CN110263804A - A kind of medical image dividing method based on safe semi-supervised clustering - Google Patents
A kind of medical image dividing method based on safe semi-supervised clustering Download PDFInfo
- Publication number
- CN110263804A CN110263804A CN201910371366.2A CN201910371366A CN110263804A CN 110263804 A CN110263804 A CN 110263804A CN 201910371366 A CN201910371366 A CN 201910371366A CN 110263804 A CN110263804 A CN 110263804A
- Authority
- CN
- China
- Prior art keywords
- sample
- marker samples
- confidence level
- medical image
- point
- 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
Classifications
-
- 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
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
Abstract
The present invention discloses a kind of medical image dividing method based on safe semi-supervised clustering, is related to semi-supervised FCM cluster and density peaks clustering method.Firstly, constructing Local map using k- near neighbor method, figure regular terms is obtained.Secondly, utilizing the confidence level of FCM cluster and Density Clustering method estimation label and unmarked sample.Then, the confidence level weighted sum for introducing sample in the objective function of former semi-supervised FCM clustering method obtains the objective function of safe Novel semi-supervised based on the regular terms of Local map.Finally, obtaining cluster result by iteration optimization subordinated-degree matrix and cluster centre.The present invention solves the problems, such as the safe handling of marker samples, while solving the problems, such as the safe handling of unmarked sample, improves accuracy and robustness to medical image segmentation.
Description
Technical field
The present invention relates to a kind of medical image dividing method based on semi-supervised clustering refers in particular to a kind of based on safety half
The medical image dividing method of supervision clustering, belongs to the field of data mining based on medical image.
Background technique
With the continuous development of visualization technique, modern medicine has increasingly be unable to do without the information processing of medical image, doctor
Learn image clinical diagnosis, in terms of just playing an important role.Medical image based on semi-supervised clustering point
Segmentation method integrates limited manual oversight information, i.e., clicks limited several points on the image to identify the pass between corresponding region
System, by these points as the sample data with label information in the medical image dividing method based on semi-supervised clustering, benefit
It is instructed to cluster with these sample datas, to improve algorithm performance, keeps image segmentation more accurate.Label in medical image
Usually completed by expert, but probably due to error flag, and medical image occur in various situations in labeling process
Often carry noise point and outlier, traditional medical image dividing method based on semi-supervised clustering in cluster process simultaneously
In terms of not accounting for two above.
In this case, the performance of traditional Novel semi-supervised may be poorer than corresponding unsupervised learning method, this
Application of the semi-supervised clustering in medical image segmentation is limited to a certain extent.In other words, flag data may be to property
Can be harmful, the noise point in Unlabeled data and outlier also have a great impact to performance at the same time.Traditional is semi-supervised
Cluster is it is generally acknowledged that priori knowledge is conducive to learning effect, however the priori knowledge being collected into (such as error flag sample and is made an uproar
Sound), it is possible to lead to the degeneration of learning performance.Xuesong Yin indicates that the priori knowledge of mistake will lead to learning performance
Decline.Based on above-mentioned two aspect, the semi-supervised learning method of design safety is meaningful.Therefore, the invention patent attempts
The mechanism that a kind of different sample has different degrees of safety is researched and developed, to realize that clustering performance is not less than former Unsupervised clustering and semi-supervised
Clustering method.
Summary of the invention
The present invention for traditional medical image dividing method based on semi-supervised clustering simultaneously do not consider marker samples and
The risk of unmarked sample may cause the shortcomings that last segmentation effect declines, and propose a kind of semi-supervised based on safety
The medical image dividing method of cluster.
