Training method, device and electronic equipment for naming model of coronary artery

Abstract

The present disclosure provides a training method, device and electronic device for a cardiac coronary vessel naming model, comprising: obtaining a three-dimensional coordinate set of vascular midline points of a first sample image; based on the three-dimensional coordinate set, confirming at least one of the bifurcation attribute, the number of the vascular segment to which it belongs, the vascular shape and the relative position in the heart of each vascular midline point of the first sample image; inputting the above-mentioned features of each vascular midline point of the first sample image into a feature fusion module included in the model, and determining that the output of the feature fusion module is the auxiliary feature corresponding to each vascular midline point; inputting the three-dimensional coordinate set and the auxiliary feature corresponding to each vascular midline point into a classification module included in the model, and obtaining the predicted name of the vascular segment corresponding to each vascular midline point in the first sample image; and adjusting the parameters of the model based on the labeled name of the vascular segment included in the first sample image and the predicted name of the vascular segment.

Description

Technical Field

[0001] The present disclosure relates to the field of image processing technology, and in particular to a method, device and electronic device for naming coronary vessels of the heart. Background Art

[0002] Accurate naming of coronary vessels is of great significance for the subsequent analysis and diagnosis of coronary artery diseases. There are two main types of current technologies. One is based on artificially designed rules, which sets the characteristic area data of known types of blood vessels in an artificially set space, associates the unknown type of test blood vessel data with the space, and compares it with the known data features to obtain the category of the test blood vessel. The other is a training method based on deep learning, which uses the blood vessel features extracted by convolutional neural networks and the midline point features extracted by point clouds, and combines the neural random grammar model to output the category of blood vessels through a deep learning model. However, the accuracy and rationality are low. Summary of the invention

[0003] The present disclosure provides a training method, device and electronic equipment for a cardiac coronary vessel naming model.

[0004] According to a first aspect of the present disclosure, a method for training a cardiac coronary vessel naming model is provided, comprising:

[0005] Acquire a three-dimensional coordinate set of a blood vessel midline point of a first sample image of a training set of coronary artery images;

[0006] Based on the three-dimensional coordinate set of the blood vessel midline points of the first sample image, confirm at least one of the bifurcation attribute, the number of the blood vessel segment to which it belongs, the blood vessel shape, and the relative position in the heart of each blood vessel midline point of the first sample image;

[0007] Inputting at least one of the bifurcation attribute, the number of the corresponding blood vessel, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point in the first sample image into a feature fusion module included in the coronary vessel naming model, and determining the output of the feature fusion module as the auxiliary feature corresponding to each blood vessel midline point;

[0008] Inputting the three-dimensional coordinate set of the blood vessel midline points of the first sample image and the auxiliary features corresponding to each blood vessel midline point into the classification module included in the coronary vessel naming model to obtain the predicted name of the blood vessel segment corresponding to each blood vessel midline point in the first sample image;

[0009] Based on the labeled names of the blood vessel segments included in the first sample image and the predicted names of the blood vessel segments, the parameters of the coronary vessel naming model are adjusted.

[0010] In the above solution, the step of obtaining a set of three-dimensional coordinates of the midline points of the blood vessels of the first sample image of the training set of coronary artery images includes:

[0011] Performing blood vessel segmentation on the first sample image to obtain a blood vessel segmentation result;

[0012] Corroding the width of each blood vessel in the blood vessel segmentation result into one pixel;

[0013] Determining at least one pixel corresponding to each of the blood vessels as a blood vessel midline point of the first sample image;

[0014] It is confirmed that the three-dimensional coordinate set corresponding to the at least one pixel is a three-dimensional coordinate set of a blood vessel midline point of the first sample image.

[0015] In the above scheme, the three-dimensional coordinate set of the vascular midline points of the first sample image is used to confirm at least one of the bifurcation attribute, the number of the vascular segment to which it belongs, the vascular shape, and the relative position in the heart of each vascular midline point of the first sample image, including:

[0016] Mapping a three-dimensional coordinate set of blood vessel midline points of the first sample image into a tree structure;

[0017] Confirming that a node including at least two child nodes in the tree structure is a bifurcation point, and that a bifurcation attribute of a blood vessel midline point corresponding to the bifurcation point is a bifurcation;

[0018] Alternatively, it is confirmed that a node including at most one child node in the tree structure is a non-bifurcation point, and the difference attribute of the blood vessel midline point corresponding to the non-bifurcation point is non-bifurcation.

[0019] In the above scheme, the three-dimensional coordinate set of the vascular midline points of the first sample image is used to confirm at least one of the bifurcation attribute, the number of the vascular segment to which it belongs, the vascular shape, and the relative position in the heart of each vascular midline point of the first sample image, including:

[0020] Confirm that the bifurcation attribute is bifurcation, and at least one of the two adjacent vascular midline points belongs to the same vascular segment;

[0021] The number of the blood vessel segment to which each blood vessel midline point of the first sample image belongs is confirmed.

[0022] In the above solution, after mapping the three-dimensional coordinate set of the blood vessel midline points of the first sample image into a tree structure, the method further includes:

[0023] The tree structure is smoothed.

[0024] In the above scheme, the three-dimensional coordinate set of the vascular midline points of the first sample image is used to confirm at least one of the bifurcation attribute, the number of the vascular segment to which it belongs, the vascular shape, and the relative position in the heart of each vascular midline point of the first sample image, including:

[0025] The tangent direction of each node in the tree structure after smoothing is determined to be the blood vessel shape of the blood vessel midline point corresponding to each node.

[0026] In the above scheme, the three-dimensional coordinate set of the vascular midline points of the first sample image is used to confirm at least one of the bifurcation attribute, the number of the vascular segment to which it belongs, the vascular shape, and the relative position in the heart of each vascular midline point of the first sample image, including:

[0027] Determine a minimum distance between each blood vessel midline point of the first sample image and the left atrium surface in the first sample image, and a direction of each blood vessel midline point relative to a first identification point on the left atrium surface corresponding to the minimum distance;

[0028] Determine a minimum distance between each blood vessel midline point of the first sample image and the left ventricle surface in the first sample image, and a direction of each blood vessel midline point relative to a second identification point on the left ventricle surface corresponding to the minimum distance;

[0029] Determine a minimum distance between each blood vessel midline point of the first sample image and the right atrium surface in the first sample image, and a direction of each blood vessel midline point relative to a third marking point on the right atrium surface corresponding to the minimum distance;

[0030] Determine the minimum distance between each blood vessel centerline point of the first sample image and the right ventricular surface in the first sample image, and the direction of each blood vessel centerline point relative to a fourth marking point on the right ventricular surface corresponding to the minimum distance.

[0031] In the above scheme, the step of inputting at least one of the bifurcation attribute, the number of the corresponding blood vessel, the shape of the blood vessel and the relative position in the heart of each blood vessel midline point of the first sample image into the feature fusion module included in the coronary vessel naming model, and determining the output of the feature fusion module as the auxiliary feature corresponding to each blood vessel midline point, comprises:

[0032] Inputting at least one of the bifurcation attribute, the number of the blood vessel, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point of the first sample image into the first linear layer included in the feature fusion module, and determining the weight score corresponding to at least one of the bifurcation attribute, the number of the blood vessel, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point;

[0033] Based on the weight scores corresponding to at least one of the bifurcation attribute of each vascular midline point, the number of the blood vessel to which it belongs, the shape of the blood vessel, and the relative position in the heart, weighted summation is performed on at least one of the bifurcation attribute of each vascular midline point, the number of the blood vessel to which it belongs, the shape of the blood vessel, and the relative position in the heart to obtain a weighted summation result;

[0034] The weighted sum result is determined as the auxiliary feature corresponding to each blood vessel midline point output by the feature fusion module.

