CN114202469B - Method and device for selecting hyper-parameters of Frangi filter, electronic equipment and storage medium - Google Patents
Method and device for selecting hyper-parameters of Frangi filter, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN114202469B CN114202469B CN202111334991.3A CN202111334991A CN114202469B CN 114202469 B CN114202469 B CN 114202469B CN 202111334991 A CN202111334991 A CN 202111334991A CN 114202469 B CN114202469 B CN 114202469B
- Authority
- CN
- China
- Prior art keywords
- voxel
- coordinate axis
- candidate
- target
- initial
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 40
- 210000004204 blood vessel Anatomy 0.000 claims abstract description 39
- 238000004590 computer program Methods 0.000 claims description 11
- 238000010968 computed tomography angiography Methods 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 8
- 239000011159 matrix material Substances 0.000 claims description 8
- 238000000354 decomposition reaction Methods 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 5
- 230000011218 segmentation Effects 0.000 description 6
- 108700041286 delta Proteins 0.000 description 5
- 230000000694 effects Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 230000014509 gene expression Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 230000002792 vascular Effects 0.000 description 4
- 230000008859 change Effects 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000010187 selection method Methods 0.000 description 2
- 239000013598 vector Substances 0.000 description 2
- 238000007792 addition Methods 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000003708 edge detection Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000002601 radiography Methods 0.000 description 1
- 230000006798 recombination Effects 0.000 description 1
- 238000005215 recombination Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10081—Computed x-ray tomography [CT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30101—Blood vessel; Artery; Vein; Vascular
Abstract
The invention discloses a method and a device for selecting hyper-parameters of Frangi filters, electronic equipment and a storage medium, wherein the method comprises the following steps: selecting an initial voxel from a blood vessel region, establishing an initial space, obtaining three characteristic values of each voxel in the initial space, and selecting a target voxel according to the characteristic values; constructing a three-dimensional coordinate system according to the characteristic value of the initial voxel, wherein the three-dimensional coordinate system comprises a first coordinate axis, a second coordinate axis and a third coordinate axis; acquiring a starting point for each coordinate axis in a three-dimensional coordinate system, and acquiring candidate voxels along the positive direction of the coordinate axis; selecting candidate voxels according to the characteristic values of the target voxels, classifying the selected candidate voxels into the target voxels to form an auxiliary space, and acquiring optimal voxels corresponding to the coordinate axis according to the auxiliary space; and based on the first coordinate axis and the second coordinate axis, obtaining the coordinate of the optimal voxel of the third coordinate axis as the hyperparameter of the Frangi filter.
Description
Technical Field
The invention relates to the field of image technical processing, in particular to a method for selecting hyper-parameters of a Frangi filter.
Background
The automatic or semi-automatic segmentation of blood vessels in CT angiography images has been a hot problem, and there are two approaches, namely, conventional algorithm and deep learning, to solve the problem. The blood vessel extraction through deep learning has the advantages of high accuracy, high speed and the like, but a large amount of labeled data is needed in the initial model training process. Therefore, it is important to reconstruct blood vessels efficiently.
The conventional blood vessel reconstruction method mainly comprises three methods, namely segmentation based on the statistical characteristics of an image, segmentation based on the morphological characteristics of the image and region growth. Among them, the region growing method is widely used for various kinds of software because of its simplicity and effectiveness.
In order to improve the quality of vascular reconstruction, a classical frani filter is generally applied to vascular enhancement before the region grows, and the frani filter is also called a frani filter, is an edge detection enhancement filter and has a good effect on vascular enhancement. However, constructing a Frangi filter requires the user to manually specify 3 hyper-parameters, and if the user is not familiar with the theory of Frangi filters, then multiple experiments are required to select the appropriate hyper-parameters. For a complete blood vessel, the characteristics of the blood vessel, such as the thickness, etc., change dramatically with the change of the observed position, which means that the user cannot quickly and accurately select a group of general hyper-parameters, and can only adjust the values of the hyper-parameters many times according to the position to obtain the appropriate hyper-parameters, which undoubtedly adds a small burden to the user.
Disclosure of Invention
The invention provides a method and a device for selecting hyper-parameters of a Frangi filter, electronic equipment and a storage medium, which at least solve the technical problems in the prior art.
The invention provides a Frangi filter hyper-parameter selection method, which comprises the following steps:
selecting an initial voxel from a blood vessel region on a CT angiography image, establishing an initial space with the initial voxel as a center, acquiring three characteristic values of each voxel in the initial space, and selecting a target voxel from the voxels according to the characteristic values to obtain a target voxel set;
determining an origin to construct a three-dimensional coordinate system according to the three characteristic values of the initial voxel, wherein the three-dimensional coordinate system comprises a first coordinate axis, a second coordinate axis and a third coordinate axis;
acquiring a collection starting point of each coordinate axis in a three-dimensional coordinate system, and collecting candidate voxels at intervals of a preset length in the positive direction of the coordinate axis to obtain a candidate voxel set;
selecting a candidate voxel from the candidate voxel set according to the three eigenvalues of the target voxel, comparing any candidate voxel in the candidate voxel set with each target voxel in the target voxel set, selecting the candidate voxel to be classified into the target voxel set if the comparison result of the candidate voxel and any target voxel meets a set condition, forming an auxiliary space taking the initial voxel as a starting point, calculating the auxiliary space corresponding to the selected candidate voxel every time the selected candidate voxel is added in the target voxel set, and polling the candidate voxel set in sequence to obtain a plurality of sequentially arranged auxiliary spaces;
calculating difference values between adjacent auxiliary spaces, wherein each difference value corresponds to two adjacent selected candidate voxels including an ith candidate voxel and an (i + 1) th candidate voxel, and determining the ith candidate voxel corresponding to the maximum difference value as an optimal voxel corresponding to the coordinate axis;
the acquisition starting point of the first coordinate axis is the origin, the acquisition starting point of the second coordinate axis is determined based on the optimal voxel of the first coordinate axis, and the acquisition starting point of the third coordinate axis is determined based on the optimal voxel of the second coordinate axis;
and acquiring the coordinate of the optimal voxel of the third coordinate axis as a hyperparameter of the Frangi filter.
