Bridge vessel segmentation extraction method, device, system and readable storage medium

Abstract

The application provides a bridge vessel segmentation and extraction method, device, system and readable storage medium, wherein position information of an inlet and position information of an outlet of a bridge vessel are acquired based on coronary angiography image data, and a segmentation result of the bridge vessel is acquired; then the segmentation result is screened by using the position information of the inlet and the position information of the outlet of the bridge vessel, so that the bridge vessel is extracted. The application screens the segmentation result by using the position information of the inlet and the position information of the outlet of the bridge vessel, optimizes the segmentation result, improves the accuracy of the extracted bridge vessel, and further improves the accuracy of the prognosis evaluation of the bridge vessel. Moreover, the application can be automatically completed without interaction with a user, and has higher segmentation efficiency and consistency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to a bridge blood vessel segmentation and extraction method, device, system and readable storage medium. BACKGROUND

[0002] Coronary artery bypass grafting (CABG) is one of the main surgical treatments for coronary heart disease. With the increasing incidence of coronary heart disease, the number of cases of coronary artery bypass grafting is increasing year by year. Coronary artery bypass grafting is a surgical method that uses the patient's own blood vessels (such as thoracic internal artery, lower limb saphenous vein, and radial artery) or artificial blood vessels to bypass the obstruction and continue to flow forward. It can also be used to open several side holes in a vein to be connected to several coronary arteries, which is called sequential bypass or serpentine bridge.

[0003] The common postoperative complication of coronary artery bypass grafting is bridge blood vessel stenosis or occlusion, so postoperative follow-up evaluation of the patency of the bridge blood vessel is of great significance, and the evaluation of the long-term prognosis of the arteriovenous bridge blood vessel is particularly important. In postoperative evaluation, the patient needs to undergo angiography to obtain coronary angiography image data, and then the coronary angiography image data is processed to segment the bridge blood vessel.

[0004] However, since the bridge blood vessel is from other parts of the human body, and during the blood vessel transplantation process, the trend of the blood vessel and the flow of blood may change, and it may be connected to multiple positions, resulting in a bridge blood vessel that is different from ordinary blood vessels and has specificity and diversity. The conventional segmentation method cannot accurately segment the bridge blood vessel from the coronary angiography image data. In addition, the postoperative age of the bridge blood vessel further increases the difficulty of segmentation. For some bridge blood vessels that have been operated for a long time, due to the insufficient concentration of contrast agent and the thinness of the blood vessel itself (which may have stenosis, plaque and other lesions), it is easy to occur. The situation of unclear imaging leads to further increase in the difficulty of bridge blood vessel segmentation, resulting in low accuracy of the evaluation of the prognosis of the bridge blood vessel. SUMMARY

[0005] The purpose of the present application is to provide a bridge blood vessel segmentation and extraction method, device, system and readable storage medium to solve the problem that the existing segmentation method cannot accurately segment the bridge blood vessel from the coronary angiography image data.

[0006] In order to achieve the above-mentioned purpose, the present application provides a bridge blood vessel segmentation and extraction method, comprising:

[0007] Obtaining coronary angiography image data;

[0008] acquire position information of an inlet and position information of an outlet of the bridge vessel based on the coronary angiography image data, and acquire a segmentation result of the bridge vessel; and

[0009] screen the segmentation result by using the position information of the inlet and the position information of the outlet of the bridge vessel to extract the bridge vessel.

[0010] Optionally, the position information of the inlet and the position information of the outlet of the bridge vessel are coordinate points of the inlet and the outlet of the bridge vessel or Gaussian kernels representing the position of the inlet and the position of the outlet of the bridge vessel.

[0011] Optionally, the segmentation result includes a plurality of blood vessels, and the step of screening the segmentation result by using the position information of the inlet and the position information of the outlet of the bridge vessel includes:

[0012] superimposing the position information of the inlet and the position information of the outlet of the bridge vessel on the segmentation result;

[0013] comparing each of the blood vessels with the position information of the inlet and the position information of the outlet of the bridge vessel, regarding the blood vessel whose position of the inlet and / or the position of the outlet coincides with the position of the inlet and / or the position of the outlet of the bridge vessel as a determined blood vessel, and determining that the determined blood vessel belongs to the bridge vessel; and

[0014] judging whether the determined blood vessel is continuous between the inlet and the outlet of the bridge vessel, and regarding the determined blood vessel as the bridge vessel when it is determined to be continuous.

[0015] Optionally, regarding the blood vessel whose position of the inlet and position of the outlet do not coincide with the position of the inlet and the position of the outlet of the bridge vessel as a pending blood vessel, and when it is determined to be discontinuous, the step of screening the segmentation result by using the position information of the inlet and the position information of the outlet of the bridge vessel further includes:

[0016] obtaining a distance threshold value between the bridge vessel and the aorta based on the type of the bridge vessel, removing the pending blood vessel whose distance to the aorta is greater than the distance threshold value; and / or obtaining a trend of the pending blood vessel relative to the heart based on a height difference between the inlet and the outlet of the pending blood vessel, removing the pending blood vessel with a transverse trend relative to the heart; and / or obtaining a position of the pending blood vessel relative to the heart based on a distance between the pending blood vessel and the left ventricle and the right ventricle, removing the pending blood vessel closer to the right ventricle;

[0017] determining that the remaining pending blood vessel belongs to the bridge vessel; and

[0018] connecting the determined blood vessel and the remaining pending blood vessel to obtain the bridge vessel.

