CN114708259B - CTA (computed tomography angiography) -based head and neck vascular stenosis detection method, device, equipment and medium - Google Patents
CTA (computed tomography angiography) -based head and neck vascular stenosis detection method, device, equipment and medium Download PDFInfo
- Publication number
- CN114708259B CN114708259B CN202210570301.2A CN202210570301A CN114708259B CN 114708259 B CN114708259 B CN 114708259B CN 202210570301 A CN202210570301 A CN 202210570301A CN 114708259 B CN114708259 B CN 114708259B
- Authority
- CN
- China
- Prior art keywords
- blood vessel
- mask
- section
- point
- central axis
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Geometry (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The application provides a CTA-based head and neck vascular stenosis detection method, which comprises the following steps: acquiring head and neck CTA image data to be predicted, and respectively taking the head and neck CTA image data as the input of a blood vessel segmentation model and a non-blood vessel segmentation model, and acquiring a predicted blood vessel mask output by the blood vessel segmentation model, a predicted skeleton mask output by the non-blood vessel segmentation model and a predicted vascular lesion mask; removing the predicted skeleton mask and the predicted vascular lesion mask from the predicted vascular mask to obtain an actual vascular mask; calculating and correcting according to the predicted blood vessel mask to obtain coordinates of each axis point; classifying and segmenting the blood vessels based on the corrected coordinates of the middle axis points; calculating to obtain the section area of the blood vessel corresponding to each middle axis point based on the actual blood vessel mask corresponding to each section of blood vessel, and obtaining the diameter of the continuous blood vessel with directional continuity; the narrowest part and the largest diameter are determined according to the diameters of the continuous blood vessels, and further, the stenosis degree is determined. The method improves the accuracy of the blood vessel segmentation and the accuracy of the blood vessel stenosis detection.
Description
Technical Field
The application relates to the technical field of computers, in particular to a method, a device, equipment and a medium for detecting head and neck vascular stenosis based on CTA.
Background
The head and neck CTA (including intracranial CTA and/or neck CTA and/or head and neck CTA) is commonly used for cerebrovascular disease examination such as vascular occlusion, deformity or soft plaque, has strong time and space screen resolution, and can carry out reasonable observation on the head and neck vessels in multiple directions. Head and neck vascular analysis requires accurate segmentation of arterial vessels. There are therefore many conventional algorithms, and AI algorithms have also recently emerged.
However, the detection and analysis of the stenosis of the blood vessel are not accurate because the blood vessel is directly segmented based on AI under the influence of the skull, the concentration of the contrast agent, and the vascular lesion.
Disclosure of Invention
Based on the above, the application provides a method, an apparatus, a computing device and a storage medium for detecting head and neck vascular stenosis based on CTA, which realizes accurate segmentation of blood vessels and accurate detection of vascular stenosis.
In order to achieve the above object, a first aspect of the present application provides a CTA-based method for detecting head and neck vascular stenosis, comprising:
acquiring head and neck CTA image data to be predicted;
taking the head and neck CTA image data as the input of a blood vessel segmentation model, and obtaining a predicted blood vessel mask output by the blood vessel segmentation model, wherein the predicted blood vessel mask comprises a blood vessel mask and a blood vessel pathological change mask;
taking the head and neck CTA image data as the input of a non-vascular segmentation model, and obtaining a predicted bone mask and a predicted vascular lesion mask output by the non-vascular segmentation model;
removing the predicted bone mask and the predicted vascular lesion mask from the predicted vascular mask to obtain an actual vascular mask;
calculating to obtain coordinates of each central axis point on the central axis of the blood vessel according to the predicted blood vessel mask, and performing fitting correction based on each central axis point coordinate to obtain corrected coordinates of each central axis point;
classifying the blood vessels based on the corrected coordinates of the central axis points, wherein each class corresponds to one blood vessel;
for each blood vessel, extracting a corresponding blood vessel section mask along the central axis of the blood vessel, and segmenting the blood vessel based on the extracted number of connected domains to obtain a blood vessel segmentation result;
for each section of blood vessel, calculating to obtain the section area of the blood vessel corresponding to each middle axis point based on the actual blood vessel mask corresponding to each section of blood vessel, and obtaining the diameters of all blood vessels with directional continuity;
taking the coordinate point with the minimum diameter as the narrowest part of each section of blood vessel, and taking the coordinate point with the maximum diameter as the maximum diameter of each section of blood vessel;
and calculating the stenosis degree of each section of the blood vessel according to the narrowest part and the maximum diameter.
To achieve the above object, a second aspect of the present application provides a CTA-based head and neck vascular stenosis detecting apparatus, comprising:
the acquisition module is used for acquiring head and neck CTA image data to be predicted;
the first prediction module is used for taking the head and neck CTA image data as the input of a blood vessel segmentation model and obtaining a predicted blood vessel mask output by the blood vessel segmentation model, wherein the predicted blood vessel mask comprises a blood vessel mask and a blood vessel pathological change mask;
the second prediction module is used for taking the head-neck CTA image data as the input of a non-vascular segmentation model, and acquiring a predicted bone mask and a predicted vascular lesion mask output by the non-vascular segmentation model;
a removing module for removing the predicted bone mask and the predicted vascular lesion mask from the predicted vascular mask to obtain an actual vascular mask;
the correction module is used for calculating to obtain coordinates of each middle axis point on a central axis of the blood vessel according to the predicted blood vessel mask, and performing fitting correction based on each middle axis point coordinate to obtain corrected coordinates of each middle axis point;
the classification module is used for classifying blood vessels based on the corrected coordinates of the middle axis points, and each class corresponds to one blood vessel;
the segmentation module is used for extracting a corresponding blood vessel section mask along the central axis of each blood vessel and segmenting the blood vessels based on the extracted number of connected domains to obtain a blood vessel segmentation result;
the first calculation module is used for calculating and obtaining the section area of the blood vessel corresponding to each middle axis point based on the actual blood vessel mask corresponding to each section of the blood vessel so as to obtain the diameters of all the blood vessels with direction continuity;
and the second calculation module is used for taking the coordinate point with the minimum diameter as the narrowest part of each section of blood vessel, taking the coordinate point with the maximum diameter as the maximum diameter of each section of blood vessel, and calculating the stenosis degree of each section of blood vessel according to the narrowest part and the maximum diameter.
To achieve the above object, a third aspect of the present application provides a computer-based device, comprising a memory and a processor, the memory storing a computer program, the computer program, when executed by the processor, causing the processor to perform the steps of:
acquiring head and neck CTA image data to be predicted;
taking the head and neck CTA image data as the input of a blood vessel segmentation model, and obtaining a predicted blood vessel mask output by the blood vessel segmentation model, wherein the predicted blood vessel mask comprises a blood vessel mask and a blood vessel pathological change mask;
taking the head and neck CTA image data as the input of a non-vascular segmentation model, and obtaining a predicted bone mask and a predicted vascular lesion mask output by the non-vascular segmentation model;
removing the predicted bone mask and the predicted vascular lesion mask from the predicted vascular mask to obtain an actual vascular mask;
calculating to obtain coordinates of each central axis point on the central axis of the blood vessel according to the predicted blood vessel mask, and performing fitting correction based on each central axis point coordinate to obtain corrected coordinates of each central axis point;
classifying the blood vessels based on the corrected coordinates of the central axis points, wherein each class corresponds to one blood vessel;
for each blood vessel, extracting a corresponding blood vessel section mask along the central axis of the blood vessel, and segmenting the blood vessel based on the extracted number of connected domains to obtain a blood vessel segmentation result;
for each section of blood vessel, calculating to obtain the section area of the blood vessel corresponding to each middle axis point based on the actual blood vessel mask corresponding to each section of blood vessel, and obtaining the diameters of all blood vessels with directional continuity;
taking the coordinate point with the minimum diameter as the narrowest part of each section of blood vessel, and taking the coordinate point with the maximum diameter as the maximum diameter of each section of blood vessel;
and calculating the stenosis degree of each section of the blood vessel according to the narrowest part and the maximum diameter.