[0035] In the above scheme, the method further includes:

[0036] Perform sample enhancement on the coronary artery image training set.

[0037] According to a second aspect of the present disclosure, a method for naming coronary vessels is provided, which uses the above-mentioned coronary vessel naming model and includes:

[0038] Acquire a three-dimensional coordinate set of the vascular midline points included in the coronary artery image to be named;

[0039] Based on the three-dimensional coordinate set of the vascular midline points included in the to-be-named cardiac coronary vessel image, confirm at least one of the bifurcation attribute, the number of the vascular segment to which it belongs, the vascular shape, and the relative position in the heart of each vascular midline point of the to-be-named cardiac coronary vessel image;

[0040] Inputting at least one of the bifurcation attribute, the number of the corresponding vessel, the shape of the vessel and the relative position in the heart of each vessel midline point of the coronary vessel image to be named into the feature fusion module included in the coronary vessel naming model, and determining the output of the feature fusion module as the auxiliary feature corresponding to each vessel midline point;

[0041] The three-dimensional coordinate set of the vascular midline points of the to-be-named cardiac coronary vessel image and the auxiliary features corresponding to each vascular midline point are input into the classification module included in the cardiac coronary vessel naming model to obtain the name of the vascular segment corresponding to each vascular midline point in the to-be-named cardiac coronary vessel image.

[0042] According to a third aspect of the present disclosure, a training device for a coronary vessel naming model is provided, comprising:

[0043] A first acquisition unit, used to acquire a three-dimensional coordinate set of a blood vessel midline point of a first sample image of a training set of coronary artery images;

[0044] A first confirmation unit is used to confirm at least one of a bifurcation attribute, a number of a corresponding vascular segment, a vascular shape, and a relative position in the heart of each vascular midline point of the first sample image based on a set of three-dimensional coordinates of the vascular midline points of the first sample image;

[0045] A first feature fusion unit is used to input at least one of the bifurcation attribute, the number of the corresponding blood vessel, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point of the first sample image into a feature fusion module included in the coronary vessel naming model, and determine that the output of the feature fusion module is the auxiliary feature corresponding to each blood vessel midline point;

[0046] a prediction unit, configured to input the three-dimensional coordinate set of the vascular midline points of the first sample image and the auxiliary features corresponding to each vascular midline point into the classification module included in the coronary vessel naming model, and obtain a predicted name of the vascular segment corresponding to each vascular midline point in the first sample image;

[0047] An adjusting unit is used to adjust the parameters of the coronary vessel naming model based on the labeled names of the vessel segments included in the first sample image and the predicted names of the vessel segments.

[0048] According to a fourth aspect of the present disclosure, a device for naming coronary vessels is provided, using the above-mentioned coronary vessel naming model, the device comprising:

[0049] A second acquisition unit is used to acquire a three-dimensional coordinate set of a blood vessel midline point included in the coronary blood vessel image to be named;

[0050] A second confirmation unit is used to confirm at least one of a bifurcation attribute, a number of a corresponding vascular segment, a vascular shape, and a relative position in the heart of each vascular midline point of the to-be-named cardiac coronary vessel image based on a set of three-dimensional coordinates of the vascular midline points included in the to-be-named cardiac coronary vessel image;

[0051] A second feature fusion unit is used to input at least one of the bifurcation attribute, the number of the corresponding blood vessel, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point of the coronary artery image to be named into the feature fusion module included in the coronary artery naming model, and determine the output of the feature fusion module as the auxiliary feature corresponding to each blood vessel midline point;

[0052] The naming unit is used to input the three-dimensional coordinate set of the vascular midline points of the to-be-named cardiac coronary vessel image and the auxiliary features corresponding to each vascular midline point into the classification module included in the cardiac coronary vessel naming model to obtain the name of the vascular segment corresponding to each vascular midline point in the to-be-named cardiac coronary vessel image.

[0053] According to a fifth aspect of the present disclosure, there is provided an electronic device, including:

[0054] at least one processor; and

[0055] a memory communicatively connected to the at least one processor; wherein,

[0056] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method described in the present disclosure.

[0057] According to a sixth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute the method described in the present disclosure.

[0058] The training method of the cardiac coronary vessel naming model disclosed in the present invention obtains a three-dimensional coordinate set of the vascular midline points of the first sample image of the cardiac coronary vessel image training set; based on the three-dimensional coordinate set of the vascular midline points of the first sample image, confirms at least one of the bifurcation attribute, the number of the corresponding vascular segment, the vascular shape, and the relative position in the heart of each vascular midline point of the first sample image; inputs at least one of the bifurcation attribute, the number of the corresponding vascular segment, the vascular shape, and the relative position in the heart of each vascular midline point of the first sample image into a feature fusion module included in the cardiac coronary vessel naming model, and determines that the output of the feature fusion module is the auxiliary feature corresponding to each vascular midline point; the three-dimensional coordinate set of the vascular midline points of the first sample image is used as the auxiliary feature set of the vascular midline points; The auxiliary features corresponding to each vascular midline point are input into the classification module included in the cardiac coronary vessel naming model to obtain the predicted name of the vascular segment corresponding to each vascular midline point in the first sample image; based on the labeled name of the vascular segment included in the first sample image and the predicted name of the vascular segment, the parameters of the cardiac coronary vessel naming model are adjusted, and the bifurcation attribute of each vascular midline point, the number of the vascular segment to which it belongs, the vascular shape and the relative position in the heart are used as the input parameters of the cardiac coronary vessel naming model, the essential feature of vascular naming, the cardiac anatomical structure, can be fully utilized to train the cardiac coronary vessel naming model, so that the subsequent use of the cardiac coronary vessel naming model for cardiac coronary vessel naming can be more accurate and reasonable.

[0059] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] The above and other objects, features and advantages of the exemplary embodiments of the present disclosure will become readily understood by reading the detailed description below with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present disclosure are shown in an exemplary and non-limiting manner, in which:

[0061] In the drawings, the same or corresponding reference numerals represent the same or corresponding parts.

[0062] Figure 1 An optional flow chart of a method for training a cardiac coronary vessel naming model provided by an embodiment of the present disclosure is shown;

[0063] Figure 2 An optional flow chart of a method for naming coronary vessels of the present disclosure is shown;

[0064] Figure 3 Another optional flow chart of the training method of the coronary vessel naming model provided by the embodiment of the present disclosure is shown;

[0065] Figure 4 A flowchart of a method for training a cardiac coronary vessel naming model provided by an embodiment of the present disclosure is shown;

[0066] Figure 5 A schematic diagram of a bifurcation point and a blood vessel segment provided by an embodiment of the present disclosure is shown;

[0067] Figure 6 An optional structural schematic diagram of a training device for a cardiac coronary vessel naming model provided by an embodiment of the present disclosure is shown;

[0068] Figure 7 An optional structural schematic diagram of a cardiac coronary vessel naming device provided by an embodiment of the present disclosure is shown;

[0069] Figure 8 A schematic diagram of the structure of an electronic device according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0070] In order to make the purpose, features, and advantages of the present disclosure more obvious and easy to understand, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present disclosure.

[0071] In the related technologies, the naming technology of coronary arteries is mainly based on two methods. One is to manually design rules, but this method does not generalize well enough, there are large differences between images, and the rules are not robust enough to changes in images; the other is a deep learning method based on point cloud and neural random grammar model, but this method is relatively complex. And most importantly, neither of the above two methods utilizes the essential characteristics of vascular naming. In practice, the main basis for doctors to name blood vessels is the blood supply area of ​​the blood vessels. For example, the left anterior descending branch LAD mainly supplies part of the left ventricle, the anterior wall of the right ventricle and the ventricular septum, and the obtuse marginal branch OM mainly supplies the lateral part of the left ventricle. None of the existing methods utilize this essential feature of vascular naming, so they cannot achieve very ideal results.

