Cerebrovascular feature extraction method and system for image processing

Abstract

The invention provides a cerebrovascular feature extraction method and system for image processing. The method comprises the following steps: acquiring a plurality of to-be-used images and a plurality of corresponding reference images; performing alignment processing on the to-be-used image and the reference image to obtain a subtraction image; dividing a plurality of first geometric feature points in the subtraction image into a central feature point and a plurality of first to-be-screened feature points, selecting edge noise points, and obtaining a first de-noised image comprising second to-be-screened feature points; dividing the first de-noised image into a to-be-checked image and a reference image, and obtaining a final image based on the to-be-checked image and the reference image; and acquiring a cerebrovascular feature map through the final image. According to the method, edge noise points located at the position of the brain contour are filtered out through the center feature points, after the second to-be-screened feature points in the to-be-checked image are mapped into different reference images based on a coaxial rotation image acquisition rule, noise points in the second to-be-screened feature points are filtered out, and it is ensured that the final image accurately represents the overall form of the cerebral vessel.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method and system for extracting cerebrovascular features through image processing. Background Art

[0002] As people age, the probability of "stroke" and "infarction" caused by cerebrovascular disease gradually increases. The vascular morphology, such as the shape, position, distribution, and thickness changes of blood vessels, is the gold standard for judging the cause and type of brain disease. It is also an important reference for neurointerventional surgery and navigation.

[0003] With the continuous development of medical imaging technology, the characteristic images of cerebrovascular tissue formed by digital subtraction angiography can better reflect the morphology and structure of cerebrovascular vessels. Therefore, digital subtraction angiography has also become one of the main examination methods for cerebrovascular lesions.

[0004] However, in the process of obtaining cerebrovascular characteristic images through digital subtraction angiography technology, the patient's swallowing and other movements, as well as the cross-mapping between different organs in the brain, will cause a large number of noise points in the obtained cerebrovascular characteristic images, making it difficult to extract cerebrovascular features and to accurately grasp the morphology of the cerebrovascular vessels. Summary of the invention

[0005] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a method and system for extracting cerebrovascular features by image processing, aiming to solve the technical problem that in the prior art, cerebrovascular feature images are obtained by digital subtraction angiography technology, and there are many noise points in the imaging process, making it difficult to extract cerebrovascular features and difficult to accurately grasp the morphology of cerebrovascular vessels.

[0006] In order to achieve the above objectives, in a first aspect, an embodiment of the present application provides a method for extracting cerebrovascular features by image processing, comprising the following steps:

[0007] Acquire an initial image set and an angiography image set corresponding to the patient's brain, wherein the initial image set includes a plurality of reference images formed by coaxial rotation, and the angiography image set includes a plurality of standby images formed by coaxial rotation, and the plurality of standby images correspond one-to-one to the plurality of reference images;

[0008] Performing alignment processing on the standby image and the reference image, and acquiring a plurality of subtraction images based on the standby image and the reference image;

[0009] Acquire a plurality of first geometric feature points of the subtracted image, separate the plurality of first geometric feature points into a central feature point and a plurality of first feature points to be screened, and select edge noise points from the plurality of first feature points to be screened based on the central feature point to acquire a plurality of first denoised images including second feature points to be screened;

[0010] Selecting one of the first denoised images as a to-be-checked image, selecting the remaining first denoised images as reference images, and filtering second to-be-screened feature points in the to-be-checked image based on a rotation angle between the to-be-checked image and the reference image to obtain a final image;

[0011] The final image is sequentially subjected to Gaussian filtering, Laplace filtering and Gabor filtering to obtain a first processed image, a second processed image and a third processed image, respectively, and a cerebral vascular feature map is obtained based on the final image, the first processed image, the second processed image, the third processed image and a feature extraction neural network.

[0012] Further, the step of acquiring an initial image set and an angiography image set corresponding to the patient's brain, wherein the initial image set includes a plurality of reference images formed by coaxial rotation, and the angiography image set includes a plurality of standby images formed by coaxial rotation, comprises:

[0013] When the contrast agent has not reached the cerebral blood vessels, the C-arm is controlled to rotate several times at a preset angle with the center line of the detection bed as the axis to obtain several reference images, and the several reference images form an initial image set;

[0014] When the contrast agent reaches the cerebral blood vessels, the C-arm is controlled to rotate several times at the preset angle with the center line of the detection bed as the axis to obtain several standby images, and the several standby images form a contrast image set.