In an implementation, the obtaining three feature values of each voxel in the initial space, and selecting a target voxel from the voxels according to the feature values includes:
calculating a Hessian matrix of each voxel, and performing eigenvalue decomposition on the Hessian matrix to obtain a first initial eigenvalue, a second initial eigenvalue and a third initial eigenvalue of which absolute values are sorted from small to large;
and if the second initial characteristic value or the third initial characteristic value is judged to be less than or equal to 0, marking the corresponding voxel as the target voxel.
In one embodiment, the method further comprises: determining a first target feature value, a second target feature value and a third target feature value of the target voxel based on the first initial feature value, the second initial feature value and the third initial feature value of the target voxel according to the following formulas:
a is a first initial characteristic value, b is a second initial characteristic value, c is a third initial characteristic value, R a Is a first target characteristic value, said R b And T is a third target characteristic value.
In an embodiment, the constructing a three-dimensional coordinate system according to the three characteristic value determination origins of the initial voxels includes:
setting a first target characteristic value of the initial voxel to be n1, a second target characteristic value to be n2 and a third target characteristic value to be n 3;
and constructing a three-dimensional coordinate system with (0, n2, 0) as an origin, n1 as a first coordinate axis length, 1-n2 as a second coordinate axis length and n3 as a third coordinate axis length.
In one embodiment, the method further comprises: obtaining coordinates of candidate voxels, wherein the coordinates comprise a first coordinate, a second coordinate and a third coordinate, the first coordinate corresponds to the first coordinate axis, the second coordinate corresponds to the second coordinate axis, and the third coordinate corresponds to the third coordinate axis;
the set conditions are that a first target characteristic value of the target voxel is larger than a first coordinate of the candidate voxel, a second target characteristic value of the target voxel is smaller than a second coordinate of the candidate voxel, and a third target characteristic value of the target voxel is larger than a third coordinate of the candidate voxel.
Another aspect of the present invention provides a device for selecting hyper-parameters of a Frangi filter, comprising:
the acquisition module is used for selecting an initial voxel from a blood vessel region on a CT angiography image, establishing an initial space with the initial voxel as a center, acquiring three characteristic values of each voxel in the initial space, and selecting a target voxel from the voxels according to the characteristic values to obtain a target voxel set;
the calculation module is used for determining an origin to construct a three-dimensional coordinate system according to the three characteristic values of the initial voxel, and the three-dimensional coordinate system comprises a first coordinate axis, a second coordinate axis and a third coordinate axis;
the calculation module is also used for acquiring a collection starting point of each coordinate axis in the three-dimensional coordinate system and collecting candidate voxels every preset length along the positive direction of the coordinate axis to obtain a candidate voxel set;
the calculation module is further configured to select a candidate voxel from the candidate voxel set according to the three feature values of the target voxel, compare the candidate voxel with each target voxel in the target voxel set for any candidate voxel in the candidate voxel set, select the candidate voxel to be classified into the target voxel set if a comparison result of the candidate voxel and any target voxel meets a set condition, so as to form an auxiliary space relative to the initial space, calculate an auxiliary space corresponding to the selected candidate voxel every time a selected candidate voxel is added to the target voxel set, and sequentially poll the candidate voxel set to obtain a plurality of sequentially arranged auxiliary spaces;
calculating difference values between adjacent auxiliary spaces, wherein each difference value corresponds to two adjacent selected candidate voxels including an ith candidate voxel and an (i + 1) th candidate voxel, and determining the ith candidate voxel corresponding to the maximum difference value as an optimal voxel corresponding to the coordinate axis;
the three-dimensional coordinate system comprises a first coordinate axis, a second coordinate axis and a third coordinate axis, wherein the acquisition starting point of the first coordinate axis is the origin, the acquisition starting point of the second coordinate axis is determined based on the optimal voxel of the first coordinate axis, and the acquisition starting point of the third coordinate axis is determined based on the optimal voxel of the second coordinate axis;
and the processing module is used for acquiring the coordinates of the optimal voxel of the third coordinate axis as the hyperparameter of the Frangi filter.
Yet another aspect of the present invention provides an electronic device, including: the storage stores a computer program executable by the processor, and the processor realizes the above-mentioned hyper-parameter selection method of Frangi filter when executing the computer program.
In another aspect, the present invention provides a storage medium, where a computer program is stored on the storage medium, and when the computer program is read and executed, the method for selecting hyper-parameters of a franli filter is implemented.
Based on the scheme, the method is mainly improved in the superparameter selection process when the Frangi filter enhances the blood vessels, a point is selected from a CT angiography image to serve as an initial voxel, the characteristic value of the minimum unit voxel of a three-dimensional image is utilized to carry out analysis and transformation, a target voxel, namely a part of a potential blood vessel, is obtained, a set condition is established, collected candidate voxels are continuously brought into the range of the target voxel, namely the potential blood vessel, and a final superparameter result is obtained through iteration. Without setting hyper-parameters manually, the selection process is associated with parameters representing the blood vessel characteristics, namely the characteristic values of voxels, so that the effect of self-adaptive collection of local voxels of the blood vessel is realized, and the efficiency of selecting the hyper-parameters is improved under the condition of ensuring the accuracy of selecting the hyper-parameters.