[0072] The present disclosure provides a training method for a cardiac coronary vessel naming model, which can take into account the essential characteristics of vessel naming, use the point cloud data of the vessel to represent the relationship between the vessel and the cardiac anatomical structure, and further obtain the prediction result of the vessel naming based on the neural network, so that the subsequent use of the cardiac coronary vessel naming model to name the cardiac coronary vessels can be more accurate and reasonable.

[0073] Figure 1 An optional flow chart of a method for training a cardiac coronary vessel naming model provided by an embodiment of the present disclosure is shown, and will be described according to each step.

[0074] Step S101, obtaining a set of three-dimensional coordinates of blood vessel midline points of a first sample image of a training set of coronary artery images.

[0075] In some embodiments, a training device for a cardiac coronary vessel naming model (hereinafter referred to as the first device) performs vessel segmentation on the first sample image to obtain a vessel segmentation result; corrodes the width of each vessel in the vessel segmentation result into one pixel; determines that at least one pixel corresponding to each vessel is a vessel midline point of the first sample image; and confirms that the three-dimensional coordinate set corresponding to the at least one pixel is a three-dimensional coordinate set of the vessel midline point of the first sample image. The cardiac coronary vessel image training set includes at least one cardiac coronary artery CTA image, and the first sample image is any sample image in the cardiac coronary vessel image training set.

[0076] In some embodiments, the first device may also perform sample enhancement on the coronary artery image training set. Optionally, the first device may perform sample enhancement on the coronary artery image training set by at least one of random rotation, random flipping and random translation.

[0077] Step S102, based on the three-dimensional coordinate set of the vascular midline points of the first sample image, confirm at least one of the bifurcation attribute, the number of the vascular segment to which it belongs, the vascular shape, and the relative position in the heart of each vascular midline point of the first sample image.

[0078] In some embodiments, the first device maps the three-dimensional coordinate set of the blood vessel midline points of the first sample image into a tree structure; confirms that a node in the tree structure that includes at least two child nodes is a bifurcation point, and the bifurcation attribute of the blood vessel midline point corresponding to the bifurcation point is bifurcation; or, confirms that a node in the tree structure that includes at most one child node is a non-bifurcation point, and the difference attribute of the blood vessel midline point corresponding to the non-bifurcation point is non-bifurcation.

[0079] In specific implementation, the first device can construct a tree structure based on the minimum spanning tree algorithm, and define the distance between two nodes in the tree structure based on the Euclidean distance between two adjacent vascular centerline points. In the tree structure, each node corresponds to a vascular centerline point; each node has its corresponding parent node and child node; if a node has more than one (at least two) child nodes, the node is a bifurcation point, and the bifurcation attribute of the corresponding vascular centerline point is bifurcation; if a node has less than one (at most 1, i.e., 1 or 0) child nodes, the node is a non-bifurcation point, and the bifurcation attribute of the corresponding vascular centerline point is non-bifurcation. Optionally, 1 can be used to represent the bifurcation attribute of the vascular centerline point as bifurcation, and 0 can be used to represent the bifurcation attribute of the vascular centerline point as non-bifurcation.

[0080] In some embodiments, the first device confirms that the bifurcation attribute is bifurcation, and at least one vascular midline point between two adjacent vascular midline points belongs to the same vascular segment; confirms the number of the vascular segment to which each vascular midline point of the first sample image belongs. The two adjacent vascular midline points refer to two adjacent bifurcation points. Further, the bifurcation attribute is bifurcation, and at least one vascular midline point between two adjacent vascular midline points belongs to the same vascular segment, which means that at least one vascular midline point between two adjacent bifurcation points belongs to the same vascular segment; the numbering of the vascular segment can be determined in any way, such as randomly numbering the vascular segment, numbering the vascular segment according to a certain strategy, etc. The specific numbering of the vascular segment does not affect the final naming result, and its more important significance lies in confirming the vascular midline points belonging to the same vascular segment.

[0081] In some embodiments, after the first device maps the three-dimensional coordinate set of the blood vessel midline points of the first sample image into a tree structure, it can also perform smoothing processing on the tree structure.

[0082] In specific implementation, the first device can perform smoothing on the tree structure based on the Bezier curve, and in the calculation process of the Bezier curve, the tangent direction of each vascular centerline point can be obtained. Since the erosion algorithm is used when obtaining the vascular centerline point (or the vascular centerline composed of multiple vascular centerline points), the obtained vascular centerline is not perfect, and there are some outliers (offset vascular centerline points) that deviate from the vascular center. In this way, smoothing the tree structure can correct the outliers.

[0083] In some embodiments, the first device determines that the tangent direction of each node in the tree structure after smoothing is the blood vessel shape of the blood vessel midline point corresponding to each node.

[0084] In some embodiments, the first device determines the minimum distance between each vascular centerline point of the first sample image and the left atrial surface in the first sample image, and the direction of each vascular centerline point relative to a first identification point of the left atrial surface corresponding to the minimum distance; determines the minimum distance between each vascular centerline point of the first sample image and the left ventricular surface in the first sample image, and the direction of each vascular centerline point relative to a second identification point of the left ventricular surface corresponding to the minimum distance; determines the minimum distance between each vascular centerline point of the first sample image and the right atrial surface in the first sample image, and the direction of each vascular centerline point relative to a third identification point of the right atrial surface corresponding to the minimum distance; determines the minimum distance between each vascular centerline point of the first sample image and the right ventricular surface in the first sample image, and the direction of each vascular centerline point relative to a fourth identification point of the right ventricular surface corresponding to the minimum distance.

[0085] In a specific implementation, the first device determines the shortest distance between each blood vessel midline point of the first sample image and the surface of the cardiac structure (left atrium, left ventricle, right atrium and right ventricle), and the direction of each blood vessel midline point relative to the identification point on the surface of the cardiac structure corresponding to the minimum distance. The direction can be determined based on the three-dimensional coordinates of the identification point and the three-dimensional coordinates of the blood vessel midline point. Assume that the three-dimensional coordinates of the surface identification point of the j-th cardiac structure (j is an integer from 1 to 4, representing the cardiac structure) from the i-th blood vessel midline point is (a ij ,b ij ,c ij ), the three-dimensional coordinates of the i-th blood vessel midline point are Then the direction β of the i-th blood vessel midline point relative to the marking point is ij It can be determined by the following formula:

[0086]

[0087] Step S103, inputting at least one of the bifurcation attribute, the number of the corresponding blood vessel, the shape of the blood vessel and the relative position in the heart of each blood vessel midline point of the first sample image into a feature fusion module included in the coronary vessel naming model, and determining that the output of the feature fusion module is the auxiliary feature corresponding to each blood vessel midline point.

[0088] In some embodiments, the first device inputs at least one of the bifurcation attributes, the number of the corresponding blood vessel, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point of the first sample image into a feature fusion module included in the coronary vessel naming model, and determines that the output of the feature fusion module is the auxiliary feature corresponding to each blood vessel midline point.