[0015] Furthermore, the step of performing alignment processing on the standby image and the reference image includes:

[0016] Acquire a plurality of second geometric feature points of the standby image and a plurality of third geometric feature points of the reference image;

[0017] Comparing the second geometric feature point with the third geometric feature point for similarity, so as to select a second registration point from a plurality of the second geometric feature points, and select a third registration point corresponding to the second registration point from a plurality of the third geometric feature points;

[0018] The standby image is aligned to the reference image based on the second registration point and the third registration point.

[0019] Furthermore, the step of acquiring the subtraction image based on the standby image and the reference image includes:

[0020] Acquire a first imaging intensity of the reference image, and acquire a second imaging intensity of the standby image;

[0021] An adjusted imaging intensity is determined based on the first imaging intensity and the second imaging intensity, and a subtraction image is formed by the adjusted imaging intensity.

[0022] Furthermore, the formula for obtaining the first imaging intensity is:

[0023] Q j =aQ z exp(-(X g H g +X r H r +X b H b )),

[0024] Among them, Q j represents the first imaging intensity, a represents a constant, Q z Indicates the contrast agent imaging intensity, X g represents the bone absorption coefficient, H g Indicates bone thickness, X r represents the soft tissue absorption coefficient, H r represents the thickness of soft tissue, X b represents the vascular absorption coefficient, H b Indicates the thickness of blood vessels;

[0025] The acquisition formula of the second imaging intensity is:

[0026] Q d =aQ z exp(-(X g H g +X r H r +X bz H b )),

[0027] Among them, Q d represents the second imaging intensity, X bz Represents the vascular absorption coefficient after injection of contrast agent.

[0028] Furthermore, the step of selecting edge noise points from a plurality of the first feature points to be screened based on the central feature point includes:

[0029] Obtaining the Euclidean distance between the first feature point to be screened and the central feature point;

[0030] Determine whether the Euclidean distance is greater than a distance threshold;

[0031] If the Euclidean distance is greater than the distance threshold, the first feature point to be screened is determined to be an edge noise point.

[0032] Furthermore, the step of filtering the second feature points to be screened in the image to be checked based on the rotation angle between the image to be checked and the reference image is specifically:

[0033] Based on the rotation angle of the coaxial rotation, mapping a second feature point to be screened in the image to be checked to a plurality of the reference images to form a mapping point;

[0034] Acquire a second feature point to be screened adjacent to the mapping point in the reference image to obtain a neighboring point;

[0035] Determine whether the distance between the neighboring point and the mapping point is less than a distance threshold, and if the distance between the neighboring point and the mapping point is less than the distance threshold, determine the neighboring point as a corresponding point;

[0036] The number of corresponding points in the reference images is obtained, and it is determined whether the number is less than a number threshold. If the number is less than the number threshold, the second feature point to be screened in the image to be checked is filtered.

[0037] Furthermore, the feature extraction neural network includes a first convolutional neural network, a second convolutional neural network, a third convolutional neural network and a fourth convolutional neural network, and the step of acquiring a cerebrovascular feature map based on the final image, the first processed image, the second processed image, the third processed image and the feature extraction neural network includes:

[0038] Using the final image as an input value of the first convolutional neural network, using the first processed image as an input value of the second convolutional neural network, using the second processed image as an input value of the third convolutional neural network, and using the third processed image as an input value of the fourth convolutional neural network to obtain a first feature map, a second feature map, a third feature map, and a fourth feature map;

[0039] Feature fusion is performed on the first feature map, the second feature map, the third feature map, and the fourth feature map to obtain a cerebrovascular feature map.

