CN113822853A - Cervical vertebra MR image curvature measuring method based on superpixel segmentation - Google Patents

Cervical vertebra MR image curvature measuring method based on superpixel segmentation Download PDF

Info

Publication number
CN113822853A
CN113822853A CN202110864656.8A CN202110864656A CN113822853A CN 113822853 A CN113822853 A CN 113822853A CN 202110864656 A CN202110864656 A CN 202110864656A CN 113822853 A CN113822853 A CN 113822853A
Authority
CN
China
Prior art keywords
image
clustering
segmentation
super
color
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
Application number
CN202110864656.8A
Other languages
Chinese (zh)
Inventor
田应仲
诸俊杰
李龙
胡慧娟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Shanghai for Science and Technology
Original Assignee
University of Shanghai for Science and Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by University of Shanghai for Science and Technology filed Critical University of Shanghai for Science and Technology
Priority to CN202110864656.8A priority Critical patent/CN113822853A/en
Publication of CN113822853A publication Critical patent/CN113822853A/en
Pending legal-status Critical Current

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
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/20—Image enhancement or restoration using local operators
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/70—Denoising; Smoothing
    • 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
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/64—Analysis of geometric attributes of convexity or concavity
    • 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
    • G06T7/66—Analysis of geometric attributes of image moments or centre of gravity
    • 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
    • G06T2207/10088—Magnetic resonance imaging [MRI]
    • 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
    • G06T2207/20024—Filtering details
    • G06T2207/20028—Bilateral filtering
    • 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
    • G06T2207/20172—Image enhancement details
    • G06T2207/20192—Edge enhancement; Edge preservation
    • 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
    • G06T2207/30008—Bone
    • G06T2207/30012—Spine; Backbone

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Probability & Statistics with Applications (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Geometry (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Quality & Reliability (AREA)
  • Health & Medical Sciences (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a cervical spine MR image curvature measuring method based on superpixel segmentation, which comprises the following steps: spinal cord candidate points are determined by comparing longitudinal gray gradient average values in adjacent regions, and a cervical vertebra MR image corresponding to an optimal image layer is determined according to the integrity of the spinal cord region; removing noise and strong anisotropic interference in the MR image by adopting bilateral filtering; adopting K-means clustering on the multi-resolution image, combining the pixel points with similar positions and characteristics to form super pixels, completing image segmentation and reducing the image complexity; combining the small super pixels to form a new super pixel area by adopting a DBSCAN clustering method, wherein the new super pixel area is overlapped with the vertebral body; selecting a vertebral body C2-C7 from the super-pixel combination result in a man-machine interaction mode for segmentation; and calculating the mass center of each vertebral body, and measuring the curvature by adopting a method for measuring the lordosis of the cervical vertebra by using the mass center of each vertebral body. The invention realizes the semi-automatic cervical curvature measurement of the cervical spine MR image, can greatly save the precious time of medical experts and improve the working efficiency.

