CN116563296A - Identification method for abdomen CT image - Google Patents
Identification method for abdomen CT image Download PDFInfo
- Publication number
- CN116563296A CN116563296A CN202310848342.8A CN202310848342A CN116563296A CN 116563296 A CN116563296 A CN 116563296A CN 202310848342 A CN202310848342 A CN 202310848342A CN 116563296 A CN116563296 A CN 116563296A
- Authority
- CN
- China
- Prior art keywords
- cluster
- image
- abdomen
- satisfaction
- pixel points
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 45
- 210000001015 abdomen Anatomy 0.000 title claims abstract description 44
- 239000006185 dispersion Substances 0.000 claims abstract description 23
- 210000000056 organ Anatomy 0.000 claims abstract description 13
- 238000012545 processing Methods 0.000 claims abstract description 6
- 238000003708 edge detection Methods 0.000 claims abstract description 5
- 238000010606 normalization Methods 0.000 claims description 14
- 238000013528 artificial neural network Methods 0.000 claims description 8
- 238000003064 k means clustering Methods 0.000 claims description 8
- 238000004422 calculation algorithm Methods 0.000 claims description 7
- 238000001914 filtration Methods 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 2
- 230000003187 abdominal effect Effects 0.000 claims 6
- 230000008569 process Effects 0.000 abstract description 5
- 238000002591 computed tomography Methods 0.000 description 39
- 230000011218 segmentation Effects 0.000 description 8
- 238000003709 image segmentation Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000013507 mapping Methods 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000004204 blood vessel Anatomy 0.000 description 1
- 238000007621 cluster analysis Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 210000000232 gallbladder Anatomy 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 210000004185 liver Anatomy 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000000496 pancreas Anatomy 0.000 description 1
- 238000002560 therapeutic procedure Methods 0.000 description 1
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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G06T7/66—Analysis of geometric attributes of image moments or centre of gravity
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10081—Computed x-ray tomography [CT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Medical Informatics (AREA)
- Multimedia (AREA)
- Databases & Information Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Geometry (AREA)
- Quality & Reliability (AREA)
- Radiology & Medical Imaging (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the field of image data processing, in particular to an identification method for an abdomen CT image. Acquiring an abdomen CT gray-scale image and clustering pixel points to obtain a plurality of clusters; performing edge detection on each cluster to obtain the number of edge pixel points in each cluster; obtaining the dispersion degree of each cluster; acquiring satisfaction degree of each cluster, and taking the cluster with satisfaction degree smaller than a threshold value as a first cluster; acquiring a centroid adjustment distance of a first cluster, updating the centroid, re-clustering by using an updated clustering center point, sequentially iterating until satisfaction of the clusters meets a preset condition, and acquiring a plurality of clusters at the moment; dividing the abdomen CT gray-scale image according to the plurality of clusters, and carrying out organ identification according to the pixel points in the divided image. According to the invention, the cluster center point is adjusted, and the satisfaction degree of each cluster is obtained as the adjusting judgment basis, so that the problem of local optimum in the clustering process can be effectively solved.
Description
Technical Field
The invention relates to the technical field of image data processing, in particular to an identification method for an abdomen CT image.
Background
With increasing use of medical images in medical treatment, segmentation of CT (computed tomography) images is increasingly important in the field of medical image processing because of the fact that the CT (computed tomography) images can assist in therapy and assist in judgment, and in various means of image segmentation, neural networks have too high use cost because of the need of training a model with a large amount of data, so that image segmentation is needed by some unsupervised learning methods, and image segmentation and object extraction are main application aspects of cluster analysis, which mainly aim at distinguishing and classifying given things according to similarity among things, so that elements in each class have the same characteristics as possible and characteristic differences among different clusters are as large as possible, wherein a K-means clustering algorithm is used as one of the simplest and fast clustering algorithms.
In practical application, the expansibility and efficiency of the K-means clustering algorithm are ideal when the large data volume is found, but when the CT image is processed by the K-means clustering method, the situation that partial optimization is possibly involved or empty class is generated during the initialization of a clustering center due to the K-means algorithm, so that sample points in partial clusters in the image are few, the positions in the image are quite scattered, the segmentation of the image is seriously affected due to the fact that the situation that the edge points are more and the like, the recognition efficiency is poor or the recognition accuracy is poor when the image is recognized through a neural network later, and serious consequences can be caused in the medical field.