[0040] In a second aspect, an embodiment of the present application provides a cerebrovascular feature extraction system for image processing, which is applied to the cerebrovascular feature extraction method for image processing as described in the first aspect above, and the system includes:

[0041] An acquisition module, used for acquiring an initial image set and an angiography image set corresponding to the patient's brain, wherein the initial image set includes a plurality of reference images formed by coaxial rotation, and the angiography image set includes a plurality of standby images formed by coaxial rotation, and the plurality of standby images correspond one to one with the plurality of reference images;

[0042] An adjustment module, used for performing alignment processing on the standby image and the reference image, and acquiring a plurality of subtraction images based on the standby image and the reference image;

[0043] A first selection module is used to obtain a plurality of first geometric feature points of the subtracted image, separate the plurality of first geometric feature points into a central feature point and a plurality of first feature points to be screened, and select edge noise points from the plurality of first feature points to be screened based on the central feature point to obtain a plurality of first denoised images including second feature points to be screened;

[0044] A second selection module is used to select one of the first denoised images as a to-be-checked image, select the remaining first denoised images as reference images, and filter second feature points to be screened in the to-be-checked image based on a rotation angle between the to-be-checked image and the reference image to obtain a final image;

[0045] An extraction module is used to perform Gaussian filtering, Laplace filtering and Gabor filtering on the final image in sequence to obtain a first processed image, a second processed image and a third processed image respectively, and obtain a cerebral vascular feature map based on the final image, the first processed image, the second processed image, the third processed image and a feature extraction neural network.

[0046] In a third aspect, an embodiment of the present application provides a computer, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for extracting cerebrovascular features by image processing as described in the first aspect above is implemented.

[0047] In a fourth aspect, an embodiment of the present application provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for extracting cerebrovascular features by image processing as described in the first aspect above.

[0048] Compared with the prior art, the beneficial effects of the present invention are as follows: by aligning the image to be used with the reference image and forming the subtraction image, the influence of non-cerebrovascular tissues can be eliminated to a certain extent, and the contrast between cerebrovascular and non-cerebrovascular tissues can be enhanced; by obtaining the central feature point and judging the distance between the first feature point to be screened and the central feature point, the edge noise point located at the position of the brain contour is filtered out, and the influence of noise is reduced to a certain extent; based on the acquisition rule of coaxial rotation in the initial image set and the angiography image set, it can be clearly seen that the position change of the second feature point to be screened in different first denoised images is small, therefore, by mapping the second feature point to be screened in the image to be checked in different reference images, it can be determined whether the corresponding second feature point to be screened exists in the adjacent area of ​​the mapping position in the different images to be checked, so as to determine whether the second feature point to be screened in the image to be checked is a noise point, thereby further reducing the influence of noise on the extraction of cerebrovascular features, ensuring that the final image accurately represents the overall morphology of the cerebrovascular, and then completing the extraction of cerebrovascular features through the feature extraction neural network. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 Flow chart of the method for extracting cerebrovascular features by image processing in the first embodiment of the present invention;

[0050] Figure 2 It is a structural block diagram of a cerebrovascular feature extraction system for image processing in a second embodiment of the present invention;

[0051] The following specific implementation manner will further illustrate the present invention in conjunction with the above-mentioned drawings. DETAILED DESCRIPTION

[0052] In order to facilitate the understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings. Several embodiments of the present invention are given in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive.

[0053] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be a central element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0054] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0055] See also Figure 1 The first embodiment of the present invention provides a method for extracting cerebrovascular features through image processing, comprising the following steps:

[0056] S10: acquiring an initial image set and an angiography image set corresponding to the patient's brain, wherein the initial image set includes a plurality of reference images formed by coaxial rotation, and the angiography image set includes a plurality of standby images formed by coaxial rotation, and the plurality of standby images correspond one to one with the plurality of reference images;

[0057] The step S10 comprises:

[0058] S110: When the contrast agent has not reached the cerebral blood vessels, the C-arm is controlled to rotate several times at a preset angle with the center line of the detection bed as the axis to obtain several reference images, and the several reference images form an initial image set;

[0059] S120: When the contrast agent reaches the cerebral blood vessels, the C-shaped arm is controlled to rotate at the preset angle for a number of times with the center line of the detection bed as the axis to obtain a number of standby images, and the number of standby images form a contrast image set;

[0060] It should be noted that when acquiring a plurality of the reference images and a plurality of the standby images respectively, the C-shaped arm rotates the same number of times and the preset angle of each rotation is the same. After a plurality of rotations, the C-shaped arm rotates 360°. In this embodiment, the preset angle is 10°. It is ensured that the reference images correspond to the standby images one by one.

