CN115423826A - Method and device for training lumen segmentation model, electronic equipment and storage medium - Google Patents
Method and device for training lumen segmentation model, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN115423826A CN115423826A CN202211150263.1A CN202211150263A CN115423826A CN 115423826 A CN115423826 A CN 115423826A CN 202211150263 A CN202211150263 A CN 202211150263A CN 115423826 A CN115423826 A CN 115423826A
- Authority
- CN
- China
- Prior art keywords
- lumen
- image
- stenosis
- initial
- labeling
- 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.)
- Pending
Links
- 230000011218 segmentation Effects 0.000 title claims abstract description 127
- 238000000034 method Methods 0.000 title claims abstract description 70
- 238000012549 training Methods 0.000 title claims abstract description 62
- 238000002372 labelling Methods 0.000 claims abstract description 145
- 208000031481 Pathologic Constriction Diseases 0.000 claims abstract description 132
- 230000036262 stenosis Effects 0.000 claims abstract description 132
- 208000037804 stenosis Diseases 0.000 claims abstract description 132
- 238000001514 detection method Methods 0.000 claims abstract description 22
- 238000005457 optimization Methods 0.000 claims description 28
- 238000012545 processing Methods 0.000 claims description 22
- 238000004590 computer program Methods 0.000 claims description 16
- 210000004351 coronary vessel Anatomy 0.000 description 17
- 238000010586 diagram Methods 0.000 description 12
- 210000004204 blood vessel Anatomy 0.000 description 11
- 238000004891 communication Methods 0.000 description 8
- 238000000605 extraction Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000012216 screening Methods 0.000 description 4
- 239000000243 solution Substances 0.000 description 4
- 238000013473 artificial intelligence Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000002591 computed tomography Methods 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000003491 array Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 239000002872 contrast media Substances 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000010790 dilution Methods 0.000 description 1
- 239000012895 dilution Substances 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 239000011521 glass Substances 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003902 lesion Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000002595 magnetic resonance imaging Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 230000001953 sensory effect Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000002966 stenotic effect Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
-
- 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
-
- 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
- G06T2207/30204—Marker
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (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)
- Image Analysis (AREA)
Abstract
The invention discloses a method and a device for training a lumen segmentation model, electronic equipment and a storage medium. The method comprises the following steps: acquiring a lumen sample image and an initial lumen labeling image corresponding to the lumen sample image; based on the lumen center line corresponding to the initial lumen marking image, performing stenosis detection on the initial lumen marking image, and determining a lumen stenosis section; filling the lumen stenosis to obtain a stenosis filling image; merging the narrow filling image and the initial lumen labeling image to obtain a target lumen labeling image; and performing iterative training on the lumen segmentation model to be trained based on the lumen sample image and the target lumen labeling image corresponding to the lumen sample image to obtain the trained lumen segmentation model. Through above-mentioned technical scheme, the problem of lumen segmentation interrupt has been solved.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a method and a device for training a lumen segmentation model, electronic equipment and a storage medium.
Background
With the development of artificial intelligence technologies such as machine learning, the application of artificial intelligence algorithms in the field of medical image processing is more and more extensive.
In the process of implementing the invention, the inventor finds that at least the following technical problems exist in the prior art: the existing coronary artery automatic segmentation method of medical images has the problems of segmentation interruption and poor blood vessel saturation.
Disclosure of Invention
The invention provides a method and a device for training a lumen segmentation model, electronic equipment and a storage medium, which are used for solving the problems of lumen image segmentation interruption and poor blood vessel saturation.
According to an aspect of the present invention, there is provided a method for training a lumen segmentation model, including:
acquiring a lumen sample image and an initial lumen labeling image corresponding to the lumen sample image;
based on the lumen center line corresponding to the initial lumen marking image, performing stenosis detection on the initial lumen marking image, and determining a lumen stenosis section;
filling the narrow section of the lumen to obtain a narrow filling image;
merging the narrow filling image and the initial lumen labeling image to obtain a target lumen labeling image;
and carrying out iterative training on the lumen segmentation model to be trained based on the lumen sample image and the target lumen labeling image corresponding to the lumen sample image to obtain the trained lumen segmentation model.
According to another aspect of the present invention, there is provided a device for training a lumen segmentation model, comprising:
the system comprises a sample acquisition module, a detection module and a display module, wherein the sample acquisition module is used for acquiring a lumen sample image and an initial lumen labeling image corresponding to the lumen sample image;
a lumen narrow section determining module, configured to perform narrow detection on the initial lumen labeling image based on a lumen center line corresponding to the initial lumen labeling image, and determine a lumen narrow section;
the lumen stenosis filling module is used for filling the lumen stenosis section to obtain a stenosis filling image;
the image merging module is used for merging the narrow filling image and the initial lumen labeling image to obtain a target lumen labeling image;
and the segmentation model training module is used for carrying out iterative training on the lumen segmentation model to be trained on the basis of the lumen sample image and the target lumen labeling image corresponding to the lumen sample image to obtain the trained lumen segmentation model.
According to another aspect of the present invention, there is provided an electronic apparatus including:
at least one processor; and
a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,
the memory stores a computer program executable by the at least one processor, the computer program being executable by the at least one processor to enable the at least one processor to perform a method of training a lumen segmentation model according to any embodiment of the present invention.
According to another aspect of the present invention, there is provided a computer-readable storage medium storing computer instructions for causing a processor to implement a method for training a lumen segmentation model according to any one of the embodiments of the present invention when executed.
According to the technical scheme of the embodiment of the invention, the lumen sample image and the initial lumen labeling image corresponding to the lumen sample image are obtained; based on the lumen center line corresponding to the initial lumen marking image, performing stenosis detection on the initial lumen marking image, and determining a lumen stenosis section; filling the narrow section of the lumen to obtain a narrow filling image; merging the narrow filling image and the initial lumen labeling image to obtain a target lumen labeling image; and performing iterative training on the lumen segmentation model to be trained based on the lumen sample image and the target lumen labeling image corresponding to the lumen sample image to obtain the trained lumen segmentation model. Above-mentioned technical scheme fills the lumen narrow section through detecting the lumen narrow section, increases the saturation of lumen to can avoid cutting apart the interrupt phenomenon because of the lumen that the stenosis arouses, guarantee the integrality of lumen.