Description

Cervical vertebra MR image curvature measuring method based on superpixel segmentation
Technical Field
The invention belongs to the field of cervical vertebra curvature measurement, and particularly relates to a cervical vertebra MR image curvature measurement method based on superpixel segmentation.
Background
The quantitative measurement of cervical vertebra is a practical means for clinical diagnosis and evaluation of cervical spondylosis. The cervical curvature is one of the important indexes for judging cervical spondylosis, and has important significance for the clinical treatment of cervical spondylosis. Most of the current cervical curvature measuring methods are judged based on medical images, and they mainly depend on the experience of medical experts and their manual operation ability. Manual measurement requires a significant amount of valuable time by a medical professional and human error. Even experienced medical professionals have difficulty in reliably and repeatably measuring the curvature of the cervical spine.
Disclosure of Invention
Aiming at the defects of long time consumption, poor consistency, low repeatability and the like of the traditional manual measurement, the invention provides the cervical vertebra MR image curvature measuring method based on superpixel segmentation, and the semi-automatic measurement of the cervical vertebra curvature can be realized.
In order to achieve the purpose, the invention adopts the technical scheme that:
a cervical vertebra MR image curvature measuring method based on superpixel segmentation is characterized by comprising the following operation steps:
step 1, optimal layer selection:
determining a cervical vertebra MR image corresponding to the optimal image layer by comparing the integrity of spinal cord regions in different image layers;
step 2, pretreatment:
bilateral filtering is carried out on the selected optimal image, and noise and strong anisotropic interference in the MR image are removed;
and 3, super-pixel segmentation:
performing K-means clustering on the multi-resolution image, combining the pixel points with similar positions and characteristics to form super pixels, completing image segmentation, and reducing the image complexity;
and 4, super-pixel combination:
due to the limitation of a search range, the superpixels obtained in the last step are often small areas, a DBSCAN clustering method is adopted to combine the small superpixels to form a new superpixel area, and the new superpixel area is overlapped with the vertebral body;
step 5, C2-C7 segmentation of vertebral body:
the doctor selects vertebral bodies C2-C7 from the super-pixel combination result in the step 4 in a man-machine interaction mode for segmentation;
step 6, measuring the curvature of the cervical vertebra:
and (5) solving the mass center of each vertebral body based on the vertebral body segmentation result in the step (5), and then measuring the curvature by adopting a method for measuring the lordosis of the cervical vertebra by using the mass center of the vertebral body.
Preferably, in the optimal layer selection in step 1, comparing longitudinal gray gradient averages in adjacent regions in the same layer to determine whether the central pixel point is a spinal cord candidate point, taking a maximum spinal cord region formed by the final spinal cord candidate point, and then taking a layer corresponding to the maximum height of the spinal cord regions in different layers as the optimal layer.
Preferably, the preprocessing in step 2 is to process the image in the Lab color space by using bilateral filtering, so as to smooth the image while preserving the edges of the image, and remove noise and strong anisotropic interference of the MR image.
Preferably, in the step 3, superpixel segmentation is performed, the number k of superpixels is initialized, a multi-resolution image is realized by using an image pyramid, clustering centers are distributed on the image which is sampled once, and the sampling distances of the rows of the clustering centers are
Figure BDA0003187114190000021
N is the number of image pixel points and the line sampling distance
Figure BDA0003187114190000022
Traversing up-sampled imagesSetting the pixel point with the minimum characteristic value as the same type as the clustering center, then respectively updating the position and color information of the clustering center by using the position characteristic and color characteristic average value of the super-pixel, remapping the obtained new clustering center to the original image, traversing the pixel points in the original image and calculating the color and position characteristics of the pixel points and the clustering center, setting the pixel point with the minimum characteristic value as the same type as the clustering center, updating the clustering center, and repeating iteration. Initial clustering is carried out on the images sampled once, so that the global attributes of the images can be better captured while the initial clustering speed is improved. The calculation formula of the colors and the positions of the characterization pixel points and the clustering centers is as follows:
Figure BDA0003187114190000023
wherein i represents a cluster center, j represents a pixel point, dcIndicating the color distance, dsRepresenting the distance of the position, D representing the total distance between the pixel point and the cluster center, Li、ai、biColor feature, x, representing cluster centeri、yiFeatures indicating the position of the cluster center, Lj、aj、bjRepresenting the colour characteristics, x, of the pixel pointsj、yjThe position characteristics of the pixel points are represented, and w represents a weighting factor between the color and the position difference.