To achieve the above object, a fourth aspect of the present application provides a computer-readable storage medium comprising: a computer program is stored which, when executed by a processor, causes the processor to perform the steps of:
acquiring head and neck CTA image data to be predicted;
taking the head-neck CTA image data as the input of a blood vessel segmentation model, and acquiring a predicted blood vessel mask output by the blood vessel segmentation model, wherein the predicted blood vessel mask comprises a blood vessel mask and a blood vessel pathological change mask;
taking the head and neck CTA image data as the input of a non-vascular segmentation model, and obtaining a predicted bone mask and a predicted vascular lesion mask output by the non-vascular segmentation model;
removing the predicted bone mask and the predicted vascular lesion mask from the predicted vascular mask to obtain an actual vascular mask;
calculating to obtain coordinates of each central axis point on the central axis of the blood vessel according to the predicted blood vessel mask, and performing fitting correction based on each central axis point coordinate to obtain corrected coordinates of each central axis point;
classifying the blood vessels based on the corrected coordinates of the central axis points, wherein each class corresponds to one blood vessel;
for each blood vessel, extracting a corresponding blood vessel section mask along the central axis of the blood vessel, and segmenting the blood vessel based on the extracted number of connected domains to obtain a blood vessel segmentation result;
for each section of blood vessel, calculating to obtain the section area of the blood vessel corresponding to each middle axis point based on the actual blood vessel mask corresponding to each section of blood vessel, and obtaining the diameters of all blood vessels with directional continuity;
taking the coordinate point with the minimum diameter as the narrowest part of each section of blood vessel, and taking the coordinate point with the maximum diameter as the maximum diameter of each section of blood vessel;
and calculating the stenosis degree of each section of the blood vessel according to the narrowest part and the maximum diameter.
Firstly, head and neck CTA image data are respectively used as the input of a blood vessel segmentation model and a non-blood vessel segmentation model, wherein the blood vessel segmentation model is used for outputting a predicted blood vessel mask comprising a blood vessel pathological change mask, and the non-blood vessel segmentation model is used for outputting a predicted bone mask and a predicted blood vessel pathological change mask; then, the predicted skeleton mask and the predicted vascular lesion mask are removed from the predicted vascular mask to obtain an actual vascular mask, so that the influence of vascular lesions and bones can be eliminated, an accurate vascular mask is extracted, and the accurate vascular stenosis detection is facilitated. In addition, in order to make the learning of the blood vessel segmentation model a more regular and smooth segmentation task and reduce the difficulty of learning the blood vessel segmentation model, the blood vessel mask segmentation learned by the blood vessel segmentation model comprises a blood vessel pathological change mask.
Furthermore, in order to ensure that the correct blood vessel trend is extracted, the blood vessel central axis is extracted based on the predicted blood vessel mask when the blood vessel central axis is extracted, and compared with a mode of directly extracting the central axis based on the actual blood vessel mask, the mode can more accurately extract the central axis of the blood vessel, because the blood flow of the actual blood vessel is in a narrow or completely blocked state due to pathological changes, and the continuity and size rules of the blood vessel on the image are greatly deteriorated only by looking at the actual blood vessel mask.
In order to obtain a more accurate central axis, fitting and correcting each central axis point on the extracted central axis of the blood vessel to obtain corrected coordinates of each central axis point, so that the more accurate central axis is obtained;
in order to accurately analyze the vascular stenosis, the blood vessels are classified, segmented and analyzed on the basis of each segment of blood vessel, so that the stenosis can be accurately detected.
In addition, for each section of blood vessel, in order to analyze the blood vessel stenosis more accurately, the diameters of all blood vessels with directional continuity are firstly calculated, then the narrowest part and the maximum diameter are analyzed according to the diameters of the blood vessels with directional continuity, and then the stenosis degree of each section of blood vessel is calculated according to the narrowest part and the maximum diameter. The method realizes accurate segmentation of the blood vessel and accurate detection of the vascular stenosis.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Wherein:
FIG. 1 is a flow chart of a CTA based method for detecting stenosis in a head and neck vessel in one embodiment;
FIG. 2 is a flow diagram of a method of training data acquisition in one embodiment;
FIG. 3 is a block diagram of a CTA-based stenosis detection apparatus for a head and neck blood vessel in an embodiment;
FIG. 4 is a diagram illustrating an internal structure of a computer device according to an embodiment.
Detailed Description
It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, 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 application.
It is noted that the terms "comprises," "comprising," and "having" and any variations thereof in the description and claims of this application and the drawings described above are intended to cover non-exclusive inclusions. For example, a process, method, system, article, or apparatus that comprises a list of steps or elements is not limited to only those steps or elements listed, but may alternatively include other steps or elements not listed, or inherent to such process, method, article, or apparatus. In the claims, the description and the drawings of the specification of the present application, relational terms such as "first" and "second", and the like, may be used solely to distinguish one entity/action/object from another entity/action/object without necessarily requiring or implying any actual such relationship or order between such entities/actions/objects.
Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is explicitly and implicitly understood by one skilled in the art that the embodiments described herein can be combined with other embodiments.
As shown in fig. 1, a CTA-based method for detecting head and neck vascular stenosis is proposed, which comprises:
Among them, CTA (CT angiography) is obtained by scanning with peripheral intravenous injection contrast agent and is commonly used for diagnosing vascular diseases based on three-dimensional imaging. The head and neck CTA image data may be head CTA data, neck CTA data, or both head and neck CTA data.
And 104, taking the head and neck CTA image data as the input of a blood vessel segmentation model, and obtaining a predicted blood vessel mask output by the blood vessel segmentation model, wherein the blood vessel segmentation model is obtained by training based on blood vessels and blood vessel pathological change masks as labels, and the predicted blood vessel mask comprises the blood vessel mask and the blood vessel pathological change mask.
The blood vessel segmentation model is used for identifying and outputting a predicted blood vessel mask according to head and neck CTA image data, the blood vessel segmentation model is marked by a blood vessel and a blood vessel pathological change mask during training, and the predicted blood vessel mask obtained by the obtained blood vessel segmentation model during prediction also comprises the blood vessel mask and the blood vessel pathological change mask.
And step 106, taking the head and neck CTA image data as the input of the non-vascular segmentation model, and obtaining a predicted bone mask and a predicted vascular lesion mask output by the non-vascular segmentation model.
