CN108280826A - A kind of solid nodules feature extraction of CT lungs Asia and detection method - Google Patents
A kind of solid nodules feature extraction of CT lungs Asia and detection method Download PDFInfo
- Publication number
- CN108280826A CN108280826A CN201810063319.7A CN201810063319A CN108280826A CN 108280826 A CN108280826 A CN 108280826A CN 201810063319 A CN201810063319 A CN 201810063319A CN 108280826 A CN108280826 A CN 108280826A
- Authority
- CN
- China
- Prior art keywords
- lung
- candidate
- asia
- feature extraction
- feature
- 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
- G06T7/13—Edge detection
-
- 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
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 invention discloses a kind of solid nodules feature extraction of CT lungs Asia and detection methods, include the following steps:Obtain lung's CT images;Lung's CT images of acquisition are pre-processed;Candidate nodule is carried out to pretreated lung's CT images to divide to obtain Candidate Set;Feature extraction is carried out to Candidate Set and obtains feature set;Lesion classification is carried out according to feature set;Lung neoplasm testing result is obtained according to lesion classification results.In the present invention, feature extraction is carried out in several ways, convenient for more comprehensively carrying out category filter, and application environment feature extraction, doubtful Lung neoplasm is accounted for the relationships such as other positions and the position of other doubtful Lung neoplasms in the picture, the sensitivity for improving classification solves existing lung Asia solid nodules computer detection method there are sensitivity low, the problems such as false sun rate is high.
Description
Technical field
The present invention relates to technical field of medical image processing more particularly to a kind of solid nodules feature extractions of CT lungs Asia and inspection
Survey method.
Background technology
In the latest 20 years, lung cancer has become the lethal most cancer in the whole world, and the U.S. dies of the case of cancer within 2012
In only lung cancer just account for sum 28%, so for the early detection of lung cancer --- Lung neoplasm detection just it is most important.
Since two thousand seven, it is more and more suggested for the computer-aided detection system of lung solid nodules, and by
Gradual perfection.But for lung Asia solid nodules, be difficult to capture in CT, therefore certainly for the computer of lung Asia solid nodules
The research of dynamic detection method is also less, and that there are sensitivity is low for existing lung Asia solid nodules computer detection method, and false sun rate is high
The problems such as, prodigious difficulty is brought to the auxiliary detection of lung Asia solid nodules, there are larger harmfulness.
Invention content
It is an object of the invention to:A kind of solid nodules feature extraction of CT lungs Asia and detection method are provided, existing lung is solved
Sub- solid nodules feature extraction is not comprehensive, and then leads to that computer detection method sensitivity is low, false positive rate is high.
The technical solution adopted by the present invention is as follows:
A kind of CT lungs Asia solid nodules feature extracting method, includes the following steps:
S1:Obtain lung's CT images;
S2:Lung's CT images of acquisition are pre-processed;
S3:Candidate nodule is carried out to pretreated lung's CT images to divide to obtain Candidate Set;
S4:Feature extraction is carried out to Candidate Set and obtains feature set.
Further, the step S2 is as follows:
S201:Lung's CT images are split using bivalve value dividing method;
S202:Excess pixel, which is removed, by edge detection method obtains area-of-interest.
Further, the step S3 is as follows:
S301:Area-of-interest is split using morphological method;
S302:It rejects the vascular system contacted with area-of-interest and obtains Candidate Set.
Further, the feature extraction includes:Brightness extraction, texture feature extraction, Shape Feature Extraction, environment
Feature extraction.
Further, the environmental characteristic extraction includes 4 kinds of modes:
Mode one:Candidate Set all pixels are calculated to the distance on lung boundary and bronchial tree, and calculate average value, standard
Difference, minimum value and maximum value are used as environmental characteristic;
Mode two:Minimum rectangle frame is set around lung, and constructs three-dimensional x, y, z coordinate, to obtain candidate
Relative position, and calculate the distance of minimum rectangle lower left bezel corner and the distance to lung Asia solid nodules center;
Mode three:Customized bounding box around segmentation candidates region is divided, wherein bounding box is divided for tracheae
Or a part for blood vessel segmentation,
Absolute overlapping=N5
Wherein N4For the quantity of candidate regions pixel, N5It is the pixel quantity in bounding box;
It will absolutely be overlapped and be used as environmental characteristic with relative superposition rate;
Mode four:Calculate Candidate Set in candidate nodule quantity, and calculate candidate nodule around distance 30mm and 50mm with
The distance of the quantity of interior candidate nodule and nearest candidate nodule is used as environmental characteristic.
A kind of CT lungs Asia solid nodules detection method, includes the following steps:
S1:Lesion classification is carried out according to feature set;
S2:Lung neoplasm testing result is obtained according to lesion classification results.
Further, the step S1 is as follows:
S101:Linear discriminant ratio calculation is carried out to feature;
S102:Choose maximum 5 features of ratio;
S103:Classified using 5 features of linear classifier pair;
S104:Classified to residue character using 10 graders of GentleBoost.
In conclusion by adopting the above-described technical solution, the beneficial effects of the invention are as follows:
1, in the present invention, by being detected automatically to lung Asia solid nodules, existing lung Asia solid nodules computer is solved
The problems such as that there are sensitivity is low for detection method, and false sun rate is high, improves detection accuracy.
2, in the present invention, feature extraction is carried out in several ways, convenient for more comprehensively carrying out category filter, improves inspection
Look into sensitivity.
3, the present invention in, application environment feature extraction, to doubtful Lung neoplasm in the picture with other positions and it is other doubt
The relationships such as the position like Lung neoplasm account for, and improve the sensitivity of classification.
Description of the drawings
Fig. 1 is flow chart of steps of the present invention;
Fig. 2 is that feature of present invention extracts classification chart;
Fig. 3 is four kinds of pixel region dividing modes of the invention;
Fig. 4 is lesion classification process figure of the present invention.
Specific implementation mode
In order to make the purpose , technical scheme and advantage of the present invention be clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that described herein, specific examples are only used to explain the present invention, not
For limiting the present invention.
Embodiment 1
As shown in Figure 1, a kind of CT lungs Asia solid nodules feature extracting method, includes the following steps:
S1:Obtain lung's CT images;
S2:Lung's CT images of acquisition are pre-processed;
S3:Candidate nodule is carried out to pretreated lung's CT images to divide to obtain Candidate Set;
S4:Feature extraction is carried out to Candidate Set and obtains feature set.
The present invention carries out feature extraction for lung Asia solid nodules, and then is convenient for detection and judges, solves existing lung
The problems such as that there are sensitivity is low for sub- solid nodules computer detection method, and false sun rate is high, improves detection accuracy.
Embodiment 2
On the basis of embodiment 1, the step S2 is as follows:
S201:Lung's CT images are split using bivalve value dividing method;
S202:Excess pixel, which is removed, by edge detection method obtains area-of-interest.
Lung's CT images are split using the method for double threshold segmentations, and gas removing pipe, blood vessel are gone by edge detection
Equal excess pixels, tentatively obtain the coarse segmentation about doubtful nodule candidate.
Embodiment 3
On the basis of embodiment 1 or 2, the step S3 is as follows:
S301:Area-of-interest is split using morphological method;
S302:It rejects the vascular system contacted with area-of-interest and obtains Candidate Set.
Doubtful nodule candidate is divided, area-of-interest is further divided using more accurate morphological method,
The vascular system contacted is rejected simultaneously, to obtain final Candidate Set.
Embodiment 4
As shown in Fig. 2, on the basis of embodiment 1, the feature extraction includes:Brightness extraction, textural characteristics carry
It takes, the extraction of Shape Feature Extraction, environmental characteristic.
Further, the environmental characteristic extraction includes 4 kinds of modes:
Mode one:Candidate Set all pixels are calculated to the distance on lung boundary and bronchial tree, and calculate average value, standard
Difference, minimum value and maximum value are used as environmental characteristic;
Mode two:Minimum rectangle frame is set around lung, and constructs three-dimensional x, y, z coordinate, to obtain candidate
Relative position, and calculate the distance of minimum rectangle lower left bezel corner and the distance to lung Asia solid nodules center;
Mode three:Customized bounding box around segmentation candidates region is divided, wherein bounding box is divided for tracheae
An or part for blood vessel segmentation.
Absolute overlapping=N5
Wherein N4For the quantity of candidate regions pixel, N5It is the pixel quantity in bounding box;
It will absolutely be overlapped and be used as environmental characteristic with relative superposition rate;
Mode four:Calculate Candidate Set in candidate nodule quantity, and calculate candidate nodule around distance 30mm and 50mm with
The distance of the quantity of interior candidate nodule and nearest candidate nodule is used as environmental characteristic.
Further, the brightness extracting mode is as follows:
As shown in figure 3, four kinds of different modes, which are respectively adopted, carries out set of pixels division, mode A1:Using segmentation candidates region
Pixel quantity as set of pixels;Mode A2:Minimum rectangle frame is built around segmentation candidates region, set of pixels is minimum
Pixel in rectangular shaped rim;Mode A3:It is extended around on the basis of segmentation candidates region, extended mode is with 3x3x3
Rectangular pixels extend, and set of pixels is the pixel extended in rear region;Mode A4:Expand around on the basis of segmentation candidates region
Exhibition, extended mode are extended with the rectangular pixels of 5x5x5, and set of pixels is the pixel extended in rear region;
It uses size for the standard histogram of 50HU to divide region to pixel for statistical analysis, and calculates maximum value, most
Small value, average value, standard deviation, entropy are to feature extraction.
The texture feature extraction mode is as follows:
Texture feature extraction A2 in such a way that pixel in brightness extraction divides, lays equal stress on and is sampled as 16x16x16
With two area-of-interests of 32x32x32;
The two area-of-interest carry out offices of 16x16x16 and 32x32x32 of 2D local binaries method to resampling are used first
Portion's two-value method calculates, and carries out binaryzation to 8 fields around each pixel, uses size for the standard histogram of 1HU
It is for statistical analysis, and maximum value, minimum value, average value, standard deviation, entropy are calculated to feature extraction.
Then haar wavelet transform method is used to be decomposed into 4 frequency bands to the 32x32x32 area-of-interests transformation of resampling,
4 frequency bands constitute 4 voxel regions, are counted wherein carrying out standard histogram to 3 regions from high frequency band
Analysis, and maximum value, minimum value, average value, standard deviation, entropy are calculated to feature extraction.
The Shape Feature Extraction mode is as follows:
Four kinds of modes are respectively adopted and carry out shape feature calculating:
Mode B1:Volume spheric region S identical with candidate region is defined at candidate region center, globoid calculates
Formula:
Wherein V1It is the pixel volume that spheric region S corresponds to candidate region, VSIt is spheric region S pixel volumes.
Mode B2:Bounding box is arranged in A2 by the way of pixel divides in brightness extracts, and marker x, y, z is sat
Mark, calculation formula:
Wherein N1It is the quantity of candidate regions pixel, N2It is the quantity of pixel in bounding box.
Mode B3:Bounding box is arranged in A2 by the way of pixel divides in brightness extracts, and its x, y, z is marked to sit
Mark, calculation formula:
Wherein N1It is the quantity of candidate regions pixel, N3It is out to out bounding box (max (dimx, dimy, dimz)) interior pixel
Quantity.
Mode B4:Bounding box is arranged in A2 by the way of pixel divides in brightness extracts, and its x, y, z is marked to sit
Mark, calculation formula:
Wherein V3 is the volume of bounding box.
Embodiment 5
As shown in figs. 1 and 4, a kind of CT lungs Asia solid nodules detection method, includes the following steps:
S1:Lesion classification is carried out according to feature set;
S2:Lung neoplasm testing result is obtained according to lesion classification results.
Further, the step S1 is as follows:
S101:Linear discriminant ratio calculation is carried out to feature;
S102:Choose maximum 5 features of ratio;
S103:Classified using 5 features of linear classifier pair;
S104:Classified to residue character using 10 graders of GentleBoost.
Lesion classification is carried out on the basis of feature extraction, classification schemes are classified using dual-stage, classified in the first stage
When only utilize 5 features, the acquisition process of this 5 features is:To all feature calculation Fisher linear discriminants ratios got,
5 features of maximum 5 features of ratio as the first stage are taken, linear classification is then used on the basis of this 5 features
Device classifies to candidate.First stage is substantially carried out the rejecting of preliminary classification and false positive candidate, then by remaining time
Choosing carries out second stage classification.Second stage classification uses 10 graders of GentleBoost, all using what is got before
Feature classifies to first stage remaining candidate, and obtains final lesion classification results.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all essences in the present invention
All any modification, equivalent and improvement etc., should all be included in the protection scope of the present invention made by within refreshing and principle.
Claims (7)
1. a kind of CT lungs Asia solid nodules feature extracting method, which is characterized in that include the following steps:
S1:Obtain lung's CT images;
S2:Lung's CT images of acquisition are pre-processed;
S3:Candidate nodule is carried out to pretreated lung's CT images to divide to obtain Candidate Set;
S4:Feature extraction is carried out to Candidate Set and obtains feature set.
2. a kind of CT lungs Asia solid nodules feature extracting method according to claim 1, which is characterized in that the step S2 tools
Steps are as follows for body:
S201:Lung's CT images are split using bivalve value dividing method;
S202:Excess pixel, which is removed, by edge detection method obtains area-of-interest.
3. a kind of CT lungs Asia according to claim 1 or claim 2 solid nodules feature extracting method, which is characterized in that the step
S3 is as follows:
S301:Area-of-interest is split using morphological method;
S302:It rejects the vascular system contacted with area-of-interest and obtains Candidate Set.
4. a kind of CT lungs Asia solid nodules feature extracting method according to claim 1, which is characterized in that the feature extraction
Including:Brightness extraction, texture feature extraction, Shape Feature Extraction, environmental characteristic extraction.
5. according to a kind of CT lungs Asia solid nodules feature extracting method of claim 1 or 4, which is characterized in that the environment
Feature extraction includes 4 kinds of modes:
Mode one:Candidate Set all pixels are calculated to the distance on lung boundary and bronchial tree, and calculate average value, standard deviation, most
Small value and maximum value are used as environmental characteristic;
Mode two:Minimum rectangle frame is set around lung, and constructs three-dimensional x, y, z coordinate, to obtain candidate phase
To position, and calculate the distance of minimum rectangle lower left bezel corner and the distance to lung Asia solid nodules center;
Mode three:Customized bounding box around segmentation candidates region is divided, wherein bounding box be tracheae divide or
A part for blood vessel segmentation,
Absolute overlapping=N5
Wherein N4For the quantity of candidate regions pixel, N5It is the pixel quantity in bounding box;
It will absolutely be overlapped and be used as environmental characteristic with relative superposition rate;
Mode four:The quantity of candidate nodule in Candidate Set is calculated, and is calculated around candidate nodule within distance 30mm and 50mm
The distance of the quantity of candidate nodule and nearest candidate nodule is used as environmental characteristic.
6. a kind of CT lungs Asia solid nodules detection method, which is characterized in that include the following steps:
S1:Lesion classification is carried out according to feature set;
S2:Lung neoplasm testing result is obtained according to lesion classification results.
7. a kind of CT lungs Asia solid nodules detection method according to claim 6, which is characterized in that the step S1 is specifically walked
It is rapid as follows:
S101:Linear discriminant ratio calculation is carried out to feature;
S102:Choose maximum 5 features of ratio;
S103:Classified using 5 features of linear classifier pair;
S104:Classified to residue character using 10 graders of GentleBoost.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810063319.7A CN108280826A (en) | 2018-01-23 | 2018-01-23 | A kind of solid nodules feature extraction of CT lungs Asia and detection method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810063319.7A CN108280826A (en) | 2018-01-23 | 2018-01-23 | A kind of solid nodules feature extraction of CT lungs Asia and detection method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108280826A true CN108280826A (en) | 2018-07-13 |
Family
ID=62804597
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810063319.7A Pending CN108280826A (en) | 2018-01-23 | 2018-01-23 | A kind of solid nodules feature extraction of CT lungs Asia and detection method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108280826A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110415206A (en) * | 2019-04-08 | 2019-11-05 | 上海墩庐生物医学科技有限公司 | A method of identification adenocarcinoma of lung infiltrates parting |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102842132A (en) * | 2012-07-12 | 2012-12-26 | 上海联影医疗科技有限公司 | CT pulmonary nodule detection method |
EP3095376A1 (en) * | 2014-02-07 | 2016-11-23 | Hiroshima University | Endoscopic image diagnosis support system |
CN106780460A (en) * | 2016-12-13 | 2017-05-31 | 杭州健培科技有限公司 | A kind of Lung neoplasm automatic checkout system for chest CT image |
-
2018
- 2018-01-23 CN CN201810063319.7A patent/CN108280826A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102842132A (en) * | 2012-07-12 | 2012-12-26 | 上海联影医疗科技有限公司 | CT pulmonary nodule detection method |
EP3095376A1 (en) * | 2014-02-07 | 2016-11-23 | Hiroshima University | Endoscopic image diagnosis support system |
CN106780460A (en) * | 2016-12-13 | 2017-05-31 | 杭州健培科技有限公司 | A kind of Lung neoplasm automatic checkout system for chest CT image |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110415206A (en) * | 2019-04-08 | 2019-11-05 | 上海墩庐生物医学科技有限公司 | A method of identification adenocarcinoma of lung infiltrates parting |
CN110415206B (en) * | 2019-04-08 | 2023-12-15 | 上海墩庐生物医学科技有限公司 | Method for identifying lung adenocarcinoma infiltration typing |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107358258B (en) | SAR image target classification based on NSCT double CNN channels and selective attention mechanism | |
Gllavata et al. | Text detection in images based on unsupervised classification of high-frequency wavelet coefficients | |
CN107944353B (en) | SAR image change detection method based on contour wave BSPP network | |
WO2022141145A1 (en) | Object-oriented high-resolution remote sensing image multi-scale segmentation method and system | |
Prajapati et al. | Brain tumor detection by various image segmentation techniques with introduction to non negative matrix factorization | |
CN102360503B (en) | SAR (Specific Absorption Rate) image change detection method based on space approach degree and pixel similarity | |
Wan et al. | Wavelet-based statistical features for distinguishing mitotic and non-mitotic cells in breast cancer histopathology | |
Jin et al. | Improved direction estimation for Di Zenzo's multichannel image gradient operator | |
CN104933719B (en) | One kind integration segment spacing is from detection image edge method | |
CN106447686A (en) | Method for detecting image edges based on fast finite shearlet transformation | |
US9715724B2 (en) | Registration of CAD data with SEM images | |
CN108985357A (en) | The hyperspectral image classification method of set empirical mode decomposition based on characteristics of image | |
Subramanyam et al. | Different image segmentation techniques for dental image extraction | |
Han et al. | Segmenting images with complex textures by using hybrid algorithm | |
Visalaxi et al. | Lesion extraction of endometriotic images using open computer vision | |
CN108280826A (en) | A kind of solid nodules feature extraction of CT lungs Asia and detection method | |
CN106023166B (en) | The detection method and device of dangerous object hidden by human body in microwave image | |
Ngau et al. | Bottom-up visual saliency map using wavelet transform domain | |
Sengar et al. | Analysis of 2D-gel images for detection of protein spots using a novel non-separable wavelet based method | |
CN108460383B (en) | Image significance refinement method based on neural network and image segmentation | |
Xu et al. | Hierarchical matching for automatic detection of masses in mammograms | |
Shahin et al. | Breast cancer detection based on dynamic template matching | |
Jia-Cun et al. | Object-oriented method of land cover change detection approach using high spatial resolution remote sensing data | |
Berbar et al. | Masses classification using discrete cosine transform and wavelet-based directional filter bank for breast cancer diagnosis | |
Wei et al. | An algorithm for segmentation of lung ROI by mean-shift clustering combined with multi-scale HESSIAN matrix dot filtering |
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: 20180713 |