[0061] S20: performing alignment processing on the standby image and the reference image, and acquiring a plurality of subtraction images based on the standby image and the reference image;

[0062] The step S20 comprises:

[0063] S210: Acquire a plurality of second geometric feature points of the standby image and a plurality of third geometric feature points of the reference image;

[0064] In this embodiment, the second geometric feature points and the third geometric feature points are extracted from the standby image and the reference image respectively by using an edge detection algorithm. The edge detection algorithm has many applications and will not be described in detail here.

[0065] S220: performing a similarity comparison between the second geometric feature point and the third geometric feature point to select a second registration point from a plurality of the second geometric feature points, and select a third registration point corresponding to the second registration point from a plurality of the third geometric feature points;

[0066] Since the standby image corresponds to the reference image, the only difference is whether contrast agent is injected into the cerebral blood vessels. Therefore, in the standby image and the reference image, there is a certain similarity between the second geometric feature points and the third geometric feature points in the same area, and the second registration point and the third registration point can be determined.

[0067] S230: Aligning the standby image to the reference image based on the second registration point and the third registration point.

[0068] S240: Acquire a first imaging intensity of the reference image, and acquire a second imaging intensity of the standby image;

[0069] The acquisition formula of the first imaging intensity is:

[0070] Q j =aQ z exp(-(X g H g +X r H r +X b H b )),

[0071] Among them, Q j represents the first imaging intensity, a represents a constant, Q z Indicates the contrast agent imaging intensity, X g represents the bone absorption coefficient, H g Indicates bone thickness, X r represents the soft tissue absorption coefficient, H r represents the thickness of soft tissue, X b represents the vascular absorption coefficient, H b Indicates the thickness of blood vessels;

[0072] The acquisition formula of the second imaging intensity is:

[0073] Q d =aQ z exp(-(X g H g +Xr H r +X bz H b )),

[0074] Among them, Q d represents the second imaging intensity, X bz Represents the vascular absorption coefficient after injection of contrast agent.

[0075] S250: determining an adjusted imaging intensity based on the first imaging intensity and the second imaging intensity, and forming a subtraction image by the adjusted imaging intensity;

[0076] The acquisition formula for adjusting the imaging intensity is:

[0077] Q t =Q d -Q j ,

[0078] Among them, Q t Indicates adjusting the imaging intensity.

[0079] By acquiring the subtraction image, other brain tissues are eliminated and the influence of cerebral blood vessels is retained.

[0080] S30: acquiring a plurality of first geometric feature points of the subtracted image, segmenting the plurality of first geometric feature points into a central feature point and a plurality of first feature points to be screened, and selecting edge noise points from the plurality of first feature points to be screened based on the central feature point, so as to acquire a plurality of first denoised images including second feature points to be screened;

[0081] Generally, the distribution of cerebral blood vessels extends from the central area of ​​the brain to the peripheral area of ​​the brain. In this embodiment, the central feature point is obtained by manual calibration.

[0082] The step S30 comprises:

[0083] S310: Obtaining the Euclidean distance between the first feature point to be screened and the central feature point;

[0084] S320: Determine whether the Euclidean distance is greater than a distance threshold;

[0085] S330: If the Euclidean distance is greater than the distance threshold, determining that the first feature point to be screened is an edge noise point;

[0086] By judging the Euclidean distance between the first feature point to be screened and the central feature point, it can be judged whether the first feature point to be screened is located in the edge area of ​​the brain based on the position of the central feature point, thereby avoiding the capture of feature points representing the brain contour mixed with the feature points of the cerebral blood vessels during the acquisition process of the first geometric feature point, thereby affecting the accuracy of the subsequent extraction of cerebral blood vessel features.