The non-vascular segmentation model is used for identifying and obtaining a predicted bone mask and a predicted vascular lesion mask according to head and neck CTA image data. During the training of the non-vascular segmentation model, the labeling of the skeleton mask and the labeling of the vascular lesion mask are respectively obtained by training as expected outputs. The reason why the individual bone segmentation is provided here is that in the CTA image, blood vessels and bones are highlighted, and the false positive segmentation result can be avoided by providing the individual bone segmentation.
And step 108, removing the predicted skeleton mask and the predicted vascular lesion mask from the predicted vascular mask to obtain an actual vascular mask.
Wherein the predicted bone mask and the predicted vascular lesion mask are removed from the predicted vascular mask in order to obtain an accurate actual vascular mask. The more accurate blood vessel mask can be obtained by removing the mask for predicting the blood vessel pathological changes, and then the segmentation result of false positive caused by adhesion of blood vessels and bones can be avoided by removing the mask for predicting the bones.
And 110, calculating to obtain coordinates of each central axis point on the central axis of the blood vessel according to the predicted blood vessel mask, and performing fitting correction based on each central axis point coordinate to obtain corrected coordinates of each central axis point.
In order to extract an accurate central axis, the central axis is extracted based on a predicted blood vessel mask instead of an actual blood vessel mask, because the blood flow of an actual blood vessel is in a narrow or completely blocked state due to a lesion, and the continuity and the size rule of the blood vessel on an image are greatly deteriorated only by looking at the actual blood vessel mask. Therefore, in order to extract a more accurate central axis point, the central axis point is extracted based on a predicted blood vessel mask, the predicted blood vessel mask comprises a blood vessel pathological change part, and when the central axis is extracted, an inaccurate central axis cannot be extracted due to the fact that pathological changes exist in blood vessels.
Based on the obtained predicted blood vessel mask, an initial central axis of the blood vessel is firstly extracted, the initial central axis is composed of central axis point coordinates, and in order to enable the extracted central axis to be more accurate, ellipse fitting correction is carried out on each central axis point coordinate on the basis of the initial central axis, so that each corrected central axis point coordinate is obtained. This results in a corrected, relatively accurate central axis.
And step 112, classifying the blood vessels based on the corrected coordinates of the central axis points, wherein each class corresponds to one blood vessel.
The classification means that spatially separated blood vessels are classified into different classes, and each class corresponds to one blood vessel. In one embodiment, based on the corrected axial point coordinates of the blood vessel, taking the axial point coordinates of the bottommost layer as an initial point p1, and calculating a point which is closest to the initial point in three-dimensional distance to serve as a related point p 2; then, taking P2 as an initial point, finding a point P3 which is the closest to P2 in three-dimensional distance, and so on, and classifying the points with the association relationship into a class P1. Then, for the central axis point coordinates which are not included in the A1, the initial points are re-confirmed to perform the calculation of the association points, the obtained points with association relations are classified as A2, and then the classification calculation is continued on the points which are not included in the A1 and the A2 until all the central axis point coordinates are classified.
And step 114, extracting a corresponding blood vessel section mask along the central axis of each blood vessel, and segmenting the blood vessels based on the extracted connected domain quantity to obtain a blood vessel segmentation result.
For the subsequent stenosis detection, a blood vessel is segmented, that is, the part where the branch exists is divided into different segments. And respectively extracting the blood vessel section masks corresponding to each axis point along the central axis of the blood vessel, calculating the number of connected domains in the mask region of the blood vessel section, and when the number of the connected domains of the plurality of blood vessel section masks continuously extracted along the central axis direction of the blood vessel is more than or equal to 2, regarding the continuous region as a segmented region of the blood vessel. Assuming that a region where a plurality of connected domains are not formed before the continuous region is the nth segment of blood vessel, n +1 and n +2 segments of blood vessels are obtained after 2 connected domains are formed.
And step 116, calculating to obtain the section area of the blood vessel corresponding to each middle axis point based on the actual blood vessel mask corresponding to each section of the blood vessel, and calculating to obtain the diameter of the continuous blood vessel with directional continuity based on the section area of the blood vessel corresponding to each middle axis point.
After the blood vessels are segmented, calculating the section area of the blood vessel corresponding to each middle axis point according to the corresponding actual blood vessel mask for each section of the blood vessel, then calculating the corresponding diameter according to the section area of each blood vessel, and obtaining the diameter of the continuous blood vessel with directional continuity according to the calculated diameter corresponding to each middle axis point, wherein the diameter of the continuous blood vessel is obtained by arranging the diameters corresponding to a plurality of middle axis points according to the direction of the blood vessel.
And step 118, taking the coordinate point with the smallest diameter as the narrowest part of each section of blood vessel, and taking the coordinate point with the largest diameter as the largest diameter of each section of blood vessel.
Wherein for each segment of the vessel the smallest diameter and the largest diameter are found.
And step 120, calculating the stenosis degree of each section of the blood vessel according to the narrowest part and the maximum diameter.
In one embodiment, the ratio of the diameter of the narrowest point to the maximum diameter is taken as the degree of stenosis of the vessel at that point. In another embodiment, the degree of stenosis is measured as the ratio of the maximum diameter minus the diameter at the narrowest point and then the maximum diameter.
Firstly, respectively inputting head and neck CTA image data as a blood vessel segmentation model and a non-blood vessel segmentation model, wherein the blood vessel segmentation model is used for outputting a predicted blood vessel mask comprising a blood vessel pathological change mask, and the non-blood vessel segmentation model is used for outputting a predicted bone mask and a predicted blood vessel pathological change mask; then, the predicted skeleton mask and the predicted vascular lesion mask are removed from the predicted vascular mask to obtain an actual vascular mask, so that the influence of vascular lesions and bones can be eliminated, an accurate vascular mask is extracted, and the accurate vascular stenosis detection is facilitated. In addition, in order to make the learning of the blood vessel segmentation model a more regular and smooth segmentation task and reduce the difficulty of learning the blood vessel segmentation model, the blood vessel mask segmentation learned by the blood vessel segmentation model comprises a blood vessel pathological change mask.
Furthermore, in order to ensure that the correct blood vessel trend is extracted, the blood vessel central axis is extracted based on the predicted blood vessel mask when the blood vessel central axis is extracted, and compared with a mode of directly extracting the central axis based on an actual blood vessel mask, the mode can more accurately extract the blood vessel central axis, because if a lesion exists in a blood vessel, abnormality is easy to occur when the blood vessel central axis is extracted based on the actual blood vessel mask.
In order to obtain a more accurate central axis, fitting and correcting each central axis point on the extracted central axis of the blood vessel to obtain corrected coordinates of each central axis point, so that the more accurate central axis is obtained;
in order to accurately analyze the vascular stenosis, the blood vessels are classified, segmented and analyzed based on each segment of blood vessel, so that the stenosis can be accurately detected.
In addition, for each section of blood vessel, in order to analyze the blood vessel stenosis more accurately, the diameters of all blood vessels with directional continuity are firstly calculated, then the narrowest part and the maximum diameter are analyzed according to the diameters of the blood vessels with directional continuity, and then the stenosis degree of each section of blood vessel is calculated according to the narrowest part and the maximum diameter. The method realizes accurate segmentation of the blood vessel and accurate detection of the vascular stenosis.