[0089] In a specific implementation, the first device can input at least one of the bifurcation attribute, the number of the blood vessel to which it belongs, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point of the first sample image into the first linear layer included in the feature fusion module, and determine the weight score corresponding to at least one of the bifurcation attribute, the number of the blood vessel to which it belongs, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point; normalize the weight score corresponding to at least one of the bifurcation attribute, the number of the blood vessel to which it belongs, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point, and determine the weight score corresponding to at least one of the bifurcation attribute, the number of the blood vessel to which it belongs, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point. The method comprises the steps of: obtaining a weight coefficient corresponding to at least one of the bifurcation attribute of each vascular midline point, the number of the blood vessel to which it belongs, the shape of the blood vessel, and the relative position in the heart; performing weighted summation on at least one of the bifurcation attribute, the number of the blood vessel to which it belongs, the shape of the blood vessel, and the relative position in the heart of each vascular midline point based on the weight coefficient corresponding to at least one of the bifurcation attribute, the number of the blood vessel to which it belongs, the shape of the blood vessel, and the relative position in the heart of each vascular midline point to obtain a weighted summation result; and determining the weighted summation result as the auxiliary feature corresponding to each vascular midline point output by the feature fusion module.

[0090] Step S104, inputting the three-dimensional coordinate set of the vascular midline points of the first sample image and the auxiliary features corresponding to each vascular midline point into the classification module included in the coronary vessel naming model to obtain the predicted name of the vascular segment corresponding to each vascular midline point in the first sample image.

[0091] In some embodiments, the first device inputs the three-dimensional coordinate set of the vascular midline points of the first sample image and the auxiliary features corresponding to each vascular midline point into the classification module included in the coronary vessel naming model to obtain the predicted name of the vascular segment corresponding to each vascular midline point in the first sample image.

[0092] Step S105 : adjusting the parameters of the coronary vessel naming model based on the annotated names of the vessel segments included in the first sample image and the predicted names of the vessel segments.

[0093] In some embodiments, the first device adjusts parameters of the coronary vessel naming model based on the labeled names of the vessel segments included in the first sample image and the predicted names of the vessel segments.

[0094] In a specific implementation, the first device can determine a training loss function based on a cross entropy loss function, the labeled name of the vascular segment, and the predicted name of the vascular segment, and adjust the parameters of the coronary vessel naming model based on the training loss function.

[0095] In this way, through the training method of the cardiac coronary vessel naming model provided by the embodiment of the present disclosure, a three-dimensional coordinate set of the vascular midline points of the first sample image of the cardiac coronary vessel image training set is obtained; based on the three-dimensional coordinate set of the vascular midline points of the first sample image, at least one of the bifurcation attributes, the number of the corresponding vascular segment, the vascular shape and the relative position in the heart of each vascular midline point of the first sample image is confirmed; at least one of the bifurcation attributes, the number of the corresponding vascular segment, the vascular shape and the relative position in the heart of each vascular midline point of the first sample image is input into the feature fusion module included in the cardiac coronary vessel naming model, and the output of the feature fusion module is determined to be the auxiliary feature corresponding to each vascular midline point; the vascular midline points of the first sample image are input into the feature fusion module included in the cardiac coronary vessel naming model, and the output of the feature fusion module is determined to be the auxiliary feature corresponding to each vascular midline point. The three-dimensional coordinate set and the auxiliary features corresponding to each vascular midline point are input into the classification module included in the cardiac coronary vessel naming model to obtain the predicted name of the vascular segment corresponding to each vascular midline point in the first sample image; based on the labeled names of the vascular segments included in the first sample image and the predicted names of the vascular segments, the parameters of the cardiac coronary vessel naming model are adjusted, and the bifurcation attributes of each vascular midline point, the number of the vascular segment to which it belongs, the vascular shape and the relative position in the heart are used as input parameters of the cardiac coronary vessel naming model, so that the essential feature of vascular naming, which is the anatomical structure of the heart, can be fully utilized to train the cardiac coronary vessel naming model, so that the subsequent use of the cardiac coronary vessel naming model for cardiac coronary vessel naming can be more accurate and reasonable.

[0096] Figure 2An optional flow chart of the method for naming coronary vessels of the heart provided in an embodiment of the present disclosure is shown, and will be explained according to each step.

[0097] Step S201, obtaining a set of three-dimensional coordinates of the midline points of the blood vessels of the coronary artery image to be named.

[0098] In some embodiments, a cardiac coronary vessel naming device (hereinafter referred to as the second device) performs vessel segmentation on the cardiac coronary vessel image to be named to obtain a vessel segmentation result; corrodes the width of each vessel in the vessel segmentation result into one pixel; determines that at least one pixel corresponding to each vessel is a vessel midline point of the cardiac coronary vessel image to be named; and confirms that the three-dimensional coordinate set corresponding to the at least one pixel is the three-dimensional coordinate set of the vessel midline point of the cardiac coronary vessel image to be named.

[0099] Step S202, based on the three-dimensional coordinate set of the vascular midline points of the to-be-named cardiac coronary vascular image, confirm at least one of the bifurcation attribute, the number of the vascular segment to which it belongs, the vascular shape, and the relative position in the heart of each vascular midline point of the to-be-named cardiac coronary vascular image.

[0100] In some embodiments, the second device maps the three-dimensional coordinate set of the vascular midline points of the to-be-named coronary artery image into a tree structure; confirms that a node in the tree structure that includes at least two child nodes is a bifurcation point, and the bifurcation attribute of the vascular midline point corresponding to the bifurcation point is bifurcation; or, confirms that a node in the tree structure that includes at most one child node is a non-bifurcation point, and the difference attribute of the vascular midline point corresponding to the non-bifurcation point is non-bifurcation.

[0101] In specific implementation, the second device can construct a tree structure based on the minimum spanning tree algorithm, and define the distance between two nodes in the tree structure based on the Euclidean distance between two adjacent vascular centerline points. In the tree structure, each node corresponds to a vascular centerline point; each node has its corresponding parent node and child node; if a node has more than one (at least two) child nodes, the node is a bifurcation point, and the bifurcation attribute of the corresponding vascular centerline point is bifurcation; if a node has less than one (at most 1, i.e., 1 or 0) child nodes, the node is a non-bifurcation point, and the bifurcation attribute of the corresponding vascular centerline point is non-bifurcation. Optionally, 1 can be used to represent the bifurcation attribute of the vascular centerline point as bifurcation, and 0 can be used to represent the bifurcation attribute of the vascular centerline point as non-bifurcation.

[0102] In some embodiments, the second device confirms that the bifurcation attribute is bifurcation, and at least one vascular midline point between two adjacent vascular midline points belongs to the same vascular segment; confirms the number of the vascular segment to which each vascular midline point of the to-be-named coronary vascular image belongs. Wherein, the two adjacent vascular midline points refer to two adjacent bifurcation points, and further, the bifurcation attribute is bifurcation, and at least one vascular midline point between two adjacent vascular midline points belongs to the same vascular segment, which means that at least one vascular midline point between two adjacent bifurcation points belongs to the same vascular segment.

[0103] In some embodiments, after the second device maps the three-dimensional coordinate set of the vascular midline points of the to-be-named coronary vessel image into a tree structure, it may also perform smoothing processing on the tree structure.

[0104] In specific implementation, the second device can perform smoothing on the tree structure based on the Bezier curve, and in the calculation process of the Bezier curve, the tangent direction of each vascular centerline point can be obtained. Since the erosion algorithm is used when obtaining the vascular centerline point (or the vascular centerline composed of multiple vascular centerline points), the obtained vascular centerline is not perfect, and there are some outliers (offset vascular centerline points) that deviate from the vascular center. In this way, smoothing the tree structure can correct the outliers.

[0105] In some embodiments, the second device determines that the tangent direction of each node in the tree structure after smoothing is the blood vessel shape of the blood vessel midline point corresponding to each node.