Firstly, the present invention constructs Local map using k- near neighbor method, figure regular terms is obtained.Secondly, using FCM cluster and it is close
Spend the confidence level of clustering method estimation label and unmarked sample.Then, in the objective function of former semi-supervised FCM clustering method
The confidence level weighted sum of sample is introduced based on the regular terms of Local map, obtains the objective function of safe Novel semi-supervised.Most
Afterwards, cluster result is obtained by iteration optimization subordinated-degree matrix and cluster centre.Technical solution: one kind is semi-supervised poly- based on safety
The medical image dividing method of class, the method includes the steps of:
Step 1: input marking and unmarked medical image data set;
Step 2: FCM cluster is carried out to data set, obtains the prediction label of data set;
Step 3: using density peaks clustering method, highly denser by the local density of unmarked sample and with having
Degree point minimum range obtain the confidence level of unmarked sample, by marker samples same tag sample Cu Zhong local density with
And the confidence level of marker samples is being obtained with the minimum range with higher density point, and confidence level is normalized;
Step 4: construction Local map, it is therefore an objective to be the output of neighbouring sample by the low marker samples export-restriction of confidence level;
Step 5: information is integrated, and constructs objective function;
Step 6: iterative optimization method solving optimization problem is used;
Step 7: determining the classification of unmarked sample, realizes medical image segmentation.
Compared with traditional Novel semi-supervised, the present invention using between sample density and distance measure sample
The low marker samples of confidence level are limited to the output of neighbour's sample by construction Local map, so that each sample by confidence level
It can safely and reasonably be used, cluster more accurate and robust.The present invention solves the problems, such as the safe handling of marker samples, simultaneously
It solves the problems, such as the safe handling of unmarked sample, improves accuracy and robustness to medical image segmentation.
Detailed description of the invention
Fig. 1 is present invention specific implementation flow chart.
Specific embodiment
In conjunction with Figure of description, the present invention is furture elucidated, it should be understood that these embodiments are merely to illustrate the present invention and do not have to
In limiting the scope of the invention, after reading this disclosure, those skilled in the art repair to various equivalent forms of the invention
Change and falls within claim limited range appended by the application.
Objects and advantages in order to better illustrate the present invention, with reference to the accompanying drawing 1 and embodiment to the method for the present invention
Embodiment is described in further details.
Step 1: input marking and unmarked medical image data set;
Input the marker samples subset of medical image data set: Xl=[x1,...,xl], corresponding label is yk∈
{ 1 ..., c }, unmarked sample set: Xu=[xl+1,...,xn]。
Step 2: FCM cluster is carried out to data set, obtains the prediction label of data set;
It clusters to obtain the prediction label of data set by FCM:
Using Kuhn-Munkres algorithm by prediction labelIt is mapped to equivalent labelWith it is given
LabelIt is consistent.
Step 3: using density peaks clustering method, highly denser by the local density of unmarked sample and with having
Degree point minimum range obtain the confidence level of unmarked sample, by marker samples same tag sample Cu Zhong local density with
And the confidence level of marker samples is being obtained with the minimum range with higher density point, and confidence level is normalized;
The local density of unmarked sample:
Wherein, j=[1,2 ..., n], k=[l+1 ..., n], dist (k, j) are point xkWith xjEuclidean distance, dcFor
Distance is truncated.
Unmarked sample and the minimum range with higher density point:
And for the data point with maximal density:
Unmarked sample confidence level: γk=ρk/δk (4)
Unmarked sample confidence level normalization:
Local density of the marker samples in same tag sample cluster:
Wherein, jy=[1,2 ... q], k '=[1,2 ..., l], jyIndicate sample set in marker samples point xk′Label
The set of identical sample.
Marker samples in same tag sample cluster with higher density point minimum range:
And for the data point with maximal density:
Marker samples confidence level:
The normalization of marker samples confidence level:
Step 4: construction k- neighbour Local map, it is therefore an objective to by the low marker samples export-restriction of confidence level for adjacent to sample
Output;
The local neighborhood figure of marker samples is constructed, then Local map side right W=[wk′r]n×nIt calculates are as follows:
Wherein, Np(xk′) refer to xk′P data of arest neighbors, xk′For marker samples point, xrFor neighbour's sample point, σ indicates high
The width parameter of this kernel function.
Step 5: information is integrated, and constructs objective function.