As shown in fig. 2, in one embodiment, the training data of the vessel segmentation model and the non-vessel segmentation model are obtained as follows:
the head and neck CTA image data (CT blood vessel imaging) is manually marked by a professional doctor to respectively obtain the MASKs of the blood vessel MASK _ A, the blood vessel lesion MASK _ B and the skeleton MASK _ C and the MASK MASK _ D of the blood vessel and the blood vessel lesion. Vascular + vasculopathy masks are meant to include vascular masks and vasculopathy masks.
Some blood vessels are in a state of stenosis or complete occlusion due to lesion, and if only the blood vessel MASK _ a is learned, the continuity and the size regularity of the blood vessel on the image are much deteriorated. In order to enable a subsequent blood vessel segmentation model to learn a more regular and smooth segmentation task and reduce the difficulty of model learning, a mask of a blood vessel and a blood vessel pathological change is taken as a learning object (corresponding to the blood vessel segmentation model), and then a blood vessel pathological change part and a bone part are independently learned (a non-blood vessel segmentation model).
For the blood vessel segmentation model, in order to make the data for learning and training more accurate and regular, the mask of blood vessel + vascular lesion needs to be processed, and the blood vessel segmentation model is trained based on the processed mask of blood vessel + vascular lesion, and the specific processing mode is as follows.
And step 204, extracting a central axis point of the central axis based on the labeling MASK _ D of the blood vessel and the blood vessel pathological change MASK, and fitting the extracted central axis point to obtain a corrected blood vessel central axis point coordinate.
The central axis point of the central axis is extracted more accurately based on the blood vessel and the vascular lesion mask, and then the extracted central axis point is subjected to fitting correction, so that more accurate central axis point coordinates can be obtained. The fitting may be an ellipse fitting, but of course, other fitting methods may be used.
And step 206, classifying the blood vessels based on the corrected coordinates of the axis points in the blood vessels, wherein each class corresponds to one blood vessel.
The classification means classifying the spatially separated blood vessels into different classes, and each class corresponds to one blood vessel.
And step 208, aiming at each blood vessel, determining a blood vessel section MASK and a blood vessel pathological change section MASK corresponding to each middle axis on the blood vessel middle axis based on the labeling MASK _ A of the blood vessel MASK and the labeling MASK _ B of the blood vessel pathological change MASK.
The labeling MASK _ A based on the blood vessel MASK can extract a blood vessel section MASK corresponding to each central axis point; the labeling MASK _ B based on the blood vessel lesion MASK can extract the blood vessel lesion cross section MASK (if any) corresponding to each central axis point, the blood vessel lesion cross section MASK can be empty, and the corresponding blood vessel lesion cross section MASK only exists in the central axis point where the lesion exists, that is, not every central axis point can correspond to the blood vessel lesion MASK.
Wherein, knowing the blood vessel cross-section mask and the blood vessel pathological change cross-section mask corresponding to each central axis point, it should be noted that the blood vessel pathological change cross-section mask can be empty, and only the place where pathological change exists will have the corresponding blood vessel pathological change cross-section mask, so as to determine the target training blood vessel cross-section mask corresponding to the central axis point, and the target training blood vessel cross-section mask is obtained according to the combined action of the blood vessel cross-section mask and the blood vessel pathological change cross-section mask (for example, taking the union of the two). And obtaining the target training blood vessel mask according to the sequential position relation of the target training blood vessel section masks corresponding to the central axis points.
The target training blood vessel mask obtained by processing is used as expected output to train the blood vessel segmentation model, and the blood vessel segmentation model obtained in the way can extract a relatively complete and regular prediction blood vessel mask.
The training data of the blood vessel segmentation model is not the blood vessel and the blood vessel pathological change mask which are marked manually, but the blood vessel and the blood vessel pathological change mask which are marked manually are further processed, and the complete and regular target training blood vessel mask can be extracted through processing, so that the learning of the blood vessel segmentation model is more regular and smooth in segmentation task, and the model learning difficulty is greatly reduced.
In one embodiment, the determining, for each blood vessel, a blood vessel cross-section MASK and a blood vessel lesion cross-section MASK corresponding to each central axis on a central axis of the blood vessel based on the labeling MASK _ a of the blood vessel MASK and the labeling MASK _ B of the blood vessel lesion MASK includes: aiming at each blood vessel, aiming at any central axis point coordinate P on the central axis of the blood vessel n Obtaining the coordinate P of the previous middle axis point n-1 And the coordinate P of the latter middle axis point n+1 Calculating the coordinates of the former middle axis point and the latter middle axis point to obtain the coordinate P of the middle axis point n Corresponding blood vessel direction vectors and normal planes; by said central axis point coordinate P n Obtaining the coordinate P corresponding to the central axis point from the labeling MASK _ A of the blood vessel MASK and the labeling MASK _ B of the blood vessel pathological change MASK by the corresponding blood vessel direction vector and the normal plane n The blood vessel section mask and the blood vessel pathological change section mask.
Wherein, a complete blood vessel corresponds to a group of medial axis coordinates P; within the coordinate set P, for any point P n Passing through its former medial axis point coordinate P n-1 And afterA central axis point coordinate P n+1 And calculating to obtain the corresponding central axis point coordinate P n The vessel direction vector and the normal plane. By P n The coordinate, direction vector and normal vector of (1) can be obtained from MASK _ A and MASK _ B to obtain the coordinate P corresponding to the central axis point n The blood vessel cross section mask yA and the blood vessel lesion cross section mask yB. Compared with the traditional mode, the blood vessel section mask and the blood vessel pathological section mask corresponding to each central axis point are determined, so that the blood vessel section mask and the blood vessel pathological section mask can be extracted more accurately.
In one embodiment, the generating a complete and regular target training blood vessel mask corresponding to the whole blood vessel based on the blood vessel section mask and the blood vessel lesion section mask corresponding to each central axis point includes: vascular masks and vascular lesion masks comprising:
taking the central axis point coordinate corresponding to the blood vessel section mask with the largest area as a starting point, and taking the blood vessel section mask with the largest area as a standard section mask;
obtaining a blood vessel section mask and a blood vessel pathological change section mask corresponding to the coordinates of the adjacent central axis points of the starting point;
taking a union set of the blood vessel section mask and the blood vessel pathological change section mask corresponding to the coordinates of the adjacent central axis points, and taking a projection overlapping area of the union set and the standard section mask in the central axis direction as a target blood vessel section mask of the adjacent central axis points, wherein the target blood vessel section mask comprises: a blood vessel section mask and a blood vessel pathological change section mask;
and taking the target blood vessel section mask corresponding to the adjacent middle axis points as a new standard section mask, taking the middle axis point coordinates corresponding to the new standard section mask as a new initial point, and entering the step of extracting the blood vessel section mask and the blood vessel lesion section mask of the adjacent middle axis point coordinates of the initial point until the target blood vessel section mask corresponding to each middle axis point of the whole blood vessel is extracted and obtained, wherein the target blood vessel section mask corresponding to each middle axis point of the whole blood vessel forms the target training blood vessel mask corresponding to the whole blood vessel.