[0087] S40: selecting one of the first denoised images as a to-be-checked image, selecting the remaining first denoised images as reference images, and filtering second to-be-screened feature points in the to-be-checked image based on a rotation angle between the to-be-checked image and the reference image to obtain a final image;

[0088] It can be understood that since the C-arm axially rotates to obtain the initial image set and the angiography image set, the subtraction images generated by the initial image set and the angiography image set are also generated by the axial rotation of the C-arm. Therefore, the feature points in the subtraction images have spatial position correlation, that is, the same feature point corresponding to the cerebral blood vessels in the subtraction images can form a kind of ring structure along the axial direction of the C-arm. Based on this principle,

[0089] The step S40 comprises:

[0090] S410: Based on the rotation angle of the coaxial rotation, mapping a second feature point to be screened in the image to be checked to a plurality of reference images to form a mapping point;

[0091] S420: Acquire a second feature point to be screened adjacent to the mapping point in the reference image to acquire a neighboring point;

[0092] S430: Determine whether the distance between the neighboring point and the mapping point is less than a distance threshold, and if the distance between the neighboring point and the mapping point is less than the distance threshold, determine the neighboring point as a corresponding point;

[0093] It can be understood that if the distance between the neighboring point and the mapping point is greater than the distance threshold, it is determined that the second feature point to be screened in the image to be checked is a noise point and needs to be filtered.

[0094] S440: Obtain the number of corresponding points in the reference images, determine whether the number is less than a number threshold, and if the number is less than the number threshold, filter the second feature point to be screened in the image to be checked.

[0095] That is, if the second feature point to be screened of the image to be checked corresponds to the cerebrovascular feature, then in all the reference images, the number of the corresponding points should be greater than the number threshold. If it is less than, it does not meet the spatial correlation between the image to be checked and the reference image, and is determined to be a noise point. It can be understood that the above determination is performed on all the second feature points to be screened in the image to be checked to complete the noise filtering in the image to be checked.

[0096] S50: performing Gaussian filtering, Laplace filtering and Gabor filtering on the final image in sequence to obtain a first processed image, a second processed image and a third processed image respectively, and obtaining a cerebrovascular feature map based on the final image, the first processed image, the second processed image, the third processed image and a feature extraction neural network;

[0097] Since cerebral blood vessels vary in thickness, it is difficult to accurately find the features of cerebral blood vessels of different thicknesses if the cerebral blood vessel feature map is directly extracted through the final image. In this embodiment, through the Gaussian filtering process, the Laplace filtering process and the Gabor filtering process, different morphological features can be emphasized for different detailed features of the cerebral blood vessels, thereby more accurately completing the extraction of cerebral blood vessel features.

[0098] The feature extraction neural network includes a first convolutional neural network, a second convolutional neural network, a third convolutional neural network and a fourth convolutional neural network, and the network structures of the first convolutional neural network, the second convolutional neural network, the third convolutional neural network and the fourth convolutional neural network are the same. It can be understood that the first convolutional neural network, the second convolutional neural network, the third convolutional neural network and the fourth convolutional neural network are all trained convolutional neural networks, and the training is completed by taking the original image as the input value and the cerebrovascular feature map as the output value.

[0099] The step S50 comprises:

[0100] S510: using the final image as an input value of the first convolutional neural network, using the first processed image as an input value of the second convolutional neural network, using the second processed image as an input value of the third convolutional neural network, and using the third processed image as an input value of the fourth convolutional neural network to obtain a first feature map, a second feature map, a third feature map, and a fourth feature map;

[0101] S520: Perform feature fusion on the first feature map, the second feature map, the third feature map, and the fourth feature map to obtain a cerebrovascular feature map.

[0102] In this embodiment, the feature values ​​of the same area in the first feature map, the second feature map, the third feature map and the fourth feature map are weighted averaged to form a replacement feature value of the area. After the replacement feature values ​​of all areas are acquired, the cerebrovascular feature map is formed.

[0103] By aligning the image to be used with the reference image and forming the subtraction image, the influence of non-cerebral vascular tissues can be eliminated to a certain extent, and the contrast between cerebral blood vessels and non-cerebral blood vessels can be enhanced; by obtaining the central feature point and judging the distance between the first feature point to be screened and the central feature point, the edge noise point located at the brain contour position is filtered out, and the influence of noise is reduced to a certain extent; based on the acquisition rule of coaxial rotation in the initial image set and the angiography image set, it can be clearly seen that the position change of the second feature point to be screened in different first denoised images is small. Therefore, by mapping the second feature point to be screened in the image to be checked in different reference images, it can be determined whether the corresponding second feature point to be screened exists in the adjacent area of ​​the mapping position in different images to determine whether the second feature point to be screened in the image to be checked is a noise point, thereby further reducing the influence of noise on the extraction of cerebrovascular features, ensuring that the final image accurately represents the overall morphology of the cerebrovascular vessels, and then completing the extraction of cerebrovascular features through the feature extraction neural network.