[0019] Optionally, before obtaining the distance threshold between the bridging vessel and the aorta based on the type of the bridging vessel, the step of screening the segmentation result using the position information of the inlet and the position information of the outlet of the bridging vessel further comprises:

[0020] Obtaining the volume and / or length of the candidate vessels, removing the candidate vessels with a volume less than a first predetermined value and / or a length less than a second predetermined value; and / or, obtaining the distance between the candidate vessels and the aorta, removing the candidate vessels with a distance greater than a third predetermined value from the aorta, the third predetermined value being greater than each of the distance thresholds.

[0021] Optionally, inputting the coronary angiography image data into a trained first deep neural network to obtain the position information of the inlet and the position information of the outlet of the bridging vessel; and / or,

[0022] Inputting the coronary angiography image data into a trained second deep neural network to obtain the segmentation result of the bridging vessel.

[0023] Optionally, the coronary angiography image data comprises CT coronary angiography image data or DSA coronary angiography image data; and / or, the first deep neural network comprises an hourglass deep neural network, a V-net deep neural network or a U-net deep neural network; and / or, the second deep neural network comprises a V-net deep neural network or a U-net deep neural network.

[0024] The present application also provides a bridging vessel segmentation and extraction device, comprising:

[0025] A data supply module for supplying coronary angiography image data;

[0026] A position acquisition module for acquiring position information of the inlet and position information of the outlet of the bridging vessel;

[0027] A segmentation module for obtaining the segmentation result of the bridging vessel; and,

[0028] A screening module for screening the segmentation result using the position information of the inlet and the position information of the outlet of the bridging vessel to extract the bridging vessel.

[0029] The present application also provides a bridging vessel segmentation and extraction system comprising a processor and a memory, the memory having instructions stored thereon, when the instructions are executed by the processor, the bridging vessel segmentation and extraction method is implemented.

[0030] The application further provides a non-transitory computer-readable storage medium, which stores instructions, and the instructions, when executed, implement the bridge vessel segmentation and extraction method.

[0031] In the bridge vessel segmentation and extraction method, device, system and readable storage medium provided by the application, the position information of the inlet and the position information of the outlet of the bridge vessel are acquired based on the coronary angiography image data, and the segmentation result of the bridge vessel is acquired; and the segmentation result is screened by using the position information of the inlet and the position information of the outlet of the bridge vessel to extract the bridge vessel. The application screens the segmentation result by using the position information of the inlet and the position information of the outlet of the bridge vessel, optimizes the segmentation result, improves the accuracy of the extracted bridge vessel, and further improves the accuracy of the prognosis evaluation of the bridge vessel. Moreover, the application can be automatically completed without user interaction, and has higher segmentation efficiency and consistency. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 A flowchart of the bridge vessel segmentation and extraction method provided by the first embodiment of the application;

[0033] Figure 2a A schematic diagram of the single-layer hourglass deep neural network model provided by the first embodiment of the application;

[0034] Figure 2b A schematic diagram of the double-layer hourglass deep neural network model provided by the first embodiment of the application;

[0035] Figure 3 A schematic diagram of the segmentation result provided by the first embodiment of the application;

[0036] Figure 4 A schematic diagram of the position information of the inlet and the position information of the outlet of the bridge vessel superimposed on the segmentation result provided by the first embodiment of the application;

[0037] Figure 5 A schematic diagram of the finally extracted bridge vessel provided by the first embodiment of the application;

[0038] Figure 6 A schematic diagram of the bridge vessel extracted by the bridge vessel segmentation and extraction method provided by the first embodiment of the application in actual application;

[0039] Figure 7 A schematic diagram of the segmentation result provided by the second embodiment of the application;

[0040] Figure 8 A schematic diagram of the position information of the inlet and the position information of the outlet of the bridge vessel superimposed on the segmentation result provided by the second embodiment of the application;

[0041] Figure 9 A schematic diagram of the finally extracted bridge blood vessels provided for Embodiment Two of the present application;

[0042] Figure 10 A structural block diagram of the bridge blood vessel segmentation and extraction device provided for Embodiment Three of the present application;

[0043] In the drawings, reference numerals are used:

[0044] 10 - data supply module; 20 - position acquisition module; 30 - segmentation module; 40 - screening module;

[0045] a, b, c, d, h, i, j, k, l, m, n - blood vessels; e, f, o, p - Gaussian kernel. DETAILED DESCRIPTION

[0046] The specific embodiments of the present application will be described in more detail below with reference to the accompanying drawings. The advantages and features of the present application will be more apparent according to the following description. It should be noted that the drawings are very simplified and use non-precise proportions, only to facilitate, clarify the purpose of assisting the description of the embodiments of the present application.

[0047] Embodiment One

[0048] Figure 1 A flowchart of the bridge blood vessel segmentation and extraction method provided for the present embodiment. As shown in Figure 1 The bridge blood vessel segmentation and extraction method provided by the present embodiment includes:

[0049] Step S100: acquiring coronary angiography image data;

[0050] Step S200: acquiring position information of an inlet and position information of an outlet of a bridge blood vessel based on the coronary angiography image data, and acquiring a segmentation result of the bridge blood vessel; and

[0051] Step S300: screening the segmentation result using the position information of the inlet and the position information of the outlet of the bridge blood vessel, to extract the bridge blood vessel.

