CN102122356A - Computer-aided method for distinguishing ultrasound endoscope image of pancreatic cancer - Google Patents
Computer-aided method for distinguishing ultrasound endoscope image of pancreatic cancer Download PDFInfo
- Publication number
- CN102122356A CN102122356A CN 201110063144 CN201110063144A CN102122356A CN 102122356 A CN102122356 A CN 102122356A CN 201110063144 CN201110063144 CN 201110063144 CN 201110063144 A CN201110063144 A CN 201110063144A CN 102122356 A CN102122356 A CN 102122356A
- Authority
- CN
- China
- Prior art keywords
- feature
- features
- image
- ultrasound endoscope
- pancreatic cancer
- 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
Abstract
The invention relates to a computer-aided method for distinguishing the ultrasound endoscope image of pancreatic cancer, providing a method for extracting and classifying the textural features of the ultrasound endoscope image of pancreatic cancer. The computer-aided method can be used for the computer-aided analysis of the ultrasound endoscope image of pancreatic cancer. 9 general classes which totally comprise 69 textural features are extracted from the ultrasound endoscope image of pancreatic cancer by a digital image processing algorithm. Class spacing is adopted to serve as a separability criterion to preliminarily screen the features; then, a chronological progress search algorithm is used for further screening the features; and the features are classified by a support vector machine. The method is realized by extracting the textural features of the ultrasound endoscope image via the classifier, various objective quantized diagnostic indexes and a method for correctly describing and explaining the ultrasound endoscope image are built, and the accuracy on the ultrasound endoscope early diagnosis of the pancreatic cancer is improved.
Description
Technical field
The invention belongs to the Medical Instruments technical field, particularly relate to the method that a kind of area of computer aided is differentiated cancer of pancreas EUS image.
Background technology
Cancer of pancreas is one of ten big malignant tumours of World Health Organization's announcement, is the great illness that seriously jeopardizes human health.In recent years, the incidence of disease of cancer of pancreas rises year by year.The cancer of pancreas development is rapid, dangerous, and clinical characters is that the course of disease is short, progress is fast, mortality ratio is high.
Pancreas belongs to the retroperitoneal organ, and the position is hidden, and the neighbouring relationship complexity has many vitals and blood vessel on every side.Therefore, the early detection of cancer of pancreas, early diagnosis, be a difficult problem that perplexs medical circle for a long time.
(endoscopic uhrasonography EUS) has obtained in countries in the world using widely since early eighties comes out in endoscopic ultrasonography.EUS is placed in the scope top with the novel high-frequency ultrasonic probe, after scope is inserted body cavity, by the form in the scope Direct observation chamber, simultaneously can carry out real-time ultrasonic scanning again, reach the ultrasonoscopy of adjacent organ on every side with the histologic characteristics that obtains the pipeline level, thereby further improved scope and Ultrasonic Diagnosis level.Because the insertion probe near pathology, shortens sound travel and has reduced acoustic attenuation, so can obviously improve picture resolution, finds tiny focus.Because its good display capabilities, EUS is applied in the diagnosis of pancreatic disease at present more and more at large.
Studies show that at present EUS has bigger diagnostic value to cancer of pancreas, the possibility of cancer of pancreas can be pointed out by pancreas endoscopic ultrosonograph, is convenient to early diagnosis.Yet it is bigger influenced by doctors experience and subjective factor based on the diagnosis of endoscopic ultrosonograph, and different doctors' accuracy rate of diagnosis can be different; Some trickle variations of image are difficult to be discovered by naked eyes; (EUS guided-finenee-dle aspiration, EUS-FNA) carrying out cytolgical examination has necessarily traumatic and use the following fine needle aspiration suction of EUS guiding art.Therefore, developing objective, the reliable and noninvasive cancer of pancreas EUS method of early diagnosis of a cover, is problem demanding prompt solution.
Summary of the invention
Technical matters to be solved by this invention provides a kind of method with Digital Image Processing and pattern classification, is applied to the computer-aided diagnosis of cancer of pancreas EUS.Realize by the textural characteristics and the sorter that extract endoscopic ultrosonograph, create various objective, the diagnosis indexs that quantize and correct description and explain the method for endoscopic ultrosonograph, improve the accuracy of cancer of pancreas EUS early diagnosis.