It should be understood that the statements in this section do not necessarily identify key or critical features of the embodiments of the present invention, nor do they necessarily limit the scope of the invention. Other features of the present invention will become apparent from the following description.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a flowchart of a training method of a lumen segmentation model according to an embodiment of the present invention;
FIG. 2 is a coronary artery lumen artificial labeling image according to an embodiment of the present invention;
fig. 3 is a lumen centerline corresponding to an artificial labeling image of a coronary lumen provided in an embodiment of the present invention;
FIG. 4 is a flowchart of a method for training a lumen segmentation model according to a second embodiment of the present invention;
FIG. 5 is a lumen label optimization image provided according to the second embodiment of the invention;
FIG. 6 is a narrow fill image according to a second embodiment of the present invention;
FIG. 7 is a diagram illustrating a target lumen labeling image according to a second embodiment of the present invention;
FIG. 8 is a flowchart of a method for training a lumen segmentation model according to a third embodiment of the present invention;
FIG. 9 is a plot of a linear fit of lumen diameters provided in accordance with a third embodiment of the present invention;
FIG. 10 is a schematic structural diagram of a training apparatus for a lumen segmentation model according to a fourth embodiment of the present invention;
fig. 11 is a schematic structural diagram of image merging according to the fourth embodiment of the present invention;
fig. 12 is a preliminarily predicted coronary artery segmentation result image according to the fourth embodiment of the present invention;
fig. 13 is a final coronary artery segmentation result image according to the fourth embodiment of the present invention;
FIG. 14 is a schematic structural diagram of a training system for a lumen segmentation model according to a fifth embodiment of the present invention;
fig. 15 is a schematic structural diagram of an electronic device implementing the method for training a lumen segmentation model according to an embodiment of the present invention.
Detailed Description
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious 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.
It should be noted that the terms "initial", "target", and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in other sequences than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
Example one
Fig. 1 is a flowchart of a method for training a lumen segmentation model according to an embodiment of the present invention, where the method is applicable to a case where a coronary image is automatically segmented, and the method may be performed by a device for training the lumen segmentation model, where the device for training the lumen segmentation model may be implemented in a form of hardware and/or software, and the device for training the lumen segmentation model may be configured in a computer terminal. As shown in fig. 1, the method includes:
s110, acquiring a lumen sample image and an initial lumen labeling image corresponding to the lumen sample image.
In this embodiment, the lumen sample image refers to an image used for training a lumen segmentation model. The lumen sample image may be two-dimensional, three-dimensional, or multi-dimensional, and is not limited herein. For example, the lumen sample image may include a Computed Tomography (CT) image, a Magnetic Resonance Imaging (MRI), and the like, in the case of a coronary medical image, a coronary lumen in the coronary medical image is a solid structure on a space enclosed by a closed curved surface, and has the following characteristics: as the distance increases, the smaller the lumen diameter; coronary lesions are frequently stenotic, which can be caused by plaques.
The initial lumen marking image corresponding to the lumen sample image is an artificially marked lumen image. Typically, the initial lumen labeling image may be a coronary lumen artificial labeling image.
Specifically, one or more lumen sample images and an initial lumen labeling image corresponding to the lumen sample image may be called from a preset storage path of the electronic device, or other associated devices, or a cloud server, where an obtaining manner of the data is not limited.
And S120, based on the lumen center line corresponding to the initial lumen marking image, performing stenosis detection on the initial lumen marking image, and determining a lumen stenosis section.
After the initial lumen labeling image is obtained, the centerline of the initial lumen labeling image can be extracted to obtain a lumen centerline, wherein the lumen centerline refers to a spatial curve distributed in the center of the lumen. In the present embodiment, the centerline extraction method includes, but is not limited to, a skeletonization method, a region growing method, and the like. For example, the initial lumen labeling image may be a binary image, and the skeletonization method is an image processing algorithm for reducing foreground pixels as much as possible on the premise of keeping the connectivity of a foreground region of the binary image unchanged, and finally obtaining a skeleton of the image.
Exemplarily, a coronary artery lumen artificial labeling image is taken as an example, fig. 2 is a coronary artery lumen artificial labeling image provided by the embodiment of the present invention, and fig. 3 is a lumen center line corresponding to the coronary artery lumen artificial labeling image provided by the embodiment of the present invention.
In the present embodiment, the luminal narrowing section refers to a narrowed portion in a blood vessel. It can be understood that, the initial lumen labeling image is subjected to stenosis detection, one or more lumen stenosis sections can be obtained, and then the lumen label value of the lumen stenosis sections is filled, so that the lumen is restored to the normal lumen size, and the phenomenon of lumen segmentation interruption caused by stenosis is avoided.
Specifically, for any center line point in the center lines of the tube cavities, the corresponding tube cavity diameter of the center line point in the initial tube cavity labeling image is determined, and the tube cavity narrow section is determined according to the size of the tube cavity diameter.
S130, filling the narrow section of the lumen to obtain a narrow filling image.
In the present embodiment, the stenosis filling image refers to an image in which a luminal stenosis is filled to a normal luminal size.
Specifically, the lumen diameter of the lumen narrow section can be determined, the lumen diameter of the lumen narrow section is recovered to the normal lumen diameter, the filling of the lumen narrow section is realized, the lumen is recovered to the normal lumen diameter, the saturation of the lumen is improved, and the phenomenon of lumen segmentation interruption caused by stenosis is avoided.
And S140, merging the narrow filling image and the initial lumen labeling image to obtain a target lumen labeling image.
In this embodiment, the target lumen labeling image may be a combined image of the stenosis filling image and the initial lumen labeling image, in other words, the target lumen labeling image is an image subjected to stenosis filling processing. By using the target lumen labeling image for model training, the phenomenon of lumen segmentation interruption caused by stenosis can be reduced.
S150, carrying out iterative training on the lumen segmentation model to be trained based on the lumen sample image and the target lumen labeling image corresponding to the lumen sample image to obtain the trained lumen segmentation model.