Preferably, the super-pixel combination in step 4 adopts a DBSCAN clustering method, and defaults to sequentially set unmarked super-pixels as seed super-pixels from left to right and from bottom to top of the image, and calculate the color feature similarity between adjacent unmarked super-pixels and seed super-pixels, and if the feature value is less than the threshold, combine the super-pixels and seed super-pixels. All superpixels are traversed until all superpixels are marked. The new super-pixel regions formed by merging will coincide with the vertebral bodies. The calculation formula for representing the similarity of the color features is as follows:
Dc＝(Ln-Lm)2+(an-am)2+(bn-bm)2 (2)
where m represents a seed superpixel, n represents an adjacent superpixel, Lm、am、bmRepresenting the color characteristics of the seed superpixel, Ln、an、bnRepresenting the color characteristics of neighboring superpixels.
Compared with the prior art, the invention has the following obvious and prominent substantive characteristics and remarkable technical progress:
1. the spinal cord region of the image layer is determined by comparing the longitudinal gray gradient average values of adjacent regions, then the heights of the spinal cord regions corresponding to different image layers are compared, and the image layer with the maximum height value is used as the optimal image layer;
2. according to the invention, the cervical vertebra MR image is preprocessed in the Lab color space by adopting bilateral filtering, so that the image can be smoothed while the edge is kept, and the noise and strong anisotropic interference of the MR image are removed;
3. the method comprises the steps of realizing a multi-resolution image by adopting an image pyramid, distributing a clustering center on the image sampled at the last time, traversing pixel points in the image sampled at the last time, calculating color and position characteristics of the pixel points and the clustering center, then updating the clustering center, remapping the obtained new clustering center to an original image, traversing the pixel points in the original image, calculating color and position characteristics of the pixel points and the clustering center, updating the clustering center, and repeating iteration; initial clustering is carried out on the image sampled once, so that the global attribute of the image can be better captured while the initial clustering speed is improved;
4. the invention adopts a DBSCAN clustering method to combine small superpixels with similar color characteristics to form a new superpixel area, cuts off a vertebral body C2-C7 by adopting a man-machine interaction mode, and measures the curvature of the cervical vertebra by using a vertebral body centroid measurement cervical lordosis method.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
Fig. 2 is a process diagram of extracting spinal cord region under a certain layer of the method of the present invention.
FIG. 3 is a schematic representation of the pretreatment results of the process of the present invention.
Fig. 4 is a view of a vertebral body segmentation process of the method of the present invention.
Fig. 5 is a schematic view of the cervical curvature measurement of the method of the present invention.
Detailed Description
The invention will be further described with reference to the accompanying drawings and preferred embodiments:
the first embodiment is as follows:
referring to fig. 1, a method for measuring curvature of a cervical spine MR image based on superpixel segmentation comprises the following steps:
step 1, optimal layer selection:
determining a cervical vertebra MR image corresponding to the optimal image layer by comparing the integrity of spinal cord regions in different image layers;
step 2, pretreatment:
bilateral filtering is carried out on the selected optimal image, and noise and strong anisotropic interference in the MR image are removed;
and 3, super-pixel segmentation:
performing K-means clustering on the multi-resolution image, combining the pixel points with similar positions and characteristics to form super pixels, completing image segmentation, and reducing the image complexity;
and 4, super-pixel combination:
due to the limitation of a search range, the superpixels obtained in the last step are often small areas, a DBSCAN clustering method is adopted to combine the small superpixels to form a new superpixel area, and the new superpixel area is overlapped with the vertebral body;
step 5, C2-C7 segmentation of vertebral body:
the doctor selects vertebral bodies C2-C7 from the super-pixel combination result in the step 4 in a man-machine interaction mode for segmentation;
step 6, measuring the curvature of the cervical vertebra:
and (5) solving the mass center of each vertebral body based on the vertebral body segmentation result in the step (5), and then measuring the curvature by adopting a method for measuring the lordosis of the cervical vertebra by using the mass center of the vertebral body.
The curvature measuring method of the cervical vertebra MR image based on superpixel segmentation can realize semi-automatic measurement of the curvature of the cervical vertebra.
Example two:
this embodiment is substantially the same as the first embodiment, and is characterized in that:
in the step 1, the optimal layer selection is performed by comparing average longitudinal gray gradients in adjacent regions in the same layer, determining whether the central pixel point is a spinal cord candidate point, taking a maximum spinal cord region formed by the final spinal cord candidate point, and then taking a layer corresponding to the maximum height of the spinal cord regions in different layers as an optimal layer.
In the preprocessing in the step 2, the image is processed in the Lab color space by adopting bilateral filtering, the image is smoothed while the edge of the image is kept, and noise and strong anisotropic interference of the MR image are removed.
In the step 3, superpixel segmentation is performed, the number k of superpixels is initialized, a multi-resolution image is realized by adopting an image pyramid, clustering centers are distributed on the image which is sampled once, and the row sampling distance of the clustering centers is
Figure BDA0003187114190000041
N is the number of image pixel points and the line sampling distance
Figure BDA0003187114190000042
Traversing pixel points in the up-sampled image, calculating the color and position characteristics of the pixel points and a clustering center, setting the pixel points corresponding to the minimum characteristic value as the same type as the clustering center, then respectively updating the position and color information of the clustering center by using the position characteristics and color characteristic average values of the super pixels, remapping the obtained new clustering center to the original image, traversing the pixel points in the original image, calculating the color and position characteristics of the pixel points and the clustering center, setting the pixel points corresponding to the minimum characteristic value as the same type as the clustering center, updating the clustering center, and repeating iteration; initial clustering is carried out on the image sampled at the last time, and the global attribute of the image is better captured while the initial clustering speed is improved; the calculation formula of the colors and the positions of the characterization pixel points and the clustering centers is as follows:
Figure BDA0003187114190000043
wherein i represents a cluster center, j represents a pixel point, dcIndicating the color distance, dsRepresenting the distance of the position, D representing the total distance between the pixel point and the cluster center, Li、ai、biColor feature, x, representing cluster centeri、yiFeatures indicating the position of the cluster center, Lj、aj、bjRepresenting the colour characteristics, x, of the pixel pointsj、yjThe position characteristics of the pixel points are represented, and w represents a weighting factor between the color and the position difference.
Combining the super pixels in the step 4, adopting a DBSCAN clustering method, defaulting to sequentially set unmarked super pixels as seed super pixels from left to right and from bottom to top of the image, calculating the color feature similarity between adjacent unmarked super pixels and the seed super pixels, and combining the super pixels and the seed super pixels if the feature value is less than a threshold value; traversing all the super pixels until all the super pixels are marked; combining the formed new super pixel areas to be superposed with the vertebral body; the calculation formula for representing the similarity of the color features is as follows:
Dc＝(Ln-Lm)2+(an-am)2+(bn-bm)2 (2)
where m represents a seed superpixel, n represents an adjacent superpixel, Lm、am、bmRepresenting the color characteristics of the seed superpixel, Ln、an、bnRepresenting the color characteristics of neighboring superpixels.
In the embodiment, the spinal cord regions of the image layers are determined by comparing the longitudinal gray gradient average values of adjacent regions, then the heights of the spinal cord regions corresponding to different image layers are compared, and the image layer with the maximum height value is used as the optimal image layer; according to the embodiment, the cervical vertebra MR image is preprocessed in the Lab color space by adopting bilateral filtering, so that the image can be smoothed while the edge is kept, and noise and strong anisotropic interference of the MR image are removed; in the embodiment, an image pyramid is adopted to realize a multi-resolution image, a clustering center is distributed on an image sampled once, pixel points in the image sampled once are traversed, the color and position characteristics of the pixel points and the clustering center are calculated, then the clustering center is updated, the obtained new clustering center is mapped to an original image again, the pixel points in the original image are traversed, the color and position characteristics of the pixel points and the clustering center are calculated, the clustering center is updated, and iteration is repeated; initial clustering is carried out on the image sampled once, so that the global attribute of the image can be better captured while the initial clustering speed is improved; in the embodiment, a DBSCAN clustering method is adopted to combine small superpixels with similar color characteristics to form a new superpixel region, a vertebral body C2-C7 is segmented in a man-machine interaction mode, and a cervical vertebra curvature is measured by using a vertebral body centroid measurement cervical lordosis method.
Example three:
this embodiment is substantially the same as the above embodiment, and is characterized in that:
in this embodiment, a group of cervical spine TI-MR images is used as an object, and a cervical spine curvature measurement method is shown in fig. 1, and a cervical spine MR image curvature measurement method based on superpixel segmentation includes the following steps:
step 1, optimal layer selection:
referring to fig. 2(a), the average longitudinal gray level gradients in adjacent regions in the same image layer are compared to determine whether the central pixel point is a spinal cord candidate point, see fig. 2 (b); taking the maximum spinal cord region formed by the final spinal cord candidate points, referring to fig. 2(c), and then taking the layer corresponding to the maximum height of the spinal cord region of different layers as an optimal layer;
step 2, pretreatment:
processing the image in the Lab color space by adopting bilateral filtering, smoothing the image while keeping the edge of the image, and removing noise and strong anisotropic interference in the MR image, as shown in FIG. 3;
and 3, super-pixel segmentation:
initializing 200 super-pixel numbers, realizing a multi-resolution image by adopting an image pyramid, distributing a clustering center on the image sampled once, traversing pixel points in the image sampled once, calculating color and position characteristics of the pixel points and the clustering center, setting the pixel point corresponding to the minimum characteristic value as the same type as the clustering center, then updating the clustering center, remapping the obtained new clustering center to an original image, traversing the pixel points in the original image, calculating color and position characteristics of the pixel points and the clustering center, setting the pixel point corresponding to the minimum characteristic value as the same type as the clustering center, updating the clustering center, repeating iteration, and finishing super-pixel segmentation, wherein the reference is shown in figure 4 (a);
and 4, super-pixel combination:
and adopting a DBSCAN clustering method, defaulting to sequentially set unmarked superpixels as seed superpixels from left to right and from bottom to top of the image, calculating the color feature similarity between adjacent unmarked superpixels and the seed superpixels, and merging the superpixels and the seed superpixels if the feature value is less than a threshold value. Traversing all the super pixels until all the super pixels are marked; the new super-pixel regions formed by merging will coincide with the vertebral bodies, as shown in fig. 4 (b);
step 5, C2-C7 segmentation of vertebral body:
the doctor selects the vertebral body C2-C7 from the super-pixel combination result in the step 4 in a man-machine interaction mode for segmentation, and the segmentation is shown in FIG. 4 (C);
step 6, measuring the curvature of the cervical vertebra:
and (5) solving the mass center of each vertebral body based on the vertebral body segmentation result in the step (5), and then measuring the curvature by adopting a method for measuring the lordosis of the cervical vertebra by using the mass center of the vertebral body, which is shown in figure 5.
Aiming at the defects of long time consumption, poor consistency, low repeatability and the like of the traditional manual measurement, the curvature measurement method of the cervical vertebra MR image based on the super-pixel segmentation can realize the semi-automatic measurement of the curvature of the cervical vertebra.
In summary, the above embodiment is based on the curvature measuring method of the cervical spine MR image by superpixel segmentation. The operation steps are as follows: spinal cord candidate points are determined by comparing longitudinal gray gradient average values in adjacent regions, and a cervical vertebra MR image corresponding to an optimal image layer is determined according to the integrity of the spinal cord region; removing noise and strong anisotropic interference in the MR image by adopting bilateral filtering; adopting K-means clustering on the multi-resolution image, combining the pixel points with similar positions and characteristics to form super pixels, completing image segmentation and reducing the image complexity; combining the small super pixels to form a new super pixel area by adopting a DBSCAN clustering method, wherein the new super pixel area is overlapped with the vertebral body; selecting a vertebral body C2-C7 from the super-pixel combination result in a man-machine interaction mode for segmentation; and calculating the mass center of each vertebral body, and measuring the curvature by adopting a method for measuring the lordosis of the cervical vertebra by using the mass center of each vertebral body. The above embodiment realizes the semi-automatic cervical curvature measurement of the cervical spine MR image, can greatly save the precious time of medical experts, and improves the working efficiency of the medical experts.
The embodiments of the present invention have been described with reference to the accompanying drawings, but the present invention is not limited to the embodiments, and various changes and modifications can be made according to the purpose of the invention, and any changes, modifications, substitutions, combinations or simplifications made according to the spirit and principle of the technical solution of the present invention shall be equivalent substitutions, as long as the purpose of the present invention is met, and the present invention shall fall within the protection scope of the present invention without departing from the technical principle and inventive concept of the present invention.