[0104] See also Figure 2 The second embodiment of the present invention provides a cerebrovascular feature extraction system for image processing, which is applied to the cerebrovascular feature extraction method for image processing in the above-mentioned embodiment, which has been described and will not be repeated. As used below, the terms "module", "unit", "subunit" and the like can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0105] The system comprises:

[0106] An acquisition module 10 is used to acquire an initial image set and an angiography image set corresponding to the patient's brain, wherein the initial image set includes a plurality of reference images formed by coaxial rotation, and the angiography image set includes a plurality of standby images formed by coaxial rotation, and the plurality of standby images correspond to the plurality of reference images in a one-to-one manner;

[0107] The acquisition module 10 includes:

[0108] The first unit is used for controlling the C-arm to rotate several times at a preset angle with the center line of the detection bed as the axis when the contrast agent has not reached the cerebral blood vessels, so as to obtain several reference images, wherein the several reference images form an initial image set;

[0109] The second unit is used for controlling the C-shaped arm to rotate several times at the preset angle with the center line of the detection bed as the axis when the contrast agent reaches the cerebral blood vessels, so as to obtain several standby images, wherein the several standby images form a contrast image set;

[0110] An adjustment module 20, configured to perform alignment processing on the standby image and the reference image, and obtain a plurality of subtraction images based on the standby image and the reference image;

[0111] The adjustment module 20 includes:

[0112] A third unit is used to obtain a plurality of second geometric feature points of the standby image and a plurality of third geometric feature points of the reference image;

[0113] A fourth unit is used to compare the second geometric feature point with the third geometric feature point for similarity, so as to select a second registration point from a plurality of the second geometric feature points, and select a third registration point corresponding to the second registration point from a plurality of the third geometric feature points;

[0114] A fifth unit, configured to align the standby image to the reference image based on the second registration point and the third registration point;

[0115] A sixth unit, configured to obtain a first imaging intensity of the reference image and a second imaging intensity of the standby image;

[0116] A seventh unit, configured to determine an adjusted imaging intensity based on the first imaging intensity and the second imaging intensity, and form a subtraction image by using the adjusted imaging intensity;

[0117] A first selection module 30 is used to obtain a plurality of first geometric feature points of the subtracted image, separate the plurality of first geometric feature points into a central feature point and a plurality of first feature points to be screened, and select edge noise points from the plurality of first feature points to be screened based on the central feature point to obtain a plurality of first denoised images including second feature points to be screened;

[0118] The first selection module 30 includes:

[0119] An eighth unit is used to obtain the Euclidean distance between the first feature point to be screened and the central feature point;

[0120] A ninth unit is used to determine whether the Euclidean distance is greater than a distance threshold;

[0121] A tenth unit, configured to determine that the first feature point to be screened is an edge noise point if the Euclidean distance is greater than the distance threshold;

[0122] A second selection module 40 is used to select one of the first denoised images as a to-be-checked image, select the remaining first denoised images as reference images, and filter the second to-be-screened feature points in the to-be-checked image based on the rotation angle between the to-be-checked image and the reference image to obtain a final image;

[0123] The second selection module 40 includes:

[0124] An eleventh unit is used for mapping a second feature point to be screened in the image to be checked to a plurality of reference images based on a rotation angle of the coaxial rotation to form a mapping point;

[0125] A twelfth unit is used to obtain a second feature point to be screened adjacent to the mapping point in the reference image to obtain a neighboring point;

[0126] A thirteenth unit is used to determine whether the distance between the neighboring point and the mapping point is less than a distance threshold, and if the distance between the neighboring point and the mapping point is less than the distance threshold, the neighboring point is determined as a corresponding point;

[0127] A fourteenth unit is used to obtain the number of corresponding points in the reference images, determine whether the number is less than a number threshold, and if the number is less than the number threshold, filter the second feature point to be screened in the image to be checked;

