CN108346140A - Based on the Otsu lung images dividing methods for improving PSO - Google Patents
Based on the Otsu lung images dividing methods for improving PSO Download PDFInfo
- Publication number
- CN108346140A CN108346140A CN201810021908.9A CN201810021908A CN108346140A CN 108346140 A CN108346140 A CN 108346140A CN 201810021908 A CN201810021908 A CN 201810021908A CN 108346140 A CN108346140 A CN 108346140A
- Authority
- CN
- China
- Prior art keywords
- particle
- otsu
- pso
- fitness
- lung images
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
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
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration by the use of histogram techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30061—Lung
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Apparatus For Radiation Diagnosis (AREA)
- Image Analysis (AREA)
Abstract
The present invention is based on the Otsu lung images dividing methods for improving PSO to belong to technical field of image segmentation；This method is to be combined PSO algorithms and two dimension Otsu algorithms, and area-of-interest is extracted from lung images；Optimizing processing is carried out to image by improved PSO algorithms first, obtains globally optimal solution；Then using the globally optimal solution as optimal threshold, according to two-dimentional Otsu algorithm principles, it is target, that is, area-of-interest of the present invention that gray value, which is more than optimal threshold, and it is background that gray value, which is less than optimal threshold, to which image segmentation is come out；The present invention is based on the Otsu lung images dividing methods for improving PSO, can improve the accuracy rate and efficiency of lung images segmentation.
Description
Technical field
Belong to technical field of image segmentation based on the Otsu lung images dividing methods for improving PSO.
Background technology
Today's society is the society of economic, culture and high and new technology rapid development.People are rapidly developing the same of new science and technology
When, also increasingly pay close attention to itself living environment and medical condition.Health increasingly obtains everyone cordial concern.
Since in recent years, the importance of lung cancer therapy is self-evident, and the accuracy of lung images segmentation and segmentation are imitated
Rate is vital to the treatment of lung cancer.National Cancer Center publication《2012 China's tumour registration annual reports》It has been shown that, the whole nation
Primary tumour registration area Incidence is lung cancer, dead primary and lung cancer.
With the development of computer technology, image segmentation is all developed and is widely used in many fields, in medicine
Application clinically be even more be more and more obvious with it is important.Suitable medical image cutting method is found to clinical diagnosis and treatment
All it is of great significance.
Between in the past few decades, image segmentation has played the effect to become more and more important in medical imaging, extensive
Ground is applied in the every field of medical research.As medical imaging is more and more significant in the effect of clinical diagnosis and treatment, doctor
Learn image segmentation just at must medical image analysis field a challenge research topic.Medical image segmentation is to determine doctor
Can image be learned provide the critical issue of reliable basis in clinic diagnosis.
Lung CT image segmentation be exactly lung is extracted from CT images, for clinic with pathological study provide reliably according to
According to.In recent years, the application of lung CT image segmentation clinically is more and more extensive, has become doctor and carries out liver function, pathology
With the important means of anatomical study.And lung boundary is determined as the prerequisite of quantitative analysis, has very high researching value:
(1) it is conducive to further determine that the volume size and lesion degree of lesion, is convenient for the formulation therapeutic scheme of doctor promptly and accurately.(2)
Extraction lung is conducive to doctor and carries out diagnosing and treating.
Lung CT image has the characteristics that gray scale is uneven, complicated, therefore the selection of dividing method will directly affect lung
The segmentation precision and speed of portion's CT images.And traditional lung images dividing method existence time complexity height, segmentation precision are low
The problems such as, therefore the method for seeking precisely quick segmentation lung images becomes research hotspot and difficult point in recent years.
Invention content
It is an object of the invention to the deficiencies for above-mentioned existing method, it is proposed that based on the Otsu lungs figure for improving PSO
As dividing method, and then improve the accuracy and segmentation efficiency of segmentation.
The object of the present invention is achieved like this：
It is a kind of to be combined PSO algorithms and two dimension Otsu algorithms based on the Otsu lung images dividing methods for improving PSO,
Area-of-interest is extracted from lung images；Optimizing processing is carried out to image by PSO algorithms, obtains globally optimal solution as most
Good threshold value is split image using Otsu algorithms.