Disclosure of Invention
In order to solve the problems that in the prior art, partial optimization occurs when a CT image is identified by using a K-means clustering algorithm, so that the subsequent image identification efficiency is poor and the identification precision has errors, the invention provides an identification method for an abdomen CT image, which comprises the following steps: acquiring an abdomen CT gray-scale image and clustering pixel points to obtain a plurality of clusters; performing edge detection on each cluster to obtain the number of edge pixel points in each cluster; obtaining the dispersion degree of each cluster; acquiring satisfaction degree of each cluster, and taking the cluster with satisfaction degree smaller than a threshold value as a first cluster; acquiring a centroid adjustment distance of a first cluster; updating the mass center of the first cluster, re-clustering by using the updated cluster center points, and sequentially iterating until the satisfaction degree of the clusters meets the preset condition, thereby obtaining a plurality of clusters at the moment; dividing the abdomen CT gray-scale image according to the plurality of clusters, and carrying out organ identification according to the pixel points in the divided image. According to the invention, the cluster center point is adjusted, and the satisfaction degree of each cluster is obtained as the adjusting judgment basis, so that the problem of local optimum in the clustering process can be effectively solved.
The invention adopts the following technical scheme that the method for identifying the abdomen CT image comprises the following steps:
acquiring an abdomen CT gray-scale image, and performing edge detection on the abdomen CT gray-scale image to obtain edge pixel points in the abdomen CT gray-scale image;
clustering pixel points in the abdomen CT gray-scale image to obtain a plurality of cluster types, and obtaining the number of edge pixel points in each cluster type;
acquiring the dispersion degree of each cluster according to the distance between every two pixel points in each cluster;
acquiring satisfaction of each cluster according to the number of pixel points in each cluster, the number of edge pixel points and the dispersion degree of each cluster, and taking the cluster with satisfaction smaller than a threshold value as a first cluster;
acquiring a centroid adjustment distance of each first cluster according to the distance between the centroid of each first cluster and the centroid of the cluster with the adjacent satisfaction degree larger than a threshold value and the satisfaction degree of each first cluster;
adjusting the position of the mass center of each first cluster according to the mass center adjusting distance of each first cluster, and clustering again by using the adjusted mass center, and iterating sequentially until the satisfaction degree of each cluster is not changed after re-clustering, so as to obtain a plurality of final clusters in the abdomen CT gray level image;
dividing the abdomen CT gray-scale image according to the final cluster, obtaining a plurality of divided areas, and carrying out organ identification on each divided area by using a neural network.
Further, a method for identifying an abdomen CT image, the method for obtaining satisfaction of each cluster comprises:
respectively carrying out normalization processing on the number of pixel points in each cluster, the number of edge pixel points and the dispersion degree of each cluster;
obtaining the product of the number of edge points in each cluster and the dispersion degree of each cluster after normalization;
and obtaining the satisfaction degree of each cluster according to the ratio of the number of the pixel points in each cluster after normalization to the product.
Further, a method for identifying an abdomen CT image, the method for obtaining the centroid adjustment distance of each first cluster class includes:
acquiring cluster classes with satisfaction degree larger than a threshold value adjacent to each first cluster class, and acquiring the maximum distance between the mass center of each first cluster class and the mass center of the cluster class with satisfaction degree larger than the threshold value adjacent to the first cluster class;
and obtaining the centroid adjustment distance of each first cluster according to the satisfaction degree of each first cluster and the maximum distance between the centroid of each first cluster and the cluster centroid of which the adjacent satisfaction degree is larger than the threshold value.
Further, the identification method for the abdomen CT image is to cluster again with the adjusted mass center, iterate in sequence, and further comprises the following steps:
setting an average threshold and a lowest threshold, and acquiring satisfaction of all clusters after each iteration;
stopping iteration when the satisfaction average value of all clusters after iteration is smaller than an average threshold value;
and stopping iteration when the satisfaction degree of the cluster exists after iteration is smaller than the lowest threshold value.
Further, the method for identifying the abdomen CT image further comprises the following steps after the abdomen CT gray-scale image is acquired:
and carrying out noise reduction treatment on the abdomen CT gray-scale image by using Gaussian filtering.