[0106] In some embodiments, the second device determines the minimum distance between each vascular centerline point of the to-be-named cardiac coronary vessel image and the left atrium surface in the to-be-named cardiac coronary vessel image, and the direction of each vascular centerline point relative to a first identification point on the left atrium surface corresponding to the minimum distance; determines the minimum distance between each vascular centerline point of the to-be-named cardiac coronary vessel image and the left ventricle surface in the to-be-named cardiac coronary vessel image, and the direction of each vascular centerline point relative to a second identification point on the left ventricle surface corresponding to the minimum distance; determines the minimum distance between each vascular centerline point of the to-be-named cardiac coronary vessel image and the right atrium surface in the to-be-named cardiac coronary vessel image, and the direction of each vascular centerline point relative to a third identification point on the right atrium surface corresponding to the minimum distance; determines the minimum distance between each vascular centerline point of the to-be-named cardiac coronary vessel image and the right ventricle surface in the to-be-named cardiac coronary vessel image, and the direction of each vascular centerline point relative to a fourth identification point on the right ventricle surface corresponding to the minimum distance.

[0107] Step S203, inputting at least one of the bifurcation attribute, the number of the corresponding blood vessel, the shape of the blood vessel and the relative position in the heart of each blood vessel midline point of the coronary artery image to be named into the feature fusion module included in the coronary artery naming model, and determining the output of the feature fusion module as the auxiliary feature corresponding to each blood vessel midline point.

[0108] In some embodiments, the second device inputs at least one of the bifurcation attributes, the number of the corresponding blood vessel, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point of the coronary artery image to be named into a feature fusion module included in the coronary artery naming model, and determines that the output of the feature fusion module is the auxiliary feature corresponding to each blood vessel midline point.

[0109] In a specific implementation, the second device can input at least one of the bifurcation attribute, the number of the blood vessel to which it belongs, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point of the to-be-named heart coronary vessel image into the first linear layer included in the feature fusion module, determine the weight score corresponding to at least one of the bifurcation attribute, the number of the blood vessel to which it belongs, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point; normalize the weight score corresponding to at least one of the bifurcation attribute, the number of the blood vessel to which it belongs, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point, and determine Determine the weight coefficient corresponding to at least one of the bifurcation attribute, the number of the blood vessel, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point; based on the weight coefficient corresponding to at least one of the bifurcation attribute, the number of the blood vessel, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point, perform weighted summation on at least one of the bifurcation attribute, the number of the blood vessel, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point to obtain a weighted summation result; determine the weighted summation result as the auxiliary feature corresponding to each blood vessel midline point output by the feature fusion module.

[0110] Step S204, inputting the three-dimensional coordinate set of the vascular midline points of the to-be-named cardiac coronary vessel image and the auxiliary features corresponding to each vascular midline point into the classification module included in the cardiac coronary vessel naming model, and obtaining the name of the vascular segment corresponding to each vascular midline point in the to-be-named cardiac coronary vessel image.

[0111] In some embodiments, the second device inputs the three-dimensional coordinate set of the vascular midline points of the cardiac coronary vessel image to be named and the auxiliary features corresponding to each vascular midline point into the classification module included in the cardiac coronary vessel naming model to obtain the name of the vascular segment corresponding to each vascular midline point in the cardiac coronary vessel image to be named.

[0112] In this way, through the cardiac coronary vessel naming method provided by the embodiment of the present disclosure, the bifurcation attributes of each vascular midline point, the number of the vascular segment to which it belongs, the vascular shape and the relative position in the heart are used as input parameters of the cardiac coronary vessel naming model, and the essential feature of vascular naming, which is the anatomical structure of the heart, can be fully utilized for vascular naming, so that the subsequent use of the cardiac coronary vessel naming model for vascular naming can be more accurate and reasonable.

[0113] Figure 3 Another optional flow chart of the training method of the coronary vessel naming model provided by the embodiment of the present disclosure is shown; Figure 4 A flow chart of a method for training a cardiac coronary vessel naming model provided by an embodiment of the present disclosure is shown, and will be described according to each step.

[0114] Step S301, obtaining a training set of coronary artery images.

[0115] In some optional embodiments, the first device may collect a first threshold number of coronary artery CTA images, and randomly divide the first threshold number of coronary artery CTA images into a training set, a validation set, and a test set according to a first ratio.

[0116] The training set is used to train the cardiac coronary vessel naming model, the validation set is used to select the best cardiac coronary vessel naming model, and the test set is used to evaluate the final effect of the cardiac coronary vessel naming model.

[0117] Optionally, the first threshold value can be set according to actual conditions or experimental results, such as 2000; the first ratio can be set according to actual conditions or experimental results, such as 8:1:1. Those skilled in the art should understand that the specific values ​​of the first threshold value and the first ratio are only examples and are not intended to limit the present disclosure.

[0118] In some embodiments, the first device may also perform sample enhancement on the coronary artery image training set. Optionally, the first device may perform sample enhancement on the coronary artery image training set by at least one of random rotation, random flipping and random translation.

[0119] Step S302: extracting blood vessel structure.

[0120] In some embodiments, the first device may use a morphological erosion operation to erode the width of each blood vessel in the blood vessel segmentation result to 1 pixel based on the blood vessel segmentation result of the first sample image, and obtain a point set consisting of points on the center line of the blood vessel (i.e., a three-dimensional coordinate set of the blood vessel centerline points), and the three-dimensional coordinate set may be recorded as where xi ,y i ,z i They represent the coordinates of the i-th vascular midline point in the x, y, and z dimensions, respectively, and N represents the number of vascular midline points.

[0121] Step S303: anatomical structure representation.

[0122] In some embodiments, the first device transforms the three-dimensional coordinate set of the blood vessel midline points of the first sample image into a minimum spanning tree algorithm. Mapping is a tree structure. The distance between two nodes in the tree structure is defined based on the Euclidean distance between two adjacent vascular centerline points. In the tree structure, each node corresponds to a vascular centerline point; each node has its corresponding parent node and child node; if a node has more than one (at least two) child nodes, the node is a bifurcation point, and the bifurcation attribute of the corresponding vascular centerline point is bifurcation; if a node has less than one (at most 1, i.e., 1 or 0) child nodes, the node is a non-bifurcation point, and the bifurcation attribute of the corresponding vascular centerline point is non-bifurcation. Optionally, 1 can be used to represent the bifurcation attribute of the vascular centerline point as bifurcation, and 0 can be used to represent the bifurcation attribute of the vascular centerline point as non-bifurcation.

[0123] The three-dimensional coordinate set of the blood vessel midline point can be divided into multiple blood vessel segments according to the bifurcation point. Figure 5 A schematic diagram of a bifurcation point and a blood vessel segment provided by an embodiment of the present disclosure is shown. Figure 5 The bifurcation point shown has two child nodes, which are the intersections of vessel segments 1, 2, and 3. Optionally, you can use Characterizes the bifurcation properties of each midline point, where g i =1 means that the i-th blood vessel midline point is a bifurcation point, g i = 0 means that the i-th vessel midline point is a non-bifurcation point; optionally, Used to indicate the number (serial number) of the vascular segment where each vascular midline point is located, k i The value of is a positive integer.

[0124] In the logic of vascular naming, vascular midline points on the same vascular segment are generally of the same category (same name), so the vascular segment where the midline point is located and whether the midline point is a bifurcation point both provide important features for vascular naming.