In this embodiment, feature extraction may be performed on the lumen sample image through the lumen segmentation model, and model parameters in the lumen segmentation model are trained based on the extracted features, and by continuously adjusting the model parameters, a distance deviation between an output result of the model and a target lumen labeling image corresponding to the lumen sample image is gradually reduced and tends to be stable.
For example, the lumen segmentation model to be trained may be a deep neural network model. When the path sampling is trained, the target lumen labeling image can be uniformly sampled. The loss function of the lumen segmentation model may semantically segment a loss function, such as BCE (Binary Cross entry), cross Entropy, etc.
On the basis of the foregoing embodiment, optionally, after obtaining the trained lumen segmentation model, the method further includes: acquiring at least one image to be segmented; inputting the image to be segmented into the trained lumen segmentation model to obtain a prediction segmentation image corresponding to the image to be segmented;
processing the maximum connected region of the prediction segmentation image to obtain a maximum connected region image;
and performing plaque removal processing on the maximum connected region image to obtain a target segmentation image.
Specifically, an image to be segmented is input into a trained lumen segmentation model as input data; the lumen segmentation model carries out prediction segmentation on an image to be segmented to obtain a prediction segmentation image, and the prediction segmentation image can be displayed in a binarization form, wherein an element 1 is a coronary artery blood vessel, and an element 0 is a background. After the prediction segmentation image is obtained, the maximum connected region processing can be carried out on the prediction segmentation image to obtain a maximum connected region image, and the screening of the non-maximum connected region is realized; further, after the maximum connected region image is obtained, the patch of the maximum connected region image can be removed, and a complete target segmentation image with high saturation can be obtained and output.
According to the technical scheme of the embodiment of the invention, the lumen sample image and the initial lumen labeling image corresponding to the lumen sample image are obtained; based on the lumen center line corresponding to the initial lumen marking image, performing stenosis detection on the initial lumen marking image, and determining a lumen stenosis section; filling the narrow section of the lumen to obtain a narrow filling image; merging the narrow filling image and the initial lumen labeling image to obtain a target lumen labeling image; and performing iterative training on the lumen segmentation model to be trained based on the lumen sample image and the target lumen labeling image corresponding to the lumen sample image to obtain the trained lumen segmentation model. Above-mentioned technical scheme fills the lumen narrow section through detecting the lumen narrow section, increases the saturation of lumen to can avoid cutting apart the interrupt phenomenon because of the lumen that the stenosis arouses, guarantee the integrality of lumen.
Example two
Fig. 4 is a flowchart of a method for training a lumen segmentation model according to a second embodiment of the present invention, and the method of this embodiment may be combined with various alternatives in the method for training a lumen segmentation model provided in the foregoing embodiments. The method for training the lumen segmentation model provided by the embodiment is further optimized. Optionally, after the acquiring the lumen sample image and the initial lumen labeling image corresponding to the lumen sample image, the method further includes: updating the lumen label value of the initial lumen labeling image based on the distance from the center line point to the starting point of the lumen center line for any center line point in the lumen center line corresponding to the initial lumen labeling image to obtain a lumen label optimized image; correspondingly, the merging the stenosis filling image and the initial lumen labeling image to obtain a target lumen labeling image includes: and merging the narrow filling image and the lumen label optimization image to obtain a target lumen labeling image.
As shown in fig. 4, the method includes:
s210, obtaining a lumen sample image and an initial lumen labeling image corresponding to the lumen sample image.
S220, updating the lumen label value of the initial lumen labeling image based on the distance from the center line point to the starting point of the lumen center line for any center line point in the lumen center line corresponding to the initial lumen labeling image to obtain a lumen label optimization image.
The central line point refers to any one point in the central line of the lumen, and the starting point of the central line of the lumen corresponds to the position near the position where the diameter of the lumen is maximum. Illustratively, the coronary lumen is divided into left and right coronary arteries, each having a starting point and a plurality of terminal end points.
In this embodiment, the lumen label value refers to image information for marking a lumen, and the lumen label value may be information such as brightness, saturation, or gray value.
It should be noted that, the value of the lumen label in the initial lumen labeling image is not different, and the information of the lumen diameter and the stenosis cannot be reflected. For example, the lumen label values of the foreground within the initial lumen labeling image may all be 1.
For example, the starting point of the lumen centerline may be denoted by a, the end point may be denoted by B, the lumen label value of a may be set as a, and the lumen label value of B may be set as B; the lumen label value corresponding to the centerline point on the a-B lumen is a value within [ a, B ], e.g., the lumen label value corresponding to the centerline point on the a-B lumen may increase with increasing distance of the centerline point to the starting point. After the updating of the lumen label value of the initial lumen labeling image is completed, the lumen label optimization image with the changed lumen label value can be obtained, so that the lumens with different distances are different, the lumen segmentation model can be favorably learned, and the phenomenon of lumen segmentation interruption is relieved.
And S230, based on the lumen central line corresponding to the initial lumen marking image, performing stenosis detection on the initial lumen marking image, and determining a lumen stenosis section.
S240, filling the narrow section of the lumen to obtain a narrow filling image.
And S250, merging the narrow filling image and the lumen label optimization image to obtain a target lumen labeling image.
In this embodiment, the target lumen labeling image may be a combined image of a stenosis filling image and a lumen label optimization image, in other words, the target lumen labeling image is an image subjected to stenosis filling processing and lumen label value updating. Model training is carried out by using the target lumen labeling image, so that the phenomenon that lumen segmentation is interrupted due to stenosis or consistent lumen label values can be reduced.
For example, the lumen stenosis section in the lumen label optimization image can be replaced by a stenosis filling image, so as to obtain a target lumen labeling image. Fig. 5 is a lumen label optimization image provided by an embodiment of the present invention, fig. 6 is a stenosis filling image provided by an embodiment of the present invention, and fig. 7 is a target lumen labeling image provided by an embodiment of the present invention.
And S260, carrying out iterative training on the lumen segmentation model to be trained based on the lumen sample image and the target lumen labeling image corresponding to the lumen sample image to obtain the trained lumen segmentation model.