Objective function is as follows:
Restrictive condition is as follows:
Step 6: iterative optimization method solving optimization problem is used;
By minimizing above-mentioned optimization problem, optimal solution can be obtained.It is calculated to simplify, m value is set as 2.The present invention
Sample degree of membership and cluster centre are solved using method of Lagrange multipliers.
The degree of membership u of unmarked sampleik:
Wherein,
The degree of membership u of marker samplesik′:
Wherein,
Cluster centre vi:
Final subordinated-degree matrix U and cluster centre V is obtained by iterative calculation.WhenOr reach maximum
When the number of iterations, iteration ends, wherein t is current iteration number, and η is the threshold value of setting.
Step 7: determining the classification of unmarked sample, realizes the segmentation of medical image.
After obtaining subordinated-degree matrix U, according to degree of membership maximum principle de-fuzzy, the classification of unmarked sample is obtained, most
After carry out image segmentation, obtain result.
Claims (1)
1. a kind of medical image dividing method based on safe semi-supervised clustering, which is characterized in that this method specifically includes following
Step:
Step 1: input marking and unmarked medical image data set;
Input the marker samples subset of medical image data set: Xl=[x1,...,xl], corresponding label is yk∈ { 1 ..., c },
Unmarked sample set: Xu=[xl+1,...,xn];
Step 2: FCM cluster is carried out to data set, obtains the prediction label of data set;
It clusters to obtain the prediction label of data set by FCM:
Using Kuhn-Munkres algorithm by prediction labelIt is mapped asMake map tagsIt is calibrated with giving
Sign ykIt is consistent in classification;
Step 3: using density peaks clustering method, by the local density of unmarked sample and with higher density point
Minimum range obtain the confidence level of unmarked sample, by marker samples same tag sample Cu Zhong local density and
The confidence level of marker samples is obtained with the minimum range with higher density point, and confidence level is normalized;
The local density of unmarked sample:
Wherein, j=[1,2 ..., n], k=[l+1 ..., n], dist (k, j) are point xkWith xjEuclidean distance, dcFor truncation
Distance;
Unmarked sample and the minimum range with higher density point:
And for the data point with maximal density:
Unmarked sample confidence level: γk=ρk/δk (4)
Unmarked sample confidence level normalization:
Local density of the marker samples in same tag sample cluster:
Wherein, jy=[1,2 ... q], k '=[1,2 ..., l], jyIndicate sample set in marker samples point xk′Label is identical
The set of sample;
Marker samples in same tag sample cluster with higher density point minimum range:
And for the data point with maximal density:
Marker samples confidence level:
The normalization of marker samples confidence level:
Step 4: construction k- neighbour Local map, it is therefore an objective to by the low marker samples export-restriction of confidence level for adjacent to the defeated of sample
Out;
The local neighborhood figure of marker samples is constructed, then Local map side right W=[wk′r]n×nIt calculates are as follows:
Wherein, Np(xk′) refer to xk′P data of arest neighbors, xk′For marker samples point, xrFor neighbour's sample point, σ indicates Gaussian kernel
The width parameter of function;
Step 5: information is integrated, and constructs objective function;
Objective function is as follows:
Restrictive condition is as follows:
Step 6: iterative optimization method solving optimization problem is used;
By minimizing above-mentioned optimization problem, optimal solution can be obtained;It is calculated to simplify, m value is set as 2;The present invention uses
Method of Lagrange multipliers solves sample degree of membership and cluster centre;
The degree of membership of marker samples:
Wherein,
The degree of membership u of unmarked sampleir:
Wherein,
Cluster centre vi:
Final subordinated-degree matrix U and cluster centre V is obtained by iterative calculation;WhenOr reach greatest iteration