After the blood vessel section mask corresponding to each central axis point is known, the blood vessel section mask with the largest area is obtained, the central axis point coordinate corresponding to the blood vessel section mask with the largest area is taken as a starting point, then the blood vessel section mask with the largest area is taken as a standard section mask, generally speaking, the position of the central axis point corresponding to the blood vessel section mask with the largest area does not have a blood vessel pathological section mask, and therefore the blood vessel section mask with the largest area can be directly taken as the standard section mask.
Then, a blood vessel section mask yA and a blood vessel pathological change section mask yB (if any) corresponding to the coordinates of the adjacent central axis points of the starting point are obtained, then a union set of the blood vessel section mask yA and the blood vessel pathological change section mask yB is taken, and then the projection overlapping area of the union set and the standard section mask in the central axis direction is used as the target blood vessel section mask of the adjacent central axis points. By analogy, the target blood vessel section mask corresponding to each middle axis point is obtained after the cycle iteration is completed. It should be noted that the blood vessel lesion cross section mask corresponding to the central axis point may be empty, that is, there is no corresponding blood vessel lesion cross section mask. By the method, the regular and complete target training blood vessel mask is extracted, so that a follow-up model can learn a more regular and smooth segmentation task, and the model learning difficulty is reduced.
In one embodiment, the method further comprises: when the stenosis degree is larger than a preset threshold value, determining that a stenosed blood vessel exists; calculating the nearest distance between the narrowest part of the blood vessel and the blood vessel lesion mask; and if the nearest distance is smaller than the preset distance, the vascular lesion is taken as a stenosis reason.
Wherein, in one embodiment, the blood vessel with the stenosis rate less than 70% is regarded as the blood vessel with the stenosis, the nearest distance between the narrowest part of the blood vessel and the lesion mask of the blood vessel is calculated, and if the distance is less than 10 pixels, the lesion is regarded as the cause of the stenosis. Whether the vascular lesion is the cause of the stenosis can be determined by calculating the closest distance between the vascular lesion mask and the narrowest part, so that the efficiency and the accuracy of the vascular stenosis detection are greatly improved.
In one embodiment, the calculating to obtain coordinates of each medial axis point on a central axis of the blood vessel according to the predicted blood vessel mask, and performing fitting correction based on each medial axis point coordinate to obtain corrected coordinates of each medial axis point includes: calculating to obtain an initial central axis of the blood vessel by adopting a three-dimensional skeleton extraction method based on the predicted blood vessel mask; traversing the coordinates of a middle axis point on the initial central axis from the bottom of the cephalic neck to the brain along the Z-axis direction of the CTA image; aiming at each central axis point coordinate, taking the central axis point coordinate as a center, extracting a predicted blood vessel mask within a preset range, and obtaining a local blood vessel mask corresponding to the central axis point coordinate; and extracting contour points of the local blood vessel mask, performing ellipse fitting based on the extracted contour points, and fitting the center axis point coordinates to center point coordinates of an ellipse to obtain corrected center axis point coordinates.
The three-dimensional skeleton extraction method is used for extracting the coordinates of the center axis points to the initial center axis, and ellipse fitting correction is carried out on the coordinates of each center axis point, so that the corrected center axis point coordinates are more accurate.
In one embodiment, the classifying the blood vessels based on the corrected coordinates of the medial axis points, each class corresponding to one blood vessel, includes: obtaining the coordinates of the central axis point at the bottommost layer in the blood vessel as a reference point; calculating a point closest to the three-dimensional distance of the reference point, taking the point as a correlation point of the reference point, and classifying the reference point and the correlation point into one class; updating the associated point of the reference point to a new reference point, entering the step of calculating the point with the three-dimensional closest distance to the reference point as the associated point of the reference point until the last point of the blood vessel is iterated, and classifying all the iterated points into one class; and for the central axis points which are not iterated, obtaining the central axis point at the bottommost layer in the central axis points which are not iterated, taking the central axis point as a reference point, and entering the step of calculating the point with the three-dimensional closest distance to the reference point as the associated point of the reference point until all the central axis points are classified.
In order to classify the blood vessels, a point-by-point clustering mode is provided, the same class is classified into one blood vessel, and segmentation is conveniently carried out on the basis of a classification result.
In an embodiment, for each blood vessel, extracting a corresponding blood vessel cross-section mask along a central axis of the blood vessel, and segmenting the blood vessel based on the extracted number of connected components to obtain a blood vessel segmentation result, including: sequentially extracting a plurality of blood vessel section masks along the central axis direction of a blood vessel; determining the number of connected domains contained in each blood vessel section mask; and if the number of the connected domains contained in the plurality of blood vessel section masks is more than or equal to 2, performing blood vessel segmentation on the continuous region forming the plurality of connected domains.
And determining whether to segment the blood vessel according to the number of connected domains in the blood vessel section mask. Specifically, the number of connected domains in the binary blood vessel cross-section mask region is calculated, when the number of three blood vessel cross-section mask connected domains extracted continuously along the central axis direction of the blood vessel is greater than or equal to 2, the continuous region is regarded as a segmented region of the blood vessel, the region without a plurality of connected domains is the nth segment of the blood vessel, and the plurality of connected domains are formed into the nth +1 and the nth +2 segments of the blood vessel respectively.
In an embodiment, for each segment of blood vessel, calculating a blood vessel cross-sectional area corresponding to each central axis point based on an actual blood vessel mask corresponding to each segment of blood vessel, and calculating a continuous blood vessel diameter with directional continuity based on the blood vessel cross-sectional area corresponding to each central axis point, includes: calculating to obtain the section area of the blood vessel corresponding to each middle axis point based on the actual blood vessel mask; taking the diameter of the approximate circle which is equal to the cross-sectional area of the blood vessel as the diameter of the blood vessel corresponding to the cross section of the corresponding blood vessel; obtaining the diameters of continuous blood vessels with directional continuity according to the sequence of the axial points in the blood vessels; the method further comprises the following steps: and calculating the diameter change gradient of the diameters of the continuous blood vessels, and removing abnormal data with the front and rear gradient changes larger than a preset gradient.
Extracting a blood vessel section mask corresponding to each central axis point based on the actual blood vessel mask, calculating to obtain the blood vessel section area of the blood vessel section mask, then taking the diameter of an approximate circle which is equal to the blood vessel section area as the blood vessel diameter of the corresponding blood vessel section, obtaining the continuous blood vessel diameter with directional continuity according to the position sequence of the central axis points, calculating the gradient R' of the change of the diameter, and removing abnormal data with large reverse gradient in the front and the back. By calculating the diameter of the blood vessel, after abnormal data are eliminated, the diameter of the blood vessel at each position of the whole blood vessel can be obtained, and then the narrowest position and the maximum diameter can be calculated, so that the stenosis of the blood vessel can be more accurately detected and evaluated.