The technical solution adopted for the present invention to solve the technical problems is: provide a kind of area of computer aided to differentiate the method for cancer of pancreas EUS image, comprise the following steps:
(1) settles scope on novel high-frequency ultrasonic probe top, obtain endoscopic ultrosonograph, it is preserved with Windows bitmap format BMP form;
(2) according to ROIs, intercepting rectangle subgraph extracts the textural characteristics that is used for pancreas EUS image classification among each ROIs, and is screened by computing machine;
(3) choose the characteristics combination that the feature of class spacing maximum obtains as preliminary screening, preferred order forward search algorithm carries out feature selecting then;
(4), by the sorter Support Vector Machine image is classified automatically, and calculate its susceptibility, specificity, positive predictive value, negative predictive value and accuracy rate according to the feature of extracting in the image.
Described texture feature extraction is to carry out texture analysis according to ROIs, extraction 9 big classes, 69 features are used for pattern classification feature, each feature is done normalization simultaneously, and described 9 big category features comprise first-order statistics measure feature, gray level co-occurrence matrixes feature, gray scale difference statistical nature, field gray scale difference matrix character, Laws texture measurement features, fractal characteristic, Fourier power spectrum characteristic, invariant moment features and wavelet character.
Described class spacing maximum is chosen 25 features as initial characteristics, choosing 12 in described 25 features is characterized as the gray level co-occurrence matrixes feature and comprises: energy, contrast, covariance, average and, variance and, entropy and, third moment, consistance, fractal characteristic comprises that fractal dimension feature, second order multifractal dimension spy, Laws texture energy estimate feature.
Described first-order statistics measure feature comprises 1) average; 2) standard deviation; 3) smoothness; 4) third moment; 5) Fourth-order moment; 6) consistance; 7) entropy.
Described gray level co-occurrence matrixes feature comprises 1) energy; 2) contrast; 3) auto-correlation; 4) relevant; 5) covariance; 6) unfavourable balance square; 7) entropy; 8) average and; 9) variance and; 10) entropy and; 11) Cha variance; 12) difference coefficient; 13) third moment; 14) Fourth-order moment; 15) consistance; 16) absolute value; 17) maximum probability.
Described gray scale difference statistical nature uses the first-order statistics amount of image local gray-scale value.
Described neighborhood gray scale difference matrix character comprises 1) roughness; 2) contrast; 3) degree of rarefication; 4) complexity; 5) texture dynamics.
Described Laws texture energy is estimated feature and is selected LL for use, EE, LE, ES, LS for nuclear, ask for filtering after image energy as textural characteristics.
Described fractal characteristic is selected fractal dimension feature and multifractal Dimension Characteristics for use, is estimated by difference box counting method.
Described Fourier power spectrum characteristic adopt discrete Fourier transform (DFT) axially with angle and and use it for textural characteristics.
Described invariant moment features extracts all insensitive 7 two dimension invariant moment of translation, convergent-divergent, mirror image and rotation as textural characteristics.
Described wavelet character is done 3 layer scattering wavelet transformations to image, extracts the entropy of each subgraph and statistic as textural characteristics.
Beneficial effect
The present invention realizes by the textural characteristics and the sorter that extract endoscopic ultrosonograph, creates various objective, the diagnosis indexs that quantize and correct description and explains the method for endoscopic ultrosonograph, improves the accuracy of cancer of pancreas EUS early diagnosis.
Description of drawings
Fig. 1 is a schematic flow sheet of the present invention.
Embodiment
Below in conjunction with specific embodiment, further set forth the present invention.Should be understood that these embodiment only to be used to the present invention is described and be not used in and limit the scope of the invention.Should be understood that in addition those skilled in the art can make various changes or modifications the present invention after the content of having read the present invention's instruction, these equivalent form of values fall within the application's appended claims institute restricted portion equally.