When number, iteration ends, wherein t is current iteration number, and η is the threshold value of setting;
Step 7: determining the classification of unmarked sample, realizes the segmentation of medical image;
After obtaining subordinated-degree matrix U, according to degree of membership maximum principle de-fuzzy, the classification of unmarked sample is obtained, it is most laggard
Row image segmentation obtains result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910371366.2A CN110263804B (en) | 2019-05-06 | 2019-05-06 | Medical image segmentation method based on safe semi-supervised clustering |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910371366.2A CN110263804B (en) | 2019-05-06 | 2019-05-06 | Medical image segmentation method based on safe semi-supervised clustering |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110263804A true CN110263804A (en) | 2019-09-20 |
CN110263804B CN110263804B (en) | 2021-08-03 |
Family
ID=67914306
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910371366.2A Active CN110263804B (en) | 2019-05-06 | 2019-05-06 | Medical image segmentation method based on safe semi-supervised clustering |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110263804B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111612735A (en) * | 2020-04-08 | 2020-09-01 | 杭州电子科技大学 | Lung nodule image classification method based on information fusion safety semi-supervised clustering |
CN111898704A (en) * | 2020-08-17 | 2020-11-06 | 腾讯科技(深圳)有限公司 | Method and device for clustering content samples |
CN113780750A (en) * | 2021-08-18 | 2021-12-10 | 同济大学 | Medical risk assessment method and device based on medical image segmentation |
CN115131610A (en) * | 2022-06-13 | 2022-09-30 | 西北工业大学 | Robust semi-supervised image classification method based on data mining |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080267471A1 (en) * | 2007-04-25 | 2008-10-30 | Siemens Corporate Research, Inc | Automatic partitioning and recognition of human body regions from an arbitrary scan coverage image |
CN104156438A (en) * | 2014-08-12 | 2014-11-19 | 德州学院 | Unlabeled sample selection method based on confidence coefficients and clustering |
CN104881687A (en) * | 2015-06-02 | 2015-09-02 | 四川理工学院 | Magnetic resonance image classification method based on semi-supervised Gaussian mixed model |
CN105825226A (en) * | 2016-03-11 | 2016-08-03 | 江苏畅远信息科技有限公司 | Association-rule-based distributed multi-label image identification method |
CN106447676A (en) * | 2016-10-12 | 2017-02-22 | 浙江工业大学 | Image segmentation method based on rapid density clustering algorithm |
CN106611418A (en) * | 2016-03-29 | 2017-05-03 | 四川用联信息技术有限公司 | Image segmentation algorithm |
CN107341812A (en) * | 2017-07-04 | 2017-11-10 | 太原理工大学 | A kind of sequence Lung neoplasm image partition method based on super-pixel and Density Clustering |
CN108629783A (en) * | 2018-05-02 | 2018-10-09 | 山东师范大学 | Image partition method, system and medium based on the search of characteristics of image density peaks |
CN109409400A (en) * | 2018-08-28 | 2019-03-01 | 西安电子科技大学 | Merge density peaks clustering method, image segmentation system based on k nearest neighbor and multiclass |
-
2019
- 2019-05-06 CN CN201910371366.2A patent/CN110263804B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080267471A1 (en) * | 2007-04-25 | 2008-10-30 | Siemens Corporate Research, Inc | Automatic partitioning and recognition of human body regions from an arbitrary scan coverage image |
CN104156438A (en) * | 2014-08-12 | 2014-11-19 | 德州学院 | Unlabeled sample selection method based on confidence coefficients and clustering |
CN104881687A (en) * | 2015-06-02 | 2015-09-02 | 四川理工学院 | Magnetic resonance image classification method based on semi-supervised Gaussian mixed model |
CN105825226A (en) * | 2016-03-11 | 2016-08-03 | 江苏畅远信息科技有限公司 | Association-rule-based distributed multi-label image identification method |
CN106611418A (en) * | 2016-03-29 | 2017-05-03 | 四川用联信息技术有限公司 | Image segmentation algorithm |
CN106447676A (en) * | 2016-10-12 | 2017-02-22 | 浙江工业大学 | Image segmentation method based on rapid density clustering algorithm |