Further, a method for identifying an abdomen CT image, the method for clustering the pixel points in the abdomen CT gray-scale image comprises the following steps:
k initial cluster centers are randomly selected, and the K initial cluster types are obtained by clustering pixel points in the abdomen CT gray-scale image by using a K-means clustering algorithm.
The beneficial effects of the invention are as follows: according to the clustering method, the number of the pixels in each cluster, the number of the edge pixels and the dispersion degree of the pixels in the clusters are combined after clustering to obtain the satisfaction degree of each cluster, so that the clustering condition of each cluster can be accurately represented, verification is carried out on each cluster according to the satisfaction degree of each cluster, the positions of mass centers of the clusters which do not meet the conditions are screened out, the convergence condition of the clusters is verified again, and compared with the traditional clustering, whether the clusters are optimal in global consideration or not is verified through the satisfaction degree, the calculation is simple, the effect is clear, and the local optimal condition of a clustering result can be effectively avoided.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions of the prior art, the drawings which are used in the description of the embodiments or the prior art will be briefly described, it being obvious that the drawings in the description below are only some embodiments of the invention, and that other drawings can be obtained according to these drawings without inventive faculty for a person skilled in the art.
Fig. 1 is a schematic structural diagram of an identification method for an abdomen CT image according to an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
As shown in fig. 1, a schematic structural diagram of an identification method for an abdomen CT image according to an embodiment of the present invention is provided, including:
101. and acquiring abdomen CT gray level images and clustering to obtain a plurality of clusters.
After acquiring the abdomen CT gray-scale image, the method further comprises the following steps:
and performing Gaussian filtering treatment on the abdomen CT gray-scale image.
The method acquires the voxel image of the abdomen through the computer tomography technology, intercepts the section image which needs to be subjected to image segmentation, and then converts the acquired section image into the gray level image.
K-means clustering is carried out on the obtained original gray image on a gray number axis, and initialized K samples are selected as initial clustering centersFor each sample in the dataset +.>And calculating the distances from the cluster center to k cluster centers and dividing the cluster center into classes corresponding to the cluster centers with the smallest distances, thereby obtaining a plurality of cluster classes.
102. And acquiring edge pixel points in each cluster, and acquiring the dispersion degree of each cluster.
The method for obtaining the dispersion degree of each cluster comprises the following steps:
obtaining the distance between any two pixel points in each cluster;
and obtaining the dispersion degree of each cluster according to the average value of the distances from each pixel point to other pixel points in each cluster.
Firstly, obtaining the average distance from each pixel point in the cluster to other pixel points in the cluster in the original image space:
wherein ,for the ith cluster set, p is given by +.>Denoted as->Pixels in cluster different from j, < >>Respectively represents +.>Coordinates of->Represents the i-th cluster set +.>The j-th pixel point of (a)>Respectively represent dot->In image +.>Coordinates of->Is cluster->Number of middle pixel points +.>And representing the average distance between the jth pixel point and other pixel points in the cluster.
Each pixel point in a cluster corresponds to an average distance, and then the dispersion degree of the cluster can be judged by the average distance of all the pixel points, and the expression is:
wherein ,is->Middle->Average distance of each pixel point in the image from other pixel points, +.>Is->Number of middle pixel points +.>For the i-th cluster class->Mapping the degree of dispersion in each cluster class to a normalized index:
wherein ,representing cluster->And e represents an exponential function based on e.
In the clustering process, it is generally not desirable that the positions of the pixels in the same cluster are too dispersed in the original image, especially when the number of pixels in a certain cluster is relatively small and the distribution is also very dispersed, which means that the points in the cluster are most likely noise points or edge points, so that the dispersion degree of all the pixels in a cluster can be determined by the average distance from each pixel in the cluster to other pixels in the cluster in the original image space.
103. And acquiring satisfaction degree of each cluster, and acquiring the cluster with satisfaction degree smaller than a threshold value as a first cluster.
The method for obtaining the satisfaction degree of each cluster comprises the following steps:
respectively carrying out normalization processing on the number of pixel points in each cluster, the number of edge pixel points and the dispersion degree of each cluster;
obtaining the product of the number of edge points in each cluster and the dispersion degree of each cluster after normalization;
and obtaining the satisfaction degree of each cluster according to the ratio of the number of the pixel points in each cluster after normalization to the product.