[0128] An extraction module 50 is used to perform Gaussian filtering, Laplace filtering and Gabor filtering on the final image in sequence to obtain a first processed image, a second processed image and a third processed image respectively, and obtain a cerebrovascular feature map based on the final image, the first processed image, the second processed image, the third processed image and a feature extraction neural network;

[0129] The extraction module 50 comprises:

[0130] A fifteenth unit is used to use the final image as an input value of the first convolutional neural network, use the first processed image as an input value of the second convolutional neural network, use the second processed image as an input value of the third convolutional neural network, and use the third processed image as an input value of the fourth convolutional neural network to obtain a first feature map, a second feature map, a third feature map, and a fourth feature map;

[0131] The sixteenth unit is used to perform feature fusion on the first feature map, the second feature map, the third feature map and the fourth feature map to obtain a cerebrovascular feature map.

[0132] The present invention also provides a computer, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for extracting cerebrovascular features by image processing as described in the above technical solution is implemented.

[0133] The present invention also provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the method for extracting cerebrovascular features by image processing as described in the above technical solution is implemented.

[0134] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0135] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A method for extracting cerebrovascular features by image processing, characterized in that: The following steps are involved: Acquire an initial image set and an angiography image set corresponding to the patient's brain, wherein the initial image set includes a plurality of reference images formed by coaxial rotation, and the angiography image set includes a plurality of standby images formed by coaxial rotation, and the plurality of standby images correspond one-to-one to the plurality of reference images; Performing alignment processing on the standby image and the reference image, and acquiring a plurality of subtraction images based on the standby image and the reference image; Acquire a plurality of first geometric feature points of the subtracted image, separate the plurality of first geometric feature points into a central feature point and a plurality of first feature points to be screened, and select edge noise points from the plurality of first feature points to be screened based on the central feature point to acquire a plurality of first denoised images including second feature points to be screened; Selecting one of the first denoised images as a to-be-checked image, selecting the remaining first denoised images as reference images, and filtering second to-be-screened feature points in the to-be-checked image based on a rotation angle between the to-be-checked image and the reference image to obtain a final image; The final image is sequentially subjected to Gaussian filtering, Laplace filtering and Gabor filtering to obtain a first processed image, a second processed image and a third processed image, respectively, and a cerebral vascular feature map is obtained based on the final image, the first processed image, the second processed image, the third processed image and a feature extraction neural network.

2. The method for extracting cerebrovascular features by image processing according to claim 1, characterized in that: The step of acquiring an initial image set and an angiography image set corresponding to the patient's brain, wherein the initial image set includes a plurality of reference images formed by coaxial rotation, and the angiography image set includes a plurality of standby images formed by coaxial rotation comprises: When the contrast agent has not reached the cerebral blood vessels, the C-arm is controlled to rotate several times at a preset angle with the center line of the detection bed as the axis to obtain several reference images, and the several reference images form an initial image set; When the contrast agent reaches the cerebral blood vessels, the C-arm is controlled to rotate several times at the preset angle with the center line of the detection bed as the axis to obtain several standby images, and the several standby images form a contrast image set.

3. The method for extracting cerebrovascular features by image processing according to claim 1, characterized in that: The step of performing alignment processing on the standby image and the reference image comprises: Acquire a plurality of second geometric feature points of the standby image and a plurality of third geometric feature points of the reference image; Comparing the second geometric feature point with the third geometric feature point for similarity, so as to select a second registration point from a plurality of the second geometric feature points, and select a third registration point corresponding to the second registration point from a plurality of the third geometric feature points; The standby image is aligned to the reference image based on the second registration point and the third registration point.

4. The method for extracting cerebrovascular features by image processing according to claim 1, characterized in that: The step of acquiring a subtraction image based on the standby image and the reference image comprises: Acquire a first imaging intensity of the reference image, and acquire a second imaging intensity of the standby image; An adjusted imaging intensity is determined based on the first imaging intensity and the second imaging intensity, and a subtraction image is formed by the adjusted imaging intensity.