On the basis of the above embodiment, the updating the lumen label value of the initial lumen labeling image based on the distance from the centerline point to the starting point of the lumen centerline to obtain a lumen label optimized image includes: matching the distance from the center line point to the starting point of the center line of the tube cavity in a preset distance-label matching relation to obtain a distance optimization label value; and updating the lumen label value of the initial lumen labeling image based on the distance optimization label value to obtain a lumen label optimization image.
In this embodiment, the distance-label matching relationship refers to mapping relationship information including distance and lumen label values, and may be set according to prior knowledge. The distance optimization label value is a lumen label value which is in corresponding relation with the distance.
Illustratively, the starting point of the lumen centerline corresponding to the left and right coronary arteries may be detected, the lumen centerline traversed and the bifurcation and end points recorded. From the bifurcation point and the end-point, the topological relation (i.e. the relation of parent and child branches) can be determined. Further, the distance from each end point to the starting point can be calculated according to the topological relation. Further, from the starting point of the central line of the tube cavity, the central line points from the end point to the starting point are sequentially obtained, the distance from the central line point to the starting point is determined, and the distance from the central line point to the starting point of the central line of the tube cavity is matched in a preset distance-label matching relation to obtain a distance optimization label value. The distance-optimized tag value may increase with increasing distance of the centerline point to the starting point, in other words, the distance-optimized tag value is proportional to the distance. Further, the lumen label value in the initial lumen labeling image can be replaced by the distance optimization label value, so that a lumen label optimization image is obtained.
It is understood that due to the dilution of the contrast agent concentration, the longer the coronary vessel, the weaker the strength of the distal end of the vessel, and may cause the lumen segmentation to break. In the embodiment, the lumen label values which are sequentially increased from the starting point to the end point are established, so that the characteristics of the far end of the blood vessel are enhanced, the saturation of the blood vessel is increased, and the condition of breakage of the far end blood vessel is reduced.
On the basis of the above embodiment, after the obtaining of the narrow filling image, the method further includes: for any central line point in the lumen narrow section, matching the distance from the central line point in the lumen narrow section to the starting point of the central line of the lumen in a preset distance-label matching relation to obtain a narrow label value; and updating the lumen label value of the stenosis filling image based on the stenosis label value to obtain an updated stenosis filling image.
It can be understood that updating the lumen label value of the stenosis filling image can make the lumen label values of the stenosis filling image and the lumen label optimization image be unified in the same interval, so that the lumen label value in the target lumen labeling image continuously changes.
According to the technical scheme of the embodiment of the invention, a lumen sample image and an initial lumen labeling image corresponding to the lumen sample image are obtained; updating the lumen label value of the initial lumen labeling image based on the distance from the center line point to the starting point of the lumen center line to obtain a lumen label optimized image for any center line point in the lumen center line corresponding to the initial lumen labeling image; based on the lumen center line corresponding to the initial lumen marking image, performing stenosis detection on the initial lumen marking image, and determining a lumen stenosis section; filling the narrow section of the lumen to obtain a narrow filling image; merging the narrow filling image and the lumen label optimization image to obtain a target lumen labeling image; and performing iterative training on the lumen segmentation model to be trained based on the lumen sample image and the target lumen labeling image corresponding to the lumen sample image to obtain the trained lumen segmentation model. According to the technical scheme, the lumen label value of the initial lumen labeling image is updated according to the distance from the center line point to the starting point of the lumen center line, so that the lumens with different distances are different, and the lumen segmentation model is favorable for learning.
EXAMPLE III
Fig. 8 is a flowchart of a method for training a lumen segmentation model according to a third embodiment of the present invention, and the method of the present embodiment may be combined with various alternatives of the method for training a lumen segmentation model according to the third embodiment. The method for training the lumen segmentation model provided by the embodiment is further optimized. Optionally, the performing stenosis detection on the initial lumen labeling image based on the lumen centerline corresponding to the initial lumen labeling image to determine a lumen stenosis section includes: for any centerline point in the lumen centerline, determining the corresponding lumen diameter of the centerline point in the initial lumen labeling image; performing linear fitting processing on the diameters of the tube cavities corresponding to all the central line points on the central line of the tube cavity to obtain the estimated tube cavity diameters corresponding to all the central line points; determining at least one lumen stenosis based on the corresponding lumen diameter of the centerline point in the initial lumen annotation image and the estimated lumen diameter.
As shown in fig. 8, the method includes:
s310, acquiring a lumen sample image and an initial lumen labeling image corresponding to the lumen sample image.
S320, determining the corresponding lumen diameter of the centerline point in the initial lumen marking image for any centerline point in the lumen centerline corresponding to the initial lumen marking image.
In this embodiment, the initial lumen annotation image may be a three-dimensional image. Illustratively, a plurality of centerline points are extracted from the centerline of the lumen, and based on the three-dimensional initial lumen labeling image, a lumen cross-section image corresponding to each centerline point is determined. The lumen cross-section image is a planar image obtained by vertically cutting the lumen from a center line point. It will be appreciated that the cross-section of the lumen is perpendicular to the centerline, with the normal being the tangential direction of the centerline to which the centerline point corresponds. Further, the edge contour of the cross section image of the lumen can be extracted to obtain the edge contour of the lumen. Edge contour extraction methods may include, but are not limited to, canny operators, frangi filtering, and the like. Further, the average value of the maximum distance value set of any two points on the contour of the edge of the lumen is taken as the diameter of the lumen.
S330, performing linear fitting processing on the diameters of the tube cavities corresponding to all the central line points on the central line of the tube cavity to obtain the estimated tube cavity diameters corresponding to all the central line points.
Specifically, the diameters of the lumens corresponding to all the centerline points on the centerline of the lumen are subjected to linear fitting processing, and a diameter fitting result is determined, so that the estimated lumen diameter corresponding to each centerline point can be determined.
For each blood vessel in the initial lumen labeling image, statistical information of the length (distance) of the blood vessel and the lumen diameter of the blood vessel can be depicted by a scatter diagram; and performing linear fitting processing by adopting a least square method to estimate the diameter of the normal lumen. Fig. 9 is a linear fitting graph of lumen diameters according to an embodiment of the present invention, in which a horizontal axis represents a length of a blood vessel, a vertical axis represents a lumen diameter of the blood vessel, and a curve represents a corresponding lumen diameter of a centerline point in an initial lumen labeling image, that is, a true lumen diameter; the straight line represents the estimated lumen diameter, i.e. the fitted normal lumen diameter.