The above-mentioned Otsu lung images dividing methods based on improvement PSO, PSO algorithms are introduced into Otsu algorithms, are obtained
Based on improve PSO Otsu lung images dividing methods, the algorithm realize the specific steps are：
Step 1 initializes population, including position and speed, sets population as N, search space is tieed up for D, particle i
It is expressed as X in the position of D dimension spacesiD={ Xi1, Xi2..., XiD, i={ 1,2,3 ..., N }, the flying speed of each example particle
It is expressed as ViD={ Vi1, Vi2..., ViD, i={ 1,2,3 ..., N }, the optimal location that each particle passes through is Pbest, population
In the position (global optimum position) with adaptive optimal control value that is lived through of all particles be Gbest；
Step 2 calculates the grey level histogram of image；
Step 3 calculates the fitness value of each particle according to fitness function；
Step 4 updates the personal best particle Pbest of each particle and complete according to the fitness value of calculated particle
Office optimal location Gbest；
Step 5, being updated to the speed of particle, position according to more new formula, specifically more new formula is：
ViD(k+1)=w × ViD(k)+c1×r1[PbestiD-XiD(k)]+c2×r2[GbestiD-XiD(k)] (1)
XiD(k+1)=XiD(k)+ViD(k+1) (2)
In formula (1), k indicates that iterations, w indicate the inertia weight of particle, c1And c2Represent accelerated factor, r1And r2Generation
Random number between table [0,1], wherein w are used for controlling influence of the previous generation particle rapidities to contemporary particle rapidity；
Step 6 randomly generates M new particle, and is ranked up by fitness size to M+N particle of generation；
Newly generated population, is divided into fitness value using node particle Q and come by step 7, the position of setting node Q
A particles of Q (1≤Q≤N) of front and remaining M+N-Q particle；
Step 8 selects N-Q particle and adaptation using particle selection new probability formula from remaining M+N-Q particle
The Q particle that degree comes front forms new particle populations；
Step 9 meets condition (setting maximum iteration as 100) and terminates optimizing and export globally optimal solution, by population
The globally optimal solution finally searched into row threshold division and exports the image after segmentation as optimal segmenting threshold, to image.
Histogram is calculated using the integrated information of grey gradient, gray scale, distance, the specific steps are：
The Grad formula of step 1, pixel is calculated as:
Wherein, Gh(x,y),Gv(x, y) indicates level of the pixel at (x, y), vertical pixel value respectively；
Step 2, calculate apart from when, first coordinate is normalized, distance calculation formula is:
Wherein, i, j are respectively the coordinate value after being normalized, 0≤i, j≤1；
Step 3, it can thus be concluded that the calculation formula for going out histogram is:
H (i, j)=a1×G(i,j)+a2×D(i,j)+a3 (5)
Wherein, a1, a2, a3The respectively weights of gradient, distance, gray scale, and a1+a2+a3=1, a10≤a1,a2,a3≤1。
In order to improve diversity of particle swarm, on the basis of the N number of particle set originally, then M new particle is randomly generated,
And this M+N particle is calculated and sorted by fitness, then node Q is set, using node particle Q newly generated
Population is divided into two parts：
First part：Fitness value comes Q (1≤Q≤N) a particle of front,
Second part：The M+N-Q particle of fitness value rearward,
In addition, during particle is newer, in order to keep the scale of population constant, N- is selected from second part particle
Q particle and Q particle of first part form new population, and the select probability P (i) of particle i is in second part：
Wherein, fitness (i) indicates the fitness function value of particle i.
By Otsu algorithm principles, it is target, that is, area-of-interest of the present invention, gray value that gray value, which is more than optimal threshold,
It is background less than optimal threshold, to which image segmentation is come out.
Advantageous effect：
The present invention provides the Otsu lung images dividing methods based on improvement PSO, while improving segmentation precision, again
Splitting speed can be effectively improved.The case where traditional Otsu Threshold Segmentation Algorithms are divided there are target, background mistake, reason is to pass
The histogram that the Otsu Threshold Segmentation Algorithms of system use is the histogram based on gray value-neighboring mean value.To avoid target, the back of the body
The scape mistake present invention is selected based on gray scale, gradient, the Otsu Threshold Segmentation Algorithms of two-dimensional histogram apart from integrated information.Together
When, there is being easily trapped into locally optimal solution in conventional particle group's algorithm, therefore herein in order to change to particle group optimizing
Into in the case where ensureing that population scale is constant, it is proposed that the concept of node particle updates particle populations, improves particle
Diversity.Using population in two-dimentional Otsu Threshold Segmentation Algorithms optimization process, gray scale is mainly distributed on according to effective information
The characteristics of below the two-dimensional histogram of value-gradient-distance, the calculation amount of histogram is further reduced, and reduce grain
The search space of son.
Description of the drawings
Fig. 1 is that the present invention is based on the Otsu lung images dividing method flow charts for improving PSO.
