CN109035227A - The system that lung tumors detection and diagnosis is carried out to CT image - Google Patents
The system that lung tumors detection and diagnosis is carried out to CT image Download PDFInfo
- Publication number
- CN109035227A CN109035227A CN201810772131.XA CN201810772131A CN109035227A CN 109035227 A CN109035227 A CN 109035227A CN 201810772131 A CN201810772131 A CN 201810772131A CN 109035227 A CN109035227 A CN 109035227A
- Authority
- CN
- China
- Prior art keywords
- candidate
- candidate nodule
- feature
- algorithm
- lung
- 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
- 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
- G06T7/10—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
- G06T2207/30064—Lung nodule
-
- 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
- G06T2207/30096—Tumor; Lesion
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The present invention provides a kind of system for carrying out Lung neoplasm detection and diagnosis to lung CT image.The system is divided into pulmonary parenchyma segmentation module, candidate nodule detection module and candidate nodule diagnostic module；Lung is divided in module, morphology denoising is carried out to original CT image, then carries out binarization segmentation with optimal threshold method, extracts initial boundary using frontier tracing method, application boundary matching repairs algorithm and carries out contour completion, finally obtains pulmonary parenchyma using flood filling algorithm segmentation lung；In candidate nodule detection module, the candidate nodule detection algorithm based on loop filter and the candidate nodule detection algorithm based on threshold value are combined；It will include a large amount of false positives in the result of detection, preliminary false positive candidate first carried out to candidate nodule with rule and method and is eliminated, then the fuzzy super box neural network in conjunction with K mean cluster algorithm is used to diagnose as system classifiers for candidate nodule.This system provides good support for diagnosis lung cancer.
Description
Technical field
The present invention relates to technical field of image processing more particularly to a kind of lung tumors detections based on ct images and diagnosis
System.
Background technique
CT image becomes the most common means of detection lung cancer at present, with the raising of CT precision, the CT that scanning generates every time
Picture number also greatly increases, and the diagnostic work amount of radiation technician aggravates, and be easy to cause mistaken diagnosis, computer-aided diagnosis system
The use of CAD can provide effective auxiliary for doctor, subtract the workload of light medical worker, also improve the efficiency and standard of diagnosis
True rate.
And Lung neoplasm assistant diagnosis system, it generally will be by lung areas segmentation, candidate nodule detection and candidate nodule point
Class diagnoses three steps.For lung segmentation common threshold method and region growing algorithm, such method can be close in segmentation result
Accidentally segmentation leads to the information for losing the diseased regions such as tubercle to pleural nodulations, or segmentation is inaccurate, is not able to satisfy clinical diagnosis
Needs；After segmentation obtains lung parenchyma section, candidate nodule detection, either gray threshold class are carried out in pulmonary parenchyma region
Detection method or combined shape feature and gray feature detection method, or the detection method based on filtering can all make
It include a large amount of false positive results in testing result, thus after candidate nodule detection, it needs to classify to candidate nodule, eliminate
False positive is candidate；It is candidate and carry out the diagnosis of tubercle in order to eliminate false positive, it is most of all to use rule-based classification method
Or linear classification method, however these two types of classification methods have limitation in terms of classification tubercle and non-nodules, wherein based on rule
Classifier then can only distinguish the special tubercle of shape, due to the feature that candidate nodule is extracted be mostly it is nonlinear, thus
Linear classifier is unable to get satisfactory result.
Summary of the invention
In view of this, the embodiment of the present invention provides the system of a kind of lung tumors detection and diagnosis based on ct images, until
Small part solves problems of the prior art.
The system that a kind of pair of CT image provided in an embodiment of the present invention carries out lung tumors detection and diagnosis, comprising:
Pulmonary parenchyma divides module, for handling the chest CT image got, by pulmonary parenchyma region and chest region
Irrelevant region segmentation comes out outside domain and body；
Candidate nodule detection module obtains candidate nodule for detecting to the pulmonary parenchyma being partitioned into；
Tubercle diagnostic module, for carrying out feature extraction operation to all candidate nodules, and in conjunction with the fuzzy of k mean value
Super box neural network algorithm carries out classification diagnosis to tubercle.