S340, determining at least one lumen stenosis based on the corresponding lumen diameter of the centerline point in the initial lumen annotation image and the estimated lumen diameter.
Specifically, the lumen narrowing is determined based on the difference between the lumen diameter and the estimated lumen diameter.
On the basis of the above embodiment, the determining at least one lumen stenosis section based on the corresponding lumen diameter of the centerline point in the initial lumen labeling image and the estimated lumen diameter includes: the determining step for any one of the luminal narrowings comprising: and if the difference value between the lumen diameter corresponding to the continuous central line points and the estimated lumen diameter meets the preset lumen stenosis condition, determining the lumen parts corresponding to the continuous central line points as the lumen stenosis sections.
In this embodiment, the lumen stenosis condition may be that a difference between the lumen diameter corresponding to the preset number of consecutive centerline points and the estimated lumen diameter is smaller than a preset difference threshold. The preset difference threshold and the preset number may be set empirically, and are not limited herein.
And S350, filling the narrow section of the lumen to obtain a narrow filling image.
And S360, combining the narrow filling image and the initial lumen labeling image to obtain a target lumen labeling image.
And S370, performing iterative training on the lumen segmentation model to be trained based on the lumen sample image and the target lumen labeling image corresponding to the lumen sample image to obtain the trained lumen segmentation model.
According to the technical scheme of the embodiment of the invention, the lumen sample image and the initial lumen labeling image corresponding to the lumen sample image are obtained; for any centerline point in the centerline of the lumen, determining the corresponding lumen diameter of the centerline point in the initial lumen labeling image; determining an estimated lumen diameter based on the corresponding lumen diameter of the centerline point in the initial lumen labeling image; determining a lumen stenosis section based on the corresponding lumen diameter and estimated lumen diameter of the centerline point in the initial lumen labeling image; filling the narrow section of the lumen to obtain a narrow filling image; merging the narrow filling image and the initial lumen labeling image to obtain a target lumen labeling image; and performing iterative training on the lumen segmentation model to be trained based on the lumen sample image and the target lumen labeling image corresponding to the lumen sample image to obtain the trained lumen segmentation model. Above-mentioned technical scheme has realized the definite of lumen narrow section through lumen diameter and estimation lumen diameter, and is further, through filling the lumen narrow section, has increased the saturation of lumen to can avoid cutting apart the interrupt phenomenon because of the lumen that the stenosis arouses, guarantee the integrality of lumen.
Example four
Fig. 10 is a flowchart of a method for training a lumen segmentation model according to a fourth embodiment of the present invention, and the method of the present embodiment may be combined with various alternatives of the method for training a lumen segmentation model according to the fourth embodiment. The training method of the lumen segmentation model provided by the embodiment is further optimized. Optionally, the filling the lumen stenosis to obtain a stenosis filling image includes: obtaining a starting position and an ending position of the lumen stenosis; based on the starting position and the ending position of the lumen stenosis, filling the lumen diameter of the lumen stenosis to an estimated lumen diameter, resulting in a stenosis filling image.
As shown in fig. 10, the method includes:
s410, acquiring a lumen sample image and an initial lumen labeling image corresponding to the lumen sample image.
And S420, based on the lumen central line corresponding to the initial lumen marking image, performing stenosis detection on the initial lumen marking image, and determining a lumen stenosis section.
S430, acquiring the starting position and the ending position of the lumen narrow section.
S440, filling the lumen diameter of the lumen stenosis section to an estimated lumen diameter based on the starting position and the ending position of the lumen stenosis section, and obtaining a stenosis filling image.
S450, merging the narrow filling image and the initial lumen labeling image to obtain a target lumen labeling image.
And S460, performing iterative training on the lumen segmentation model to be trained based on the lumen sample image and the target lumen labeling image corresponding to the lumen sample image to obtain the trained lumen segmentation model.
In this embodiment, the center line points at the two ends of the lumen narrowing section can be respectively used as the starting position and the ending position, and the lumen diameters corresponding to the center line points in the starting position and the ending position are filled to the estimated lumen diameter, so that the lumen narrowing section is close to the normal lumen size, thereby avoiding the lumen segmentation interruption phenomenon caused by the stenosis and ensuring the integrity and the saturation of the lumen.
In some alternative embodiments, fig. 11 is a schematic structural diagram of image merging provided in the embodiments of the present invention. Taking a coronary medical image as an example, the coronary medical image can be manually marked to obtain an initial lumen marking image, and the initial lumen marking image is subjected to center line extraction to obtain a lumen center line; further, the determination of the stenosis filling image and the determination of the lumen label optimization image can be executed in parallel, specifically, for any centerline point in the lumen centerline, the distance from the centerline point to the starting point of the lumen centerline is determined, and then the lumen label value of the initial lumen labeling image is updated based on the distance from the centerline point to the starting point of the lumen centerline, so as to obtain the lumen label optimization image; determining the lumen diameter corresponding to each central line point, performing stenosis detection according to the lumen diameter corresponding to each central line point to obtain a lumen stenosis section, filling the lumen stenosis section to obtain a stenosis filling image, and combining the stenosis filling image and the lumen label optimization image to obtain a target lumen labeling image.
In some optional embodiments, after the image to be segmented of the coronary artery is input to the model of the trained lumen segmentation model, a predicted segmentation image may be obtained, where the predicted segmentation image is a preliminarily predicted coronary artery segmentation result image, and the predicted segmentation image may be displayed in a binarization form, where element 1 is a coronary artery blood vessel and element 0 is a background. After the prediction segmentation image is obtained, maximum connected region processing can be carried out on the prediction segmentation image to obtain a maximum connected region image, and non-maximum connected region screening is achieved; further, after obtaining the maximum connected region image, the patch can be removed from the maximum connected region image to obtain a target segmentation image, wherein the target segmentation image is a final coronary artery segmentation result image. Illustratively, the images may be optimized by the luminal gradient, it being understood that the derivative at the plaque location is more different than the derivative at the normal lumen, and detection of the plaque location may be achieved. Plaque removal can be performed by setting a plaque screening threshold, for example, the plaque screening threshold can be set to dark plaque <130HU, calcified plaque >350HU, and the like.