As shown in fig. 3, in one embodiment, a CTA-based head and neck vascular stenosis detection apparatus is provided, comprising:
an obtaining module 302, configured to obtain head and neck CTA image data to be predicted;
a first prediction module 304, configured to use the head and neck CTA image data as an input of a blood vessel segmentation model, and obtain a predicted blood vessel mask output by the blood vessel segmentation model, where the blood vessel segmentation model is obtained by training based on blood vessels and blood vessel lesion masks as labels, and the predicted blood vessel mask includes the blood vessel mask and the blood vessel lesion mask;
a second prediction module 306, configured to use the head and neck CTA image data as an input of a non-vascular segmentation model, and obtain a predicted bone mask and a predicted vascular lesion mask output by the non-vascular segmentation model;
a removal module 308 for removing the predicted bone mask and the predicted vascular lesion mask from the predicted vascular mask to obtain an actual vascular mask;
the correction module 310 is configured to calculate, according to the predicted vessel mask, coordinates of each medial axis point on a central axis of the vessel, and perform fitting correction based on each medial axis point coordinate to obtain corrected coordinates of each medial axis point;
a classification module 312, configured to classify blood vessels based on the corrected coordinates of each medial axis point, where each class corresponds to one blood vessel;
the segmentation module 314 is configured to extract, for each blood vessel, a corresponding blood vessel cross-section mask along a central axis of the blood vessel, and segment the blood vessel based on the extracted number of connected domains to obtain a blood vessel segmentation result;
a first calculating module 316, configured to calculate, for each segment of blood vessel, a cross-sectional area of the blood vessel corresponding to each central axis point based on an actual blood vessel mask corresponding to each segment of blood vessel, so as to obtain diameters of all blood vessels having directional continuity;
and a second calculating module 318, configured to use the coordinate point with the smallest diameter as the narrowest point of each segment of blood vessel, use the coordinate point with the largest diameter as the largest diameter of each segment of blood vessel, and calculate, according to the narrowest point and the largest diameter, the stenosis degree of each segment of blood vessel.
FIG. 4 is a diagram that illustrates an internal structure of the computer device in one embodiment. The computer device may specifically be a server or a terminal. As shown in fig. 4, the computer device includes a processor, a memory, and a network interface connected by a system bus. Wherein the memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium of the computer device has a stored operating system and may also have a stored computer program that, when executed by the processor, causes the processor to implement the CTA-based head and neck stenosis detection method described above. The internal memory may also have stored thereon a computer program that, when executed by the processor, causes the processor to perform the CTA-based head and neck stenosis detection method described above. Those skilled in the art will appreciate that the configuration shown in fig. 4 is a block diagram of only a portion of the configuration associated with the present application and does not constitute a limitation on the devices to which the present application applies, and that a particular device may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, a computer device is provided comprising a memory and a processor, the memory having stored thereon a computer program that, when executed by the processor, causes the processor to perform the steps of the CTA based head and neck stenosis detection method described above.
In one embodiment, a computer readable storage medium is provided, having a computer program stored thereon, which, when executed by a processor, causes the processor to perform the steps of the CTA based head and neck stenosis detection method described above.
It is understood that the above CTA-based method, apparatus, computer device and computer-readable storage medium for detecting stenosis of a head and neck blood vessel belong to a general inventive concept, and the embodiments are applicable to each other.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a non-volatile computer-readable storage medium, and can include the processes of the embodiments of the methods described above when the program is executed. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present application. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent application shall be subject to the appended claims.
Claims (10)
1. A CTA-based method for detecting head and neck vascular stenosis, comprising:
acquiring head and neck CTA image data to be predicted, wherein the head and neck CTA image data comprises: at least one of intracranial CTA image data, cervical CTA image data, and head-neck CTA image data;
taking the head and neck CTA image data as the input of a blood vessel segmentation model, and obtaining a predicted blood vessel mask output by the blood vessel segmentation model, wherein the blood vessel segmentation model is obtained by training based on blood vessels and blood vessel pathological change masks as labels, and the predicted blood vessel mask comprises the blood vessel mask and the blood vessel pathological change mask;
taking the head and neck CTA image data as the input of a non-vascular segmentation model, and obtaining a predicted bone mask and a predicted vascular lesion mask output by the non-vascular segmentation model;
removing the predicted bone mask and the predicted vascular lesion mask from the predicted vascular mask to obtain an actual vascular mask;
calculating to obtain coordinates of each central axis point on the central axis of the blood vessel according to the predicted blood vessel mask, and performing fitting correction based on each central axis point coordinate to obtain corrected coordinates of each central axis point;
classifying the blood vessels based on the corrected coordinates of the central axis points, wherein each class corresponds to one blood vessel;
for each blood vessel, extracting a corresponding blood vessel section mask along the central axis of the blood vessel, and segmenting the blood vessel based on the extracted number of connected domains to obtain a blood vessel segmentation result;
for each section of blood vessel, calculating to obtain the section area of the blood vessel corresponding to each middle axis point based on the actual blood vessel mask corresponding to each section of blood vessel, and calculating to obtain the diameter of the continuous blood vessel with directional continuity based on the section area of the blood vessel corresponding to each middle axis point;
taking the coordinate point with the smallest diameter in the diameters of the continuous blood vessels as the narrowest part of each section of blood vessel, and taking the coordinate point with the largest diameter as the largest diameter of each section of blood vessel;
and calculating the stenosis degree of each section of the blood vessel according to the narrowest part and the maximum diameter.
2. The method according to claim 1, wherein the training data of the vessel segmentation model and the non-vessel segmentation model are obtained as follows:
obtaining manual labeling of head and neck CTA image training data, the manual labeling comprising: the labeling MASK _ A of the blood vessel MASK, the labeling MASK _ B of the blood vessel pathological change MASK, the labeling MASK _ C of the skeleton MASK and the labeling MASK _ D of the blood vessel and blood vessel pathological change MASK;
extracting a central axis point of a central axis based on the labeling MASK _ D of the blood vessel and the blood vessel pathological change MASK, and fitting the extracted central axis point to obtain a corrected blood vessel central axis point coordinate;
classifying based on the corrected axis point coordinates of the blood vessels, wherein each class corresponds to one blood vessel;
for each blood vessel, determining a blood vessel section MASK and a blood vessel pathological change section MASK corresponding to each middle axis point on the middle axis line of the blood vessel based on the marking MASK _ A of the blood vessel MASK and the marking MASK _ B of the blood vessel pathological change MASK;
generating a complete and regular target training blood vessel mask corresponding to the whole blood vessel based on the blood vessel section mask corresponding to each middle axis point and the blood vessel pathological change section mask, wherein the target training blood vessel mask comprises: vascular masks and vascular lesion masks;
the head and neck CTA image training data is used as the input of a blood vessel segmentation model, and a target training blood vessel mask corresponding to each blood vessel is used as the expected output to train the blood vessel segmentation model;
and (3) taking the head and neck CTA image training data as the input of a non-blood vessel segmentation model, and taking the labeling MASK _ B of a blood vessel pathological change MASK and the labeling MASK _ C of a bone MASK as expected outputs to train the non-blood vessel segmentation model.
3. The method according to claim 2, wherein the determining, for each blood vessel, a blood vessel cross-section MASK and a blood vessel lesion cross-section MASK corresponding to each medial axis on a blood vessel medial axis based on the labeling MASK _ a of the blood vessel MASK and the labeling MASK _ B of the blood vessel lesion MASK comprises:
aiming at each blood vessel, aiming at any central axis point coordinate P on the central axis of the blood vessel n Obtaining the coordinate P of the previous middle axis point n-1 And the coordinate P of the latter middle axis point n+1 Calculating the coordinates of the former middle axis point and the latter middle axis point to obtain the coordinate P of the middle axis point n Corresponding blood vessel direction vectors and normal planes;
by said central axis point coordinate P n Obtaining the coordinate P corresponding to the central axis point from the labeling MASK _ A of the blood vessel MASK and the labeling MASK _ B of the blood vessel pathological MASK by the corresponding blood vessel direction vector and the normal plane n The cross-sectional mask of blood vessel, the cross-sectional mask of blood vessel lesion.