Drawings
Fig. 1 is a schematic flow chart illustrating a method for selecting hyper-parameters of a Frangi filter according to an embodiment of the present invention;
FIG. 2 is a diagram comparing the effect of the method for selecting the hyperparameter of the Frangi filter according to the embodiment of the present invention with the prior art;
fig. 3 is a schematic structural diagram of a hyper-parameter selecting apparatus for a Frangi filter according to an embodiment of the present invention.
Detailed Description
In order to make the objects, features and advantages of the present invention more obvious and understandable, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In order to improve the efficiency of selecting the hyperparameters of the Frangi filter and ensure that the accuracy of selecting the hyperparameters is kept at a standard level, as shown in FIG. 1, an embodiment of the invention provides a method for selecting the hyperparameters of the Frangi filter, which includes:
step 101, selecting an initial voxel from a blood vessel region on a CT angiography image, establishing an initial space with the initial voxel as a center, obtaining three characteristic values of each voxel in the initial space, and selecting a target voxel from the voxels according to the characteristic values to obtain a target voxel set;
the initial space is a three-dimensional region formed by outward expansion with an initial voxel as a center, and under normal conditions, the diameter of a blood vessel is smaller than 4mm, in one example, the initial space can be a cubic region, based on the general diameter of the blood vessel, the side length of the cubic region is preferably greater than or equal to 4mm, the specific range can be adaptively adjusted according to an actual image, and no limitation is made herein; the initial voxel can be selected by the user on the CT angiography image through the human-computer interaction interface acquisition.
Acquiring three eigenvalues of each voxel in an initial space, in one example, calculating a hessian matrix of each voxel, and performing eigenvalue decomposition on the hessian matrix to obtain a first initial eigenvalue, a second initial eigenvalue and a third initial eigenvalue of which absolute values are sorted from small to large;
along with the feature value decomposition, three feature vectors respectively corresponding to the three feature values are derived besides the feature values, and the geometric significance of the three feature values can be further explained by combining the direction indication of the feature vectors, wherein the first initial feature value represents the extension along the trend of the blood vessel, the second initial feature value and the third initial feature value represent the direction along the cross section of the blood vessel, and the second initial feature value and the third initial feature value respectively correspond to the short diameter and the long diameter which are perpendicular to each other on the cross section of the blood vessel because the cross section of the blood vessel is circular under the ideal condition and the cross section of the blood vessel is closer to the ellipse in the actual condition. Therefore, the three characteristic values represent three geometric meanings related to the vascular structure, and a foundation is laid for further utilizing the three characteristic values to carry out operation and comparison.
In an example, the hessian matrix of each voxel is calculated by obtaining an image gray value or an image brightness value or an image saturation value corresponding to each voxel, or other values that enable a characteristic expression to be distinguished between the voxels to be fed back, and the corresponding hessian matrix is calculated based on the values fed back by the characteristic expression.
In an example, the corresponding voxel is labeled as the target voxel by determining that the second initial eigenvalue or the third initial eigenvalue is less than or equal to 0.
When the second initial characteristic value is less than or equal to 0 or the third initial characteristic value is less than or equal to 0, it is indicated that the position of the voxel is brighter than the surrounding positions, and the voxel can be determined to be a part of the potential blood vessel, so that the voxel can be included in the target voxel.
Further, in an example, based on the first initial eigenvalue, the second initial eigenvalue, and the third initial eigenvalue of the target voxel, the first target eigenvalue, the second target eigenvalue, and the third target eigenvalue of the target voxel may be determined as follows:
a is a first initial characteristic value, b is a second initial characteristic value, c is a third initial characteristic value, R a Is a first target characteristic value, R b T is the third target feature value.
The geometric meaning of the first target characteristic value is that the cross section of the blood vessel is judged to be flat or round, the closer the first target characteristic value is to 1, the more round the cross section of the blood vessel is, and the more flat the cross section of the blood vessel is otherwise;
the geometric meaning of the second target characteristic value representation is to judge the extending structure of the blood vessel, the first initial characteristic value is approximately close to 0, the more standard the blood vessel shape is, namely, the closer to a cylinder, the larger the second target characteristic value is, the more occluded the blood vessel is, and the blood vessel does not extend forwards, or the more sphere-like the blood vessel is, the less blood vessel is;
the geometric meaning of the third target characteristic value representation is a unified unit and a standard dimension.
102, determining an origin to construct a three-dimensional coordinate system according to the three characteristic values of the initial voxel, wherein the three-dimensional coordinate system comprises a first coordinate axis, a second coordinate axis and a third coordinate axis;
setting the first target eigenvalue of the initial voxel to be n1, the second target eigenvalue to be n2, and the third target eigenvalue to be n3, in an example, a three-dimensional coordinate system with (0, n2, 0) as an origin, n1 as a first coordinate axis length, 1-n2 as a second coordinate axis length, and n3 as a third coordinate axis length is constructed based on the eigenvalues of the initial voxel. By limiting the space length of the three-dimensional coordinate system on each coordinate axis, the range of the subsequently acquired voxels to be included in the potential blood vessels is reduced, and the efficiency is improved.