Illustratively, the maximum connected component may be represented by a mask, and the final coronary segmentation result image = mask × D, where D is a circumscribed rectangular region of the maximum connected component. Fig. 12 is a diagram of a preliminary predicted coronary artery segmentation result image according to an embodiment of the present invention, and fig. 13 is a diagram of a final coronary artery segmentation result image according to an embodiment of the present invention.
According to the technical scheme of the embodiment of the invention, the lumen sample image and the initial lumen labeling image corresponding to the lumen sample image are obtained; based on the lumen center line corresponding to the initial lumen marking image, performing stenosis detection on the initial lumen marking image, and determining a lumen stenosis section; acquiring a starting position and an ending position of a lumen stenosis section; filling the lumen diameter of the lumen stenosis section to an estimated lumen diameter based on the starting position and the ending position of the lumen stenosis section to obtain a stenosis filling image; merging the narrow filling image and the initial lumen labeling image to obtain a target lumen labeling image; and performing iterative training on the lumen segmentation model to be trained based on the lumen sample image and the target lumen labeling image corresponding to the lumen sample image to obtain the trained lumen segmentation model. Above-mentioned technical scheme fills the lumen narrow section through detecting the lumen narrow section, increases the saturation of lumen to can avoid cutting apart the interrupt phenomenon because of the lumen that the stenosis arouses, guarantee the integrality of lumen.
EXAMPLE five
Fig. 14 is a schematic structural diagram of a training apparatus for a lumen segmentation model according to a fifth embodiment of the present invention. As shown in fig. 14, the apparatus includes:
a sample obtaining module 510, configured to obtain a lumen sample image and an initial lumen labeling image corresponding to the lumen sample image;
a lumen stenosis section determining module 520, configured to perform stenosis detection on the initial lumen labeling image based on a lumen centerline corresponding to the initial lumen labeling image, and determine a lumen stenosis section;
a lumen stenosis filling module 530, configured to fill the lumen stenosis section, so as to obtain a stenosis filling image;
an image merging module 540, configured to merge the narrow filling image and the initial lumen labeling image to obtain a target lumen labeling image;
and a segmentation model training module 550, configured to perform iterative training on the to-be-trained lumen segmentation model based on the lumen sample image and the target lumen labeling image corresponding to the lumen sample image, to obtain a trained lumen segmentation model.
According to the technical scheme of the embodiment of the invention, the lumen sample image and the initial lumen labeling image corresponding to the lumen sample image are obtained; based on the lumen center line corresponding to the initial lumen marking image, performing stenosis detection on the initial lumen marking image, and determining a lumen stenosis section; filling the narrow section of the lumen to obtain a narrow filling image; merging the narrow filling image and the initial lumen labeling image to obtain a target lumen labeling image; and performing iterative training on the lumen segmentation model to be trained based on the lumen sample image and the target lumen labeling image corresponding to the lumen sample image to obtain the trained lumen segmentation model. Above-mentioned technical scheme fills the lumen narrow section through detecting the lumen narrow section, increases the saturation of lumen to can avoid cutting apart the interrupt phenomenon because of the lumen that the stenosis arouses, guarantee the integrality of lumen.
In some optional embodiments, the apparatus further comprises:
a lumen label value updating module, configured to update, for any center line point in a lumen center line corresponding to the initial lumen labeling image, a lumen label value of the initial lumen labeling image based on a distance from the center line point to a starting point of the lumen center line, so as to obtain a lumen label optimized image;
an image merging module 540, further configured to:
and merging the narrow filling image and the lumen label optimization image to obtain a target lumen labeling image.
In some optional embodiments, the lumen tag value updating module is specifically configured to:
matching the distance from the center line point to the starting point of the center line of the tube cavity in a preset distance-label matching relation to obtain a distance optimization label value;
and updating the lumen label value of the initial lumen labeling image based on the distance optimization label value to obtain a lumen label optimization image.
In some alternative embodiments, the lumen narrowing section determination module 520 includes:
a lumen diameter determining unit, configured to determine, for any centerline point in a lumen centerline corresponding to the initial lumen labeling image, a lumen diameter corresponding to the centerline point in the initial lumen labeling image;
an estimated lumen diameter determining unit, configured to determine an estimated lumen diameter based on a corresponding lumen diameter of the centerline point in the initial lumen labeling image;
a lumen stenosis determining unit for determining a lumen stenosis based on the corresponding lumen diameter of the centerline point in the initial lumen labeling image and the estimated lumen diameter.
In some alternative embodiments, the lumen stenosis determination unit is specifically configured to:
and if the difference value between the lumen diameter corresponding to the plurality of continuous central line points and the estimated lumen diameter meets the preset lumen stenosis condition, determining the plurality of continuous central line points as the lumen stenosis section.
In some alternative embodiments, the lumen stenosis filling module 530 is specifically configured to:
obtaining a starting position and an ending position of the lumen stenosis;
based on the starting and ending locations of the lumen stenosis, filling the lumen diameter of the lumen stenosis to an estimated lumen diameter resulting in a stenosis filling image.
In some optional embodiments, the apparatus is further configured to:
for any central line point in the lumen narrow section, matching the distance from the central line point in the lumen narrow section to the starting point of the central line of the lumen in a preset distance-label matching relation to obtain a narrow label value;
and updating the lumen label value of the stenosis filling image based on the stenosis label value to obtain an updated stenosis filling image.
In some optional embodiments, the apparatus further comprises:
the image to be segmented acquisition module is used for acquiring at least one image to be segmented;
and the segmented image prediction module is used for inputting the image to be segmented into the trained lumen segmentation model to obtain a target segmented image corresponding to the image to be segmented.
The device for training the lumen segmentation model provided by the embodiment of the invention can execute the method for training the lumen segmentation model provided by any embodiment of the invention, and has corresponding functional modules and beneficial effects of the execution method.