[0052] Specifically, first, step S100 is performed, and angiography is performed on a postoperative patient to obtain coronary angiography image data containing a bridge blood vessel. In the present embodiment, the patient is subjected to Computed Tomography (CT) Angiography (CTA) examination, so as to obtain high-quality coronary angiography image data, and since CTA examination belongs to non-invasive examination, the safety of the patient is high.

[0053] As an optional embodiment, postoperative patients can also undergo examinations such as digital subtraction angiography (DSA) to obtain coronary angiography image data including bypass vessels; examples will not be given here.

[0054] In step S200, in this embodiment, the location information of the inlet and outlet of the bypass vessel, as well as the segmentation result of the bypass vessel, are obtained by training a neural network. Specifically, the coronary angiography image data is input into a trained first deep neural network to obtain the location information of the inlet and outlet of the bypass vessel; the coronary angiography image data is input into a trained second deep neural network to obtain the segmentation result of the bypass vessel.

[0055] First, a first deep neural network is trained. In this embodiment, the first deep neural network is an hourglass deep neural network, and the following explanation will take the training of the hourglass deep neural network as an example.

[0056] First, construct the initial hourglass deep neural network. Figure 2a This is a schematic diagram of the single-layer hourglass deep neural network model provided in this embodiment, as shown below. Figure 2a As shown, x11 represents the input image data, x12 represents the image data obtained after downsampling x11, x13 represents the image data obtained after upsampling x12, x11' represents the image data obtained after convolution, batch normalization, and ReLU activation of x11, y11 represents the image data obtained by adding x11' and x13, and y11 is the image data output by the single-layer hourglass deep neural network. Figure 2b This is a schematic diagram of the two-layer hourglass deep neural network model provided in this embodiment, as shown below. Figure 2bAs shown, x21 represents image data to be input, x22 represents image data obtained after down-sampling x21, x23 represents image data obtained after down-sampling x22, x24 represents image data obtained after up-sampling x23, x21' represents image data obtained after convolution, batch normalization and activation (ReLU) of x21, x22' represents image data obtained after convolution, batch normalization and activation (ReLU) of x22, y21 represents image data obtained after adding x22' and x24, y22 represents image data obtained after up-sampling y21, y23 represents image data obtained after adding x21' and y22, and y23 is image data output by the double-layer hourglass deep neural network. The initial hourglass deep neural network can be a single-layer hourglass deep neural network, a double-layer hourglass deep neural network or an hourglass deep neural network with other number of layers, and the present application does not make any limitation.

[0057] Next, samples for training the initial hourglass deep neural network are prepared. Specifically, a plurality of coronary angiography image data samples are obtained, and then the position information of the inlet and the position information of the outlet of the bridging vessel are marked in the coronary angiography image data samples. The marked coronary angiography image data samples are divided into a training set and a validation set, and the initial hourglass deep neural network is trained and validated, so as to obtain a trained hourglass deep neural network.

[0058] It should be noted that the coronary angiography image data samples used for training the initial hourglass deep neural network can include coronary angiography image data of various bridging vessels such as conventional venous bridging, sequential bridging, arterial bridging and Y-shaped bridging, so as to improve the accuracy of training. In addition, the bridging vessels in the coronary angiography image data samples used for training the initial hourglass deep neural network can be bridging vessels satisfying Gaussian field distribution, i.e. a blood vessel prototype with a highlighted center line and gradually darkening around the blood vessel.

[0059] Optionally, the position information of the inlet and the position information of the outlet of the bridge vessel marked in the coronary angiography image data sample can be the coordinate points of the inlet and the coordinate points of the outlet of the bridge vessel. Since the bridge vessel is a three-dimensional tubular structure, ideally, the coordinate points of the inlet and the coordinate points of the outlet of the bridge vessel should be the center points of the end face of the inlet and the end face of the outlet of the bridge vessel. In this embodiment, the position information of the inlet and the position information of the outlet of the bridge vessel marked in the coronary angiography image data sample are Gaussian kernels representing the position of the inlet and the position of the outlet of the bridge vessel. The Gaussian kernel is a sphere that satisfies Gaussian distribution, is brightest at the sphere center, and gradually darkens radially outward from the sphere center. The sphere center is the position of the inlet or the position of the outlet of the bridge vessel with the highest probability, and the probability gradually decreases radially outward from the sphere center. Compared with marking fixed coordinate points as the position information of the inlet and the position information of the outlet of the bridge vessel, in this embodiment, the entire internal region of the Gaussian kernel can be the position of the inlet and the position of the outlet of the bridge vessel, which is wider in range and reduces the requirement for labeling accuracy. Since the inlet and the outlet of the bridge vessel are not a point but a surface, using the Gaussian kernel representing the position of the inlet and the position of the outlet of the bridge vessel as the position information of the inlet and the position information of the outlet of the bridge vessel does not cause excessive adverse effects on the positioning of the inlet and the outlet of the bridge vessel.

[0060] As an optional embodiment, the first deep neural network can also be a V-net deep neural network or a U-net deep neural network or other deep neural networks capable of realizing point positioning, which will not be illustrated one by one here.