103, acquiring an acquisition starting point of each coordinate axis in a three-dimensional coordinate system, and acquiring candidate voxels every preset length in the positive direction of the coordinate axis to obtain a candidate voxel set;
in an example, the candidate voxels are acquired every preset length in the positive direction of the coordinate axis, and the method may be implemented by uniformly dividing the coordinate axis into 100 parts in the positive direction by using a plane perpendicular to the coordinate axis, where each division position is an acquisition position of one candidate voxel.
Starting from the acquisition starting point, the candidate voxels are sequentially acquired in the positive direction to obtain a candidate voxel set, for example, on the current coordinate axis, the coordinate set of the position where the candidate voxel is acquired corresponding to the coordinate axis may be (0, 0.008,0.016,0.024 … 0.792.792).
104, selecting candidate voxels from the candidate voxel set according to the three characteristic values of the target voxel, classifying the selected candidate voxels into the target voxel set to form an auxiliary space taking the initial voxel as a starting point, and acquiring optimal voxels corresponding to the coordinate axis according to the auxiliary space;
for any candidate voxel in the candidate voxel set, comparing the candidate voxel with each target voxel in the target voxel set, and selecting the candidate voxel to be classified into the target voxel set if the comparison result of the candidate voxel and any target voxel meets the set condition;
acquiring coordinates of the candidate voxels, wherein the coordinates comprise a first coordinate, a second coordinate and a third coordinate, the first coordinate corresponds to a first coordinate axis, the second coordinate corresponds to a second coordinate axis, and the third coordinate corresponds to a third coordinate axis;
for example, if the first coordinate axis is an X axis, the second coordinate axis is a Y axis, and the third coordinate axis is a Z axis, the candidate voxel has coordinates (X, Y, Z), where X corresponds to a value between [0, n1], Y corresponds to a value between [ n2, 1], and Z corresponds to a value between [0, n3 ].
In one example, when the candidate voxels are acquired every preset length along the positive direction of the first coordinate axis, the coordinates of the acquired candidate voxels are (x, 0, 0), such as (0.1, 0, 0), (0.2, 0, 0) … (n1, 0, 0);
in an example, when the candidate voxels are acquired every preset length along the positive direction of the second coordinate axis, the coordinates of the acquired candidate voxels are (0, y, 0), such as (0, n2, 0), (0, 0.6, 0) … (0, 1, 0);
in one example, when the candidate voxels are acquired every preset length along the positive direction of the second coordinate axis, the coordinates of the acquired candidate voxels are (0, 0, z), such as (0, 0, 0.2), (0, 0, 0.4) … (0, 0, n 3). The setting conditions are that a first target feature value of the target voxel is larger than a first coordinate of the candidate voxel, a second target feature value of the target voxel is smaller than a second coordinate of the candidate voxel, and a third target feature value of the target voxel is larger than a third coordinate of the candidate voxel.
Obtaining the optimal voxel corresponding to the coordinate axis according to the auxiliary space, including:
every time a selected candidate voxel is added in the target voxel set, calculating an auxiliary space corresponding to the selected candidate voxel, and polling the candidate voxel set in sequence to obtain a plurality of sequentially arranged auxiliary spaces; after the candidate voxels are selected, calculating a corresponding auxiliary space, wherein the auxiliary space is an auxiliary space taking the initial voxels as a starting point;
in an example, the initially established target voxels include a plurality of voxels, assuming that the coordinates of the initial voxels are (x0, y0, z0), taking the first coordinate axis as an example, the corresponding candidate voxel sets are sequentially polled, and finally three candidate voxels are selected to be classified into the target voxels, and the three candidate voxels and the corresponding coordinates thereof are sequentially a first candidate voxel (x1, 0, 0), a second candidate voxel (x2, 0, 0), and a third candidate voxel (x3, 0, 0);
calculating a volume f1 between a first candidate voxel with coordinates (x1, 0, 0) and an initial voxel with coordinates (x0, y0, z0) to serve as a corresponding auxiliary long space after the first candidate voxel is classified into a target voxel;
calculating a volume f2 between a second candidate voxel with the coordinate of (x2, 0, 0) and an initial voxel with the coordinate of (x0, y0, z0) as an auxiliary space corresponding to the second candidate voxel after being classified into the target voxel;
and calculating a volume f3 between the third candidate voxel with the coordinate of (x3, 0, 0) and the initial voxel with the coordinate of (x0, y0, z0) to serve as an auxiliary space corresponding to the third candidate voxel after being classified into the target voxel.
Thus, a set is obtained comprising a plurality of ordered subspaces: (f1, f2, f 3).
On the second coordinate axis and the third coordinate axis, the calculation of the auxiliary space corresponding to the candidate voxels classified into the target voxel is performed with reference to the first coordinate axis.
And calculating difference values between adjacent auxiliary spaces, wherein each difference value corresponds to two adjacent selected candidate voxels, including the ith candidate voxel and the (i + 1) th candidate voxel, and determining the ith candidate voxel corresponding to the maximum difference value as the optimal voxel corresponding to the coordinate axis.
In one example, reference is made to the set of affiliated spaces in the above example: (f1, f2, f3) calculating a difference between adjacent satellite spaces to obtain two differences, Δ 1 ═ f2-f1 and Δ 2 ═ f3-f2, respectively, wherein Δ 1 corresponds to a first candidate voxel and Δ 2 corresponds to a second candidate voxel; and comparing the sizes of the delta 1 and the delta 2, selecting the first candidate voxel as the optimal voxel corresponding to the coordinate axis if the delta 1 is larger than the delta 2, and selecting the second candidate voxel as the optimal voxel corresponding to the coordinate axis if the delta 2 is larger than the delta 1.