CN107341812A (en) * | 2017-07-04 | 2017-11-10 | 太原理工大学 | A kind of sequence Lung neoplasm image partition method based on super-pixel and Density Clustering |
CN108629783A (en) * | 2018-05-02 | 2018-10-09 | 山东师范大学 | Image partition method, system and medium based on the search of characteristics of image density peaks |
CN109409400A (en) * | 2018-08-28 | 2019-03-01 | 西安电子科技大学 | Merge density peaks clustering method, image segmentation system based on k nearest neighbor and multiclass |
Non-Patent Citations (1)
Title |
---|
何瀚志: "基于遗传密度峰值聚类的医学图像分割", 《计算机工程与设计》 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111612735A (en) * | 2020-04-08 | 2020-09-01 | 杭州电子科技大学 | Lung nodule image classification method based on information fusion safety semi-supervised clustering |
CN111898704A (en) * | 2020-08-17 | 2020-11-06 | 腾讯科技(深圳)有限公司 | Method and device for clustering content samples |
CN113780750A (en) * | 2021-08-18 | 2021-12-10 | 同济大学 | Medical risk assessment method and device based on medical image segmentation |
CN113780750B (en) * | 2021-08-18 | 2024-03-01 | 同济大学 | Medical risk assessment method and device based on medical image segmentation |
CN115131610A (en) * | 2022-06-13 | 2022-09-30 | 西北工业大学 | Robust semi-supervised image classification method based on data mining |
CN115131610B (en) * | 2022-06-13 | 2024-02-27 | 西北工业大学 | Robust semi-supervised image classification method based on data mining |
Also Published As
Publication number | Publication date |
---|---|
CN110263804B (en) | 2021-08-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110263804A (en) | A kind of medical image dividing method based on safe semi-supervised clustering | |
Tsai et al. | Image co-saliency detection and co-segmentation via progressive joint optimization | |
JP6935377B2 (en) | Systems and methods for automatic inference of changes in spatiotemporal images | |
Gao et al. | SAR image change detection based on multiscale capsule network | |
Azadi et al. | Auxiliary image regularization for deep cnns with noisy labels | |
Wang et al. | Multi-modal classification of neurodegenerative disease by progressive graph-based transductive learning | |
CN102609681B (en) | Face recognition method based on dictionary learning models | |
Xing et al. | Pixel-to-pixel learning with weak supervision for single-stage nucleus recognition in ki67 images | |
Cao et al. | A multi-kernel based framework for heterogeneous feature selection and over-sampling for computer-aided detection of pulmonary nodules | |
Al-Dulaimi et al. | Benchmarking HEp-2 specimen cells classification using linear discriminant analysis on higher order spectra features of cell shape | |
CN111553127A (en) | Multi-label text data feature selection method and device | |
Li et al. | Adaptive metric learning for saliency detection | |
CN111325750B (en) | Medical image segmentation method based on multi-scale fusion U-shaped chain neural network | |
Bicego et al. | A bioinformatics approach to 2D shape classification | |
CN109543693B (en) | Weak labeling data noise reduction method based on regularization label propagation | |
CN113408605A (en) | Hyperspectral image semi-supervised classification method based on small sample learning | |
CN112614131A (en) | Pathological image analysis method based on deformation representation learning | |
Neena et al. | Image classification using an ensemble-based deep CNN | |
CN110084314A (en) | A kind of false positive gene mutation filter method for targeted capture gene sequencing data | |
CN113344044A (en) | Cross-species medical image classification method based on domain self-adaptation | |
Seibold et al. | Self-guided multiple instance learning for weakly supervised thoracic diseaseclassification and localizationin chest radiographs | |
Liu et al. | Auxiliary segmentation method of osteosarcoma MRI image based on transformer and U-Net | |
Cho et al. | Effective pseudo-labeling based on heatmap for unsupervised domain adaptation in cell detection | |
Gu et al. | Unsupervised and semi-supervised robust spherical space domain adaptation | |
CN105426836A (en) | Single-sample face recognition method based on segmented model and sparse component analysis |
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 |