[0061] After the first deep neural network is trained, the coronary angiography image data of the postoperative patient is input into the trained first deep neural network, and the first deep neural network can automatically predict the position information of the inlet and the position information of the outlet of the bridge vessel.

[0062] It can be understood that, since in this embodiment, when training the first deep neural network, the Gaussian kernel representing the position of the inlet and the position of the outlet of the bridge vessel is marked in the coronary angiography image data sample, the position information of the inlet and the position information of the outlet of the bridge vessel obtained should also be the Gaussian kernel representing the position of the inlet and the position of the outlet of the bridge vessel.

[0063] Next, the second deep neural network is trained. In this embodiment, the second deep neural network is a V-net deep neural network, which will be illustrated below by taking training a V-net deep neural network as an example.

[0064] Firstly, an initial V-net deep neural network is constructed. The model of the initial V-net deep neural network is roughly in the shape of a V, one side of the V is a compression path, the size of the image is reduced through layer-by-layer down-sampling, the other side of the V is a decompression path, the size of the image is restored through layer-by-layer up-sampling, then the down-sampled image at the corresponding position is spliced to the up-sampled image, the fusion of global image information and local image information is realized, and finally the result is obtained through Softmax processing.

[0065] Next, samples for training the initial V-net deep neural network are prepared. Specifically, a plurality of coronary angiography image data samples are obtained, and then the area where the bridge vessel is located is outlined in the coronary angiography image data samples. The outlined coronary angiography image data samples are divided into a training set and a validation set, the initial V-net deep neural network is trained and validated, and a trained V-net deep neural network is obtained.

[0066] It should be noted that the coronary angiography image data samples used for training the initial V-net deep neural network can include coronary angiography image data of various bridge vessels such as conventional vein bridges, sequential bridges, arterial bridges, Y-shaped bridges, etc., thereby improving the accuracy of training. Moreover, the bridge vessels in the coronary angiography image data samples used for training the initial V-net deep neural network can be bridge vessels that meet the Gaussian field distribution, i.e., the blood vessel centerline is a highlight point, and the blood vessel around the highlight point gradually darkens.

[0067] As an optional embodiment, the second deep neural network can also be a U-net deep neural network or other deep neural networks capable of realizing image segmentation, which will not be described one by one here.

[0068] After the second deep neural network is trained, the coronary angiography image data of the postoperative patient is input into the trained second deep neural network, and the second deep neural network can automatically segment the bridge vessel to obtain the segmentation result of the bridge vessel.

[0069] It should be noted that the present application is not limited to obtaining the position information of the inlet and the position information of the outlet of the bridge vessel first and then obtaining the segmentation result of the bridge vessel, but can also obtain the segmentation result of the bridge vessel first and then obtain the position information of the inlet and the position information of the outlet of the bridge vessel, or can also obtain the position information of the inlet and the position information of the outlet of the bridge vessel and obtain the segmentation result of the bridge vessel simultaneously, which will not be described in detail here.

[0070] It can be understood that, since the bridge blood vessel is a blood vessel from other part of the human body, and the trend of the blood vessel and the flow direction of the blood may change during the blood vessel transplantation, and the bridge blood vessel may be carried by multiple positions, the bridge blood vessel is different from the ordinary blood vessel and has specificity and diversity, and the second deep neural network may not accurately segment the bridge blood vessel from the coronary angiography image data, that is, there may be only part of the bridge blood vessel in the segmentation result (the bridge blood vessel is partially missing, broken, etc.), or there may be other blood vessels that do not belong to the bridge blood vessel. Based on this, step S300 is performed, and the segmentation result is screened by using the position information of the inlet and the position information of the outlet of the bridge blood vessel.

[0071] Specifically, Figure 3 A schematic diagram of the segmentation result provided for the present embodiment is shown in FIG. 4. As shown in FIG. 4, there are four blood vessels in the segmentation result, which are blood vessels a, b, c, and d. Figure 3 First, the position information of the inlet and the position information of the outlet of the bridge blood vessel are superimposed on the segmentation result.

[0072] A schematic diagram of the segmentation result provided for the present embodiment is shown in FIG. 4. As shown in FIG. 4, there are four blood vessels in the segmentation result, which are blood vessels a, b, c, and d. Figure 4 A schematic diagram of the segmentation result provided for the present embodiment is shown in FIG. 4. As shown in FIG. 4, there are four blood vessels in the segmentation result, which are blood vessels a, b, c, and d. Figure 4 In the present embodiment, the position information of the inlet of the bridge blood vessel is a Gaussian kernel e representing the position of the inlet of the bridge blood vessel, and the position information of the outlet of the bridge blood vessel is a Gaussian kernel f representing the position of the outlet of the bridge blood vessel. Figure 4 As can be seen from FIG. 5, the Gaussian kernel e and the Gaussian kernel f coincide with the inlet and the outlet of the blood vessel a, respectively, while the inlets and outlets of the blood vessels b, c, and d do not coincide with the Gaussian kernel e and the Gaussian kernel f. Therefore, it can be directly determined that the blood vessel a belongs to the bridge blood vessel, and the blood vessel a is a determined blood vessel, while whether the blood vessels b, c, and d belong to the bridge blood vessel is unknown, and the blood vessels b, c, and d are blood vessels to be determined.