Therefore, it can be understood that the volume of the auxiliary space corresponding to the selected candidate voxel is the volume from the initial voxel to the candidate voxel.
In one example, the difference between the adjacent subspaces is summarized by the following formula:
Δf i ＝f i+1 -f i
i is the selected candidate voxel order number, f i Is the volume of said auxiliary space.
As an example, assuming that the first coordinate axis is the first acquisition direction, the acquisition starting point is Q0, candidate voxels are acquired according to the above preset length, the obtained candidate voxel set is SQ { Q0, Q1, Q2, Q3, …, Qn }, for each candidate voxel in the candidate voxel set, the corresponding coordinate including the first coordinate x, the second coordinate y and the third coordinate z is obtained, and the current candidate voxel is compared with any target voxel in the target voxel set, and the following setting conditions are satisfied:
R a >x；
AndR b <y；
And T>z；
classifying the current candidate voxel into a target voxel set, wherein the candidate voxel only needs to be compared with a target voxel in the initially established target voxel set, but does not need to be compared with a selected candidate voxel added later;
wherein R is present in the above-mentioned set conditions a 、R b T, refer to step 101, R a Is a first target characteristic value, R b Obtaining a result of the second target characteristic value and T of the third target characteristic value; and the three comparison expressions in the setting conditions are required to be satisfied at the same time, the meaning of the setting conditions appearing here can be explained by referring to the meaning of the setting conditions in the step 104, that is, the setting conditions appearing here and the setting conditions described in the step 104 are different expressions of the same technical features.
And acquiring candidate voxels from Q0 to Qn in sequence according to the candidate voxel set, wherein sequence numbers 1, 2, 3, 4 and 5 are respectively known according to the sequence number of the candidate voxels selected by the assumption that Q2, Q5, Q9, Q16 and Q30 are the selected candidate voxels. Assuming that i is 1, then f i ＝f 1 ＝f Q2 ,f i+1 ＝f 2 ＝f Q5 (ii) a Assuming that i is 3, then f i ＝f 3 ＝f Q9 ,f i+1 ＝f 4 ＝f Q16 。
Step 105, wherein the acquisition starting point of the first coordinate axis is the origin, the acquisition starting point of the second coordinate axis is determined based on the optimal voxel of the first coordinate axis, and the acquisition starting point of the third coordinate axis is determined based on the optimal voxel of the second coordinate axis;
when the obtaining of the optimal voxel of the first coordinate axis is finished, iteration is started from the step 103, and the optimal voxel of the second coordinate axis is continuously obtained until the optimal voxel of the third coordinate axis is obtained.
Therefore, when the first coordinate axis is acquired, the acquisition route of the candidate voxel is overlapped with the axis of the first coordinate axis;
when the second coordinate axis is acquired, the acquisition route of the candidate voxel is parallel to the axis of the second coordinate axis, and the value of the first coordinate corresponding to the optimal voxel of the first coordinate axis is determined and is used as the distance between the acquisition route of the candidate voxel and the axis of the second coordinate axis; for example, if the optimal voxel on the first coordinate axis is Qk, the acquisition route of the candidate voxel on the second coordinate axis coincides with x ═ Qk;
when the third coordinate axis is acquired, the acquisition route of the candidate voxel is parallel to the axis of the third coordinate axis, and the distance from the optimal voxel of the second coordinate axis to the origin is determined and used as the distance between the acquisition route of the candidate voxel and the axis of the third coordinate axis; for example, if the optimal voxel of the second coordinate axis is Qp, the intersection point of the acquisition route of the candidate voxel of the third coordinate axis with the plane where the first coordinate axis and the second coordinate axis are located is Qp.
In an example, when the candidate voxels are acquired every preset length in the positive direction of the first coordinate axis, the coordinates of the acquired candidate voxels should be (x, 0, 0), and the acquisition starting point is the origin, so that the coordinates of the acquisition starting point is (0, 0, 0), and in the actual acquisition process, the acquisition starting point may be a candidate voxel or may not be a candidate voxel, which is not specifically limited herein.
For example, the specific coordinates of the acquired candidate voxels may be (0.1, 0, 0), (0.2, 0, 0) … (n1, 0, 0), and assuming that the optimal voxel of the first coordinate axis is Qk, the coordinate of Qk is (Qk, 0, 0), the value of the second coordinate axis should be located between [ n2, 1] based on the limitation of the spatial length of the three-dimensional coordinate system on each coordinate axis in step 102, so that the second coordinate of the acquisition start point of the second coordinate axis is n2, and the acquisition start point coordinate of the second coordinate axis is determined to be (Qk, n2, 0) based on the first coordinate of the optimal voxel Qk of the first coordinate axis.
Then, when the candidate voxels are acquired every preset length in the positive direction of the second coordinate axis, the coordinates of the acquired candidate voxels are (Qk, y, 0), y is greater than n2 or y is greater than or equal to n2, and the specific coordinates may be (Qk, n2, 0), (Qk, 0.5, 0) … (Qk, 1, 0);
assuming that the optimal voxel of the second coordinate axis is Qp, the coordinate of Qp is (Qk, Qp, 0), and according to the method for determining the acquisition starting point of the second coordinate axis, the acquisition starting point coordinate of the third coordinate axis is (Qk, Qp, 0) based on the second coordinate of the optimal voxel Qp of the second coordinate axis.