In the clustering of the abdomen image, each type of pixel points in the expectation represent more edge points in each organ rather than the image, so the number of the pixel points in each cluster is required to be as high as possible and the number of the edge points is required to be reduced as possible, and meanwhile, if the pixel points in the cluster are more aggregated, the obtained cluster is more desirable, therefore, the invention counts the number of the pixel points in each cluster, and the method is as follows:。
mapping the number of pixel points in each cluster as normalized indexes:
wherein ,for the normalized index of the number of pixels corresponding to the ith cluster, e is an exponential function based on e, and ++>For the number of pixels in the ith cluster class, k represents the number of clusters.
The method comprises the steps of judging edge points in an image through edge detection, detecting by using a Canny operator, returning to the image, obtaining a binary matrix after normalization, judging whether the position of the binary matrix corresponding to each point in a cluster is 1 or not, if so, considering the binary matrix as the edge points, counting the number of all the edge points in the cluster, and recordingThe number of edge points in each cluster is then mapped to a normalized index:
wherein ,for the number of edge points in the ith cluster, k in the value range of i represents the number of cluster clusters, +.>For the i-th cluster class->Normalization index of the corresponding edge point number.
The index of the number of edge points is considered according to the expected segmentation of the medical image, and the medical image segmentation aims at distinguishing blood vessels in each organ or section image, and compared with the whole segmentation, the index of the number of edge points in most of original images, which are singly gathered into one type, is low, so the invention calculates the index of cluster satisfaction by acquiring the number of edge points in clusters.
The expression for calculating satisfaction is:
wherein ,for the i-th cluster class->Normalization index of the number of corresponding pixels, < >>For the i-th cluster class->Normalization index of the corresponding number of edge points, < >>For the i-th cluster class->A normalization index of the dispersion degree, wherein k in the value range of i represents the set cluster number,/for the cluster group>Is the i cluster class in the clustering process +.>Mapping the satisfaction of each cluster class to a normalized index:
wherein ,is the->Cluster->The invention takes the normalized index as the satisfaction of the clusters, thus obtaining the satisfaction of each cluster.
The invention takes the cluster with satisfaction lower than the threshold value as the first cluster, the first cluster is clustered by randomly selecting a clustering center, and the distribution of pixel points in the cluster is more discrete and the number is less after the clustering is carried out by using a K-means clustering method, so that the situation of local optimum is considered to possibly occur in the clustering process, at the moment, the invention adjusts the centroid position of the cluster according to the satisfaction of the first cluster and re-clusters the cluster, thereby overcoming the local optimum situation and simultaneously striving for the clustering to achieve global optimum.
104. And acquiring the centroid adjustment distance of the first cluster.
The method for obtaining the centroid adjustment distance of the first cluster class comprises the following steps:
acquiring cluster classes with satisfaction degree larger than a threshold value adjacent to each first cluster class, and acquiring the maximum distance between the mass center of each first cluster class and the mass center of the cluster class with satisfaction degree larger than the threshold value adjacent to the first cluster class;
and obtaining the centroid adjustment distance of each first cluster according to the satisfaction degree of each first cluster and the maximum distance between the centroid of each first cluster and the cluster centroid of which the adjacent satisfaction degree is larger than the threshold value.
For gray image clustering on a gray axis, cluster center points all move on a number axis, and when the positions of the cluster center points are adjusted, the cluster center points need to be free from the influence of local optimization, so that iteration can be continued to find the optimal point, and therefore, the invention measures the distance and satisfaction degree of the farthest adjacent clusters.
Calculating the centroid adjustment distance by calculating the distance between the first cluster and the farthest adjacent cluster, searching the nearest adjacent cluster of the cluster, and recordingIs cluster->Center of class (centroid) of +.>Judging->One centroid point farther from the middle distance, namely: />, wherein />Respectively->It should be noted that, when the centroid adjustment distance of the first cluster is obtained, the cluster with satisfaction degree greater than the threshold needs to be selected from the nearest neighboring clusters.
The invention uses the centroid point of the farthest adjacent cluster of the first clusterAs the first cluster centroid +.>Is to acquire->And->The distance between them is->The centroid adjustment distance is +.>。
105. Updating the mass center of the first cluster, re-clustering by using the updated cluster center points, and sequentially iterating until the satisfaction degree of the clusters is not changed.