[0125] In some embodiments, after constructing the tree structure, the first device uses a Bezier curve to smooth each blood vessel segment. There are two main purposes for smoothing the blood vessel segments. One is to correct the position of individual outliers. The blood vessel centerline (composed of multiple blood vessel centerline points) obtained by the corrosion algorithm is not perfect. There are some outliers that deviate from the center of the blood vessel. Therefore, it is necessary to smooth according to the overall structure of the blood vessel to correct the outliers; the second is to obtain the tangent direction of each point. The tangent direction of each point can be obtained in the calculation of the Bezier curve. This direction is one of the important features of subsequent modeling. The three-dimensional coordinate set of the smoothed blood vessel centerline can be expressed as The tangent direction of each blood vessel midline point can be expressed as The bifurcation point information is expressed as The vessel segment information is represented as M represents the number of midline points after smoothing.

[0126] In clinical practice, the cardiac anatomical structures based on which blood vessels are named are mainly the left atrium, left ventricle, right atrium and right ventricle. In order to effectively use this information, the first device calculates the minimum value of each smoothed blood vessel midline point from the surface of each atrium (or ventricle), and the three-dimensional coordinates of the identification points on the cardiac structure can be expressed by the following formula:

[0127]

[0128] Among them, the values ​​of j are 1, 2, 3, and 4, which represent the cardiac structures, namely the left atrium, left ventricle, right atrium, and right ventricle, respectively. (a ij ,b ij ,c ij ) represents the 3D coordinates of the point in the jth cardiac structure surface point set that is closest to the i-th vascular midline point. From this, the i-th midline point can be calculated with respect to (a ij ,b ij ,c ij ) and the angle (relative direction), the distance is expressed as The angle is expressed as Thus, eleven features of each vascular midline point can be obtained, which are the distance and angle between the vascular midline point and the four cardiac structures, the tangent direction of the vascular midline point (vascular shape), whether the point is a bifurcation point (bifurcation attribute), and the vascular segment number where the vascular midline point is located (the number of the vascular segment to which it belongs).

[0129] These eleven features are an important supplement to the vascular position information, and indirectly reflect the blood supply area of ​​the blood vessel. Specifically, the tangent direction of the vascular midline point indicates the shape of the blood vessel near the vascular midline point, and together with the coordinates of the vascular midline point, reflects the overall trend of the blood vessel. Each specific type of blood vessel has its relatively fixed shape and trend, so the tangent direction of the midline point can promote the naming of the blood vessel. The distance and angle (relative position) of the vascular midline point from each atrium or ventricle jointly reflect the relative position of the vascular midline point in the heart, and the relative position with the atria (left atrium and right atrium) and ventricles (left ventricle and right ventricle), so it indirectly reflects the blood supply area of ​​the blood vessel where the vascular midline point is located, and can therefore also play an auxiliary judgment role in the naming of the blood vessels.

[0130] Therefore, the above eleven features are used as auxiliary features of each vascular midline point, and together with the three-dimensional coordinates of the midline point, they are used as the input of the coronary vessel naming model to predict the vascular naming.

[0131] Step S304: neural network prediction.

[0132] In some embodiments, the coronary vessel naming model can be a neural network. The input of the coronary vessel naming model is the smoothed three-dimensional coordinates of the vessel centerline points and the eleven auxiliary features corresponding to each vessel centerline point calculated. Therefore, the input dimension of each vessel centerline point is twelve dimensions, which contains rich vessel prior information. Optionally, a PointNet++ network is used for feature extraction and final vessel classification.

[0133] It should be noted that in the feature extraction process, an adaptive feature fusion module is added to allow the coronary artery naming model to focus on more important features among the eleven auxiliary features. Specifically, the auxiliary features extracted by the coronary artery naming model are F i ∈R C , respectively represent the eleven auxiliary features defined, and C is the number of channels. These features pass through a linear layer (the first linear layer) to obtain the corresponding weight score w i =Linear(F i ), and then the normalized weight coefficient is obtained through the softmax formula Afterwards, the auxiliary features are fused according to the weight coefficient to obtain the final auxiliary features The auxiliary feature F is the final feature obtained by adaptively fusing eleven auxiliary features in the cardiac coronary vessel naming model. It is more accurate and effective than the features before fusion. As a new auxiliary feature, F and the three-dimensional coordinates of the corresponding vascular midline points are input into the cardiac coronary vessel naming model for training.

[0134] In some optional embodiments, the cardiac coronary vessel naming model can be a PointNet++ model, and the first device can set the batch size in the training process to 16, the learning rate to 1e-3, and use the stochastic gradient descent method to train the model for a total of 300 rounds. The loss function of the training uses the cross entropy loss function. During the training process, the model is saved after every 5 rounds of training (i.e., after repeating steps S101 to S105 and steps S301 to S304 for 5 rounds of training, the trained cardiac coronary vessel naming model is saved), the parameters are changed to continue training, and finally the model with the best effect obtained by the verification set verification is selected for model prediction.

[0135] Thus, the training method of the coronary artery naming model provided by the embodiment of the present disclosure makes full use of the essential feature of vascular naming, namely, the information of the cardiac anatomical structure, so as to achieve more accurate and reasonable vascular naming; the point cloud is used to represent the structure of the blood vessel, and the information of the cardiac anatomical structure is pre-calculated and input into the network as prior information. The model structure is simple, the operation speed is fast, and it is more robust than the method of manually designing rules.

[0136] Figure 6 An optional structural schematic diagram of a training device for a cardiac coronary vessel naming model provided by an embodiment of the present disclosure is shown and will be explained according to each step.

[0137] In some embodiments, the training device 600 for the cardiac coronary vessel naming model includes a first acquisition unit 601 , a first confirmation unit 602 , a first feature fusion unit 603 , a prediction unit 604 and an adjustment unit 605 .

[0138] The first acquisition unit 601 is used to acquire a set of three-dimensional coordinates of a blood vessel midline point of a first sample image of a training set of coronary artery images;

[0139] The first confirmation unit 602 is used to confirm at least one of the bifurcation attribute, the number of the blood vessel segment to which it belongs, the blood vessel shape, and the relative position in the heart of each blood vessel midline point in the first sample image based on the three-dimensional coordinate set of the blood vessel midline points in the first sample image;

[0140] The first feature fusion unit 603 is used to input at least one of the bifurcation attribute, the number of the corresponding blood vessel, the shape of the blood vessel and the relative position in the heart of each blood vessel midline point in the first sample image into the feature fusion module included in the coronary vessel naming model, and determine the output of the feature fusion module as the auxiliary feature corresponding to each blood vessel midline point;

[0141] The prediction unit 604 is used to input the three-dimensional coordinate set of the blood vessel midline points of the first sample image and the auxiliary features corresponding to each blood vessel midline point into the classification module included in the coronary vessel naming model to obtain the predicted name of the blood vessel segment corresponding to each blood vessel midline point in the first sample image;

[0142] The adjusting unit 605 is configured to adjust the parameters of the coronary vessel naming model based on the annotated names of the vessel segments included in the first sample image and the predicted names of the vessel segments.

[0143] The first acquisition unit 601 is specifically used to perform blood vessel segmentation on the first sample image to obtain a blood vessel segmentation result; corrode the width of each blood vessel in the blood vessel segmentation result into one pixel; determine that at least one pixel corresponding to each blood vessel is a blood vessel midline point of the first sample image; and confirm that the three-dimensional coordinate set corresponding to the at least one pixel is the three-dimensional coordinate set of the blood vessel midline point of the first sample image.

[0144] The first confirmation unit 602 is specifically configured to map the three-dimensional coordinate set of the blood vessel midline points of the first sample image into a tree structure;

[0145] Confirming that a node including at least two child nodes in the tree structure is a bifurcation point, and that a bifurcation attribute of a blood vessel midline point corresponding to the bifurcation point is a bifurcation;

[0146] Alternatively, it is confirmed that a node including at most one child node in the tree structure is a non-bifurcation point, and the difference attribute of the blood vessel midline point corresponding to the non-bifurcation point is non-bifurcation.