Specific implementation mode
The specific embodiment of the invention is described in further detail below in conjunction with the accompanying drawings.
The present embodiment based on the Otsu lung images dividing methods for improving PSO, as shown in flow chart 1, this method is by PSO
Algorithm and two dimension Otsu algorithms are combined, and area-of-interest is extracted from lung images；Image is sought by PSO algorithms
Excellent processing is obtained globally optimal solution as optimal threshold, is split to image using Otsu algorithms.
The above-mentioned Otsu lung images dividing methods based on improvement PSO, PSO algorithms are introduced into Otsu algorithms, are obtained
Based on improve PSO Otsu lung images dividing methods, the algorithm realize the specific steps are：
Step 1 initializes population, including position and speed, sets population as N, search space is tieed up for D, particle i
It is expressed as X in the position of D dimension spacesiD={ Xi1, Xi2..., XiD, i={ 1,2,3 ..., N }, the flight speed of each example particle
Degree is expressed as ViD={ Vi1, Vi2..., ViD, i={ 1,2,3 ..., N }, the optimal location that each particle passes through is Pbest, kind
The position (global optimum position) with adaptive optimal control value that all particles are lived through in group is Gbest；
Step 2 calculates the grey level histogram of image；
Step 3 calculates the fitness value of each particle according to fitness function；
Step 4 updates the personal best particle Pbest of each particle and complete according to the fitness value of calculated particle
Office optimal location Gbest；
Step 5, being updated to the speed of particle, position according to more new formula, specifically more new formula is：
ViD(k+1)=w × ViD(k)+c1×r1[PbestiD-XiD(k)]+c2×r2[GbestiD-XiD(k)] (1)
XiD(k+1)=XiD(k)+ViD(k+1) (2)
In formula (1), k indicates that iterations, w indicate the inertia weight of particle, c1And c2Represent accelerated factor, r1And r2Generation
Random number between table [0,1], wherein w are used for controlling influence of the previous generation particle rapidities to contemporary particle rapidity；
Step 6 randomly generates M new particle, and is ranked up by fitness size to M+N particle of generation；
Newly generated population, is divided into fitness value using node particle Q and come by step 7, the position of setting node Q
A particles of Q (1≤Q≤N) of front and remaining M+N-Q particle；
Step 8 selects N-Q particle and adaptation using particle selection new probability formula from remaining M+N-Q particle
The Q particle that degree comes front forms new particle populations；
Step 9 meets condition (setting maximum iteration as 100) and terminates optimizing and export globally optimal solution, by population
The globally optimal solution finally searched into row threshold division and exports the image after segmentation as optimal segmenting threshold, to image.
Histogram is calculated using the integrated information of grey gradient, gray scale, distance, the specific steps are：
The Grad formula of step 1, pixel is calculated as:
Wherein, Gh(x,y),Gv(x, y) indicates level of the pixel at (x, y), vertical pixel value respectively；
Step 2, calculate apart from when, first coordinate is normalized, distance calculation formula is:
Wherein, i, j are respectively the coordinate value after being normalized, 0≤i, j≤1；
Step 3, it can thus be concluded that the calculation formula for going out histogram is:
H (i, j)=a1×G(i,j)+a2×D(i,j)+a3 (5)
Wherein, a1, a2, a3The respectively weights of gradient, distance, gray scale, and a1+a2+a3=1, a10≤a1,a2,a3≤1。
In order to improve diversity of particle swarm, on the basis of the N number of particle set originally, then M new particle is randomly generated,
And this M+N particle is calculated and sorted by fitness, then node Q is set, using node particle Q newly generated
Population is divided into two parts：
First part：Fitness value comes Q (1≤Q≤N) a particle of front,
Second part：The M+N-Q particle of fitness value rearward,
In addition, during particle is newer, in order to keep the scale of population constant, N- is selected from second part particle
Q particle and Q particle of first part form new population, and the select probability P (i) of particle i is in second part：
Wherein, fitness (i) indicates the fitness function value of particle i.
By Otsu algorithm principles, it is target, that is, area-of-interest of the present invention, gray value that gray value, which is more than optimal threshold,
It is background less than optimal threshold, to which image segmentation is come out.
Claims (5)
1. a kind of based on the Otsu lung images dividing methods for improving PSO, which is characterized in that calculate PSO algorithms and two dimension Otsu
Method is combined, and area-of-interest is extracted from lung images；Optimizing processing is carried out to image by PSO algorithms, obtains the overall situation most
Excellent solution is used as optimal threshold, is split to image using Otsu algorithms.