[0073] Next, it is determined whether the determined blood vessel between the inlet and the outlet of the bridge blood vessel is continuous. If the determined blood vessel between the inlet and the outlet of the bridge blood vessel is continuous, it indicates that the determined blood vessel between the inlet and the outlet of the bridge blood vessel is complete and has no missing or broken part, and the determined blood vessel can be directly output as the bridge blood vessel. Specifically, please continue to refer to FIG. 6. Figure 4 As can be seen from FIG. 6, it is determined whether the blood vessel a between the Gaussian kernel e and the Gaussian kernel f is continuous. Since the blood vessel a is a whole continuous blood vessel, it can be determined that the blood vessel a is the complete bridge blood vessel, and the blood vessels b, c, and d do not belong to the bridge blood vessel.

[0074] Finally, the blood vessel a is output as the bridge blood vessel, Figure 5A schematic diagram of the finally extracted bridge vessel is shown.

[0075] It should be noted that the bridge vessel has only one inlet, but the bridge vessel does not necessarily have only one outlet, and can have two or more outlets, so when determining whether the determined blood vessels between the inlet and the outlet of the bridge vessel are continuous, it is necessary to determine whether the determined blood vessels between the inlet and each outlet of the bridge vessel are continuous.

[0076] Figure 6 The schematic diagram of the bridge vessel extracted by the segmentation and extraction method of the bridge vessel provided in the embodiment in actual application is shown in FIG. 6. Figure 6 As can be seen from FIG. 6, since the advantages of the traditional segmentation algorithm and the deep learning algorithm are combined, the segmentation and extraction method of the bridge vessel provided in the embodiment can more accurately segment the bridge vessel, and is compatible with the specificity and diversity of the bridge vessel.

[0077] Embodiment Two

[0078] The difference from the first embodiment is that in the embodiment, the determined blood vessels between the inlet and the outlet of the bridge vessel are discontinuous, indicating that the bridge vessel has missing and broken situations. Figure 7 The schematic diagram of the segmentation result provided in the embodiment is shown in FIG. 7. Figure 7 As shown in FIG. 7, there are seven blood vessels in the segmentation result, which are blood vessels h, i, j, k, l, m and n.

[0079] First, the position information of the inlet and the position information of the outlet of the bridge vessel are superimposed on the segmentation result. Figure 8 The schematic diagram of the superimposed position information of the inlet and the position information of the outlet of the bridge vessel on the segmentation result provided in the embodiment is shown in FIG. 8. Figure 8 In FIG. 8, the position information of the inlet and the position information of the outlet of the bridge vessel are Gaussian kernel o representing the position of the inlet of the bridge vessel and Gaussian kernel p representing the position of the outlet of the bridge vessel, respectively. Figure 7 As can be seen from FIG. 8, the Gaussian kernel o and the Gaussian kernel p coincide with the inlet of the blood vessel h and the outlet of the blood vessel j respectively, and the inlets and outlets of the blood vessels i, k, l, m and n do not coincide with the Gaussian kernel o and the Gaussian kernel p. Therefore, it can be directly determined that the blood vessel h and the blood vessel j belong to the bridge vessel, and the blood vessel h and the blood vessel j are determined blood vessels, and at this time, whether the blood vessels i, k, l, m and n belong to the bridge vessel is unknown, and the blood vessels i, k, l, m and n are undetermined blood vessels.

[0080] Next, it is determined whether the determined blood vessel between the inlet and the outlet of the bridge blood vessel is continuous. If the determined blood vessel between the inlet and the outlet of the bridge blood vessel is continuous, it indicates that the determined blood vessel between the inlet and the outlet of the bridge blood vessel is complete and has no missing or broken part, and the determined blood vessel can be directly output as the bridge blood vessel. Specifically, please continue to refer to Figure 7 Since the blood vessel h and the blood vessel j are not a whole continuous blood vessel, it can be determined that the blood vessel h and the blood vessel j are not the complete bridge blood vessel.

[0081] Then, it is needed to determine whether the pending blood vessel belongs to the bridge blood vessel.

[0082] Specifically, the volume and / or length of each pending blood vessel can be obtained first. If the volume of the pending blood vessel is less than a first predetermined value and / or the length of the pending blood vessel is less than a second predetermined value, it indicates that the pending blood vessel is likely to be a noise point and has a smaller probability of belonging to the bridge blood vessel. The pending blood vessel with the volume less than the first predetermined value and / or the length less than the second predetermined value is removed, so as to simplify the subsequent calculation. In this embodiment, the volume and / or length of the blood vessels i, k, l, m and n are obtained. Since the volume of the blood vessel l is less than the first predetermined value and the length of the blood vessel l is less than the second predetermined value, it is considered that the blood vessel l is extremely likely to not belong to the bridge blood vessel, and the blood vessel l is removed from the pending blood vessels.

[0083] Further, the distance between each pending blood vessel and the aorta can be obtained first. If the distance between the pending blood vessel and the aorta is greater than a third predetermined value, it indicates that the distance between the pending blood vessel and the aorta is too far, and the pending blood vessel has a smaller probability of belonging to the bridge blood vessel. The pending blood vessel with the distance greater than the third predetermined value is removed, so as to simplify the subsequent calculation. In this embodiment, the distance between the blood vessels i, k, m and n and the aorta is obtained. Since the distance between the blood vessel n and the aorta is greater than the third predetermined value, it is considered that the blood vessel n is extremely likely to not belong to the bridge blood vessel, and the blood vessel n is removed from the pending blood vessels.

