CN105740875A - Pulmonary nodule multi-round classification method based on multi-scale three-dimensional block feature extraction - Google Patents
Pulmonary nodule multi-round classification method based on multi-scale three-dimensional block feature extraction Download PDFInfo
- Publication number
- CN105740875A CN105740875A CN201610154156.4A CN201610154156A CN105740875A CN 105740875 A CN105740875 A CN 105740875A CN 201610154156 A CN201610154156 A CN 201610154156A CN 105740875 A CN105740875 A CN 105740875A
- Authority
- CN
- China
- Prior art keywords
- tuberosity
- feature
- feature extraction
- classification
- grader
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/03—Recognition of patterns in medical or anatomical images
- G06V2201/031—Recognition of patterns in medical or anatomical images of internal organs
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The invention provides a pulmonary nodule multi-round classification method based on multi-scale three-dimensional block feature extraction. The concept of a multi-scale mask is introduced in a nodule mask of feature extraction; feature extraction is performed on nodules by utilizing blocks of different scales, and the extracted features are inputted to a first classifier; then the classified false positive nodules are inputted to a second classifier; and finally subsets of the true positive nodules in the two classifiers are combined. Detailed information in an image can be better captured by focusing on multi-scale feature extraction of the image and the classification aspect of the image so that representative feature sets can be generated. New feature pack models and classification models can be specifically applied to classification and retrieval of medical image, remote sensing image, network image and other digital image data.
Description
Technical field
The invention belongs to computer vision field, more specifically say, relate to a kind of Lung neoplasm based on multiple dimensioned three-dimensional bits feature extraction and take turns classification and Detection method more.
Background technology
The annual cancer new cases of China are 3,120,000 examples, and every year because cancer mortality is more than 2,000,000 examples, wherein dead maximum cancer kind is pulmonary carcinoma.The clinical stages when cure rate of pulmonary carcinoma and diagnosis, is closely related, and 5 years survival rates of the patients with lung cancer of early stage are more than 90%, and the survival rate of I phase patients with lung cancer reduces to 60%, and the year survival rate of the patients with lung cancer of II to IV phase drops to 5% from 40%.Therefore, " early find, early diagnosis, early treatment " it is the key carrying patients with lung cancer survival rate.But it is not promising that the pulmonary carcinoma of only 15% is discovered in an early phase.But when there is coughing the ill symptoms do not healed for a long time with hemoptysis etc., Zai Qu hospital checks, often just own through being advanced lung cancer.Therefore, how to accomplish that " early finding " and " early diagnosis " is the important topic needing research.
Detection Lung neoplasm CADe system be generally made up of five subsystems: obtain, pretreatment, segmentation, nodule detection and reduce false positive.The prime responsibility obtaining subsystem obtains medical picture exactly.Public data storehouse can be used to development, training and checking CADe system.It can be used for training of medical student equally, and it can make comparisons between different CADe systems.Main preconditioning technique has: medium filtering, boostfiltering, and contrast limited adaptive histogram equalization strengthens automatically, Wiener filtering, fast Fourier transform, wavelet transformation, anti-geometrical attenuation, erosion filter, smothing filtering and noise correction.
Two main methods of segmentation lung images are: based on the segmentation of the segmentation of threshold value and deformable model.Based in the segmentation of threshold value, the threshold value of a brightness implements lock out operation.The major type of deformable model that segmentation pulmonary picture uses has: active contour and the deformable model based on level set.It is that the best bet of segmentation tumor region is because the border found is very perfectly thin and connects that region increases.Other technology splitting Lung neoplasm in history has: cylindrical and spherical filter, morphology operations, thresholding, multiple gray level thresholding and connection element analysis.The main feature extracted has: the brightness value of pixel, morphology, texture, fractal.Main grader has: linear discriminant analysis, rule-based, cluster, markov random file, artificial neural network, support vector machine (SVM), and the neutral net (MTANNs) of a large amount of training, dual threshold cuts.
Summary of the invention
It is an object of the invention to design a kind of extraction that Lung neoplasm carries out feature from different scale, and representative feature is put in two-wheeled grader, the result of two-wheeled classification is done a merging, to improve last sensitivity.
For achieving the above object, a kind of Lung neoplasm based on multiple dimensioned three-dimensional bits feature extraction of the present invention takes turns classification and Detection method more, mainly include herein below: in the tuberosity mask of feature extraction, introduce the concept of multiple dimensioned mask, tuberosity is carried out feature extraction by the block utilizing different scale, and the feature of extraction is put in first grader, then put in second grader by what classify again for false-positive tuberosity again, finally merge the subset of true positives tuberosity in two graders.
Know-why is as it is shown in figure 1, concrete techniqueflow is as follows:
Step one: the acquisition of picture, it is intended to utilize public data storehouse LIDC, and the pretreatment carrying out being correlated with directly can be processed by matlab;
Step 2: original image is carried out primary segmentation first by threshold value, secondly use the element of disc structure that the pulmonary of coarse segmentation is carried out Two-dimensional morphology opening operation, then three-dimensional hexa-atomic element is used to connect the part extracting maximum volume, this part is exactly the pulmonary of segmentation, it is then used by Hole filling algorithms to repair hole, finally employs three dimensional morphology closed operation and come the edge of refine pulmonary;
Step 3: first by a series of threshold values, original image is carried out Threshold segmentation, then the result of segmentation and pulmonary's mask are carried out logical AND operation, then carry out Two-dimensional morphology opening operation with the element of a series of disc structures tuberosity to having split again, finally merge all of middle nodule mask and form last nodule mask;
Step 4: feature extraction, it is intended to utilize more traditional features, and extract based on the block that different scale is three-dimensional.Feature mainly has three major types: geometry, brightness, Gradient Features, and all these feature contains two dimension and three-dimensional.Last again by the feature that SFS is next preferably a series of;
Step 5: the total two-wheeled classification of grader classification one, the first round carries out two classification with SVM, second takes turns and again classifies to being categorized as false-positive tuberosity in the first round with FLD grader, finally the true positives tuberosity in result is merged as last result.
Accompanying drawing explanation
Fig. 1 is the theory diagram that a kind of Lung neoplasm based on multiple dimensioned three-dimensional bits feature extraction of the present invention takes turns classification and Detection method more.
Fig. 2 is the technical scheme figure that a kind of Lung neoplasm based on multiple dimensioned three-dimensional bits feature extraction of the present invention takes turns classification and Detection method more.
Detailed description of the invention
Below in conjunction with accompanying drawing, the specific embodiment of the present invention is described, in order to those skilled in the art is more fully understood that the present invention.Requiring particular attention is that, in the following description, it is possible to the detailed description of the known function and design of desalinating main contents of the present invention will be left in the basket.
In the present embodiment, a kind of feature bag construction method based on multiscale analysis of the present invention mainly includes following link: 1. picture obtain, 2. lung segmentation, 3. nodule segmentation, 4. Multi resolution feature extraction, 5. two take turns classification.
The acquisition of picture, it is intended to utilize public data storehouse LIDC, and the pretreatment carrying out being correlated with directly can be processed by matlab.
The segmentation of pulmonary, the content mainly utilizing threshold method relevant with morphology is split.Respectively original image is carried out primary segmentation first by two threshold values, secondly use the element of disc structure that the pulmonary of coarse segmentation is carried out Two-dimensional morphology opening operation, then using three-dimensional hexa-atomic element to connect the part extracting maximum volume, this part is exactly the pulmonary of segmentation.Subsequently employ Hole filling algorithms to repair these holes, finally employ three dimensional morphology closed operation and come the edge of refine pulmonary.
The segmentation of Lung neoplasm, morphologic method is adopted to split, first by a series of threshold values, original image is carried out Threshold segmentation, then the result of segmentation and pulmonary's mask are carried out logical AND operation, then carry out opening operation with the element of a series of disc structures tuberosity to having split again and remove some residual error structures, finally merge all of middle nodule mask and form last nodule mask.
Feature extraction, it is intended to utilizing more traditional features, feature mainly has three major types: geometry, brightness, Gradient Features.First to the nodular feature bounding box split, then carry out expanding 3 pixels on the basis of bounding box, enable the tuberosity major part of not precisely segmentation to include.Owing to the precisely segmentation of tuberosity has a definite limitation, so not carrying out feature extraction on the tuberosity mask split, and carrying out feature extraction in the three-dimensional bits expanded, enabling tuberosity major part to be included.All these feature contains two dimension and three-dimensional, omnibearing can represent tuberosity.Finally select the algorithm of excellent namely dimensionality reduction to carry out preferably a series of feature by feature again, intend adopting advance algorithm SFS to carry out preferred feature.
Grader is classified, first the feature that above-mentioned link is extracted is put in SVM classifier and is gone all of tuberosity is carried out two classification, then put in FLD grader be categorized as intubercular characteristic set to improve sensitivity again again this part tuberosity is carried out two classification, finally constitute last result double classification result being divided into the subset of tuberosity merge.
A kind of Lung neoplasm based on multiple dimensioned three-dimensional bits feature extraction of the present invention is taken turns classification and Detection method more and is had the following characteristics that
The present invention proposes the model of a kind of new feature extraction and the method for classification, the nodule detection sensitivity of whole system, the model of new feature extraction and the method for classification be can better improve and the classification of the Digital Image Data such as medical image, remote sensing images, network image, retrieval etc. can be applicable to.
Although above the illustrative detailed description of the invention of the present invention being described; so that those skilled in the art understand the present invention; it is to be understood that; the invention is not restricted to the scope of detailed description of the invention; to those skilled in the art; as long as various changes limit and in the spirit and scope of the present invention determined, these changes are apparent from, and all utilize the innovation and creation of present inventive concept all at the row of protection in appended claim.
Claims (2)
1. a kind of Lung neoplasm based on multiple dimensioned three-dimensional bits feature extraction of the present invention takes turns classification and Detection method more, mainly include herein below: in the tuberosity mask of feature extraction, introduce the concept of multiple dimensioned mask, tuberosity is carried out feature extraction by the block utilizing different scale, and the feature of extraction is put in first grader, then put in second grader by what classify again for false-positive tuberosity again, finally merge the subset of true positives tuberosity in two graders.Technical scheme is as follows:
Step one: the acquisition of picture, it is intended to utilize public data storehouse LIDC, and the pretreatment carrying out being correlated with directly can be processed by matlab;
Step 2: original image is carried out primary segmentation first by threshold value, secondly use the element of disc structure that the pulmonary of coarse segmentation is carried out Two-dimensional morphology opening operation, then three-dimensional hexa-atomic element is used to connect the part extracting maximum volume, this part is exactly the pulmonary of segmentation, it is then used by Hole filling algorithms to repair hole, finally employs three dimensional morphology closed operation and come the edge of refine pulmonary;
Step 3: first by a series of threshold values, original image is carried out Threshold segmentation, then the result of segmentation and pulmonary's mask are carried out logical AND operation, then carry out Two-dimensional morphology opening operation with the element of a series of disc structures tuberosity to having split again, finally merge all of middle nodule mask and form last nodule mask;
Step 4: feature extraction, it is intended to utilize more traditional features, and extract based on the block that different scale is three-dimensional.Feature mainly has three major types: geometry, brightness, Gradient Features, and all these feature contains two dimension and three-dimensional.Last again by the feature that SFS is next preferably a series of;
Step 5: the total two-wheeled classification of grader classification one, the first round carries out two classification with SVM, second takes turns and again classifies to being categorized as false-positive tuberosity in the first round with FLD grader, finally the true positives tuberosity in result is merged as last result.
2. a kind of Lung neoplasm based on multiple dimensioned three-dimensional bits feature extraction according to claim 1 takes turns classification and Detection method more, it is characterized in that introducing multiscale analysis theory in tuberosity feature extraction, the Lung neoplasm split is utilized to build the block of different scale, the extraction of feature is carried out afterwards on different blocks, then feature is put in first grader, then putting into be categorized into intubercular feature in second grader, the true positives tuberosity finally merged in double classification is end product.
Characteristic main in the present invention is the introducing multiscale analysis theory when feature extraction, and introduces two graders at tuberosity sorting phase.Specifically include that the feature extraction mask that (1) utilizes the nodular feature split multiple dimensioned.(2) first with svm grader, institute's nodosity being classified, then with FLD grader, are categorized as intubercular feature the first round and again classify, finally merging two times result is end product.
First to the nodular feature bounding box split, then carry out expanding 3 pixels on the basis of bounding box, enable the tuberosity major part of not precisely segmentation to include.Owing to the precisely segmentation of tuberosity has a definite limitation, so not carrying out feature extraction on the tuberosity mask split, and carrying out feature extraction in the three-dimensional bits expanded, enabling tuberosity major part to be included.
First the feature that above-mentioned link is extracted is put in SVM classifier and is gone all of tuberosity is carried out two classification, then put in FLD grader be categorized as intubercular characteristic set to improve sensitivity again again this part tuberosity is carried out two classification, finally constitute last result double classification result being divided into the subset of tuberosity merge.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610154156.4A CN105740875A (en) | 2016-03-17 | 2016-03-17 | Pulmonary nodule multi-round classification method based on multi-scale three-dimensional block feature extraction |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610154156.4A CN105740875A (en) | 2016-03-17 | 2016-03-17 | Pulmonary nodule multi-round classification method based on multi-scale three-dimensional block feature extraction |
Publications (1)
Publication Number | Publication Date |
---|---|
CN105740875A true CN105740875A (en) | 2016-07-06 |
Family
ID=56251767
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610154156.4A Pending CN105740875A (en) | 2016-03-17 | 2016-03-17 | Pulmonary nodule multi-round classification method based on multi-scale three-dimensional block feature extraction |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105740875A (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106250701A (en) * | 2016-08-10 | 2016-12-21 | 华东理工大学 | Computer software realizes the system and method for Lung neoplasm degree of danger classification |
CN107274402A (en) * | 2017-06-27 | 2017-10-20 | 北京深睿博联科技有限责任公司 | A kind of Lung neoplasm automatic testing method and system based on chest CT image |
CN108171694A (en) * | 2017-12-28 | 2018-06-15 | 深圳开立生物医疗科技股份有限公司 | Nodule detection methods, system and equipment based on convolutional neural networks |
CN108230323A (en) * | 2018-01-30 | 2018-06-29 | 浙江大学 | A kind of Lung neoplasm false positive screening technique based on convolutional neural networks |
CN108573485A (en) * | 2017-03-08 | 2018-09-25 | 卡西欧计算机株式会社 | Identification device, recognition methods and program recorded medium |
CN109636817A (en) * | 2018-11-30 | 2019-04-16 | 华中科技大学 | A kind of Lung neoplasm dividing method based on two-dimensional convolution neural network |
CN111199548A (en) * | 2018-11-19 | 2020-05-26 | 零氪医疗智能科技(广州)有限公司 | System and method for segmenting nodules in lung CT image |
CN111369530A (en) * | 2020-03-04 | 2020-07-03 | 浙江明峰智能医疗科技有限公司 | CT image pulmonary nodule rapid screening method based on deep learning |
CN116452898A (en) * | 2023-06-16 | 2023-07-18 | 中国人民大学 | Lung adenocarcinoma subtype identification method and device based on image histology and deep learning |
-
2016
- 2016-03-17 CN CN201610154156.4A patent/CN105740875A/en active Pending
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106250701B (en) * | 2016-08-10 | 2019-02-01 | 华东理工大学 | The system and method for Lung neoplasm degree of danger classification are realized in computer software |
CN106250701A (en) * | 2016-08-10 | 2016-12-21 | 华东理工大学 | Computer software realizes the system and method for Lung neoplasm degree of danger classification |
CN108573485A (en) * | 2017-03-08 | 2018-09-25 | 卡西欧计算机株式会社 | Identification device, recognition methods and program recorded medium |
CN108573485B (en) * | 2017-03-08 | 2022-10-14 | 卡西欧计算机株式会社 | Identification device, identification method, and program storage medium |
CN107274402A (en) * | 2017-06-27 | 2017-10-20 | 北京深睿博联科技有限责任公司 | A kind of Lung neoplasm automatic testing method and system based on chest CT image |
CN108171694A (en) * | 2017-12-28 | 2018-06-15 | 深圳开立生物医疗科技股份有限公司 | Nodule detection methods, system and equipment based on convolutional neural networks |
CN108230323B (en) * | 2018-01-30 | 2021-03-23 | 浙江大学 | Pulmonary nodule false positive screening method based on convolutional neural network |
CN108230323A (en) * | 2018-01-30 | 2018-06-29 | 浙江大学 | A kind of Lung neoplasm false positive screening technique based on convolutional neural networks |
CN111199548A (en) * | 2018-11-19 | 2020-05-26 | 零氪医疗智能科技(广州)有限公司 | System and method for segmenting nodules in lung CT image |
CN109636817A (en) * | 2018-11-30 | 2019-04-16 | 华中科技大学 | A kind of Lung neoplasm dividing method based on two-dimensional convolution neural network |
CN109636817B (en) * | 2018-11-30 | 2020-10-30 | 华中科技大学 | Lung nodule segmentation method based on two-dimensional convolutional neural network |
CN111369530B (en) * | 2020-03-04 | 2021-04-02 | 浙江明峰智能医疗科技有限公司 | CT image pulmonary nodule rapid screening method based on deep learning |
CN111369530A (en) * | 2020-03-04 | 2020-07-03 | 浙江明峰智能医疗科技有限公司 | CT image pulmonary nodule rapid screening method based on deep learning |
CN116452898A (en) * | 2023-06-16 | 2023-07-18 | 中国人民大学 | Lung adenocarcinoma subtype identification method and device based on image histology and deep learning |
CN116452898B (en) * | 2023-06-16 | 2023-10-17 | 中国人民大学 | Lung adenocarcinoma subtype identification method and device based on image histology and deep learning |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105740875A (en) | Pulmonary nodule multi-round classification method based on multi-scale three-dimensional block feature extraction | |
Xie et al. | Automated pulmonary nodule detection in CT images using deep convolutional neural networks | |
CN102663426B (en) | Face identification method based on wavelet multi-scale analysis and local binary pattern | |
Zhang et al. | A dense u-net with cross-layer intersection for detection and localization of image forgery | |
Vaishnavi et al. | Application of local invariant symmetry features to detect and localize image copy move forgeries | |
Elaskily et al. | Deep learning based algorithm (ConvLSTM) for copy move forgery detection | |
Wang et al. | A new keypoint-based copy-move forgery detection for color image | |
Guedri et al. | Indexing and images retrieval by content | |
CN106600584A (en) | Tsallis entropy selection-based suspected pulmonary nodule detection method | |
Nguyen et al. | Completed local structure patterns on three orthogonal planes for dynamic texture recognition | |
Mehrjardi et al. | A survey on deep learning-based image forgery detection | |
Zou et al. | Distinguishable zero-watermarking scheme with similarity-based retrieval for digital rights Management of Fundus Image | |
Xu et al. | Mammographic mass segmentation using multichannel and multiscale fully convolutional networks | |
Velliangira et al. | A novel forgery detection in image frames of the videos using enhanced convolutional neural network in face images | |
CN109191436A (en) | The low-dose CT Lung neoplasm detection algorithm of view-based access control model conspicuousness spectrum residual error method | |
Stucker et al. | Supervised outlier detection in large-scale MVS point clouds for 3D city modeling applications | |
CN117789185A (en) | Automobile oil hole gesture recognition system and method based on deep learning | |
Al-Qershi et al. | Copy-move forgery detection using on locality sensitive hashing and k-means clustering | |
Shao et al. | Generative image inpainting with salient prior and relative total variation | |
CN104504715A (en) | Image segmentation method based on local quaternion-moment characteristic | |
Zhu et al. | Foreground object sensing for saliency detection | |
Ehret et al. | How to reduce anomaly detection in images to anomaly detection in noise | |
Zhang et al. | Multicontext 3D residual CNN for false positive reduction of pulmonary nodule detection | |
Dhar et al. | Interval type-2 fuzzy set and human vision based multi-scale geometric analysis for text-graphics segmentation | |
Resmi et al. | A novel segmentation based copy-move forgery detection in digital images |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20160706 |