2. according to claim 1 based on the Otsu lung images dividing methods for improving PSO, which is characterized in that calculate PSO
Method is introduced into Otsu algorithms, obtains the specific steps realized based on the Otsu lung images dividing methods for improving PSO, the algorithm
For：
Step 1 initializes population, including position and speed, sets population as N, and search space is tieed up for D, and particle i is tieed up in D
The position in space is expressed as XiD={ Xi1,Xi2,…,XiD, the flying speed of i={ 1,2,3 ..., N }, each example particle are expressed as
ViD={ Vi1,Vi2,…,ViD, i={ 1,2,3 ..., N }, the optimal location that each particle passes through is Pbest, all grains in population
The position (global optimum position) with adaptive optimal control value that son is lived through is Gbest；
Step 2 calculates the grey level histogram of image；
Step 3 calculates the fitness value of each particle according to fitness function；
Step 4, according to the fitness value of calculated particle, update each particle personal best particle Pbest and it is global most
Excellent position Gbest；
Step 5, being updated to the speed of particle, position according to more new formula, specifically more new formula is：
ViD(k+1)=w × ViD(k)+c1×r1[PbestiD-XiD(k)]+c2×r2[GbestiD-XiD(k)] (1)
XiD(k+1)=XiD(k)+ViD(k+1) (2)
In formula (1), k indicates that iterations, w indicate the inertia weight of particle, c1And c2Represent accelerated factor, r1And r2Represent [0,
1] random number between, wherein w are used for controlling influence of the previous generation particle rapidities to contemporary particle rapidity；
Step 6 randomly generates M new particle, and is ranked up by fitness size to M+N particle of generation；
Newly generated population is divided into fitness value using node particle Q and comes front by step 7, the position of setting node Q
A particles of Q (1≤Q≤N) and remaining M+N-Q particle；
Step 8 selects N-Q particle and fitness to come using particle selection new probability formula from remaining M+N-Q particle
Q particle of front forms new particle populations；
Step 9 meets condition (setting maximum iteration as 100) and terminates optimizing and export globally optimal solution, and population is final
The globally optimal solution searched into row threshold division and exports the image after segmentation as optimal segmenting threshold, to image.
3. according to claim 2 based on the Otsu lung images dividing methods for improving PSO, which is characterized in that utilize ash
Gradient, gray scale, the integrated information of distance calculate histogram, the specific steps are：
The Grad formula of step 1, pixel is calculated as:
Wherein, Gh(x,y),Gv(x, y) indicates level of the pixel at (x, y), vertical pixel value respectively；
Step 2, calculate apart from when, first coordinate is normalized, distance calculation formula is:
Wherein, i, j are respectively the coordinate value after being normalized, 0≤i, j≤1；
Step 3, it can thus be concluded that the calculation formula for going out histogram is:
H (i, j)=a1×G(i,j)+a2×D(i,j)+a3 (5)
Wherein, a1, a2, a3The respectively weights of gradient, distance, gray scale, and a1+a2+a3=1, a10≤a1,a2,a3≤1。
4. according to the Otsu lung images dividing methods based on improvement PSO described in claim 2, which is characterized in that in order to improve
Diversity of particle swarm on the basis of the N number of particle set originally, then randomly generates M new particle, and to this M+N particle
It is calculated and is sorted by fitness, then node Q is set, newly generated population is divided into two parts using node particle Q：
First part：Fitness value comes Q (1≤Q≤N) a particle of front,
Second part：The M+N-Q particle of fitness value rearward,
In addition, during particle is newer, in order to keep the scale of population constant, N-Q are selected from second part particle
Particle and Q particle of first part form new population, and the select probability P (i) of particle i is in second part：
Wherein, fitness (i) indicates the fitness function value of particle i.