[0084] Of course, in other embodiments, the steps of removing the pending blood vessel with the volume less than the first predetermined value and / or the length less than the second predetermined value and removing the pending blood vessel with the distance greater than the third predetermined value from the aorta can be omitted.

[0085] Further, the bridge vessel can be an arterial bridge or a venous bridge, the maximum distance between the arterial bridge and the aorta is different from the maximum distance between the venous bridge and the aorta, the distance threshold between the bridge vessel and the aorta can be obtained based on the type of the bridge vessel, if the distance between the to-be-determined vessel and the aorta is greater than the distance threshold, it indicates that the probability of the to-be-determined vessel belonging to the bridge vessel is extremely small, and the to-be-determined vessel with a distance greater than the distance threshold from the aorta can be removed. Specifically, since the inlet of the venous bridge is at the ascending aorta and the inlet of the arterial bridge is at the aortic arch and its branches, the type of the bridge vessel can be determined according to the position of the determined vessel or the position information of the inlet of the bridge vessel. In this embodiment, whether the bridge vessel is an arterial bridge or a venous bridge can be determined according to the position of the inlet of the vessel a or the Gaussian kernel o, and after determining the type of the bridge vessel, the corresponding distance threshold can be obtained. Next, the distances between the vessels i, k and m and the aorta are obtained, and since the distances between the vessels i, k and m and the aorta are all less than or equal to the distance threshold, the vessels i, k and m are all retained in this step.

[0086] It should be noted that the distance threshold is different from the third predetermined value, and the third predetermined value can be greater than each distance threshold, thereby being used for initially excluding the vessels far away from the aorta.

[0087] Further, since the bridge vessel has a longitudinal trend relative to the heart, the trend of the to-be-determined vessel relative to the heart can be obtained based on the height difference between the inlet and the outlet of the to-be-determined vessel, if the trend of the to-be-determined vessel relative to the heart is a transverse trend, it indicates that the probability of the to-be-determined vessel belonging to the bridge vessel is extremely small, and the to-be-determined vessel with a transverse trend relative to the heart can be removed. In this embodiment, the height difference between the inlet and the outlet of the vessel i, k and m is calculated, and since the height difference between the inlet and the outlet of the vessel k is greater than a fourth predetermined value, it is considered that the vessel k has a transverse trend relative to the heart and has a smaller probability of belonging to the bridge vessel, and the vessel k is removed from the to-be-determined vessels.

[0088] Further, since the bridge vessel is closer to the left ventricle than to the right ventricle, the position of the to-be-determined vessel relative to the heart can be obtained based on the distances between the to-be-determined vessel and the left ventricle and the right ventricle, and the to-be-determined vessel closer to which ventricle can be determined according to the distances between the to-be-determined vessel and the left ventricle and the right ventricle, if the to-be-determined vessel is closer to the right ventricle, it indicates that the probability of the to-be-determined vessel belonging to the bridge vessel is extremely small, and the to-be-determined vessel closer to the right ventricle can be removed. In this embodiment, the distances between the vessels i and m and the left ventricle and the right ventricle are calculated, and since the vessel m is closer to the right ventricle, it is considered that the vessel m has a smaller probability of belonging to the bridge vessel, and the vessel m is removed from the to-be-determined vessels.

[0089] Next, it is determined whether the remaining candidate vessels belong to the bridge vessel. In this embodiment, the vessels i remaining are determined to belong to the bridge vessel.

[0090] It should be noted that in this embodiment, the steps of removing the candidate vessels with a distance from the aorta greater than the distance threshold, removing the candidate vessels with a transverse trend relative to the heart, and removing the candidate vessels closer to the right ventricle are performed sequentially, but it should be understood that these three steps can also be performed simultaneously or in a different order, or some steps can be omitted, which will not be described in detail here.

[0091] Further, since the determined vessels are not continuous with the remaining candidate vessels, the determined vessels need to be connected with the remaining candidate vessels. In this embodiment, the vessels h, i, and j have been determined to belong to the bridge vessel, while the remaining vessels do not belong to the bridge vessel. However, the vessels h, i, and j are not a complete continuous vessel, there is a missing segment between the vessel h and the vessel i, and there is a missing segment between the vessel i and the vessel j. Therefore, the vessel h is connected with the vessel i, and the vessel i is connected with the vessel j to form a complete vessel output. Figure 9 A schematic diagram of the final extracted bridge vessel is shown, Figure 9 The dashed line is a missing segment completed by the algorithm.

[0092] It should be noted that in this embodiment, after superimposing the position information of the inlet and the position information of the outlet of the bridge vessel on the segmentation result, there can be a case where part or all of the Gaussian kernels do not coincide with the inlets and outlets of any vessels. At this time, after connecting the determined vessels with the remaining candidate vessels, the Gaussian kernels can also be connected with adjacent candidate vessels to complete the bridge vessel.

[0093] Optionally, the region growing algorithm, threshold segmentation algorithm, or gradient segmentation algorithm can be used to connect the determined vessels with the remaining candidate vessels, but it should not be limited thereto.