The present invention includes the following step:
(1) settles scope on novel high-frequency ultrasonic probe top, obtain endoscopic ultrosonograph, it is preserved with Windows bitmap format BMP form;
(2) according to ROIs, intercepting rectangle subgraph extracts the textural characteristics that is used for pancreas EUS image classification among each ROIs, and is screened by computing machine;
(3) choose the characteristics combination that the feature of class spacing maximum obtains as preliminary screening, preferred order forward search algorithm carries out feature selecting then;
(4), by the sorter Support Vector Machine image is classified automatically, and calculate its susceptibility, specificity, positive predictive value, negative predictive value and accuracy rate according to the feature of extracting in the image.
Described texture feature extraction is to carry out texture analysis according to ROIs, extraction 9 big classes, 69 features are used for pattern classification feature, each feature is done normalization simultaneously, and described 9 big category features comprise first-order statistics measure feature, gray level co-occurrence matrixes feature, gray scale difference statistical nature, field gray scale difference matrix character, Laws texture measurement features, fractal characteristic, Fourier power spectrum characteristic, invariant moment features and wavelet character.
Described class spacing maximum is chosen 25 features as initial characteristics, choosing 12 in described 25 features is characterized as the gray level co-occurrence matrixes feature and comprises: energy, contrast, covariance, average and, variance and, entropy and, third moment, consistance, fractal characteristic comprises that fractal dimension feature, second order multifractal dimension spy, Laws texture energy estimate feature.
Described first-order statistics measure feature comprises 1) average; 2) standard deviation; 3) smoothness; 4) third moment; 5) Fourth-order moment; 6) consistance; 7) entropy.
Described gray level co-occurrence matrixes feature comprises 1) energy; 2) contrast; 3) auto-correlation; 4) relevant; 5) covariance; 6) unfavourable balance square; 7) entropy; 8) average and; 9) variance and; 10) entropy and; 11) Cha variance; 12) difference coefficient; 13) third moment; 14) Fourth-order moment; 15) consistance; 16) absolute value; 17) maximum probability.
Described gray scale difference statistical nature uses the first-order statistics amount of image local gray-scale value.
Described neighborhood gray scale difference matrix character comprises 1) roughness; 2) contrast; 3) degree of rarefication; 4) complexity; 5) texture dynamics.
Described Laws texture energy is estimated feature and is selected LL for use, EE, LE, ES, LS for nuclear, ask for filtering after image energy as textural characteristics.
Described fractal characteristic is selected fractal dimension feature and multifractal Dimension Characteristics for use, is estimated by difference box counting method.
Described Fourier power spectrum characteristic adopt discrete Fourier transform (DFT) axially with angle and and use it for textural characteristics.
Described invariant moment features extracts all insensitive 7 two dimension invariant moment of translation, convergent-divergent, mirror image and rotation as textural characteristics.
Described wavelet character is done 3 layer scattering wavelet transformations to image, extracts the entropy of each subgraph and statistic as textural characteristics.
With 216 routine case random division is training set and test set, and training set 108 examples (cancer 76 examples, non-cancer 32 examples), test set 108 examples (cancer 77 examples, non-cancer 31 examples) are used the training set training classifier, and test set is tested.For reducing the limited error that causes of experiment sample, random experiments have been carried out altogether 50 times, the accuracy that finally draws svm classifier is (97.98 ± 1.237) %, susceptibility is (94.324-0.0354) %, specificity is (99.454-0.0102) %, positive predictive value (98.654-0.0251) %, negative predictive value are (97.774-0.0137) %.
Claims (3)
1. the method for an area of computer aided differentiation cancer of pancreas EUS image comprises the following steps:
(1) settles scope on novel high-frequency ultrasonic probe top, obtain endoscopic ultrosonograph, it is preserved with Windows bitmap format BMP form;
(2) according to ROIs, intercepting rectangle subgraph extracts the textural characteristics that is used for pancreas EUS image classification among each ROIs, and is screened by computing machine;
(3) choose the characteristics combination that the feature of class spacing maximum obtains as preliminary screening, preferred order forward search algorithm carries out feature selecting then;
(4), by the sorter Support Vector Machine image is classified automatically, and calculate its susceptibility, specificity, positive predictive value, negative predictive value and accuracy rate according to the feature of extracting in the image.