Claims (5)

1. A cervical vertebra MR image curvature measuring method based on superpixel segmentation is characterized by comprising the following steps:
step 1, optimal layer selection:
determining a cervical vertebra MR image corresponding to the optimal image layer by comparing the integrity of spinal cord regions in different image layers;
step 2, pretreatment:
bilateral filtering is carried out on the selected optimal image, and noise and strong anisotropic interference in the MR image are removed;
and 3, super-pixel segmentation:
performing K-means clustering on the multi-resolution image, combining the pixel points with similar positions and characteristics to form super pixels, completing image segmentation, and reducing the image complexity;
and 4, super-pixel combination:
due to the limitation of the search range, the superpixels obtained in the last step are often small areas, so that a new superpixel area is formed by combining the small superpixels by adopting a DBSCAN clustering method, and the new superpixel area is overlapped with the vertebral body;
step 5, C2-C7 segmentation of vertebral body:
the doctor selects vertebral bodies C2-C7 from the super-pixel combination result in the step 4 in a man-machine interaction mode for segmentation;
step 6, measuring the curvature of the cervical vertebra:
and (5) solving the mass center of each vertebral body based on the vertebral body segmentation result in the step (5), and then measuring the curvature by adopting a method for measuring the lordosis of the cervical vertebra by using the mass center of the vertebral body.
2. The method for measuring curvature of MR image of cervical vertebra based on superpixel segmentation as claimed in claim 1, wherein: in the step 1, the optimal layer selection is performed by comparing the average longitudinal gray gradients in adjacent areas in the same layer to determine whether the central pixel point is a spinal cord candidate point, taking the maximum spinal cord area formed by the final spinal cord candidate point, and then taking the layer corresponding to the maximum height of the spinal cord area in different layers as the optimal layer.
3. The method for measuring curvature of MR image of cervical vertebra based on superpixel segmentation as claimed in claim 1, wherein: the preprocessing in the step 2 is to process the image in the Lab color space by adopting bilateral filtering, so that the image can be smoothed while the edge of the image is kept, and noise and strong anisotropic interference of the MR image are removed.
4. The method for measuring curvature of MR image of cervical vertebra based on superpixel segmentation as claimed in claim 1, wherein: in the step 3, superpixel segmentation is performed, the number k of superpixels is initialized, a multi-resolution image is realized by adopting an image pyramid, clustering centers are distributed on the image which is sampled once, and the row sampling distance of the clustering centers is
Figure FDA0003187114180000011
(N is the number of image pixels), line sampling distance
Figure FDA0003187114180000012
Traversing pixel points in the up-sampled image, calculating the color and position characteristics of the pixel points and the clustering center, setting the pixel points corresponding to the minimum characteristic values to be the same as the clustering center, then respectively updating the position and color information of the clustering center by using the position characteristics and color characteristic average values of the super pixels, remapping the obtained new clustering center to the original image, traversing the pixel points in the original image, calculating the color and position characteristics of the pixel points and the clustering center, setting the pixel points corresponding to the minimum characteristic values to be the same as the clustering center, updating the clustering center, and repeating iteration. Initial clustering is carried out on the images sampled once, so that the global attributes of the images can be better captured while the initial clustering speed is improved. The calculation formula of the colors and the positions of the characterization pixel points and the clustering centers is as follows:
Figure FDA0003187114180000021
wherein i represents a cluster center, j represents a pixel point, dcIndicating the color distance, dsRepresenting the distance of the position, D representing the total distance between the pixel point and the cluster center, Li、ai、biColor feature, x, representing cluster centeri、yiFeatures indicating the position of the cluster center, Lj、aj、bjRepresenting the colour characteristics, x, of the pixel pointsj、yjThe position characteristics of the pixel points are represented, and w represents a weighting factor between the color and the position difference.
5. The method for measuring curvature of MR image of cervical vertebra based on superpixel segmentation as claimed in claim 1, wherein: and 4, combining the super pixels in the step 4, adopting a DBSCAN clustering method, defaulting to sequentially set unmarked super pixels as seed super pixels from left to right and from bottom to top of the image, calculating the color feature similarity between adjacent unmarked super pixels and the seed super pixels, and combining the super pixels and the seed super pixels if the feature value is less than a threshold value. All superpixels are traversed until all superpixels are marked. The new super-pixel regions formed by merging will coincide with the vertebral bodies. The calculation formula for representing the similarity of the color features is as follows:
Dc＝(Ln-Lm)2+(an-am)2+(bn-bm)2 (2)
where m represents a seed superpixel, n represents an adjacent superpixel, Lm、am、bmRepresenting the color characteristics of the seed superpixel, Ln、an、bnRepresenting the color characteristics of neighboring superpixels.
CN202110864656.8A 2021-07-29 2021-07-29 Cervical vertebra MR image curvature measuring method based on superpixel segmentation Pending CN113822853A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110864656.8A CN113822853A (en) 2021-07-29 2021-07-29 Cervical vertebra MR image curvature measuring method based on superpixel segmentation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110864656.8A CN113822853A (en) 2021-07-29 2021-07-29 Cervical vertebra MR image curvature measuring method based on superpixel segmentation