Example six
FIG. 15 illustrates a block diagram of an electronic device 10 that may be used to implement embodiments of the present invention. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing devices, cellular phones, smart phones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the inventions described and/or claimed herein.
As shown in fig. 15, the electronic device 10 includes at least one processor 11, and a memory communicatively connected to the at least one processor 11, such as a Read Only Memory (ROM) 12, a Random Access Memory (RAM) 13, and the like, wherein the memory stores a computer program executable by the at least one processor, and the processor 11 may perform various appropriate actions and processes according to the computer program stored in the Read Only Memory (ROM) 12 or the computer program loaded from the storage unit 18 into the Random Access Memory (RAM) 13. In the RAM 13, various programs and data necessary for the operation of the electronic apparatus 10 may also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other via a bus 14. An input/output (I/O) interface 15 is also connected to the bus 14.
A number of components in the electronic device 10 are connected to the I/O interface 15, including: an input unit 16 such as a keyboard, a mouse, or the like; an output unit 17 such as various types of displays, speakers, and the like; a storage unit 18 such as a magnetic disk, an optical disk, or the like; and a communication unit 19 such as a network card, modem, wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information/data with other devices via a computer network, such as the internet, and/or various telecommunication networks.
The processor 11 may be a variety of general and/or special purpose processing components having processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), various specialized Artificial Intelligence (AI) computing chips, various processors running machine learning model algorithms, a Digital Signal Processor (DSP), and any suitable processor, controller, microcontroller, or the like. The processor 11 performs the various methods and processes described above, such as a method of training a lumen segmentation model, the method comprising:
acquiring a lumen sample image and an initial lumen labeling image corresponding to the lumen sample image;
based on the lumen center line corresponding to the initial lumen marking image, performing stenosis detection on the initial lumen marking image, and determining a lumen stenosis section;
filling the narrow section of the lumen to obtain a narrow filling image;
merging the narrow filling image and the initial lumen labeling image to obtain a target lumen labeling image;
and performing iterative training on the lumen segmentation model to be trained based on the lumen sample image and the target lumen labeling image corresponding to the lumen sample image to obtain the trained lumen segmentation model.
In some embodiments, the method of training the lumen segmentation model may be implemented as a computer program tangibly embodied in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and/or installed onto the electronic device 10 via the ROM 12 and/or the communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the method of training a lumen segmentation model described above may be performed. Alternatively, in other embodiments, the processor 11 may be configured by any other suitable means (e.g. by means of firmware) to perform a method of training the lumen segmentation model.
Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuitry, field Programmable Gate Arrays (FPGAs), application Specific Integrated Circuits (ASICs), application Specific Standard Products (ASSPs), system on a chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs that are executable and/or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, receiving data and instructions from, and transmitting data and instructions to, a storage system, at least one input device, and at least one output device.
A computer program for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the computer programs, when executed by the processor, cause the functions/acts specified in the flowchart and/or block diagram block or blocks to be performed. A computer program can execute entirely on a machine, partly on a machine, as a stand-alone software package partly on a machine and partly on a remote machine or entirely on a remote machine or server.
In the context of the present invention, a computer-readable storage medium may be a tangible medium that can contain, or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. A computer readable storage medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. Alternatively, the computer readable storage medium may be a machine readable signal medium. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, 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 compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
To provide for interaction with a user, the systems and techniques described here can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which a user may provide input to the electronic device. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic, speech, or tactile input.
The systems and techniques described here can be implemented in a computing system that includes a back-end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), wide Area Networks (WANs), blockchain networks, and the internet.
The computing system may include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, also called a cloud computing server or a cloud host, and is a host product in a cloud computing service system, so that the defects of high management difficulty and weak service expansibility in the traditional physical host and VPS service are overcome.
It should be understood that various forms of the flows shown above may be used, with steps reordered, added, or deleted. For example, the steps described in the present invention may be executed in parallel, sequentially, or in different orders, and are not limited herein as long as the desired result of the technical solution of the present invention can be achieved.
The above-described embodiments should not be construed as limiting the scope of the invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made in accordance with design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (11)
1. A method for training a lumen segmentation model, comprising:
acquiring a lumen sample image and an initial lumen labeling image corresponding to the lumen sample image;
based on the lumen center line corresponding to the initial lumen marking image, performing stenosis detection on the initial lumen marking image, and determining a lumen stenosis section;
filling the lumen stenosis to obtain a stenosis filling image;
merging the narrow filling image and the initial lumen labeling image to obtain a target lumen labeling image;
and carrying out iterative training on the lumen segmentation model to be trained based on the lumen sample image and the target lumen labeling image corresponding to the lumen sample image to obtain the trained lumen segmentation model.
2. The method of claim 1, wherein after the obtaining the lumen sample image and an initial lumen annotation image corresponding to the lumen sample image, the method further comprises:
updating the lumen label value of the initial lumen labeling image based on the distance from the centerline point to the starting point of the lumen centerline for any centerline point in the lumen centerline corresponding to the initial lumen labeling image to obtain a lumen label optimized image;
correspondingly, the merging the stenosis filling image and the initial lumen labeling image to obtain a target lumen labeling image includes:
and combining the narrow filling image and the lumen label optimization image to obtain a target lumen labeling image.
3. The method according to claim 2, wherein the updating the lumen label value of the initial lumen labeling image based on the distance from the centerline point to the starting point of the lumen centerline, resulting in a lumen label optimized image, comprises:
matching the distance from the center line point to the starting point of the center line of the tube cavity in a preset distance-label matching relation to obtain a distance optimization label value;
and updating the lumen label value of the initial lumen labeling image based on the distance optimization label value to obtain a lumen label optimization image.
4. The method according to claim 1, wherein the performing stenosis detection on the initial lumen labeling image based on the lumen centerline corresponding to the initial lumen labeling image to determine a lumen stenosis section comprises:
determining the corresponding lumen diameter of the centerline point in the initial lumen marking image for any centerline point in the lumen centerline corresponding to the initial lumen marking image;
performing linear fitting processing on the diameters of the tube cavities corresponding to all the central line points on the central line of the tube cavity to obtain the estimated tube cavity diameters corresponding to all the central line points;
determining at least one lumen stenosis based on the corresponding lumen diameter of the centerline point in the initial lumen annotation image and the estimated lumen diameter.