5. according to the Otsu lung images dividing methods based on improvement PSO described in claim 2, which is characterized in that calculated by Otsu
Method principle, it is target, that is, area-of-interest of the present invention that gray value, which is more than optimal threshold, and it is the back of the body that gray value, which is less than optimal threshold,
Scape, to which image segmentation is come out.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810021908.9A CN108346140A (en) | 2018-01-10 | 2018-01-10 | Based on the Otsu lung images dividing methods for improving PSO |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810021908.9A CN108346140A (en) | 2018-01-10 | 2018-01-10 | Based on the Otsu lung images dividing methods for improving PSO |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108346140A true CN108346140A (en) | 2018-07-31 |
Family
ID=62961414
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810021908.9A Pending CN108346140A (en) | 2018-01-10 | 2018-01-10 | Based on the Otsu lung images dividing methods for improving PSO |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108346140A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109872330A (en) * | 2019-01-25 | 2019-06-11 | 安徽理工大学 | A kind of two-dimentional Otsu Fast image segmentation method for improving lion group's optimization |
| CN111863232A (en) * | 2020-08-06 | 2020-10-30 | 罗春华 | Remote disease intelligent diagnosis system based on block chain and medical image |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106600606A (en) * | 2016-12-19 | 2017-04-26 | 上海电气自动化设计研究所有限公司 | Ship painting profile detection method based on image segmentation |
-
2018
- 2018-01-10 CN CN201810021908.9A patent/CN108346140A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106600606A (en) * | 2016-12-19 | 2017-04-26 | 上海电气自动化设计研究所有限公司 | Ship painting profile detection method based on image segmentation |
Non-Patent Citations (2)
| Title |
|---|
| 刘桂红: "《一种改进粒子群优化算法的Otsu图像阈值分割方法》", 《计算机科学》 * |
| 申铉京: "《基于一维Otsu的多阈值医学图像分割算法》", 《吉林大学学报（理学版）》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109872330A (en) * | 2019-01-25 | 2019-06-11 | 安徽理工大学 | A kind of two-dimentional Otsu Fast image segmentation method for improving lion group's optimization |
| CN109872330B (en) * | 2019-01-25 | 2022-10-14 | 安徽理工大学 | Two-dimensional Otsu rapid image segmentation method for improving lion group optimization |
| CN111863232A (en) * | 2020-08-06 | 2020-10-30 | 罗春华 | Remote disease intelligent diagnosis system based on block chain and medical image |
| CN111863232B (en) * | 2020-08-06 | 2021-02-19 | 深圳市柯尼达巨茂医疗设备有限公司 | Remote disease intelligent diagnosis system based on block chain and medical image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Gu et al. | Classification of bacterial and viral childhood pneumonia using deep learning in chest radiography | |
| Men et al. | More accurate and efficient segmentation of organs‐at‐risk in radiotherapy with convolutional neural networks cascades | |
| Liu et al. | Automatic organ segmentation for CT scans based on super-pixel and convolutional neural networks | |
| Hua et al. | Segmenting brain tumor using cascaded V-Nets in multimodal MR images | |
| CN109034221A (en) | A kind of processing method and its device of cervical cytology characteristics of image | |
| CN107977952A (en) | Medical image cutting method and device | |
| Guo et al. | Multi-level semantic adaptation for few-shot segmentation on cardiac image sequences | |
| CN108648182B (en) | Breast cancer nuclear magnetic resonance image tumor region segmentation method based on molecular subtype | |
| CN108897984A (en) | Based on correlation analysis between CT images group feature and lung cancer gene expression | |
| CN107705305A (en) | A kind of method and device of Medical Image Processing | |
| CN108898602A (en) | A kind of FCM medical image cutting method based on improvement QPSO | |
| Wiktorowicz et al. | Calcium Pattern Assessment in Patients with Severe Aortic Stenosis via the Chou’s 5-Steps Rule | |
| CN110189307B (en) | Pulmonary nodule detection method and system based on multi-model fusion | |
| CN107730542A (en) | Cone beam computed tomography image corresponds to and method for registering | |
| CN110751651A (en) | MRI pancreas image segmentation method based on multi-scale migration learning | |
| CN108537779A (en) | The method of vertebra segmentation and centroid detection based on cluster | |
| Yin et al. | Prediction of EGFR mutation status based on 18F-FDG PET/CT imaging using deep learning-based model in lung adenocarcinoma | |
| CN108346140A (en) | Based on the Otsu lung images dividing methods for improving PSO | |
| Weimin et al. | Enhancing Liver Segmentation: A Deep Learning Approach with EAS Feature Extraction and Multi-Scale Fusion | |
| Zhu et al. | Preliminary clinical study of the differences between interobserver evaluation and deep convolutional neural network-based segmentation of multiple organs at risk in CT images of lung cancer | |
| CN104915989B (en) | Blood vessel three-dimensional dividing method based on CT images | |
| Zhang et al. | Classification of benign and malignant pulmonary nodules based on deep learning | |
| Wu et al. | Automatic lung segmentation in CT images using dilated convolution based weighted fully convolutional network | |
| CN110858412B (en) | Heart coronary artery CTA model building method based on image registration | |
| Dickson et al. | An integrated 3D-sparse deep belief network with enriched seagull optimization algorithm for liver segmentation |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180731 |