Then, when the candidate voxels are acquired every preset length in the positive direction of the third coordinate axis, the coordinates of the acquired candidate voxels are (Qk, Qp, z), z is greater than 0 or z is greater than or equal to 0, and the specific coordinates may be (Qk, Qp, 0), (Qk, Qp, 0.2) … (Qk, Qp, n 3).
Step 106, obtaining the coordinate of the optimal voxel of the third coordinate axis as the hyperparameter of the Frangi filter;
and assuming that the optimal voxel of the third coordinate axis is Qj, acquiring numerical values of a first coordinate, a second coordinate and a third coordinate of the Qj as values corresponding to three hyper-parameters of the Frangi filter.
According to the method for selecting hyper-parameters of the Frangi filter provided by the embodiment, the selected hyper-parameters are good in use accuracy, in order to explain the effectiveness of the method, for the same CT radiography image, the method and artificial labels are respectively used for processing, a group of hyper-parameters of the Frangi filter are respectively obtained, the two groups of hyper-parameters are utilized for carrying out blood vessel enhancement operation, on the basis, the same steps (such as a region growing method) are used for obtaining a blood vessel segmentation result, and in order to better observe the details of the blood vessel segmentation, the local segmentation result is displayed as comparison.
As shown in FIG. 2, three comparison images of local regions are shown in the figure, the left side of each comparison image is a region for selecting the hyper-parameters by using the artificial labels, and the right side of each comparison image is a region for selecting the hyper-parameters by using the method.
An embodiment of the present invention further provides a device for selecting hyper-parameters of a Frangi filter, as shown in fig. 3, the device includes:
an acquisition module 10, configured to select an initial voxel from a blood vessel region on a CT angiography image, establish an initial space with the initial voxel as a center, acquire three feature values of each voxel in the initial space, and select a target voxel from the voxels according to the feature values to obtain a target voxel set;
a calculating module 20, configured to determine an origin according to the three feature values of the initial voxel to construct a three-dimensional coordinate system, where the three-dimensional coordinate system includes a first coordinate axis, a second coordinate axis, and a third coordinate axis:
the calculation module 20 is further configured to acquire a collection starting point of each coordinate axis in the three-dimensional coordinate system, and collect candidate voxels every preset length in a positive direction of the coordinate axis to obtain a candidate voxel set;
the calculating module 20 is further configured to select a candidate voxel from the candidate voxel set according to the three feature values of the target voxel, compare any candidate voxel in the candidate voxel set with each target voxel in the target voxel set, select the candidate voxel to be classified into the target voxel set if a comparison result of the candidate voxel and any target voxel meets a set condition, form an auxiliary space taking the initial voxel as a starting point, calculate an auxiliary space corresponding to the selected candidate voxel every time one selected candidate voxel is added to the target voxel set, and sequentially poll the candidate voxel set to obtain a plurality of sequentially arranged auxiliary spaces;
calculating difference values between adjacent auxiliary spaces, wherein each difference value corresponds to two adjacent selected candidate voxels, including the ith candidate voxel and the (i + 1) th candidate voxel, and determining the ith candidate voxel corresponding to the maximum difference value as the optimal voxel corresponding to the coordinate axis;
the three-dimensional coordinate system comprises a first coordinate axis, a second coordinate axis and a third coordinate axis, wherein the acquisition starting point of the first coordinate axis is the origin, the acquisition starting point of the second coordinate axis is determined based on the optimal voxel of the first coordinate axis, and the acquisition starting point of the third coordinate axis is determined based on the optimal voxel of the second coordinate axis;
and the processing module 30 acquires the coordinates of the optimal voxel of the third coordinate axis as the hyper-parameter of the Frangi filter.
In addition to the methods and apparatus described above, embodiments of the present application may also be a computer program product comprising computer program instructions that, when executed by a processor, cause the processor to perform the steps in the methods according to the various embodiments of the present application described in the "exemplary methods" section above of this specification.
The computer program product may be written with program code for performing the operations of embodiments of the present application in any combination of one or more programming languages, including an object oriented programming language such as Java, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device, or entirely on the remote computing device or server.
Furthermore, embodiments of the present application may also be a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, cause the processor to perform steps in a method according to various embodiments of the present application described in the "exemplary methods" section above of this specification.
The computer-readable storage medium may take any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may include, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable 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 disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
The basic principles of the present application have been described above with reference to specific embodiments, but it should be noted that advantages, effects, etc. mentioned in the present application are only examples and are not limiting, and the advantages, effects, etc. must not be considered to be possessed by various embodiments of the present application. Furthermore, the foregoing disclosure of specific details is for the purpose of illustration and description and is not intended to be limiting, since the foregoing disclosure is not intended to be exhaustive or to limit the disclosure to the precise details disclosed.
The block diagrams of devices, apparatuses, systems referred to in this application are only given as illustrative examples and are not intended to require or imply that the connections, arrangements, configurations, etc. must be made in the manner shown in the block diagrams. These devices, apparatuses, devices, systems may be connected, arranged, configured in any manner, as will be appreciated by one skilled in the art. Words such as "including," "comprising," "having," and the like are open-ended words that mean "including, but not limited to," and are used interchangeably herein. The words "or" and "as used herein mean, and are used interchangeably with, the word" and/or, "unless the context clearly dictates otherwise. The word "such as" is used herein to mean, and is used interchangeably with, the phrase "such as but not limited to".
It should also be noted that in the devices, apparatuses, and methods of the present application, the components or steps may be decomposed and/or recombined. These decompositions and/or recombinations are to be considered as equivalents of the present application.