5. The method for extracting cerebrovascular features by image processing according to claim 4, characterized in that: The acquisition formula of the first imaging intensity is: Q j =aQ z exp(-(X g H g +X r H r +X b H b )), Among them, Q j represents the first imaging intensity, a represents a constant, Q z Indicates the contrast agent imaging intensity, X g represents the bone absorption coefficient, H g Indicates bone thickness, X r represents the soft tissue absorption coefficient, H r represents the thickness of soft tissue, X b represents the vascular absorption coefficient, H b Indicates the thickness of blood vessels; The acquisition formula of the second imaging intensity is: Q d =aQ z exp(-(X g H g +X r H r +X bz H b )), Among them, Q d represents the second imaging intensity, X bz Represents the vascular absorption coefficient after injection of contrast agent.

6. The method for extracting cerebrovascular features by image processing according to claim 1, characterized in that: The step of selecting edge noise points from a plurality of the first feature points to be screened based on the central feature point comprises: Obtaining the Euclidean distance between the first feature point to be screened and the central feature point; Determine whether the Euclidean distance is greater than a distance threshold; If the Euclidean distance is greater than the distance threshold, the first feature point to be screened is determined to be an edge noise point.

7. The method for extracting cerebrovascular features by image processing according to claim 1, characterized in that: The step of filtering the second feature points to be screened in the image to be checked based on the rotation angle between the image to be checked and the reference image is specifically: Based on the rotation angle of the coaxial rotation, mapping a second feature point to be screened in the image to be checked to a plurality of the reference images to form a mapping point; Acquire a second feature point to be screened adjacent to the mapping point in the reference image to obtain a neighboring point; Determine whether the distance between the neighboring point and the mapping point is less than a distance threshold, and if the distance between the neighboring point and the mapping point is less than the distance threshold, determine the neighboring point as a corresponding point; The number of corresponding points in the reference images is obtained, and it is determined whether the number is less than a number threshold. If the number is less than the number threshold, the second feature point to be screened in the image to be checked is filtered.

8. The method for extracting cerebrovascular features by image processing according to claim 1, characterized in that: The feature extraction neural network includes a first convolutional neural network, a second convolutional neural network, a third convolutional neural network and a fourth convolutional neural network, and the step of acquiring a cerebrovascular feature map based on the final image, the first processed image, the second processed image, the third processed image and the feature extraction neural network includes: Using the final image as an input value of the first convolutional neural network, using the first processed image as an input value of the second convolutional neural network, using the second processed image as an input value of the third convolutional neural network, and using the third processed image as an input value of the fourth convolutional neural network to obtain a first feature map, a second feature map, a third feature map, and a fourth feature map; Feature fusion is performed on the first feature map, the second feature map, the third feature map, and the fourth feature map to obtain a cerebrovascular feature map.

9. A cerebrovascular feature extraction system for image processing, applied to the cerebrovascular feature extraction method for image processing as claimed in any one of claims 1 to 8, characterized in that: The system comprises: An acquisition module, used for acquiring an initial image set and an angiography image set corresponding to the patient's brain, wherein the initial image set includes a plurality of reference images formed by coaxial rotation, and the angiography image set includes a plurality of standby images formed by coaxial rotation, and the plurality of standby images correspond one to one with the plurality of reference images; An adjustment module, used for performing alignment processing on the standby image and the reference image, and acquiring a plurality of subtraction images based on the standby image and the reference image; A first selection module is used to obtain a plurality of first geometric feature points of the subtracted image, separate the plurality of first geometric feature points into a central feature point and a plurality of first feature points to be screened, and select edge noise points from the plurality of first feature points to be screened based on the central feature point to obtain a plurality of first denoised images including second feature points to be screened; A second selection module is used to select one of the first denoised images as a to-be-checked image, select the remaining first denoised images as reference images, and filter second feature points to be screened in the to-be-checked image based on a rotation angle between the to-be-checked image and the reference image to obtain a final image; An extraction module is used to perform Gaussian filtering, Laplace filtering and Gabor filtering on the final image in sequence to obtain a first processed image, a second processed image and a third processed image respectively, and obtain a cerebral vascular feature map based on the final image, the first processed image, the second processed image, the third processed image and a feature extraction neural network.

10. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method for extracting cerebrovascular features through image processing according to any one of claims 1 to 8 is implemented.

Images