Publications (1)

Publication Number Publication Date
CN113822853A true CN113822853A (en) 2021-12-21

Family

ID=78923983

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110864656.8A Pending CN113822853A (en) 2021-07-29 2021-07-29 Cervical vertebra MR image curvature measuring method based on superpixel segmentation

Country Status (1)

Country Link
CN (1) CN113822853A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116809431A (en) * 2023-08-31 2023-09-29 江苏欧港昌盛装饰材料有限公司 But defect detection's floor conveying system for production line

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116809431A (en) * 2023-08-31 2023-09-29 江苏欧港昌盛装饰材料有限公司 But defect detection's floor conveying system for production line
CN116809431B (en) * 2023-08-31 2023-11-10 江苏欧港昌盛装饰材料有限公司 But defect detection's floor conveying system for production line

Similar Documents

Publication Publication Date Title
CN107230206B (en) Multi-mode data-based 3D pulmonary nodule segmentation method for hyper-voxel sequence lung image
CN106846344B (en) A kind of image segmentation optimal identification method based on the complete degree in edge
CN110930416B (en) MRI image prostate segmentation method based on U-shaped network
CN109325942B (en) Fundus image structure segmentation method based on full convolution neural network
CN104933709B (en) Random walk CT lung tissue image automatic segmentation methods based on prior information
CN108573502B (en) Method for automatically measuring Cobb angle
CN110503649B (en) Liver segmentation method based on spatial multi-scale U-net and superpixel correction
WO2019000455A1 (en) Method and system for segmenting image
CN103310457B (en) A kind of pulmonary parenchyma dividing method based on para-curve correction convex closure
CN106780522B (en) A kind of bone marrow fluid cell segmentation method based on deep learning
CN109753997B (en) Automatic accurate robust segmentation method for liver tumor in CT image
CN109064470B (en) Image segmentation method and device based on self-adaptive fuzzy clustering
CN104392460A (en) Adherent white blood cell segmentation method based on nucleus-marked watershed transformation
CN107146228A (en) A kind of super voxel generation method of brain magnetic resonance image based on priori
CN108615239B (en) Tongue image segmentation method based on threshold technology and gray level projection
CN108876769B (en) Left auricle CT image segmentation method
CN103942780B (en) Based on the thalamus and its minor structure dividing method that improve fuzzy connectedness algorithm
CN117408929B (en) Tumor CT image area dynamic enhancement method based on image characteristics
CN109447963A (en) A kind of method and device of brain phantom identification
CN113822853A (en) Cervical vertebra MR image curvature measuring method based on superpixel segmentation
CN113191968A (en) Method for establishing three-dimensional ultrasonic image blind denoising model and application thereof
CN108665474B (en) B-COSFIRE-based retinal vessel segmentation method for fundus image
CN115100123A (en) Brain extraction method combining UNet and active contour model
CN114708165A (en) Edge perception texture filtering method combining super pixels
CN109741358B (en) Superpixel segmentation method based on adaptive hypergraph learning

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