[0147] The first confirmation unit 602 is specifically used to confirm that the bifurcation attribute is bifurcation, and at least one blood vessel midline point between two adjacent blood vessel midline points belongs to the same blood vessel segment;

[0148] The number of the blood vessel segment to which each blood vessel midline point of the first sample image belongs is confirmed.

[0149] The first confirmation unit 602 is specifically configured to perform smoothing processing on the tree structure.

[0150] The first confirmation unit 602 is specifically configured to determine the tangent direction of each node in the tree structure after smoothing, which is the blood vessel shape of the blood vessel midline point corresponding to each node.

[0151] The first confirmation unit 602 is specifically configured to determine a minimum distance between each blood vessel midline point of the first sample image and the left atrium surface in the first sample image, and a direction of each blood vessel midline point relative to a first identification point on the left atrium surface corresponding to the minimum distance;

[0152] Determine a minimum distance between each blood vessel midline point of the first sample image and the left ventricle surface in the first sample image, and a direction of each blood vessel midline point relative to a second identification point on the left ventricle surface corresponding to the minimum distance;

[0153] Determine a minimum distance between each blood vessel midline point of the first sample image and the right atrium surface in the first sample image, and a direction of each blood vessel midline point relative to a third marking point on the right atrium surface corresponding to the minimum distance;

[0154] Determine the minimum distance between each blood vessel centerline point of the first sample image and the right ventricular surface in the first sample image, and the direction of each blood vessel centerline point relative to a fourth marking point on the right ventricular surface corresponding to the minimum distance.

[0155] The first feature fusion unit 603 is specifically used to input at least one of the bifurcation attribute, the number of the blood vessel, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point of the first sample image into the first linear layer included in the feature fusion module, and determine the weight score corresponding to at least one of the bifurcation attribute, the number of the blood vessel, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point;

[0156] Based on the weight scores corresponding to at least one of the bifurcation attribute of each vascular midline point, the number of the blood vessel to which it belongs, the shape of the blood vessel, and the relative position in the heart, weighted summation is performed on at least one of the bifurcation attribute of each vascular midline point, the number of the blood vessel to which it belongs, the shape of the blood vessel, and the relative position in the heart to obtain a weighted summation result;

[0157] The weighted sum result is determined as the auxiliary feature corresponding to each blood vessel midline point output by the feature fusion module.

[0158] The first acquisition unit 601 is further used to perform sample enhancement on the coronary artery image training set.

[0159] Figure 7 An optional structural schematic diagram of the cardiac coronary vessel naming device provided in an embodiment of the present disclosure is shown, and will be explained according to each step.

[0160] In some embodiments, the cardiac coronary vessel naming device 700 includes a second acquisition unit 701 , a second confirmation unit 702 , a second feature fusion unit 703 and a naming unit 704 .

[0161] The second acquisition unit 701 is used to acquire a set of three-dimensional coordinates of the midline points of the blood vessels included in the coronary blood vessel image to be named;

[0162] The second confirmation unit 702 is used to confirm at least one of the bifurcation attribute, the number of the blood vessel segment to which it belongs, the blood vessel shape, and the relative position in the heart of each blood vessel midline point of the to-be-named heart coronary vessel image based on the three-dimensional coordinate set of the blood vessel midline points included in the to-be-named heart coronary vessel image;

[0163] The second feature fusion unit 703 is used to input at least one of the bifurcation attribute, the number of the corresponding vessel, the shape of the vessel and the relative position in the heart of each vessel midline point of the coronary vessel image to be named into the feature fusion module included in the coronary vessel naming model, and determine the output of the feature fusion module as the auxiliary feature corresponding to each vessel midline point;

[0164] The naming unit 704 is used to input the three-dimensional coordinate set of the vascular midline points of the to-be-named cardiac coronary vessel image and the auxiliary features corresponding to each vascular midline point into the classification module included in the cardiac coronary vessel naming model to obtain the name of the vascular segment corresponding to each vascular midline point in the to-be-named cardiac coronary vessel image.

[0165] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device and a readable storage medium.

[0166] Figure 8 A schematic block diagram of an example electronic device 800 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0167] like Figure 8 As shown, the electronic device 800 includes a computing unit 801, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 808 into a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the electronic device 800 can also be stored. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0168] Multiple components in the electronic device 800 are connected to the I / O interface 805, including: an input unit 806, such as a keyboard, a mouse, etc.; an output unit 807, such as various types of displays, speakers, etc.; a storage unit 808, such as a disk, an optical disk, etc.; and a communication unit 809, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 809 allows the electronic device 800 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0169] The computing unit 801 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 801 performs the various methods and processes described above, such as the training method of the cardiac coronary vessel naming model and / or the cardiac coronary vessel naming method. For example, in some embodiments, the training method of the cardiac coronary vessel naming model and / or the cardiac coronary vessel naming method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 808. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 800 via the ROM 802 and / or the communication unit 809. When the computer program is loaded into the RAM 803 and executed by the computing unit 801, one or more steps of the training method of the cardiac coronary vessel naming model and / or the cardiac coronary vessel naming method described above may be performed. Alternatively, in other embodiments, the computing unit 801 may be configured in any other appropriate manner (eg, by means of firmware) to execute the training method of the cardiac coronary vessel naming model and / or the cardiac coronary vessel naming method.

[0170] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0171] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0172] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0173] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0174] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0175] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0176] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0177] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of the present disclosure, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0178] The above is only a specific embodiment of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any person skilled in the art who is familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present disclosure, which should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. A training method for a cardiac coronary vessel naming model, characterized in that: The method comprises: Acquire a three-dimensional coordinate set of a blood vessel midline point of a first sample image of a training set of coronary artery images; Based on the three-dimensional coordinate set of the blood vessel midline points of the first sample image, confirm at least one of the bifurcation attribute, the number of the blood vessel segment to which it belongs, the blood vessel shape, and the relative position in the heart of each blood vessel midline point of the first sample image; Inputting at least one of the bifurcation attribute, the number of the corresponding blood vessel, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point in the first sample image into a feature fusion module included in the coronary vessel naming model, and determining the output of the feature fusion module as the auxiliary feature corresponding to each blood vessel midline point; Inputting the three-dimensional coordinate set of the blood vessel midline points of the first sample image and the auxiliary features corresponding to each blood vessel midline point into the classification module included in the coronary vessel naming model to obtain the predicted name of the blood vessel segment corresponding to each blood vessel midline point in the first sample image; Based on the labeled names of the blood vessel segments included in the first sample image and the predicted names of the blood vessel segments, the parameters of the coronary vessel naming model are adjusted.

2. The method according to claim 1, characterized in that The step of obtaining a set of three-dimensional coordinates of a blood vessel midline point of a first sample image of a training set of coronary artery images includes: Performing blood vessel segmentation on the first sample image to obtain a blood vessel segmentation result; Corroding the width of each blood vessel in the blood vessel segmentation result into one pixel; Determining at least one pixel corresponding to each of the blood vessels as a blood vessel midline point of the first sample image; It is confirmed that the three-dimensional coordinate set corresponding to the at least one pixel is a three-dimensional coordinate set of a blood vessel midline point of the first sample image.