[0094] It should be understood that the distance threshold, the first predetermined value, the second predetermined value, the third predetermined value, and the fourth predetermined value in this embodiment are empirical values, which can be obtained through actual tests, and will not be described one by one here.

[0095] Compared with embodiment one, this embodiment can accurately extract a complete bridge vessel even if the segmented bridge vessel has missing and / or broken segments, and can better cope with unclear contrast.

[0096] Embodiment three

[0097] The embodiment also provides a bridge vessel segmentation and extraction device. Figure 10 A structural block diagram of the bridge vessel segmentation and extraction device provided by the embodiment is shown in the figure, and the bridge vessel segmentation and extraction device comprises: Figure 10

[0098] A data supply module 10 is configured to acquire coronary angiography image data.

[0099] A position acquisition module 20 is configured to acquire position information of an inlet of the bridge vessel and position information of an outlet of the bridge vessel.

[0100] A segmentation module 30 is configured to acquire a segmentation result of the bridge vessel.

[0101] A screening module 40 is configured to screen the segmentation result by using the position information of the inlet of the bridge vessel and the position information of the outlet of the bridge vessel, so as to extract the bridge vessel.

[0102] The embodiment also provides an electronic device comprising a processor and a memory, wherein the memory stores instructions, and when the instructions are executed by the processor, the steps of the above bridge vessel segmentation and extraction method are implemented.

[0103] The processor can execute various actions and processes according to the instructions stored in the memory. Specifically, the processor can be an integrated circuit chip with signal processing capability. The processor can be a general processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a ready programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component. Various methods, steps and logic block diagrams disclosed in the embodiment of the present application can be implemented or executed. The general processor can be a microprocessor or the processor can be any conventional processor, etc., which can be an X86 architecture or an ARM architecture, etc.

[0104] ​The memory stores executable instructions that, when executed by the processor, implement the above-described method of segmenting and extracting bridge vessels. The memory can be volatile memory or non-volatile memory, or can include both volatile and non-volatile memory. The non-volatile memory can be read only memory (ROM), programmable read only memory (PROM), erasable programmable read only memory (EPROM), electrically erasable programmable read only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), used as external cache memory. By way of example, and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double-data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), Synchlink dynamic random access memory (SLDRAM), and direct

[0105] According to another aspect of embodiments of the present application, a non-transitory computer readable storage medium is provided, having stored thereon instructions that, when executed, can implement the steps of the above-described method of segmenting and extracting bridge vessels.

[0106] Similarly, the non-transitory computer readable storage medium in embodiments of the present application can be volatile memory or non-volatile memory, or can include both volatile and non-volatile memory. It should be noted that the computer readable storage medium described herein is intended to include, but not be limited to, these and any other suitable types of memory.

[0107] From the above description of the embodiments, it can be clear to those skilled in the art that the embodiments of the present application can be implemented by means of software and the necessary general purpose hardware, of course, but also by hardware alone, but in many cases the former is the better implementation. Based on such an understanding, the technical solutions of the present application, essentially or in the part that contributes to the prior art, can be embodied in the form of a software product, which can be stored in a non-transitory computer readable storage medium.

[0108] To sum up, in the bridge vessel segmentation extraction method, device, system and readable storage medium provided by the embodiment of the present application, the position information of the inlet and the position information of the outlet of the bridge vessel are obtained based on the coronary angiography image data, and the segmentation result of the bridge vessel is obtained; then the segmentation result is screened by using the position information of the inlet and the position information of the outlet of the bridge vessel, so as to extract the bridge vessel. The present application screens the segmentation result by using the position information of the inlet and the position information of the outlet of the bridge vessel, optimizes the segmentation result, improves the accuracy of the extracted bridge vessel, and further improves the accuracy of the prognosis evaluation of the bridge vessel. Moreover, the present application can be completed automatically without interaction with the user, and has higher segmentation efficiency and consistency.

[0109] It should be noted that the embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between embodiments can be referred to each other. For the system disclosed in the embodiments, the description is relatively simple because it corresponds to the method disclosed in the embodiments. The relevant parts can be referred to the description of the method.

[0110] It should also be noted that, although the present application has been disclosed as above with reference to the preferred embodiments, the above embodiments are not intended to limit the present application. Any skilled person in the art can make many possible changes and modifications to the technical solutions of the present application disclosed above, or modify equivalent embodiments with equivalent changes, without departing from the scope of the technical solutions of the present application. Therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present application, without departing from the content of the technical solutions of the present application, still belongs to the protection scope of the technical solutions of the present application.

[0111] It should also be understood that, unless specifically described or indicated, the terms "first", "second", "third" and the like in the specification are only used to distinguish different components, elements, steps and the like in the specification, and do not represent a logical relationship or sequence relationship between the components, elements, steps and the like.

[0112] It is also to be appreciated that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the scope of the present application. It must be noted that, as used herein, the articles "a", "an" and "the" are intended to include both singular and plural references unless the context clearly dictates otherwise. For example, the references "a step" or "an element" can mean one or more steps or elements, and the references "the first step" or "the first element" can mean one or more steps or elements. Also, the use of the term "about" is intended to encompass variations, for example, due to manufacturing or processing tolerances, or variations in the natural properties of a given material. In addition, the use of the term "or" is intended to encompass both exclusive and inclusive meanings of the term, unless the context clearly indicates otherwise. Furthermore, the embodiments of methods and / or apparatuses can be implemented in hardware, software, or a combination thereof.