The preset conditions met by cluster satisfaction further comprise:
setting an average threshold and a lowest threshold, and acquiring satisfaction of all clusters after each iteration;
stopping iteration when the satisfaction average value of all clusters after iteration is smaller than an average threshold value; and stopping iteration when the satisfaction degree of the cluster class after iteration is smaller than the lowest threshold value, wherein the average threshold value and the lowest threshold value can be set according to actual conditions, and the invention is not limited.
Since the satisfaction of all clusters may change after each re-clustering, when the satisfaction of the first cluster is adjusted, the satisfaction of the original non-first cluster may be reduced, and in order to prevent the non-first cluster from being converted into the first cluster due to the satisfaction adjustment of the first cluster, the invention sets two thresholds, namely an average threshold and a minimum threshold, simultaneously when considering the satisfaction of the first cluster.
The average threshold is used to define the satisfaction average value of all clusters after each iteration, that is, after the re-clustering is completed, the satisfaction of the first cluster is still changed, and then the satisfaction of the whole cluster needs to be considered, that is, if the satisfaction average value of the whole cluster is reduced to the average threshold, although the satisfaction of the first cluster is still smaller than the threshold, the clustering effect of the whole image cannot be adjusted, and this may be caused by non-ideal K value selection, so that the K value needs to be re-selected for clustering.
In the method, when the first cluster is subjected to iterative clustering, the satisfaction degree of the first cluster is continuously improved until no change occurs, if the satisfaction degree of the first cluster is reduced after iteration, the lowest threshold is set at the moment, that is, the satisfaction degree of any cluster after iteration is reduced to the lowest threshold, a plurality of clusters obtained after initial clustering are obtained and used as clusters for segmenting the image.
106. Dividing the abdomen CT gray-scale image according to the plurality of clusters, and identifying organs.
When each organ in the CT image is identified, the invention needs to select the clustering K value according to different identification purposes, namely, the clustering K value is selected through different K values, and the cluster class of the satisfactory clustering result is obtained, so that the corresponding organ identification is carried out on each region in the segmented image.
The invention identifies organs represented by segmentation clusters in the acquired abdomen CT image through a semantic segmentation neural network, and the specific content of the adopted target identification neural network is as follows:
and using the semantic segmentation Unet network, and outputting a corresponding semantic segmentation result after the cluster image is input. The image features are extracted through convolution and pooling, and then the deconvolution and pooling operation is adopted to reconstruct the image, so that the label classification result corresponding to the input image is obtained.
And collecting a plurality of organ images segmented by the corresponding CT images as a data set to train the neural network.
And (3) obtaining image label information corresponding to each image by manually labeling the category information in the images, wherein the labeled background pixel category is 0, the liver is 1, the gall bladder is 2 and the pancreas is 3.
Because of the classification task, the network employs a cross entropy loss function to supervise training.
After the neural network training is completed, the images of the segmented clusters can be sent into a network, corresponding class labels are obtained by network reasoning, organs shown by the clusters are determined, and after each organ in the CT image is identified, the images are required to be visually displayed on a corresponding display, so that a user can more intuitively check information of each part in the CT image.
According to the clustering method, the number of the pixels in each cluster, the number of the edge pixels and the dispersion degree of the pixels in the clusters are combined after clustering to obtain the satisfaction degree of each cluster, so that the clustering condition of each cluster can be accurately represented, verification is carried out on each cluster according to the satisfaction degree of each cluster, the positions of mass centers of the clusters which do not meet the conditions are screened out, the convergence condition of the clusters is verified again, and compared with the traditional clustering, whether the clusters are optimal in global consideration or not is verified through the satisfaction degree, the calculation is simple, the effect is clear, and the local optimal condition of a clustering result can be effectively avoided.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, alternatives, and improvements that fall within the spirit and scope of the invention.
Claims (7)
1. A method for identifying an abdominal CT image, comprising:
acquiring an abdomen CT gray-scale image, and performing edge detection on the abdomen CT gray-scale image to obtain edge pixel points in the abdomen CT gray-scale image;
clustering pixel points in the abdomen CT gray-scale image to obtain a plurality of cluster types, and obtaining the number of edge pixel points in each cluster type;
acquiring the dispersion degree of each cluster according to the distance between every two pixel points in each cluster;
acquiring satisfaction of each cluster according to the number of pixel points in each cluster, the number of edge pixel points and the dispersion degree of each cluster, and taking the cluster with satisfaction smaller than a threshold value as a first cluster;
acquiring a centroid adjustment distance of each first cluster according to the distance between the centroid of each first cluster and the centroid of the cluster with the adjacent satisfaction degree larger than a threshold value and the satisfaction degree of each first cluster;
adjusting the position of the mass center of each first cluster according to the mass center adjusting distance of each first cluster, and clustering again by using the adjusted mass center, and iterating sequentially until the satisfaction degree of each cluster is not changed after re-clustering, so as to obtain a plurality of final clusters in the abdomen CT gray level image;
dividing the abdomen CT gray-scale image according to the final cluster, obtaining a plurality of divided areas, and carrying out organ identification on each divided area by using a neural network.