4. The method according to claim 2, wherein the generating of the complete and regular target training blood vessel mask corresponding to the whole blood vessel based on the blood vessel section mask and the blood vessel lesion section mask corresponding to each central axis point comprises: vascular masks and vascular lesion masks comprising:
taking the central axis point coordinate corresponding to the blood vessel section mask with the largest area as a starting point, and taking the blood vessel section mask with the largest area as a standard section mask;
obtaining a blood vessel section mask and a blood vessel pathological change section mask corresponding to the coordinates of the adjacent central axis points of the starting point;
taking a union set of the blood vessel section mask and the blood vessel pathological change section mask corresponding to the coordinates of the adjacent central axis points, and taking a projection overlapping area of the union set and the standard section mask in the direction of the central axis as a target blood vessel section mask of the adjacent central axis points, wherein the target blood vessel section mask comprises: a blood vessel section mask and a blood vessel pathological change section mask;
and taking the target blood vessel section mask corresponding to the adjacent middle axis points as a new standard section mask, taking the middle axis point coordinates corresponding to the new standard section mask as a new initial point, and entering the step of extracting the blood vessel section mask and the blood vessel lesion section mask of the adjacent middle axis point coordinates of the initial point until the target blood vessel section mask corresponding to each middle axis point of the whole blood vessel is extracted and obtained, wherein the target blood vessel section mask corresponding to each middle axis point of the whole blood vessel forms the target training blood vessel mask corresponding to the whole blood vessel.
5. The method according to claim 1, wherein the calculating according to the predicted vessel mask to obtain coordinates of each medial axis point on a central axis of the vessel, and performing fitting correction based on each of the medial axis point coordinates to obtain corrected coordinates of each medial axis point comprises:
calculating to obtain an initial central axis of the blood vessel by adopting a three-dimensional skeleton extraction method based on the predicted blood vessel mask;
traversing the coordinates of a middle axis point on the initial central axis from the bottom of the cephalic neck to the brain along the Z-axis direction of the CTA image;
aiming at each central axis point coordinate, taking the central axis point coordinate as a center, extracting a predicted blood vessel mask within a preset range, and obtaining a local blood vessel mask corresponding to the central axis point coordinate;
and extracting contour points of the local blood vessel mask, performing ellipse fitting based on the extracted contour points, and fitting the center axis point coordinates to center point coordinates of an ellipse to obtain corrected center axis point coordinates.
6. The method according to claim 1, wherein said classifying the blood vessels based on the modified respective medial axis point coordinates, one blood vessel for each class, comprises:
obtaining the coordinates of the central axis point at the bottommost layer in the blood vessel as a reference point;
calculating a point closest to the three-dimensional distance of the reference point, taking the point as a correlation point of the reference point, and classifying the reference point and the correlation point into one class;
updating the associated points of the reference points to new reference points, entering a step of calculating the points with the closest three-dimensional distance to the reference points as the associated points of the reference points until the last point of the blood vessel is iterated, and classifying all the iterated points into one class;
and for the central axis points which are not iterated, obtaining the central axis point at the bottommost layer in the central axis points which are not iterated, taking the central axis point as a reference point, and entering the step of calculating the point with the three-dimensional closest distance to the reference point as the associated point of the reference point until all the central axis points are classified.
7. The method according to claim 1, wherein for each segment of blood vessel, calculating a blood vessel cross-sectional area corresponding to each central axis point based on an actual blood vessel mask corresponding to each segment of blood vessel, and calculating a continuous blood vessel diameter with directional continuity based on the blood vessel cross-sectional area corresponding to each central axis point comprises:
calculating to obtain the section area of the blood vessel corresponding to each middle axis point based on the actual blood vessel mask;
taking the diameter of the approximate circle which is equal to the cross section area of the blood vessel as the diameter of the blood vessel corresponding to the cross section of the corresponding blood vessel;
obtaining the diameters of continuous blood vessels with directional continuity according to the sequence of the axial points in the blood vessels;
the method further comprises the following steps:
and calculating the diameter change gradient of the diameters of the continuous blood vessels, and removing abnormal data with the front and rear gradient changes larger than a preset gradient.
8. A CTA-based head and neck vascular stenosis detection apparatus, comprising:
the acquisition module is used for acquiring head and neck CTA image data to be predicted;
the first prediction module is used for taking the head and neck CTA image data as the input of a blood vessel segmentation model and obtaining a predicted blood vessel mask output by the blood vessel segmentation model, wherein the blood vessel segmentation model is obtained by training based on blood vessels and blood vessel pathological change masks as labels, and the predicted blood vessel mask comprises the blood vessel mask and the blood vessel pathological change mask;
the second prediction module is used for taking the head and neck CTA image data as the input of a non-vascular segmentation model, and acquiring a predicted bone mask and a predicted vascular lesion mask output by the non-vascular segmentation model;
a removing module for removing the predicted bone mask and the predicted vascular lesion mask from the predicted vascular mask to obtain an actual vascular mask;
the correction module is used for calculating to obtain coordinates of each middle axis point on a central axis of the blood vessel according to the predicted blood vessel mask, and performing fitting correction based on each middle axis point coordinate to obtain corrected coordinates of each middle axis point;
the classification module is used for classifying blood vessels based on the corrected coordinates of the middle axis points, and each class corresponds to one blood vessel;
the segmentation module is used for extracting a corresponding blood vessel section mask along the central axis of each blood vessel and segmenting the blood vessels based on the extracted number of connected domains to obtain a blood vessel segmentation result;
the first calculation module is used for calculating and obtaining the section area of the blood vessel corresponding to each middle axis point based on the actual blood vessel mask corresponding to each section of the blood vessel so as to obtain the diameters of all the blood vessels with direction continuity;
and the second calculation module is used for taking the coordinate point with the minimum diameter as the narrowest part of each section of blood vessel, taking the coordinate point with the maximum diameter as the maximum diameter of each section of blood vessel, and calculating the stenosis degree of each section of blood vessel according to the narrowest part and the maximum diameter.
9. A computer device comprising a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the CTA based head and neck stenosis detection method of any of claims 1 to 7.