A kind of specific implementation according to an embodiment of the present invention, the pulmonary parenchyma segmentation module are also used to:
Carry out noise suppression preprocessing operation based on morphologic opening operation, clothing outside removal body in irrelevant region and
The noises such as Medical Devices；
Algorithm based on optimal threshold just divides pulmonary parenchyma, according to Otsu method threshold value；
Initial lung boundary is extracted to every CT image edge following algorithm；
It is matched using adaptive boundary and repairs algorithm reparation reparation lung outlines；
According to complete lung outlines, divide to obtain complete pulmonary parenchyma using flood filling algorithm.
A kind of specific implementation according to an embodiment of the present invention, the adaptive boundary matching repair algorithm according to lung
The size of portion boundary recess, adaptive adjustment match step-length.
A kind of specific implementation according to an embodiment of the present invention, the candidate nodule detection module, is also used to:
According to the morphological feature of tubercle, with the candidate nodule detection method based on loop filter, extraction obtains seed knot
Point carries out region growing to seed node using the region growing algorithm based on seed, and it is opposite to obtain feature for dividing candidate tubercle
Complete candidate nodule；
According to the gray feature of tubercle, with the candidate nodule detection method based on gray threshold, after extracting seed node, benefit
With global area growth algorithm dividing candidate tubercle；
False positive is carried out using rule and method to the loop filter detection candidate nodule in candidate nodule detection module to disappear
It removes；
To being disappeared based on the candidate nodule of gray threshold using rule and method progress false positive in candidate nodule detection module
It removes；
The candidate nodule result that will test is combined together.
A kind of specific implementation according to an embodiment of the present invention chooses optimal threshold using iterative algorithm, to the ash
Degree threshold value optimizes.
A kind of specific implementation according to an embodiment of the present invention, the elimination rule of formulation are R1, R2 and R3, R1, R2 and
R3 is based respectively on volume characteristic, radial features and the sphericity feature of candidate nodule；
WithThe bottom threshold of volume, diameter and sphericity is respectively represented, if candidate nodule pair
Feature is answered to be less than bottom threshold, it will be considered as false positive candidate, to be eliminated；
WithThe upper threshold of volume, diameter and sphericity is respectively represented, if candidate nodule pair
Feature is answered to be greater than upper threshold, it will be taken as false positive candidate to eliminate.
A kind of specific implementation according to an embodiment of the present invention, the threshold value be according to the characteristic information of candidate nodule into
Row setting.
A kind of specific implementation according to an embodiment of the present invention chooses the feature of 12 most common characterization tubercles, will
These features constitute vectorAs the input vector of the super box neural network classifier with compensation member, input
Classify into classifier to candidate nodule, the diagnosis for candidate nodule.
A kind of specific implementation according to an embodiment of the present invention, all features are all based on the letter such as form, shape, gray scale
What breath was calculated, the feature contains 6 gray features: the maximum value of pixel gray value, time in candidate nodule region
Select the minimum value of pixel gray value in knuckle areas, the average value of pixel gray value, candidate nodule in candidate nodule region
The standard deviation of pixel gray value, degree of skewness feature, kurtosis feature in region；
The feature further includes 6 morphological features: volume characteristic, sphericity feature, elongation percentage feature, characteristics of diameters, cube
Compactness feature, compactness feature.
A kind of specific implementation according to an embodiment of the present invention, using fuzzy super box neural network classifier, wherein super
The determination of box spreading coefficient carries out cluster operation to training sample according to k means clustering algorithm, is estimated with the size of class cluster
It arrives.
Lung tumors and diagnostic system based on ct images of the invention have the following beneficial effects: that give one complete
The CAD system of the kind segmentation of slave pulmonary parenchyma, candidate nodule detection, false-positive nodule elimination and tubercle diagnosis；Divide in pulmonary parenchyma
Module successively denoises image, is divided, being repaired, filling processing, fast and accurately obtaining pulmonary parenchyma；It is examined in candidate nodule
Survey module, loop filter detection method and threshold detection method are combined, not only can detecte solitary tubercle, to gray scale compared with
Small frosted glass and nearly blood vessel nodes can detect well, there is higher CAD detection tubercle susceptibility；To detection mould
The candidate nodule that block detects carries out false positive elimination before diagnosis, using regular method, reduces the work of follow-up diagnosis work
Amount；Finally, proposing to estimate super box neural network spreading coefficient based on k- means clustering algorithm, the automatic of classifier is improved
Property.
Detailed description of the invention
In order to illustrate the technical solution of the embodiments of the present invention more clearly, below will be to needed in the embodiment attached
Figure is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the invention, for this field
For those of ordinary skill, without creative efforts, it can also be obtained according to these attached drawings other attached drawings.
Fig. 1 is area of computer aided lung lesion detection and diagnostic system block schematic illustration；
Fig. 2 is that pulmonary parenchyma divides module flow diagram schematic diagram；
Fig. 3 is lung segmentation result figure；
Fig. 4 is candidate nodule detection module design diagram；
Fig. 5 is candidate nodule diagnostic module design diagram.
Specific embodiment
The embodiment of the present invention is described in detail with reference to the accompanying drawing.
It will be appreciated that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.Base
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts it is all its
Its embodiment, shall fall within the protection scope of the present invention.
Specific embodiments of the present invention are given below, help for the understanding of the present invention.
As shown in Figure 1, lung tumors detection based on ct images provided by the present invention and diagnostic system, comprising:
A) pulmonary parenchyma divides module, handles the chest CT image got, by pulmonary parenchyma region and chest area
And irrelevant region segmentation comes out outside body；
B) candidate nodule detection module detects the pulmonary parenchyma being partitioned into, obtains candidate nodule；
C) tubercle diagnostic module carries out feature extraction operation to all candidate nodules, and in conjunction with the fuzzy super of k mean value
Box neural network algorithm carries out classification diagnosis to tubercle.
It is of the present invention a) in pulmonary parenchyma divide module, main flow is as shown in Figure 2:
1) for original CT image, noise suppression preprocessing operation is carried out with based on morphologic opening operation, is removed outside body
The noises such as clothing and Medical Devices in irrelevant region, the characteristics of according to noise, in this embodiment, selection diameter is 5 pictures
The disc structure member of vegetarian refreshments does opening operation processing to CT image.
2) just divide pulmonary parenchyma using the algorithm based on optimal threshold, according to Otsu method threshold value, thought is choosing
A threshold value T is taken, so that the inter-class variance of segmentation result is reached maximum, while variance within clusters are minimum；
According to the threshold value T that Otsu method determines, chest CT image is obtained to segmentation and is handled, the image after threshold process
G (x, y) is defined as:
3) initial lung boundary is extracted to every CT image edge following algorithm, after edge following algorithm operates,
A series of available pixels, these pixels constitute a closed lung outlines L={ (p1,p2,...,pn)}。
4) algorithm reparation is repaired using adaptive boundary matching and repair lung outlines, mainly solve pulmonary parenchyma cutting procedure
In, the recess of boundary caused by the nearly pleural nodulations of over-segmentation；
Since lung boundary recess size is different, one can not be found and be effectively matched step-length, to the lung side of all sizes
Boundary's recess is all effective, therefore uses improved matching step-length, can adaptively be adjusted to different recess sizes；
In matching process, when finding rightest point within the scope of step-length, we connect starting point and rightest point and are recessed to boundary
It is repaired, while the intermediate point between starting point and rightest point being adjusted on connecting line segment.In an experiment, by repeatedly surveying
Examination, it is final to determine that matching step-length is 50pixels, adaptive threshold λ0It is 0.33, adaptive adjusting parameter δ=0.9.
5) finally lung outlines are filled using flood completion method according to complete lung outlines, using filling
To binary image divide in original CT and obtain pulmonary parenchyma.
Fig. 3 illustrates the result figure of lung's different location pulmonary parenchyma segmentation pilot process, and Fig. 3 (a) is initial CT input
Image；After noise suppression preprocessing, binarization segmentation, effect such as Fig. 3 (b) are carried out to image using optimal threshold algorithm；Then
It selects edge following algorithm to extract initial lung boundary, obtains lung segmentation initial results, Fig. 3 (c) using initial boundary；It utilizes
Adaptive boundary matching is repaired algorithm and is repaired to lung boundary, according to the lung boundary after reparation, is filled and is calculated using flood
Method divides lung areas, Fig. 3 (d).By the displaying divided to different location lung areas, it can be proved that our segmentation side
Method is effective to lung segmentation.
Candidate nodule detection module b) in the present invention, main flow are as shown in Figure 4:
1) according to the morphological feature of tubercle, with the candidate nodule detection method based on loop filter, extraction obtains seed
Node, process are as follows:
(1) to lung CT image f (x, y), with disk-like structure member Bd(x, y) carries out morphological dilations processing, obtains result
Scheme g(d)(x,y)；
(2) to lung CT image f (x, y), with cyclic structure member Br(x, y) carries out morphological dilations processing, obtains result
Scheme g(r)(x,y)；
(3) to image g(d)(x, y) and g(r)It is poor that (x, y) makees, and obtains image g (x, y)；
(4) for image g (x, y) indicate be two images difference in height, according to the analysis of front isolate shade position
Difference in height is bigger, thus can be handled with selected threshold T pixel each in image g (x, y) (x', y'), if g
(x', y') is greater than threshold value T, then the point is considered the seed node of candidate nodule；
And threshold value T is to calculate to assess by many experiments, can choose T is 650 as plate-like filtering image and cyclic annular filtering
The threshold value of picture altitude difference；
For accurate dividing candidate tubercle, it is raw that region is carried out to seed node using the region growing algorithm based on seed
It is long, the relatively complete candidate nodule of feature is obtained, so as to carry out further discriminant classification to candidate nodule.
2) according to the gray feature of tubercle, with the candidate nodule detection method based on gray threshold, after extracting seed node,
The selection of gray threshold chooses optimal threshold with iterative algorithm, recycles global area growth algorithm dividing candidate tubercle；
3) false positive is carried out using rule and method to the loop filter detection candidate nodule in candidate nodule detection module
It eliminates, eliminating rule is R1, R2 and R3, these three rules are based respectively on volume characteristic, radial features and the sphericity of candidate nodule
Feature needs to limit a maximum value and minimum value for the feature in each rule, uses to eliminate those false positives candidateThe bottom threshold of volume, diameter and sphericity is respectively represented, if candidate nodule character pair is less than
Bottom threshold, it will be considered as false positive candidate, to be eliminated；It is similarRespectively represent body
The upper threshold of product, diameter and sphericity, if candidate nodule character pair is greater than upper threshold, it will be taken as false positive candidate
It eliminates；
In addition, forWithThese threshold values are according to candidate
The characteristic information of tubercle is set, and in order to avoid eliminating Lung neoplasm, threshold condition is needed to relax.
4) false positive is carried out using rule and method to the candidate nodule based on gray threshold in candidate nodule detection module
It eliminates；
5) the candidate nodule result that two methods detect is combined together, it is respective not makes up two kinds of detection methods
Foot place.
Since the diameter of tubercle is generally between 3mm~30mm, but also have 40mm or more tubercle occur, this module in line with
From the point of view of improving verification and measurement ratio, we relax rule condition, takeMake
For the lower and upper limit of volume, wherein vol_pix is volume occupied by each pixel；It takes
Lower and upper limit as diameter；It takesLower and upper limit as characteristics of diameters.
Tubercle diagnostic module c) in the present invention is main innovation point of the invention, and process is as shown in Figure 5:
For the feature of Efficient Characterization candidate nodule, the feature of 12 most common characterization tubercles is summarized, all features are all
It is to be calculated based on information such as form, shape, gray scales, wherein containing 6 gray features: (the candidate nodule area Maximum
The maximum value of pixel gray value in domain), Minimum (minimum value of pixel gray value in candidate nodule region), Mean (are waited
Select the average value of pixel gray value in knuckle areas), Standard deviation (pixel gray level in candidate nodule region
The standard deviation of value), Skewness (degree of skewness feature), Kurtosis (kurtosis feature) and 6 morphological features: Volume (volume
Feature), Sphericity (sphericity feature), Elongation (elongation percentage feature), Diameter (characteristics of diameters), Cube
Compactness (cube compactness feature), Compactness (compactness feature)；
These features constitute vectorAs with compensation member super box neural network classifier input to
Amount, is input in classifier and classifies to candidate nodule, the diagnosis for candidate nodule.
And super its training process of box neural network with compensation member is as follows:
Step -1: super box spreading coefficient is determined with K mean cluster algorithm
(1) k class cluster center is initialized
(2) iteration optimization clusters, and iterative process is as follows: (a) according to distance, sample point being grouped into nearest center class, (b)
The center of each class cluster is replaced with average value, and iteration (a), (b) are until class cluster center is constant or variation is less than specified threshold.
(3) maximum class cluster size L=max { L is takeniI=1...k } as super box upper dimension bound, then with θ=L/n come
Spreading coefficient is calculated, wherein n is the dimension in space, LiFor the size of i-th of class cluster, Manhattan distance is selected to be used as size
Measurement.
Step -2: the foundation and extension of super box
(1) finding can include training sample AhSuitable super box bjOr carry out (2)；
(2) new super box is established to include this training sample.
Step -3: compensation neuron is established
According to following test, the super box extended in judgment step -1 is that neural network increases OCN nerve according to judging result
Member or CCN neuron.
(a) test is isolated
The super box b of extended operation is in checking procedure -1jWhether isolate, carries out if not isolated comprising test.
(b) comprising test
The super box b extended in testing procedure -1jWhether comprising the super box of other inhomogeneities or by the super box packet of other inhomogeneities
Contain.If test result be it is determining, establish a CCN neuron, otherwise establish an OCN neuron.
In embodiment, training sample and test sample are divided using three kinds of different ratios.30-70:30% is trained
Sample ratio, 70% is test sample ratio.70-30:70% is training sample ratio, and 30% is test sample ratio.50-
50:50% is training sample ratio, and 50% is test sample ratio.
By test, 70-30 group data can obtain better performance susceptibility, this CAD compared with Most current system,
With preferable detectability.
It should be appreciated that each section of the invention can be realized with hardware, software, firmware or their combination.
The above description is merely a specific embodiment, but scope of protection of the present invention is not limited thereto, any
In the technical scope disclosed by the present invention, any changes or substitutions that can be easily thought of by those familiar with the art, all answers
It is included within the scope of the present invention.Therefore, protection scope of the present invention should be subject to the protection scope in claims.
Claims (10)
1. the system that a kind of pair of CT image carries out lung tumors detection and diagnosis characterized by comprising
Pulmonary parenchyma divides module, for handling the chest CT image got, by pulmonary parenchyma region and chest area and
Irrelevant region segmentation comes out outside body；
Candidate nodule detection module obtains candidate nodule for detecting to the pulmonary parenchyma being partitioned into；
Tubercle diagnostic module for carrying out feature extraction operation to all candidate nodules, and obscures super box in conjunction with k mean value
Neural network algorithm carries out classification diagnosis to tubercle.
2. system according to claim 1, which is characterized in that the pulmonary parenchyma segmentation module is also used to:
Noise suppression preprocessing operation is carried out based on morphologic opening operation, clothing and medical treatment outside removal body in irrelevant region
The noises such as equipment；
Algorithm based on optimal threshold just divides pulmonary parenchyma, according to Otsu method threshold value；
Initial lung boundary is extracted to every CT image edge following algorithm；
It is matched using adaptive boundary and repairs algorithm reparation reparation lung outlines；
According to complete lung outlines, divide to obtain complete pulmonary parenchyma using flood filling algorithm.
3. system according to claim 2, it is characterised in that:
The size that algorithm is recessed according to lung boundary is repaired in the adaptive boundary matching, and adaptive adjustment matches step-length.
4. system according to claim 1, which is characterized in that the candidate nodule detection module is also used to:
According to the morphological feature of tubercle, with the candidate nodule detection method based on loop filter, extraction obtains seed node, answers
Region growing is carried out to seed node with the region growing algorithm based on seed, it is relatively complete to obtain feature for dividing candidate tubercle
Candidate nodule；
According to the gray feature of tubercle, with the candidate nodule detection method based on gray threshold, after extracting seed node, using complete
Office's region growing algorithm dividing candidate tubercle；
False positive elimination is carried out using rule and method to the loop filter detection candidate nodule in candidate nodule detection module；
False positive elimination is carried out using rule and method to the candidate nodule based on gray threshold in candidate nodule detection module；
The candidate nodule result that will test is combined together.
5. system according to claim 4, it is characterised in that:
Optimal threshold is chosen using iterative algorithm, the gray threshold is optimized.
6. system according to claim 4, it is characterised in that:
The elimination rule of formulation is R1, R2 and R3, R1, R2 and R3 be based respectively on the volume characteristic of candidate nodule, radial features and
Sphericity feature；
WithThe bottom threshold of volume, diameter and sphericity is respectively represented, if candidate nodule character pair
Less than bottom threshold, it will be considered as false positive candidate, to be eliminated；
WithThe upper threshold of volume, diameter and sphericity is respectively represented, if candidate nodule is corresponding special
Sign is greater than upper threshold, it will be taken as false positive candidate to eliminate.
7. system according to claim 6, it is characterised in that:
The threshold value is set according to the characteristic information of candidate nodule.
8. system according to claim 6, it is characterised in that:
These features are constituted vector by the feature for choosing 12 most common characterization tuberclesIt is compensated as band
The input vector of the super box neural network classifier of member, is input in classifier and classifies to candidate nodule, for candidate knot
The diagnosis of section.
9. system according to claim 8, it is characterised in that:
All features are all based on what the information such as form, shape, gray scale were calculated, and the feature contains 6 gray features:
The maximum value of pixel gray value in candidate nodule region, the minimum value of pixel gray value, candidate knot in candidate nodule region
Save the average value of pixel gray value in region, in candidate nodule region pixel gray value standard deviation, degree of skewness feature, peak
Spend feature；
The feature further includes 6 morphological features: volume characteristic, sphericity feature, elongation percentage feature, characteristics of diameters, cube closely knit
Spend feature, compactness feature.
10. system according to claim 8, it is characterised in that:
Using fuzzy super box neural network classifier, wherein the determination of super box spreading coefficient, according to k means clustering algorithm to instruction
Practice sample and carry out cluster operation, is estimated to obtain with the size of class cluster.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810772131.XA CN109035227A (en) | 2018-07-13 | 2018-07-13 | The system that lung tumors detection and diagnosis is carried out to CT image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810772131.XA CN109035227A (en) | 2018-07-13 | 2018-07-13 | The system that lung tumors detection and diagnosis is carried out to CT image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109035227A true CN109035227A (en) | 2018-12-18 |
Family
ID=64642785
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810772131.XA Pending CN109035227A (en) | 2018-07-13 | 2018-07-13 | The system that lung tumors detection and diagnosis is carried out to CT image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109035227A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109711315A (en) * | 2018-12-21 | 2019-05-03 | 四川大学华西医院 | A kind of method and device of Lung neoplasm analysis |
| CN109816655A (en) * | 2019-02-01 | 2019-05-28 | 华院数据技术(上海)有限公司 | Lung neoplasm image feature detection method based on ct images |
| CN109978880A (en) * | 2019-04-08 | 2019-07-05 | 哈尔滨理工大学 | Lung tumors CT image is carried out sentencing method for distinguishing using high dimensional feature selection |
| CN110276407A (en) * | 2019-06-26 | 2019-09-24 | 哈尔滨理工大学 | A kind of Hepatic CT staging system and classification method |
| CN113129314A (en) * | 2021-04-01 | 2021-07-16 | 上海市第一人民医院 | Intelligent image processing method for lung shadows |
| CN113538409A (en) * | 2021-08-05 | 2021-10-22 | 佳木斯大学 | Cervical cancer image region segmentation method based on fuzzy logic and ANFIS |
| CN116681701A (en) * | 2023-08-02 | 2023-09-01 | 青岛市妇女儿童医院（青岛市妇幼保健院、青岛市残疾儿童医疗康复中心、青岛市新生儿疾病筛查中心） | Children lung ultrasonic image processing method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104751178A (en) * | 2015-03-31 | 2015-07-01 | 上海理工大学 | Pulmonary nodule detection device and method based on shape template matching and combining classifier |
| CN105469419A (en) * | 2016-01-07 | 2016-04-06 | 于翠妮 | Pulmonary nodule detection method based on CT image |
-
2018
- 2018-07-13 CN CN201810772131.XA patent/CN109035227A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104751178A (en) * | 2015-03-31 | 2015-07-01 | 上海理工大学 | Pulmonary nodule detection device and method based on shape template matching and combining classifier |
| CN105469419A (en) * | 2016-01-07 | 2016-04-06 | 于翠妮 | Pulmonary nodule detection method based on CT image |
Non-Patent Citations (2)
| Title |
|---|
| JINKE WANG ET AL: "A New Pulmonary Nodules Computer-aided Detection System in Chest CT Images Based on Adaptive Fuzzy C-Means Technology", 《2015 7TH INTERNATIONAL CONFERENCE ON INTELLIGENT HUMAN-MACHINE SYSTEMS AND CYBERNETICS》 * |
| 翟志伟: "基于CT图像的肺部肿瘤检测与诊断", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109711315A (en) * | 2018-12-21 | 2019-05-03 | 四川大学华西医院 | A kind of method and device of Lung neoplasm analysis |
| CN109816655A (en) * | 2019-02-01 | 2019-05-28 | 华院数据技术(上海)有限公司 | Lung neoplasm image feature detection method based on ct images |
| CN109816655B (en) * | 2019-02-01 | 2021-05-28 | 华院计算技术(上海)股份有限公司 | Pulmonary nodule image feature detection method based on CT image |
| CN109978880A (en) * | 2019-04-08 | 2019-07-05 | 哈尔滨理工大学 | Lung tumors CT image is carried out sentencing method for distinguishing using high dimensional feature selection |
| CN110276407A (en) * | 2019-06-26 | 2019-09-24 | 哈尔滨理工大学 | A kind of Hepatic CT staging system and classification method |
| CN113129314A (en) * | 2021-04-01 | 2021-07-16 | 上海市第一人民医院 | Intelligent image processing method for lung shadows |
| CN113129314B (en) * | 2021-04-01 | 2022-08-02 | 上海市第一人民医院 | Intelligent image processing method for lung shadows |
| CN113538409A (en) * | 2021-08-05 | 2021-10-22 | 佳木斯大学 | Cervical cancer image region segmentation method based on fuzzy logic and ANFIS |
| CN116681701A (en) * | 2023-08-02 | 2023-09-01 | 青岛市妇女儿童医院（青岛市妇幼保健院、青岛市残疾儿童医疗康复中心、青岛市新生儿疾病筛查中心） | Children lung ultrasonic image processing method |
| CN116681701B (en) * | 2023-08-02 | 2023-11-03 | 青岛市妇女儿童医院（青岛市妇幼保健院、青岛市残疾儿童医疗康复中心、青岛市新生儿疾病筛查中心） | Children lung ultrasonic image processing method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109035227A (en) | The system that lung tumors detection and diagnosis is carried out to CT image | |
| CN109493954B (en) | SD-OCT image retinopathy detection system based on category distinguishing and positioning | |
| CN109325942B (en) | Fundus image structure segmentation method based on full convolution neural network | |
| Zhao et al. | Retinal vessels segmentation based on level set and region growing | |
| CN106056595B (en) | Based on the pernicious assistant diagnosis system of depth convolutional neural networks automatic identification Benign Thyroid Nodules | |
| TWI496112B (en) | Cell image segmentation method and a nuclear-to-cytoplasmic ratio evaluation method using the same | |
| US7474775B2 (en) | Automatic detection of red lesions in digital color fundus photographs | |
| Xu et al. | Automated optic disk boundary detection by modified active contour model | |
| CN109858540B (en) | Medical image recognition system and method based on multi-mode fusion | |
| CN108257135A (en) | The assistant diagnosis system of medical image features is understood based on deep learning method | |
| Garg et al. | Unsupervised curvature-based retinal vessel segmentation | |
| CN107346545A (en) | Improved confinement growing method for the segmentation of optic cup image | |
| CN108186051B (en) | Image processing method and system for automatically measuring double-apical-diameter length of fetus from ultrasonic image | |
| Melo et al. | Microaneurysm detection in color eye fundus images for diabetic retinopathy screening | |
| CN107680110B (en) | Inner ear three-dimensional level set segmentation method based on statistical shape model | |
| CN109919254B (en) | Breast density classification method, system, readable storage medium and computer device | |
| CN106204555A (en) | A kind of combination Gbvs model and the optic disc localization method of phase equalization | |
| CN104217213B (en) | A kind of medical image multistage sorting technique based on symmetric theory | |
| CN107563996A (en) | A kind of new discus nervi optici dividing method and system | |
| Suryani et al. | The hybrid method of SOM artificial neural network and median thresholding for segmentation of blood vessels in the retina image fundus | |
| CN103279960B (en) | A kind of image partition method of human body cache based on X-ray backscatter images | |
| Purnama et al. | Follicle detection on the usg images to support determination of polycystic ovary syndrome | |
| Devasia et al. | Automatic Optic Disc Localization and Segmentation using Swarm Intelligence. | |
| Krishnamoorthi et al. | Hybrid feature vector based detection of Glaucoma | |
| Devasia et al. | Automatic optic disc boundary extraction from color fundus images |
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: 20181218 |