2. The method for recognizing an abdominal CT image according to claim 1, wherein the method for obtaining the degree of dispersion of each cluster is as follows:
and obtaining the dispersion degree of each cluster according to the ratio of the sum of the average value of the distances from each pixel point to other pixel points in each cluster to the number of the pixel points in each cluster.
3. The method for recognizing an abdominal CT image according to claim 1, wherein the method for obtaining satisfaction of each cluster is as follows:
respectively carrying out normalization processing on the number of pixel points in each cluster, the number of edge pixel points and the dispersion degree of each cluster;
obtaining the product of the number of edge points in each cluster and the dispersion degree of each cluster after normalization;
and obtaining the satisfaction degree of each cluster according to the ratio of the number of the pixel points in each cluster after normalization to the product.
4. The method for recognizing an abdominal CT image according to claim 1, wherein the method for acquiring the centroid adjustment distance of each first cluster class comprises:
acquiring cluster classes with satisfaction degree larger than a threshold value adjacent to each first cluster class, and acquiring the maximum distance between the mass center of each first cluster class and the mass center of the cluster class with satisfaction degree larger than the threshold value adjacent to the first cluster class;
and obtaining the centroid adjustment distance of each first cluster according to the satisfaction degree of each first cluster and the maximum distance between the centroid of each first cluster and the cluster centroid of which the adjacent satisfaction degree is larger than the threshold value.
5. The method of claim 1, wherein the clustering is performed again with the adjusted centroids, and the iterative steps are performed sequentially, further comprising:
setting an average threshold and a lowest threshold, and acquiring satisfaction of all clusters after each iteration;
stopping iteration when the satisfaction average value of all clusters after iteration is smaller than an average threshold value;
and stopping iteration when the satisfaction degree of the cluster exists after iteration is smaller than the lowest threshold value.
6. The method for recognizing an abdominal CT image according to claim 1, further comprising, after acquiring the abdominal CT gray scale image:
and carrying out noise reduction treatment on the abdomen CT gray-scale image by using Gaussian filtering.
7. The method for identifying an abdomen CT image according to claim 1, wherein the method for clustering pixels in the abdomen CT gray image comprises:
k initial cluster centers are randomly selected, and the K initial cluster types are obtained by clustering pixel points in the abdomen CT gray-scale image by using a K-means clustering algorithm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310848342.8A CN116563296B (en) | 2023-07-12 | 2023-07-12 | Identification method for abdomen CT image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310848342.8A CN116563296B (en) | 2023-07-12 | 2023-07-12 | Identification method for abdomen CT image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116563296A true CN116563296A (en) | 2023-08-08 |
| CN116563296B CN116563296B (en) | 2023-09-05 |
Family
ID=87493272
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310848342.8A Active CN116563296B (en) | 2023-07-12 | 2023-07-12 | Identification method for abdomen CT image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116563296B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116721099A (en) * | 2023-08-09 | 2023-09-08 | 山东奥洛瑞医疗科技有限公司 | Image segmentation method of liver CT image based on clustering |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115311278A (en) * | 2022-10-11 | 2022-11-08 | 南通欧惠纺织科技有限公司 | Yarn cutting method for yarn detection |
| CN115330800A (en) * | 2022-10-14 | 2022-11-11 | 深圳市亿康医疗技术有限公司 | Automatic segmentation method for radiotherapy target area based on image processing |
| CN115423813A (en) * | 2022-11-05 | 2022-12-02 | 江苏惠汕新能源集团有限公司 | Method for detecting welding defects on surface of welded pipe |
| CN115953774A (en) * | 2023-03-08 | 2023-04-11 | 济宁安泰矿山设备制造有限公司 | Alarm display digital identification method based on machine vision |