3. The method according to claim 1, characterized in that The three-dimensional coordinate set of the blood vessel midline points based on the first sample image, confirming at least one of the bifurcation attribute, the number of the blood vessel segment to which it belongs, the blood vessel shape, and the relative position in the heart of each blood vessel midline point of the first sample image, includes: Mapping a three-dimensional coordinate set of blood vessel midline points of the first sample image into a tree structure; Confirming that a node including at least two child nodes in the tree structure is a bifurcation point, and that a bifurcation attribute of a blood vessel midline point corresponding to the bifurcation point is a bifurcation; Alternatively, it is confirmed that a node including at most one child node in the tree structure is a non-bifurcation point, and the difference attribute of the blood vessel midline point corresponding to the non-bifurcation point is non-bifurcation.

4. The method according to claim 1, characterized in that The three-dimensional coordinate set of the blood vessel midline points based on the first sample image, confirming at least one of the bifurcation attribute, the number of the blood vessel segment to which it belongs, the blood vessel shape, and the relative position in the heart of each blood vessel midline point of the first sample image, includes: Confirm that the bifurcation attribute is bifurcation, and at least one of the two adjacent vascular midline points belongs to the same vascular segment; The number of the blood vessel segment to which each blood vessel midline point of the first sample image belongs is confirmed.

5. The method according to claim 3, characterized in that: After mapping the three-dimensional coordinate set of the blood vessel midline points of the first sample image into a tree structure, the method further includes: The tree structure is smoothed.

6. The method according to claim 5, characterized in that The three-dimensional coordinate set of the blood vessel midline points based on the first sample image, confirming at least one of the bifurcation attribute, the number of the blood vessel segment to which it belongs, the blood vessel shape, and the relative position in the heart of each blood vessel midline point of the first sample image, includes: The tangent direction of each node in the tree structure after smoothing is determined to be the blood vessel shape of the blood vessel midline point corresponding to each node.

7. The method according to claim 5, characterized in that The three-dimensional coordinate set of the blood vessel midline points based on the first sample image, confirming at least one of the bifurcation attribute, the number of the blood vessel segment to which it belongs, the blood vessel shape, and the relative position in the heart of each blood vessel midline point of the first sample image, includes: Determine a minimum distance between each blood vessel midline point of the first sample image and the left atrium surface in the first sample image, and a direction of each blood vessel midline point relative to a first identification point on the left atrium surface corresponding to the minimum distance; Determine a minimum distance between each blood vessel midline point of the first sample image and the left ventricle surface in the first sample image, and a direction of each blood vessel midline point relative to a second identification point on the left ventricle surface corresponding to the minimum distance; Determine a minimum distance between each blood vessel midline point of the first sample image and the right atrium surface in the first sample image, and a direction of each blood vessel midline point relative to a third marking point on the right atrium surface corresponding to the minimum distance; Determine the minimum distance between each blood vessel centerline point of the first sample image and the right ventricular surface in the first sample image, and the direction of each blood vessel centerline point relative to a fourth marking point on the right ventricular surface corresponding to the minimum distance.

8. The method according to claim 1, characterized in that The step of inputting at least one of the bifurcation attribute, the number of the corresponding blood vessel, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point of the first sample image into a feature fusion module included in the coronary vessel naming model, and determining that the output of the feature fusion module is the auxiliary feature corresponding to each blood vessel midline point, comprises: Inputting at least one of the bifurcation attribute, the number of the blood vessel, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point of the first sample image into the first linear layer included in the feature fusion module, and determining the weight score corresponding to at least one of the bifurcation attribute, the number of the blood vessel, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point; Based on the weight scores corresponding to at least one of the bifurcation attribute of each vascular midline point, the number of the blood vessel to which it belongs, the shape of the blood vessel, and the relative position in the heart, weighted summation is performed on at least one of the bifurcation attribute of each vascular midline point, the number of the blood vessel to which it belongs, the shape of the blood vessel, and the relative position in the heart to obtain a weighted summation result; The weighted sum result is determined as the auxiliary feature corresponding to each blood vessel midline point output by the feature fusion module.

9. The method according to claim 1, characterized in that: The method further comprises: Sample enhancement is performed on the coronary artery image training set.

10. A method for naming coronary vessels, characterized in that: Using the coronary vessel naming model described in any one of claims 1 to 9, the method comprises: Acquire a three-dimensional coordinate set of the vascular midline points included in the coronary artery image to be named; Based on the three-dimensional coordinate set of the vascular midline points included in the to-be-named cardiac coronary vessel image, confirm at least one of the bifurcation attribute, the number of the vascular segment to which it belongs, the vascular shape, and the relative position in the heart of each vascular midline point of the to-be-named cardiac coronary vessel image; Inputting at least one of the bifurcation attribute, the number of the corresponding vessel, the shape of the vessel and the relative position in the heart of each vessel midline point of the coronary vessel image to be named into the feature fusion module included in the coronary vessel naming model, and determining the output of the feature fusion module as the auxiliary feature corresponding to each vessel midline point; The three-dimensional coordinate set of the vascular midline points of the to-be-named cardiac coronary vessel image and the auxiliary features corresponding to each vascular midline point are input into the classification module included in the cardiac coronary vessel naming model to obtain the name of the vascular segment corresponding to each vascular midline point in the to-be-named cardiac coronary vessel image.

11. A training device for a coronary artery naming model, characterized in that: The device comprises: A first acquisition unit, used to acquire a three-dimensional coordinate set of a blood vessel midline point of a first sample image of a training set of coronary artery images; A first confirmation unit is used to confirm at least one of a bifurcation attribute, a number of a corresponding vascular segment, a vascular shape, and a relative position in the heart of each vascular midline point of the first sample image based on a set of three-dimensional coordinates of the vascular midline points of the first sample image; A first feature fusion unit is used to input at least one of the bifurcation attribute, the number of the corresponding blood vessel, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point of the first sample image into a feature fusion module included in the coronary vessel naming model, and determine that the output of the feature fusion module is the auxiliary feature corresponding to each blood vessel midline point; a prediction unit, configured to input the three-dimensional coordinate set of the vascular midline points of the first sample image and the auxiliary features corresponding to each vascular midline point into the classification module included in the coronary vessel naming model, and obtain a predicted name of the vascular segment corresponding to each vascular midline point in the first sample image; An adjusting unit is used to adjust the parameters of the coronary vessel naming model based on the labeled names of the vessel segments included in the first sample image and the predicted names of the vessel segments.

12. A device for naming coronary vessels, characterized in that: Using the coronary vessel naming model described in any one of claims 1 to 9, the device comprises: A second acquisition unit is used to acquire a three-dimensional coordinate set of a blood vessel midline point included in the coronary blood vessel image to be named; A second confirmation unit is used to confirm at least one of a bifurcation attribute, a number of a corresponding vascular segment, a vascular shape, and a relative position in the heart of each vascular midline point of the to-be-named cardiac coronary vessel image based on a set of three-dimensional coordinates of the vascular midline points included in the to-be-named cardiac coronary vessel image; A second feature fusion unit is used to input at least one of the bifurcation attribute, the number of the corresponding blood vessel, the shape of the blood vessel, and the relative position in the heart of each blood vessel midline point of the coronary artery image to be named into the feature fusion module included in the coronary artery naming model, and determine the output of the feature fusion module as the auxiliary feature corresponding to each blood vessel midline point; The naming unit is used to input the three-dimensional coordinate set of the vascular midline points of the to-be-named cardiac coronary vessel image and the auxiliary features corresponding to each vascular midline point into the classification module included in the cardiac coronary vessel naming model to obtain the name of the vascular segment corresponding to each vascular midline point in the to-be-named cardiac coronary vessel image.

13. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 9; Alternatively, the method of claim 10 can be performed.

14. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 9; Alternatively, the method of claim 10 can be performed.

Images