The previous description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other aspects without departing from the scope of the application. Thus, the present application is not intended to be limited to the aspects shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
The foregoing description has been presented for purposes of illustration and description. Furthermore, the description is not intended to limit embodiments of the application to the form disclosed herein. While a number of example aspects and embodiments have been discussed above, those of skill in the art will recognize certain variations, modifications, alterations, additions and sub-combinations thereof.
Claims (8)
1. A method for selecting hyper-parameters of a Frangi filter is characterized by comprising the following steps:
selecting an initial voxel from a blood vessel region on a CT angiography image, establishing an initial space with the initial voxel as a center, acquiring three characteristic values of each voxel in the initial space, and selecting a target voxel from the voxels according to the characteristic values to obtain a target voxel set;
determining an origin to construct a three-dimensional coordinate system according to the three characteristic values of the initial voxel, wherein the three-dimensional coordinate system comprises a first coordinate axis, a second coordinate axis and a third coordinate axis;
acquiring a collection starting point of each coordinate axis in a three-dimensional coordinate system, and collecting candidate voxels every preset length in the positive direction of the coordinate axis to obtain a candidate voxel set;
selecting a candidate voxel from the candidate voxel set according to the three eigenvalues of the target voxel, comparing any candidate voxel in the candidate voxel set with each target voxel in the target voxel set, selecting the candidate voxel to be classified into the target voxel set if the comparison result of the candidate voxel and any target voxel meets a set condition, forming an auxiliary space taking the initial voxel as a starting point, calculating the auxiliary space corresponding to the selected candidate voxel every time the selected candidate voxel is added in the target voxel set, and polling the candidate voxel set in sequence to obtain a plurality of sequentially arranged auxiliary spaces;
calculating difference values between adjacent auxiliary spaces, wherein each difference value corresponds to two adjacent selected candidate voxels, including the ith candidate voxel and the (i + 1) th candidate voxel, and determining the ith candidate voxel corresponding to the maximum difference value as the optimal voxel corresponding to the coordinate axis;
the acquisition starting point of the first coordinate axis is the origin, the acquisition starting point of the second coordinate axis is determined based on the optimal voxel of the first coordinate axis, and the acquisition starting point of the third coordinate axis is determined based on the optimal voxel of the second coordinate axis;
and acquiring the coordinate of the optimal voxel of the third coordinate axis as a hyperparameter of the Frangi filter.
2. The method of claim 1, wherein the obtaining three eigenvalues for each voxel in the initial space and selecting a target voxel from the voxels according to the eigenvalues comprises:
calculating a Hessian matrix of each voxel, and performing eigenvalue decomposition on the Hessian matrix to obtain a first initial eigenvalue, a second initial eigenvalue and a third initial eigenvalue of which absolute values are sorted from small to large;
and if the second initial characteristic value or the third initial characteristic value is judged to be less than or equal to 0, marking the corresponding voxel as the target voxel.
3. The method of claim 2, further comprising:
determining a first target feature value, a second target feature value and a third target feature value of the target voxel based on the first initial feature value, the second initial feature value and the third initial feature value of the target voxel according to the following formulas:
a is a first initial characteristic value, b is a second initial characteristic value, c is a third initial characteristic value, R a Is a first target characteristic value, theR b And T is a third target characteristic value.
4. The method of selecting hyper-parameters of a Frangi filter according to claim 3, wherein said determining an origin from three eigenvalues of said initial voxel to construct a three dimensional coordinate system comprises:
setting a first target characteristic value of the initial voxel to be n1, a second target characteristic value to be n2 and a third target characteristic value to be n 3;
and constructing a three-dimensional coordinate system with (0, n2, 0) as an origin, n1 as a first coordinate axis length, 1-n2 as a second coordinate axis length and n3 as a third coordinate axis length.
5. The method of claim 3, further comprising:
obtaining coordinates of the candidate voxels, wherein the coordinates comprise a first coordinate, a second coordinate and a third coordinate, the first coordinate corresponds to the first coordinate axis, the second coordinate corresponds to the second coordinate axis, and the third coordinate corresponds to the third coordinate axis;
the setting condition is that a first target characteristic value of the target voxel is larger than a first coordinate of the candidate voxel, a second target characteristic value of the target voxel is smaller than a second coordinate of the candidate voxel, and a third target characteristic value of the target voxel is larger than a third coordinate of the candidate voxel.
6. A hyper-parametric extractor for a Frangi filter, comprising:
the acquisition module is used for selecting an initial voxel from a blood vessel region on a CT angiography image, establishing an initial space with the initial voxel as a center, acquiring three characteristic values of each voxel in the initial space, and selecting a target voxel from the voxels according to the characteristic values to obtain a target voxel set;
the calculation module is used for determining an origin to construct a three-dimensional coordinate system according to the three characteristic values of the initial voxel, and the three-dimensional coordinate system comprises a first coordinate axis, a second coordinate axis and a third coordinate axis;
the calculation module is further used for acquiring a collection starting point of each coordinate axis in the three-dimensional coordinate system, and collecting candidate voxels every preset length along the positive direction of the coordinate axis to obtain a candidate voxel set;
the calculation module is further configured to select a candidate voxel from the candidate voxel set according to the three feature values of the target voxel, compare the candidate voxel with each target voxel in the target voxel set for any candidate voxel in the candidate voxel set, select the candidate voxel to be classified into the target voxel set if a comparison result of the candidate voxel and any target voxel meets a set condition, form an auxiliary space taking the initial voxel as a starting point, calculate an auxiliary space corresponding to the selected candidate voxel every time one selected candidate voxel is added to the target voxel set, and sequentially poll the candidate voxel set to obtain a plurality of sequentially arranged auxiliary spaces;
calculating difference values between adjacent auxiliary spaces, wherein each difference value corresponds to two adjacent selected candidate voxels, including the ith candidate voxel and the (i + 1) th candidate voxel, and determining the ith candidate voxel corresponding to the maximum difference value as the optimal voxel corresponding to the coordinate axis;
the acquisition starting point of the first coordinate axis is the origin, the acquisition starting point of the second coordinate axis is determined based on the optimal voxel of the first coordinate axis, and the acquisition starting point of the third coordinate axis is determined based on the optimal voxel of the second coordinate axis;
and the processing module is used for acquiring the coordinate of the optimal voxel of the third coordinate axis as the hyperparameter of the Frangi filter.