2. a kind of area of computer aided according to claim 1 is differentiated the method for cancer of pancreas EUS image, it is characterized in that: described texture feature extraction is to carry out texture analysis according to ROIs, extraction 9 big classes, 69 features are used for pattern classification feature, each feature is done normalization simultaneously, and described 9 big category features comprise first-order statistics measure feature, gray level co-occurrence matrixes feature, gray scale difference statistical nature, field gray scale difference matrix character, Laws texture measurement features, fractal characteristic, Fourier power spectrum characteristic, invariant moment features and wavelet character.
3. a kind of area of computer aided according to claim 1 is differentiated the method for cancer of pancreas EUS image, it is characterized in that: described class spacing maximum is chosen 25 features as initial characteristics, choosing 12 in described 25 features is characterized as the gray level co-occurrence matrixes feature and comprises: energy, contrast, covariance, average and, variance and, entropy and, third moment, consistance, fractal characteristic comprises that fractal dimension feature, second order multifractal dimension spy, Laws texture energy estimate feature.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 201110063144 CN102122356A (en) | 2011-03-16 | 2011-03-16 | Computer-aided method for distinguishing ultrasound endoscope image of pancreatic cancer |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 201110063144 CN102122356A (en) | 2011-03-16 | 2011-03-16 | Computer-aided method for distinguishing ultrasound endoscope image of pancreatic cancer |
Publications (1)
Publication Number | Publication Date |
---|---|
CN102122356A true CN102122356A (en) | 2011-07-13 |
Family
ID=44250910
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN 201110063144 Pending CN102122356A (en) | 2011-03-16 | 2011-03-16 | Computer-aided method for distinguishing ultrasound endoscope image of pancreatic cancer |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102122356A (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103632156A (en) * | 2013-12-23 | 2014-03-12 | 中南大学 | Multi-scale neighboring dependence matrix-based method for extracting froth image texture characteristics |
CN104798105A (en) * | 2012-11-20 | 2015-07-22 | 皇家飞利浦有限公司 | Integrated phenotyping employing image texture features. |
CN106020827A (en) * | 2016-05-24 | 2016-10-12 | 福建师范大学 | Medical image texture analysis system |
CN106023188A (en) * | 2016-05-17 | 2016-10-12 | 天津大学 | Breast tumor feature selection method based on Relief algorithm |
US10035009B2 (en) | 2013-04-15 | 2018-07-31 | The Board Of Trustees Of The Leland Stanford Junior University | Systems and methods for treating pancreatic cancer |
CN109858428A (en) * | 2019-01-28 | 2019-06-07 | 四川大学 | ANA flourescent sheet automatic identifying method based on machine learning and deep learning |
CN110188788A (en) * | 2019-04-15 | 2019-08-30 | 浙江工业大学 | The classification method of cystic Tumor of Pancreas CT image based on radiation group feature |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101853493A (en) * | 2009-10-21 | 2010-10-06 | 首都医科大学 | Method for extracting multi-dimensional texture of nodi from medical images |
CN101859441A (en) * | 2010-05-25 | 2010-10-13 | 中国人民解放军第四军医大学 | Image-based computer-aided analytical method for performing non-invasive monitoring to degree of tumor-infiltrated surrounding tissue |
-
2011
- 2011-03-16 CN CN 201110063144 patent/CN102122356A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101853493A (en) * | 2009-10-21 | 2010-10-06 | 首都医科大学 | Method for extracting multi-dimensional texture of nodi from medical images |
CN101859441A (en) * | 2010-05-25 | 2010-10-13 | 中国人民解放军第四军医大学 | Image-based computer-aided analytical method for performing non-invasive monitoring to degree of tumor-infiltrated surrounding tissue |
Non-Patent Citations (1)
Title |
---|