5. The method according to claim 4, wherein the determining at least one lumen stenosis based on the corresponding lumen diameter of the centerline point in the initial lumen labeling image and the estimated lumen diameter comprises:
the determining step for any one of the luminal narrowings comprising: and if the difference value between the lumen diameter corresponding to the continuous central line points and the estimated lumen diameter meets the preset lumen stenosis condition, determining the lumen parts corresponding to the continuous central line points as the lumen stenosis sections.
6. The method of claim 1, wherein said filling the luminal stenosis resulting in a stenosis filling image comprises:
obtaining a starting position and an ending position of the lumen narrowing;
based on the starting and ending locations of the lumen stenosis, filling the lumen diameter of the lumen stenosis to an estimated lumen diameter resulting in a stenosis filling image.
7. The method of claim 6, further comprising, after said obtaining a stenosis filling image:
for any central line point in the lumen stenosis section, matching the distance from the central line point in the lumen stenosis section to the starting point of the lumen central line in a preset distance-label matching relation to obtain a stenosis label value;
and updating the lumen label value of the stenosis filling image based on the stenosis label value to obtain an updated stenosis filling image.
8. The method according to any one of claims 1-7, further comprising, after said deriving a trained lumen segmentation model:
acquiring at least one image to be segmented;
inputting the image to be segmented into the trained lumen segmentation model to obtain a predicted segmentation image corresponding to the image to be segmented;
processing the maximum connected region of the prediction segmentation image to obtain a maximum connected region image;
and performing plaque removal processing on the maximum connected region image to obtain a target segmentation image.
9. A device for training a lumen segmentation model, comprising:
the system comprises a sample acquisition module, a detection module and a display module, wherein the sample acquisition module is used for acquiring a lumen sample image and an initial lumen labeling image corresponding to the lumen sample image;
a lumen stenosis section determining module, configured to perform stenosis detection on the initial lumen labeling image based on a lumen center line corresponding to the initial lumen labeling image, and determine a lumen stenosis section;
the lumen stenosis filling module is used for filling the lumen stenosis section to obtain a stenosis filling image;
the image merging module is used for merging the narrow filling image and the initial lumen labeling image to obtain a target lumen labeling image;
and the segmentation model training module is used for carrying out iterative training on the lumen segmentation model to be trained on the basis of the lumen sample image and the target lumen labeling image corresponding to the lumen sample image to obtain the trained lumen segmentation model.
10. An electronic device, characterized in that the electronic device comprises:
at least one processor; and
a memory communicatively coupled to the at least one processor; wherein, the first and the second end of the pipe are connected with each other,
the memory stores a computer program executable by the at least one processor to enable the at least one processor to perform the method of training a lumen segmentation model according to any one of claims 1-8.
11. A computer-readable storage medium, having stored thereon computer instructions for causing a processor, when executed, to implement a method of training a lumen segmentation model according to any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211150263.1A CN115423826A (en) | 2022-09-21 | 2022-09-21 | Method and device for training lumen segmentation model, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211150263.1A CN115423826A (en) | 2022-09-21 | 2022-09-21 | Method and device for training lumen segmentation model, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115423826A true CN115423826A (en) | 2022-12-02 |
Family
ID=84204986
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211150263.1A Pending CN115423826A (en) | 2022-09-21 | 2022-09-21 | Method and device for training lumen segmentation model, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115423826A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116612102A (en) * | 2023-05-31 | 2023-08-18 | 上海博动医疗科技股份有限公司 | Vascular image processing system, vascular image processing device, and storage medium |
-
2022
- 2022-09-21 CN CN202211150263.1A patent/CN115423826A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116612102A (en) * | 2023-05-31 | 2023-08-18 | 上海博动医疗科技股份有限公司 | Vascular image processing system, vascular image processing device, and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115423826A (en) | Method and device for training lumen segmentation model, electronic equipment and storage medium | |
| CN115564787A (en) | Method and device for training lumen segmentation model, electronic equipment and storage medium | |
| CN113971728B (en) | Image recognition method, training method, device, equipment and medium for model | |
| CN113313715A (en) | Method, device, apparatus and medium for segmenting cardiac artery blood vessel | |
| CN110288572B (en) | Automatic extraction method and device for blood vessel center line | |
| CN115482261A (en) | Blood vessel registration method, device, electronic equipment and storage medium | |
| CN116843622A (en) | Vascular centerline determination system, device, and storage medium | |
| CN115861189A (en) | Image registration method and device, electronic equipment and storage medium | |
| CN115762798A (en) | Method and device for determining hemodynamic parameters, storage medium and electronic equipment | |
| CN114937149A (en) | Image processing method, image processing device, electronic equipment and storage medium | |
| CN112927319B (en) | Model training method, image processing method, device, equipment and storage medium | |
| CN115690143A (en) | Image segmentation method and device, electronic equipment and storage medium | |
| CN117496055A (en) | Vascular reconstruction method, vascular reconstruction device, electronic equipment and storage medium | |
| CN115908821A (en) | Blood vessel image segmentation method, device, electronic device and storage medium | |
| CN115880494A (en) | Blood vessel image segmentation method, device, electronic device and storage medium | |
| CN115482358B (en) | Triangular mesh curved surface generation method, device, equipment and storage medium | |
| CN115512186A (en) | Model training method and device, electronic equipment and storage medium | |
| CN115147435A (en) | Image segmentation method, image segmentation device, electronic equipment and storage medium | |
| CN116740094A (en) | Blood vessel segmentation method, blood vessel segmentation device, electronic equipment and storage medium | |
| CN117764911A (en) | Blood vessel naming method, device, equipment and medium | |
| CN114972242B (en) | Training method and device for myocardial bridge detection model and electronic equipment | |
| CN117422687A (en) | Image processing method and device, electronic equipment and storage medium | |
| CN117689680A (en) | Pulmonary artery segmentation method and device, electronic equipment and storage medium | |
| CN117274332A (en) | Contrast and intra-cavity image registration interaction method, device, equipment and medium | |
| CN115984562A (en) | Blood vessel image segmentation method, device, electronic 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 |