7. An electronic device, comprising: memory storing a computer program executable by said processor and a processor implementing the method for hyper-parametric selection of a Frangi filter according to any of the preceding claims 1 to 5 when said computer program is executed by said processor.
8. Storage medium, characterized in that it has stored thereon a computer program which, when read and executed, implements the method for the hyper-parameter selection of a Frangi filter according to any of the preceding claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111334991.3A CN114202469B (en) | 2021-11-11 | 2021-11-11 | Method and device for selecting hyper-parameters of Frangi filter, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111334991.3A CN114202469B (en) | 2021-11-11 | 2021-11-11 | Method and device for selecting hyper-parameters of Frangi filter, electronic equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114202469A CN114202469A (en) | 2022-03-18 |
| CN114202469B true CN114202469B (en) | 2022-08-19 |
Family
ID=80647390
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111334991.3A Active CN114202469B (en) | 2021-11-11 | 2021-11-11 | Method and device for selecting hyper-parameters of Frangi filter, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114202469B (en) |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102006046285A1 (en) * | 2006-09-29 | 2009-04-23 | Siemens Ag | Object vessel-like structures segmented representing method for patient, involves using determined probabilities as criterion for presence of vessel in segmentation process for representation of vessel structures |
| CN105701799B (en) * | 2015-12-31 | 2018-10-30 | 东软集团股份有限公司 | Divide pulmonary vascular method and apparatus from lung's mask image |
| EP3352135B1 (en) * | 2017-01-19 | 2019-09-11 | INESC TEC - Instituto de Engenharia de Sistemas e Computadores, Tecnologia e Ciência | Method and apparatus for segmentation of blood vessels |
-
2021
- 2021-11-11 CN CN202111334991.3A patent/CN114202469B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN114202469A (en) | 2022-03-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | A multi-view deep convolutional neural networks for lung nodule segmentation | |
| CN106127819B (en) | The method and device thereof of vessel centerline are extracted in medical image | |
| US7941462B2 (en) | Method and apparatus for classification of coronary artery image data | |
| CN108805871B (en) | Blood vessel image processing method and device, computer equipment and storage medium | |
| JP6539303B2 (en) | Transforming 3D objects to segment objects in 3D medical images | |
| Ulku et al. | Computer aided brain tumor detection with histogram equalization and morphological image processing techniques | |
| JP2015036123A (en) | Medical image processor, medical image processing method and classifier training method | |
| CN110781937B (en) | Point cloud feature extraction method based on global visual angle | |
| Saavedra et al. | Sketch-based 3D Model Retrieval using Keyshapes for Global and Local Representation. | |
| CN112927353A (en) | Three-dimensional scene reconstruction method based on two-dimensional target detection and model alignment, storage medium and terminal | |
| CN106610977B (en) | Data clustering method and device | |
| CN113168729B (en) | 3D shape matching method and device based on local reference coordinate system | |
| CN113095333A (en) | Unsupervised feature point detection method and unsupervised feature point detection device | |
| CN106469138B (en) | Word cloud generation method and device | |
| CN106683105B (en) | Image segmentation method and image segmentation device | |
| CN110268441A (en) | The method for obtaining the 3D model data of multiple components of object | |
| Dong et al. | An improved YOLOv5 network for lung nodule detection | |
| CN114202469B (en) | Method and device for selecting hyper-parameters of Frangi filter, electronic equipment and storage medium | |
| CN111292230B (en) | Spiral transformation data amplification method, system, medium and equipment in deep learning | |
| Rashidi et al. | Point cloud data cleaning and refining for 3D as-built modeling of built infrastructure | |
| Wu et al. | Semiautomatic segmentation of glioma on mobile devices | |
| US8854366B1 (en) | Automated geometric representation of discrete point sets | |
| CN111126617B (en) | Method, device and equipment for selecting fusion model weight parameters | |
| JP6253962B2 (en) | Information processing apparatus and method | |
| CN109308703B (en) | Intelligent contouring of anatomical structures using structured user click-through points |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP03 | Change of name, title or address |
Address after: Room 3011, 2nd Floor, Building A, No. 1092 Jiangnan Road, Nanmingshan Street, Liandu District, Lishui City, Zhejiang Province, 323000 Patentee after: Zhejiang Yizhun Intelligent Technology Co.,Ltd. Address before: No. 1202-1203, 12 / F, block a, Zhizhen building, No. 7, Zhichun Road, Haidian District, Beijing 100083 Patentee before: Beijing Yizhun Intelligent Technology Co.,Ltd. |
|
| CP03 | Change of name, title or address |