10. A computer-readable storage medium, having stored thereon a computer program which, when executed by a processor, causes the processor to carry out the steps of the CTA-based head and neck vascular stenosis detection method as claimed in any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210570301.2A CN114708259B (en) | 2022-05-24 | 2022-05-24 | CTA (computed tomography angiography) -based head and neck vascular stenosis detection method, device, equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210570301.2A CN114708259B (en) | 2022-05-24 | 2022-05-24 | CTA (computed tomography angiography) -based head and neck vascular stenosis detection method, device, equipment and medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114708259A CN114708259A (en) | 2022-07-05 |
| CN114708259B true CN114708259B (en) | 2022-08-16 |
Family
ID=82176677
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210570301.2A Active CN114708259B (en) | 2022-05-24 | 2022-05-24 | CTA (computed tomography angiography) -based head and neck vascular stenosis detection method, device, equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114708259B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117274280B (en) * | 2023-09-06 | 2024-08-20 | 强联智创（北京）科技有限公司 | Method, apparatus and computer readable storage medium for vessel segmentation |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106296664A (en) * | 2016-07-30 | 2017-01-04 | 上海联影医疗科技有限公司 | Vessel extraction method |
| CN109754397A (en) * | 2019-01-03 | 2019-05-14 | 东软集团股份有限公司 | Extracting method, device, terminal device and the storage medium of vessel segment center path |
| CN109872321A (en) * | 2019-02-26 | 2019-06-11 | 数坤（北京）网络科技有限公司 | A kind of hemadostewnosis detection method and equipment |
| CN111640102A (en) * | 2020-05-29 | 2020-09-08 | 复旦大学附属中山医院 | Aneurysm detection method, aneurysm detection device and computer equipment |
| CN111681226A (en) * | 2020-06-09 | 2020-09-18 | 上海联影医疗科技有限公司 | Target tissue positioning method and device based on blood vessel identification |
| CN113628193A (en) * | 2021-08-12 | 2021-11-09 | 推想医疗科技股份有限公司 | Method, device and system for determining blood vessel stenosis rate and storage medium |
| CN114266742A (en) * | 2021-12-17 | 2022-04-01 | 天津医科大学 | Method for detecting narrow region of cerebral blood vessel CTA image |
| CN114359205A (en) * | 2021-12-29 | 2022-04-15 | 推想医疗科技股份有限公司 | Head and neck blood vessel analysis method and device, storage medium and electronic equipment |
| CN114418977A (en) * | 2022-01-07 | 2022-04-29 | 乐普（北京）医疗器械股份有限公司 | Method and device for coronary angiography quantitative analysis based on angiography video |
| WO2022095895A1 (en) * | 2020-11-05 | 2022-05-12 | 杭州深睿博联科技有限公司 | Vascular stenosis analysis method and apparatus |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8526699B2 (en) * | 2010-03-12 | 2013-09-03 | Siemens Aktiengesellschaft | Method and system for automatic detection and classification of coronary stenoses in cardiac CT volumes |
| US10037603B2 (en) * | 2015-05-04 | 2018-07-31 | Siemens Healthcare Gmbh | Method and system for whole body bone removal and vascular visualization in medical image data |
-
2022
- 2022-05-24 CN CN202210570301.2A patent/CN114708259B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106296664A (en) * | 2016-07-30 | 2017-01-04 | 上海联影医疗科技有限公司 | Vessel extraction method |
| CN109754397A (en) * | 2019-01-03 | 2019-05-14 | 东软集团股份有限公司 | Extracting method, device, terminal device and the storage medium of vessel segment center path |
| CN109872321A (en) * | 2019-02-26 | 2019-06-11 | 数坤（北京）网络科技有限公司 | A kind of hemadostewnosis detection method and equipment |
| CN111640102A (en) * | 2020-05-29 | 2020-09-08 | 复旦大学附属中山医院 | Aneurysm detection method, aneurysm detection device and computer equipment |
| CN111681226A (en) * | 2020-06-09 | 2020-09-18 | 上海联影医疗科技有限公司 | Target tissue positioning method and device based on blood vessel identification |
| WO2022095895A1 (en) * | 2020-11-05 | 2022-05-12 | 杭州深睿博联科技有限公司 | Vascular stenosis analysis method and apparatus |
| CN113628193A (en) * | 2021-08-12 | 2021-11-09 | 推想医疗科技股份有限公司 | Method, device and system for determining blood vessel stenosis rate and storage medium |
| CN114266742A (en) * | 2021-12-17 | 2022-04-01 | 天津医科大学 | Method for detecting narrow region of cerebral blood vessel CTA image |
| CN114359205A (en) * | 2021-12-29 | 2022-04-15 | 推想医疗科技股份有限公司 | Head and neck blood vessel analysis method and device, storage medium and electronic equipment |
| CN114418977A (en) * | 2022-01-07 | 2022-04-29 | 乐普（北京）医疗器械股份有限公司 | Method and device for coronary angiography quantitative analysis based on angiography video |
Non-Patent Citations (2)
| Title |
|---|
| Intracranial Vascular Stenosis and Occlusive Disease: Evaluation with CT Angiography,MR Angiography, and Digital Subtraction Angiography;Suzie Bash et al.;《AJNR Am J Neuroradiol 26》;20050531;第1012-1021页 * |
| 人工智能在脑血管疾病诊疗中的相关应用;鲁伟 等;《人工智能》;20210331(第3期);第73-78页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114708259A (en) | 2022-07-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110310256B (en) | Coronary stenosis detection method, coronary stenosis detection device, computer equipment and storage medium | |
| JP6703082B2 (en) | Medical image processing apparatus for segmenting structures in medical image, method for segmenting medical image, and storage medium storing computer program for segmenting medical image | |
| US10910099B2 (en) | Segmentation, landmark detection and view classification using multi-task learning | |
| CN111815599B (en) | Image processing method, device, equipment and storage medium | |
| CN111402260A (en) | Medical image segmentation method, system, terminal and storage medium based on deep learning | |
| CN110570407B (en) | Image processing method, storage medium, and computer device | |
| CN110400626B (en) | Image detection method, image detection device, computer equipment and storage medium | |
| US11471096B2 (en) | Automatic computerized joint segmentation and inflammation quantification in MRI | |
| CN109903298B (en) | Method, system and computer storage medium for repairing blood vessel segmentation image fracture | |
| CN114972255B (en) | Image detection method and device for cerebral micro-bleeding, computer equipment and storage medium | |
| CN116681716B (en) | Method, device, equipment and storage medium for dividing intracranial vascular region of interest | |
| US11967079B1 (en) | System and method for automatically detecting large vessel occlusion on a computational tomography angiogram | |
| CN114708259B (en) | CTA (computed tomography angiography) -based head and neck vascular stenosis detection method, device, equipment and medium | |
| CN112308846A (en) | Blood vessel segmentation method and device and electronic equipment | |
| CN111128348B (en) | Medical image processing method, medical image processing device, storage medium and computer equipment | |
| CN111145200A (en) | Blood vessel center line tracking method combining convolutional neural network and cyclic neural network | |
| CN110738702B (en) | Three-dimensional ultrasonic image processing method, device, equipment and storage medium | |
| CN113160199B (en) | Image recognition method and device, computer equipment and storage medium | |
| CN114299057A (en) | Method for extracting blood vessel center line and storage medium | |
| CN111401102A (en) | Deep learning model training method and device, electronic equipment and storage medium | |
| CN111462203B (en) | DR focus evolution analysis device and method | |
| CN112651960A (en) | Image processing method, device, equipment and storage medium | |
| Delmoral et al. | Segmentation of pathological liver tissue with dilated fully convolutional networks: A preliminary study | |
| CN111696113A (en) | Method and system for monitoring a biological process | |
| CN115035145A (en) | Blood vessel and bone segmentation method and device, computer device and storage medium |
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 |