which comprises computer program codes, and when the computer program codes are run on a computer, the computer is enabled to perform any one of the aforementioned real-time image guidance methods for coronary intervention treatment.

[0121] Based on the same inventive concept, the embodiments of the present application further provide a computer readable storage medium, which stores computer program codes, and when the computer program codes are run on a computer, the computer is enabled to perform any one of the aforementioned real-time image guidance methods for coronary intervention treatment.

[0122] It should be noted that the above-mentioned embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalent ones; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A real-time image guidance method for coronary interventions, characterized in that, The method comprises the following steps: obtaining DSA image data and OCT cross-sectional image data of a patient's coronary artery; performing lesion block segmentation on static frames in the DSA image data, determining a lesion vessel segment, and extracting each target static frame of the lesion vessel segment in the OCT cross-sectional image data; determining a lipid calcification co-occurrence index of each target static frame based on pixel distribution characteristics of each layer structure region of the coronary lesion in the target static frame; determining a lesion rupture confidence of each target static frame based on structure change characteristics of the layer structure of the coronary lesion in the target static frame; determining a lesion rupture index of each target static frame based on the lipid calcification co-occurrence index and the lesion rupture confidence; performing edge detection on the target static frames, and modifying a segmentation edge screening index in the edge detection process based on the lesion rupture index to obtain an edge detection result, and constructing a three-dimensional model of the coronary artery based on the edge detection result; determining the lipid calcification co-occurrence index of each target static frame comprises: determining a lesion block lipid heterogeneity index based on pixel distribution differences in the intimal region of each layer structure region of the coronary lesion in the target static frame; determining a lesion block calcification density coefficient based on pixel distribution levels in the intimal region of each layer structure region of the coronary lesion in the target static frame; determining the lipid calcification co-occurrence index of each target static frame based on the lesion block lipid heterogeneity index and the lesion block calcification density coefficient; determining the lesion rupture confidence of each target static frame comprises: determining a lipid pool arc change rate corresponding to each target static frame based on contour changes of the intimal region in the layer structure of the coronary lesion in the target static frame; determining a fibrous cap thickness corresponding to each target static frame based on distances from boundary points of the intimal region in the layer structure of the coronary lesion in the target static frame to the boundary of the vascular structure; determining an adventitial rupture index corresponding to each target static frame based on edge curve distribution of the adventitial region in the layer structure of the coronary lesion in the target static frame; determining the lesion rupture confidence of each target static frame based on the lipid pool arc change rate and a change trend correlation between the fibrous cap thickness and the adventitial rupture index.

2. The method for real-time image guidance for coronary interventions according to claim 1, characterized in that, determining the lesion block lipid heterogeneity index comprises: determining a pixel variance in the intimal region of each layer structure region of the coronary lesion in the target static frame, and determining the pixel variance as the lesion block lipid heterogeneity index.

3. The method for real-time image guidance of coronary interventions according to any one of claims 1-2, characterized in that, The determination process of the layer structure region of the coronary lesion in each target static frame comprises: determining a distance and a pixel difference value between any two pixel points in the target static frame; determining a distance measurement improvement value between the any two pixel points based on the distance and the pixel difference value; performing clustering on all pixel points in the target static frame based on the distance measurement improvement value to obtain each clustering cluster; determining a specific type of the layer structure region corresponding to each clustering cluster according to a pixel mean value of the pixel points in each clustering cluster.

4. The method for real-time image guidance for coronary interventions according to claim 1, characterized in that, determining the lipid pool arc change rate corresponding to each target static frame comprises: determine a long axis inclination angle of a minimum circumscribed ellipse of an intimal region in the layered structure of the coronary lesion in each target static frame; determine a lipid pool arc variation rate corresponding to each target static frame based on a difference between long axis inclination angles corresponding to adjacent target static frames and a time interval between the acquisition.

5. The method for real-time image guidance of coronary interventions according to claim 1, characterized in that, determine an adventitial disruption index corresponding to each target static frame, including: determine a slope of an edge curve of an adventitial region in the layered structure of the coronary lesion in each target static frame; determine an average slope of the slopes of all the edge curves in each target static frame, and determine the average slope as the adventitial disruption index corresponding to each target static frame.

6. The method for real-time image guidance for coronary interventions according to claim 1, characterized in that, determine a lesion rupture confidence of each target static frame, including: determine a Pearson correlation coefficient between a first sequence constituted by the fibrous cap thicknesses corresponding to each adjacent target static frame of each target static frame and a second sequence constituted by the adventitial disruption indexes; determine a lesion rupture confidence of each target static frame based on the lipid pool arc variation rate and a negative correlation mapping result of the Pearson correlation coefficient.

7. The method for real-time image guidance for coronary interventions according to claim 1, characterized in that, perform edge detection on the target static frames, and modify a segmentation edge screening index in the edge detection process based on the lesion rupture index, so as to obtain an edge detection result, including: perform negative correlation normalization on the lesion rupture index to obtain an edge segmentation correction parameter; perform edge segmentation on the target static frames by using a U-Net image segmentation method, and in the edge segmentation process, correct the segmentation edge screening index by using the edge segmentation correction parameter, and obtain an edge segmentation result based on the adaptive segmentation edge screening index obtained after the correction.

8. A real-time image guidance system for coronary interventions, characterized in that The computer program product comprises a memory, a processor, and executable computer program code stored in the memory and executable on the processor, and the processor executes the computer program code to perform the real-time image guided method for coronary intervention treatment according to any one of claims 1-7.

Images