Claims

1. A method of segmenting and extracting a bridge vessel, characterized by, The method comprises: acquiring coronary angiography image data; acquiring position information of an inlet and position information of an outlet of a bridge vessel based on the coronary angiography image data and acquiring a segmentation result of the bridge vessel; and superimposing the position information of the inlet and the position information of the outlet of the bridge vessel with the segmentation result, wherein the segmentation result comprises a plurality of blood vessels; comparing each of the blood vessels with the position information of the inlet and the position information of the outlet of the bridge vessel, and determining a blood vessel whose position of the inlet and / or position of the outlet coincides with the position of the inlet and / or position of the outlet of the bridge vessel as a determined blood vessel, judging that the determined blood vessel belongs to the bridge vessel, and judging whether the determined blood vessel is continuous between the inlet and the outlet of the bridge vessel, and when it is judged that the determined blood vessel is continuous, the determined blood vessel is taken as the bridge vessel.

2. The bridge blood vessel segmentation extraction method according to claim 1, wherein, The position information of the inlet and the position information of the outlet of the bridge vessel are coordinate points of the inlet and the outlet of the bridge vessel or Gaussian kernels representing the position of the inlet and the position of the outlet of the bridge vessel.

3. The bridge blood vessel segmentation extraction method according to claim 1, wherein, The blood vessel whose position of the inlet and the position of the outlet do not coincide with the position of the inlet and the position of the outlet of the bridge vessel is taken as a blood vessel to be determined, and when it is judged that the determined blood vessel is not continuous, the step of screening the segmentation result based on the position information of the inlet and the position information of the outlet of the bridge vessel further comprises: obtaining a distance threshold value between the bridge vessel and the aorta based on the type of the bridge vessel, removing the blood vessel to be determined whose distance to the aorta is greater than the distance threshold value; and / or, obtaining the trend of the blood vessel to be determined relative to the heart based on the height difference between the inlet and the outlet of the blood vessel to be determined, removing the blood vessel to be determined which has a transverse trend relative to the heart; and / or, obtaining the position of the blood vessel to be determined relative to the heart based on the distance between the blood vessel to be determined and the left ventricle and the right ventricle, removing the blood vessel to be determined which is closer to the right ventricle; judging that the remaining blood vessel to be determined belongs to the bridge vessel; and connecting the determined blood vessel with the remaining blood vessel to be determined to obtain the bridge vessel.

4. The bridge blood vessel segmentation extraction method according to claim 3, wherein, Before obtaining the distance threshold value between the bridge vessel and the aorta based on the type of the bridge vessel, the step of screening the segmentation result based on the position information of the inlet and the position information of the outlet of the bridge vessel further comprises: acquiring the volume and / or length of the blood vessel to be determined, removing the blood vessel to be determined whose volume is less than a first predetermined value and / or length is less than a second predetermined value; and / or, acquiring the distance between the blood vessel to be determined and the aorta, removing the blood vessel to be determined whose distance to the aorta is greater than a third predetermined value, the third predetermined value being greater than each of the distance threshold values.

5. The bridge blood vessel segmentation extraction method according to claim 1, wherein, inputting the coronary angiography image data into a trained first deep neural network to acquire the position information of the inlet and the position information of the outlet of the bridge vessel; and / or, inputting the coronary angiography image data into a trained second deep neural network to acquire the segmentation result of the bridge vessel.

6. The bridge blood vessel segmentation extraction method according to claim 5, wherein, The coronary angiography image data comprises CT coronary angiography image data or DSA coronary angiography image data; and / or, the first deep neural network comprises an hourglass deep neural network, a V-net deep neural network or a U-net deep neural network; and / or, the second deep neural network comprises a V-net deep neural network or a U-net deep neural network.

7. A bridge blood vessel segmentation extraction apparatus characterized by comprising: The method comprises: a data providing module configured to acquire coronary angiography image data; a position acquiring module configured to acquire position information of an inlet and position information of an outlet of a bridge vessel; a segmentation module configured to acquire a segmentation result of the bridge vessel; and a screening module configured to superimpose the position information of the inlet and the position information of the outlet of the bridge vessel with the segmentation result, wherein the segmentation result comprises a plurality of blood vessels; and compare each of the blood vessels with the position information of the inlet and the position information of the outlet of the bridge vessel, take the blood vessel whose position of the inlet and / or position of the outlet coincides with the position of the inlet and / or the position of the outlet of the bridge vessel as a determined blood vessel, determine that the determined blood vessel belongs to the bridge vessel, and determine whether the determined blood vessel between the inlet and the outlet of the bridge vessel is continuous, and when it is determined to be continuous, take the determined blood vessel as the bridge vessel.

8. A system for segmenting and extracting bridging blood vessels, characterized by, The method comprises a processor and a memory, and the memory stores instructions, when the instructions are executed by the processor, the method for segmenting and extracting a bridge vessel according to any one of claims 1 to 6 is realized.

9. A non-transitory computer-readable storage medium, comprising: The non-transitory computer readable storage medium stores instructions, when the instructions are executed by the processor, the method for segmenting and extracting a bridge vessel according to any one of claims 1 to 6 is realized.

Images