| CN116309575A (en) * | 2023-05-19 | 2023-06-23 | 济宁众达利电气设备有限公司 | Electric plug production quality detection method based on image processing |
| CN116277973A (en) * | 2023-03-02 | 2023-06-23 | 河南大学 | 3D prints detecting system |
-
2023
- 2023-07-12 CN CN202310848342.8A patent/CN116563296B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115311278A (en) * | 2022-10-11 | 2022-11-08 | 南通欧惠纺织科技有限公司 | Yarn cutting method for yarn detection |
| CN115330800A (en) * | 2022-10-14 | 2022-11-11 | 深圳市亿康医疗技术有限公司 | Automatic segmentation method for radiotherapy target area based on image processing |
| CN115423813A (en) * | 2022-11-05 | 2022-12-02 | 江苏惠汕新能源集团有限公司 | Method for detecting welding defects on surface of welded pipe |
| CN116277973A (en) * | 2023-03-02 | 2023-06-23 | 河南大学 | 3D prints detecting system |
| CN115953774A (en) * | 2023-03-08 | 2023-04-11 | 济宁安泰矿山设备制造有限公司 | Alarm display digital identification method based on machine vision |
| CN116309575A (en) * | 2023-05-19 | 2023-06-23 | 济宁众达利电气设备有限公司 | Electric plug production quality detection method based on image processing |
Non-Patent Citations (1)
| Title |
|---|
| 王存睿;张庆灵;段晓东;刘向东;: "基于尺度聚类仿射过滤的图像拼接算法", 东北大学学报(自然科学版), no. 07 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116721099A (en) * | 2023-08-09 | 2023-09-08 | 山东奥洛瑞医疗科技有限公司 | Image segmentation method of liver CT image based on clustering |
| CN116721099B (en) * | 2023-08-09 | 2023-11-21 | 山东奥洛瑞医疗科技有限公司 | Image segmentation method of liver CT image based on clustering |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116563296B (en) | 2023-09-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Sari et al. | Unsupervised feature extraction via deep learning for histopathological classification of colon tissue images | |
| CN106803247B (en) | Microangioma image identification method based on multistage screening convolutional neural network | |
| Zhou et al. | Automatic detection of exudates in digital color fundus images using superpixel multi-feature classification | |
| US10262410B2 (en) | Methods and systems for inspecting goods | |
| Xie et al. | Automatic segmentation of dermoscopy images using self-generating neural networks seeded by genetic algorithm | |
| CN107316294B (en) | Lung nodule feature extraction method based on improved depth Boltzmann machine | |
| CN106599913B (en) | A kind of multi-tag imbalance biomedical data classification method based on cluster | |
| CN115330800B (en) | Automatic segmentation method of radiotherapy target area based on image processing | |
| He et al. | A method of green litchi recognition in natural environment based on improved LDA classifier | |
| CN108830853A (en) | A kind of melanoma aided diagnosis method based on artificial intelligence | |
| WO2013037983A1 (en) | Method and system for the automatic analysis of an image of a biological sample | |
| Song et al. | Using HOG-LBP features and MMP learning to recognize imaging signs of lung lesions | |
| CN104217213B (en) | A kind of medical image multistage sorting technique based on symmetric theory | |
| CN116563296B (en) | Identification method for abdomen CT image | |
| Wang et al. | Pixel classification based color image segmentation using quaternion exponent moments | |
| CN110610472A (en) | Computer device and method for realizing classification detection of lung nodule images | |
| CN110276764A (en) | K-Means underwater picture background segment innovatory algorithm based on the estimation of K value | |
| Zhao et al. | Separate degree based Otsu and signed similarity driven level set for segmenting and counting anthrax spores | |
| Yang et al. | Color texture segmentation based on image pixel classification | |
| CN112183237A (en) | Automatic white blood cell classification method based on color space adaptive threshold segmentation | |
| Chopra et al. | Recent studies on segmentation techniques for food recognition: A survey | |
| Behara et al. | Artificial intelligence in medical diagnostics: A review from a South African context | |
| Setiawan et al. | Maize leaf disease image classification using bag of features | |
| Sarada et al. | Spatial Intuitionistic Fuzzy C-means with Calcifications enhancement based on Nonsubsampled Shearlet Transform to detect Masses and Microcalcifications from MLO Mammograms | |
| CN117195027A (en) | Cluster weighted clustering integration method based on member selection |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |