CN109191436A - The low-dose CT Lung neoplasm detection algorithm of view-based access control model conspicuousness spectrum residual error method - Google Patents

The low-dose CT Lung neoplasm detection algorithm of view-based access control model conspicuousness spectrum residual error method Download PDF

Info

Publication number
CN109191436A
CN109191436A CN201810926369.3A CN201810926369A CN109191436A CN 109191436 A CN109191436 A CN 109191436A CN 201810926369 A CN201810926369 A CN 201810926369A CN 109191436 A CN109191436 A CN 109191436A
Authority
CN
China
Prior art keywords
candidate regions
residual error
lung neoplasm
lung
pixel
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
Application number
CN201810926369.3A
Other languages
Chinese (zh)
Inventor
张文强
盛嘉成
丁志鹏
管晖
管一晖
谭海波
李艺鸣
王华滔
徐建荣
路红
张睿
周峰
路建乡
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Fudan University
Original Assignee
Fudan University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Fudan University filed Critical Fudan University
Priority to CN201810926369.3A priority Critical patent/CN109191436A/en
Publication of CN109191436A publication Critical patent/CN109191436A/en
Pending legal-status Critical Current

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
    • 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
    • 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
    • G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
    • 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
    • G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
    • G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
    • 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
    • G06T2207/20—Special algorithmic details
    • G06T2207/20024—Filtering details
    • 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
    • G06T2207/20—Special algorithmic details
    • G06T2207/20036—Morphological image processing
    • 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
    • G06T2207/20—Special algorithmic details
    • G06T2207/20048—Transform domain processing
    • G06T2207/20056—Discrete and fast Fourier transform, [DFT, FFT]
    • 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)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Biology (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 belongs to medical image process field, specially a kind of low-dose CT Lung neoplasm detection algorithm of view-based access control model conspicuousness spectrum residual error method.The process of the algorithm is as follows: (1) pulmonary parenchyma is divided；(2) candidate regions extract: carrying out conspicuousness detection to pulmonary parenchyma part by spectrum residual error method, extract the Lung neoplasm candidate regions with vision significance；(3) feature calculation；(4) candidate regions are classified: being classified using Lung neoplasm candidate regions of the C-SVM classifier to extraction, removed false positive sample.The present invention is extracted candidate nodule and can be efficiently extracted using spectrum residual error method has cavity, and the knuckle areas of the labyrinths such as burr has high detection sensibility and low false positive, and algorithm flow has high interpretation, to meet the needs of actual medical auxiliary system.

Description

The low-dose CT Lung neoplasm detection algorithm of view-based access control model conspicuousness spectrum residual error method
Technical field
The invention belongs to computer medical image processing technology fields, and in particular to a kind of view-based access control model conspicuousness spectrum residual error The low-dose CT Lung neoplasm detection algorithm of method.
Background technique
Lung neoplasm is the important avatar of lung cancer early stage, as lung cancer in China illness rate and lethality rise, based on low The screening lung cancer of dosage CT is particularly important.But since chest CT tomography is up to 200 layers or more, pass through mankind doctor completely Check that every CT image of every generaI investigation audient is difficult to realize.So a kind of algorithm of computer aided lung nodule detection is non- Chang Guanjian.
Candidate regions extraction algorithm in traditional Lung neoplasm detection algorithm has two classes, and one kind is simply to find image based on threshold value In all high CT value regions, but since lung includes that its hetero-organization, this way such as a large amount of blood vessels, tracheae can greatly improve time Constituency quantity reduce algorithm entirety efficiency and and accuracy rate, morphological feature of the another kind of method based on tubercle similar round, but Be such methods often missing inspection have cavity, burr, ground glass structure Lung neoplasm, and this kind of tubercle is exactly most pernicious Risk, so such missing inspection must avoid.
Summary of the invention
For overcome the deficiencies in the prior art, the purpose of the present invention is to provide a kind of view-based access control model conspicuousnesses to compose residual error side The low-dose CT Lung neoplasm detection algorithm of method.The present invention can not to tubercle CT range and form make a priori assumption the case where Under, reach high sensitive and low false positive.Vision significance spectrum residual error method is applied to Lung neoplasm detection algorithm by the present invention, is made While obtaining entire algorithm has lower false positive, improve to the susceptibility for having cavity, the isostructural tubercle of burr.
The low-dose CT Lung neoplasm detection algorithm of vision significance spectrum residual error method proposed by the present invention, is calculated using spectrum residual error Method converts lung images to domain space after lung segmentation, then significant by smoothly obtaining lung images no visual The background redundancy section of property, then the two is subtracted each other into transformation back to spatial domain, Saliency maps are obtained, corresponding lung knot is obtained after binaryzation Candidate regions are saved, then are classified by feature extraction and C-SVM (support vector machines) classifier, false positive sample is removed.The present invention Technical solution be specifically described as follows.
A kind of low-dose CT Lung neoplasm detection algorithm of view-based access control model conspicuousness spectrum residual error method, the specific steps are as follows:
(1) pulmonary parenchyma is divided: pulmonary parenchyma region is partitioned on low-dose CT image based on threshold method and morphological operation；
(2) candidate regions extract: carrying out conspicuousness detection to pulmonary parenchyma region by spectrum residual error method, extract aobvious with vision The Lung neoplasm candidate regions of work property；
(3) gray scale and morphological feature feature calculation: are extracted to the Lung neoplasm candidate regions of extraction；
(4) candidate regions are classified: classified using Lung neoplasm candidate regions of the support vector machines C-SVM classifier to extraction, removal False positive sample leaves true Lung neoplasm.
In the present invention, in step (1), optimal threshold is found using Da-Jin algorithm, after binary image, searching represents pulmo Largest connected region, then modified using waveforms method, be partitioned into pulmonary parenchyma cut zone；Specific step is as follows:
1. converting 8 256 grades of grayscale images, the grey level histogram of statistical picture for CT image；
2. each section pixel number is converted to gray probability divided by total pixel number；
3. enumerating threshold value i in [0,255]；
4. calculating the total accounting w0 of foreground pixel, foreground pixel average gray u0, gray scale interval [i+ in gray scale interval [0, i] 1,255] the total accounting w1 of background pixel in, foreground pixel average gray u1；
5. calculating inter-class variance gi=w0*w1* (u0-u1) 2；
6. finding makes giMaximum i is as optimal threshold t；
7. binary image；
8. finding in bianry image maximum two connected domains as preliminary lung by contour detecting Suzuki algorithm Mask；
9. being corroded to lung's mask using the circular filter that a radius is more than or equal to 2 pixels, i.e., by 0 pixel Surrounding pixel is set as 0；
10. carrying out closed operation, i.e., first expanded using the circular filter that a radius is 5 to 10 pixels, by 1 picture Plain surrounding pixel is set as 1, is then corroded again with the filter of same size；
Fill remaining perforated in lung's mask.
In the present invention, in step (2), the detailed process for composing residual error method is as follows:
1. by compression of images to 64*64；
2. original image f (x, y) is done discrete Fourier transform to domain space, R, I indicate the real and imaginary parts of Fourier spectrum, Angle phi and power spectrum are taken out respectively | F (u, v) |:
| F (u, v) |=(R (u, v)2+ I (u, v)2)1/2
3. taking logarithm to obtain logarithmic spectrum L (f) amplitude；
4. carrying out smoothly, obtaining A (f) to L (f) using a mean filter h (f)；
5. calculating:
R (f)=L (f)-A (f)
6. residual error is composed Inverse Discrete Fourier Transform back to spatial domain G using the phase angle saved:
G (u, v)=exp (R (f (u, v)+i* φ (u, v))2
7. filtering to the result of step 6. using Gaussian filter, Saliency maps are obtained；
8. obtaining final target mask to Saliency maps binaryzation.
In the present invention, in step (3), gray scale and morphological feature include:
1. minimum CT value；
2. highest CT value；
3. CT mean value；
4. CT variance；
⑤Wherein S is area, and p is perimeter；
⑥Wherein H is region convex closure area；
7. flexibility: the ratio between minimum external elliptical long axis max_dim and short axle min_dim of candidate regions；
⑧
⑨
⑩Boundingbox_area indicates bounding box area；
The not bending moment of image.
In the present invention, in step (4), the process of the candidate regions classification is as follows:
1. classifier is trained, to categorised decision face y (x)=ωTφ (x)+b, x are sample characteristics, and ω, b are decision surface ginseng Number, enable sample point to decision surface distance be 1；
2. minimizing:
Wherein to correct classification and interior sample in the edge of decision surface, ξ is definedn=0, and other points ξn=| tn-y (xn) |, t indicates the true classification of sample, i.e. tubercle, then ξn>'s 1 is exactly by the point of misclassification；The limitation item in Optimal Decision-making face Part is tny(xn)≥1-ξn, C is punishment parameter, default setting 1.
Compared to the prior art, the beneficial effects of the present invention are the comprehensive a variety of computer vision algorithms makes of: the present invention to be formed The Lung neoplasm detection algorithm that can be applied on low-dose CT image of complete set, spectrum residual error method extracts candidate nodule can be with Efficiently extracting has cavity, and the knuckle areas of the labyrinths such as burr has high detection sensibility and low false positive, algorithm stream Journey has high interpretation, and then meets the needs of actual medical auxiliary system.
Detailed description of the invention
Fig. 1 is the low-dose CT Lung neoplasm detection algorithm flow chart that view-based access control model conspicuousness of the invention composes residual error method.
Fig. 2 is that pulmonary parenchyma cut zone flow diagram described in step 1 (b) is in embodiment 1 wherein (a) is original image Image after binaryzation is (d) image after corrosion, is (e) figure after closed operation (c) to retain the image behind largest connected domain Picture is (f) the final pulmonary parenchyma segmentation result after holes filling.
Fig. 3 is that spectrum residual error described in step 2 extracts candidate regions schematic diagram in embodiment 1, and wherein a centre circle is algorithm most final inspection The Lung neoplasm of survey, b are Saliency maps, and c is candidate regions mask.
Specific embodiment
The present invention is described in more detail with an example with reference to the accompanying drawing.
Embodiment 1
Referring to Fig.1, the present invention provides a kind of low-dose spiral CT Lung neoplasm inspection of view-based access control model conspicuousness spectrum residual error method Method of determining and calculating, the specific steps of which are as follows:
1. pulmonary parenchyma is divided:
(1) 8 256 grades of grayscale images, the ash of statistical picture are converted by Fig. 2 (a) (its data comes from LUNA16 data set) Spend histogram；
(2) each gray scale interval pixel number is converted to gray probability divided by total pixel number；
(3) threshold value i is enumerated in [0,255]；
(4) the total accounting w0 of foreground pixel, foreground pixel average gray u0, gray scale interval [i+ in gray scale interval [0, i] are calculated 1,255] the total accounting w1 of background pixel in, foreground pixel average gray u1；
(5) inter-class variance gi=w0*w1* (u0-u1) 2 is calculated；
(6) finding makes the maximum i of gi, i.e., 80 are used as optimal threshold t；
(7) binary image obtains Fig. 2 (b)；
(8) maximum two connected domains are found in bianry image as preliminary lung by Suzuki (contour detecting) algorithm Portion's mask, such as Fig. 2 (c)；
(9) lung's mask is corroded using the circular filter that a radius is 2, i.e., set 0 pixel surrounding pixel It is set to 0, such as Fig. 2 (d)；
(10) closed operation is carried out, i.e., is first expanded using the circular filter that a radius is 10, by picture around 1 pixel Element is set as 1, is then corroded again with the filter of same size, such as Fig. 2 (e)；
(11) remaining perforated in mask is filled, such as Fig. 2 (f).
2. candidate regions extract:
The Saliency maps that pulmonary parenchyma region is calculated using spectrum residual error method are waited by finding conspicuousness Objective extraction Lung neoplasm Constituency, detailed process is as follows:
(1) by compression of images to 64*64；
(2) original image f (x, y) is done into discrete Fourier transform, arrives domain space, R, I indicate the real part and void of Fourier spectrum Angle phi and power spectrum are taken out respectively in portion | F (u, v) |:
| F (u, v) |=(R (u, v)2+ I (u, v)2)1/2
(3) logarithm is taken to obtain logarithmic spectrum L (f) amplitude；
(4) L (f) is carried out using mean filter h (f) (take 3*3's rectangular) smooth；
(5) it calculates:
R (f)=L (f)-A (f)
(6) residual error is composed into Inverse Discrete Fourier Transform back to spatial domain G using the phase angle saved:
G (u, v)=exp (R (f (u, v)+i* φ (u, v))2
(7) result of (6) is filtered using Gaussian filter, obtains final Saliency maps (Fig. 3 b)；
(8) use 30 times of mean values as can be obtained by final target mask (Fig. 3 c) notable figure.
3. feature calculation
To each candidate regions of proposition, gray scale and morphological feature are calculated:
A) minimum CT value；
B) highest CT value；
C) CT mean value；
D) CT variance；
e)Wherein S is area, and p is perimeter；
f)Wherein H is region convex closure area；
G) flexibility: the ratio between minimum external elliptical long axis max_dim and short axle min_dim of candidate regions；
h)
i)
j)Boundingbox_area indicates bounding box area.
K) the not bending moment of image；
4. candidate regions are classified
Using C-SVM (support vector machines), algorithmic procedure is as follows:
(1) training classifier, to categorised decision face y (x)=ωTφ (x)+b, x are sample characteristics, and ω, b are decision surface ginseng Number, enable sample point to decision surface distance be 1；
(2) it minimizes:
Wherein to correct classification and interior sample in the edge of decision surface, ξ is definedn=0, and other points ξn=| tn-y (xn) |, t indicates the true classification of sample, i.e. tubercle, then ξn>'s 1 is exactly by the point of misclassification.The limitation item in Optimal Decision-making face Part is tny(xn)≥1-ξn, C is punishment parameter.
(3) classified using trained classifier, remove false positive sample, leave true Lung neoplasm, as opened up in Fig. 3 a circle What is shown is the sample for being determined as true tubercle by classifier.The sensibility of the method for the present invention reaches 89.68%, and false positive is 4.39/.

Claims (5)

1. a kind of low-dose CT Lung neoplasm detection algorithm of view-based access control model conspicuousness spectrum residual error method, which is characterized in that specific step It is rapid as follows:
(1) pulmonary parenchyma is divided: pulmonary parenchyma region is partitioned on low-dose CT image based on threshold method and morphological operation；
(2) candidate regions extract: carrying out conspicuousness detection to pulmonary parenchyma region by spectrum residual error method, extracting has vision significance Lung neoplasm candidate regions；
(3) gray scale and morphological feature feature calculation: are extracted to the Lung neoplasm candidate regions of extraction；
(4) candidate regions are classified: being classified using Lung neoplasm candidate regions of the support vector machines C-SVM classifier to extraction, removed false sun Property sample, leaves true Lung neoplasm.
2. detection algorithm according to claim 1, it is characterised in that: in step (1), find best threshold using Da-Jin algorithm It is worth, after binary image, finds the largest connected region for representing pulmo, then modified using waveforms method, be partitioned into pulmonary parenchyma Cut zone；Specific step is as follows:
1. converting 8 256 grades of grayscale images, the grey level histogram of statistical picture for CT image；
2. each section pixel number is converted to gray probability divided by total pixel number；
3. enumerating threshold value i in [0,255]；
4. the total accounting w0 of foreground pixel in gray scale interval [0, i] is calculated, foreground pixel average gray u0, gray scale interval [i+1, 255] the total accounting w1 of background pixel in, foreground pixel average gray u1；
5. calculating inter-class variance gi=w0*w1* (u0-u1) 2；
6. finding makes giMaximum i is as optimal threshold t；
7. binary image；
8. finding in bianry image maximum two connected domains as preliminary lung's mask by contour detecting Suzuki algorithm；
9. lung's mask is corroded using the circular filter that a radius is more than or equal to 2 pixels, i.e., it will be around 0 pixel Pixel is set as 0；
10. carrying out closed operation, i.e., first expanded using the circular filter that a radius is 5 to 10 pixels, by 1 pixel week It encloses pixel and is set as 1, then corroded again with the filter of same size；
Fill remaining perforated in lung's mask.
3. detection algorithm according to claim 1, it is characterised in that: in step (2), compose the detailed process of residual error method such as Under:
1. by compression of images to 64*64；
2. original image f (x, y) is done discrete Fourier transform to domain space, R, I indicate the real and imaginary parts of Fourier spectrum, respectively Take out angle phi and power spectrum | F (u, v) |:
| F (u, v) |=(R (u, v)2+ I (u, v)2)1/2
3. taking logarithm to obtain logarithmic spectrum L (f) amplitude；
4. carrying out smoothly, obtaining A (f) to L (f) using a mean filter h (f)；
5. calculating:
R (f)=L (f)-A (f)
6. residual error is composed Inverse Discrete Fourier Transform back to spatial domain G using the phase angle saved:
G (u, v)=exp (R (f (u, v)+i* φ (u, v))2
7. filtering to the result of step 6. using Gaussian filter, Saliency maps are obtained；
8. obtaining final target mask to Saliency maps binaryzation.
4. detection algorithm according to claim 1, it is characterised in that: in step (3), gray scale and morphological feature include:
1. minimum CT value；
2. highest CT value；
3. CT mean value；
4. CT variance；
⑤Wherein S is area, and p is perimeter；
⑥Wherein H is region convex closure area；
7. flexibility: the ratio between minimum external elliptical long axis max_dim and short axle min_dim of candidate regions；
⑧
⑨
⑩Boundingbox_area indicates bounding box area；
The not bending moment of image.
5. detection algorithm according to claim 1, it is characterised in that: in step (4), the process of the candidate regions classification is such as Under:
1. classifier is trained, to categorised decision face y (x)=ωTφ (x)+b, x are sample characteristics, and ω, b are decision surface parameter, are enabled Sample point to decision surface distance be 1；
2. minimizing:
Wherein to correct classification and interior sample in the edge of decision surface, ξ is definedn=0, and other points ξn=| tn-y(xn) |, t Indicate the true classification of sample, i.e. tubercle, then ξn>'s 1 is exactly by the point of misclassification；The restrictive condition in Optimal Decision-making face is tny (xn)≥1-ξn, C is punishment parameter, default setting 1.
CN201810926369.3A 2018-08-15 2018-08-15 The low-dose CT Lung neoplasm detection algorithm of view-based access control model conspicuousness spectrum residual error method Pending CN109191436A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810926369.3A CN109191436A (en) 2018-08-15 2018-08-15 The low-dose CT Lung neoplasm detection algorithm of view-based access control model conspicuousness spectrum residual error method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810926369.3A CN109191436A (en) 2018-08-15 2018-08-15 The low-dose CT Lung neoplasm detection algorithm of view-based access control model conspicuousness spectrum residual error method

Publications (1)

Publication Number Publication Date
CN109191436A true CN109191436A (en) 2019-01-11

Family

ID=64921801

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810926369.3A Pending CN109191436A (en) 2018-08-15 2018-08-15 The low-dose CT Lung neoplasm detection algorithm of view-based access control model conspicuousness spectrum residual error method

Country Status (1)

Country Link
CN (1) CN109191436A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109829902A (en) * 2019-01-23 2019-05-31 电子科技大学 A kind of lung CT image tubercle screening technique based on generalized S-transform and Teager attribute
CN110738235A (en) * 2019-09-16 2020-01-31 平安科技（深圳）有限公司 Pulmonary tuberculosis determination method, pulmonary tuberculosis determination device, computer device, and storage medium
CN111539928A (en) * 2020-04-20 2020-08-14 武汉轻工大学 Image seam measurement optimizing device
CN117830322A (en) * 2024-03-06 2024-04-05 慧创科仪(北京)科技有限公司 Method and device for performing significance difference analysis on near infrared data

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104599159A (en) * 2015-02-06 2015-05-06 浪潮集团有限公司 User behavior recognition method based on support vector machine principle
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
CN107230204A (en) * 2017-05-24 2017-10-03 东北大学 A kind of method and device that the lobe of the lung is extracted from chest CT image
CN107274399A (en) * 2017-06-19 2017-10-20 太原理工大学 A kind of Lung neoplasm dividing method based on Hession matrixes and 3D shape index
CN107831737A (en) * 2016-09-16 2018-03-23 霍尼韦尔有限公司 The model factory detection of mismatch using SVMs for transverse process behavior monitoring
CN108198179A (en) * 2018-01-03 2018-06-22 华南理工大学 A kind of CT medical image pulmonary nodule detection methods for generating confrontation network improvement

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104599159A (en) * 2015-02-06 2015-05-06 浪潮集团有限公司 User behavior recognition method based on support vector machine principle
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
CN107831737A (en) * 2016-09-16 2018-03-23 霍尼韦尔有限公司 The model factory detection of mismatch using SVMs for transverse process behavior monitoring
CN107230204A (en) * 2017-05-24 2017-10-03 东北大学 A kind of method and device that the lobe of the lung is extracted from chest CT image
CN107274399A (en) * 2017-06-19 2017-10-20 太原理工大学 A kind of Lung neoplasm dividing method based on Hession matrixes and 3D shape index
CN108198179A (en) * 2018-01-03 2018-06-22 华南理工大学 A kind of CT medical image pulmonary nodule detection methods for generating confrontation network improvement

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
"MAXINE TAN 等": ""A novel computer-aided lung nodule detection system for CT images"", 《MEDICAL PHYSICS》 *
XIAODI HOU 等: "《Saliency Detection: A Spectral Residual Approach》", 《2007 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION》 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109829902A (en) * 2019-01-23 2019-05-31 电子科技大学 A kind of lung CT image tubercle screening technique based on generalized S-transform and Teager attribute
CN109829902B (en) * 2019-01-23 2022-04-12 电子科技大学 Lung CT image nodule screening method based on generalized S transformation and Teager attribute
CN110738235A (en) * 2019-09-16 2020-01-31 平安科技（深圳）有限公司 Pulmonary tuberculosis determination method, pulmonary tuberculosis determination device, computer device, and storage medium
CN111539928A (en) * 2020-04-20 2020-08-14 武汉轻工大学 Image seam measurement optimizing device
CN117830322A (en) * 2024-03-06 2024-04-05 慧创科仪(北京)科技有限公司 Method and device for performing significance difference analysis on near infrared data

Similar Documents

Publication Publication Date Title
CN109522908B (en) Image significance detection method based on region label fusion
CN105956582B (en) A kind of face identification system based on three-dimensional data
CN109191436A (en) The low-dose CT Lung neoplasm detection algorithm of view-based access control model conspicuousness spectrum residual error method
CN103048329B (en) A kind of road surface crack detection method based on active contour model
CN109086714A (en) Table recognition method, identifying system and computer installation
CN103886589B (en) Object-oriented automated high-precision edge extracting method
CN108154519A (en) Dividing method, device and the storage medium of eye fundus image medium vessels
US9501823B2 (en) Methods and systems for characterizing angle closure glaucoma for risk assessment or screening
CN109461163B (en) Edge detection method for magnetic resonance standard water model
CN104794440B (en) A kind of false fingerprint detection method based on the multiple dimensioned LBP of more piecemeals
CN109978848A (en) Method based on hard exudate in multiple light courcess color constancy model inspection eye fundus image
Wen et al. Virus image classification using multi-scale completed local binary pattern features extracted from filtered images by multi-scale principal component analysis
CN116188786B (en) Image segmentation system for hepatic duct and biliary tract calculus
CN102609903B (en) A kind of method of the movable contour model Iamge Segmentation based on marginal flow
Omer et al. Preprocessing of digital mammogram image based on otsu’s threshold
CN116524269A (en) Visual recognition detection system
CN105354547A (en) Pedestrian detection method in combination of texture and color features
Mehrara et al. Novel edge detection using BP neural network based on threshold binarization
CN117765287A (en) Image target extraction method combining LWR and density clustering
CN105354823A (en) Tree-ring image edge extraction and segmentation system
Chen et al. A study of image segmentation algorithms combined with different image preprocessing methods for thyroid ultrasound images
Shanmugavadivu et al. Segmentation of pectoral muscle in mammograms using Fractal method
CN114446447A (en) Semi-automatic labeling method for chromosome karyotype image data
CN113269756A (en) Retina blood vessel segmentation method and device based on multi-scale matched filtering and particle swarm optimization
Mehrara et al. Quad-pixel edge detection using neural network

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

Application publication date: 20190111

WD01 Invention patent application deemed withdrawn after publication