《生物医学工程学进展》 20081231 蔡哲元 等 胰腺内镜超声图像纹理特征提取与分类研究 141-145页 1-3 第29卷, 第3期 * |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104798105A (en) * | 2012-11-20 | 2015-07-22 | 皇家飞利浦有限公司 | Integrated phenotyping employing image texture features. |
CN104798105B (en) * | 2012-11-20 | 2019-06-07 | 皇家飞利浦有限公司 | Using the integrated phenotype of image texture characteristic |
US10035009B2 (en) | 2013-04-15 | 2018-07-31 | The Board Of Trustees Of The Leland Stanford Junior University | Systems and methods for treating pancreatic cancer |
CN103632156A (en) * | 2013-12-23 | 2014-03-12 | 中南大学 | Multi-scale neighboring dependence matrix-based method for extracting froth image texture characteristics |
CN103632156B (en) * | 2013-12-23 | 2016-06-22 | 中南大学 | Froth images texture characteristic extracting method based on multiple dimensioned neighborhood correlation matrix |
CN106023188A (en) * | 2016-05-17 | 2016-10-12 | 天津大学 | Breast tumor feature selection method based on Relief algorithm |
CN106020827A (en) * | 2016-05-24 | 2016-10-12 | 福建师范大学 | Medical image texture analysis system |
CN106020827B (en) * | 2016-05-24 | 2019-12-10 | 福建师范大学 | Medical image texture analysis system |
CN109858428A (en) * | 2019-01-28 | 2019-06-07 | 四川大学 | ANA flourescent sheet automatic identifying method based on machine learning and deep learning |
CN109858428B (en) * | 2019-01-28 | 2021-08-17 | 四川大学 | Automatic ANA fluorescent film identification method based on machine learning and deep learning |
CN110188788A (en) * | 2019-04-15 | 2019-08-30 | 浙江工业大学 | The classification method of cystic Tumor of Pancreas CT image based on radiation group feature |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Guo et al. | Radiomics analysis on ultrasound for prediction of biologic behavior in breast invasive ductal carcinoma | |
CN111243042A (en) | Ultrasonic thyroid nodule benign and malignant characteristic visualization method based on deep learning | |
Zhu et al. | Differentiation of pancreatic cancer and chronic pancreatitis using computer-aided diagnosis of endoscopic ultrasound (EUS) images: a diagnostic test | |
Zortea et al. | A simple weighted thresholding method for the segmentation of pigmented skin lesions in macroscopic images | |
CN102122356A (en) | Computer-aided method for distinguishing ultrasound endoscope image of pancreatic cancer | |
Alilou et al. | An integrated segmentation and shape‐based classification scheme for distinguishing adenocarcinomas from granulomas on lung CT | |
CN107133638B (en) | Multi-parameter MRI prostate cancer CAD method and system based on two classifiers | |
CN103578099B (en) | The extracting method of tumor elastic characteristic based on ultrasonic elastograph imaging | |
Samah et al. | Classification of benign and malignant tumors in histopathology images | |
EP2564355A1 (en) | Malignant mass detection and classification in radiographic images | |
CN104000619A (en) | Thyroid CT image computer-aided diagnosis system and method | |
CN108241865B (en) | Ultrasound image-based multi-scale and multi-subgraph hepatic fibrosis multistage quantitative staging method | |
CN113269225A (en) | Non-invasive liver epithelium sample vascular smooth muscle lipoma image classification device based on image omics | |
Kim et al. | The recent progress in quantitative medical image analysis for computer aided diagnosis systems | |
CN104143047A (en) | Automatic tissue calibration method for IVUS gray-scale image | |
Cabral et al. | Fractal analysis of breast masses in mammograms | |
Hu et al. | Reproducibility of quantitative high‐throughput BI‐RADS features extracted from ultrasound images of breast cancer | |
Zhang et al. | Comparison of multiple feature extractors on Faster RCNN for breast tumor detection | |
Gao et al. | Segmentation of ultrasonic breast tumors based on homogeneous patch | |
CN103455821B (en) | Image analysis apparatus and method based on BI-RADS | |
CN109065150A (en) | A kind of ultrasonic tumor of breast stage division based on multi-feature extraction and Linear SVM | |
Li et al. | Computer aided diagnosis of fatty liver ultrasonic images based on support vector machine | |
Wei et al. | Automatic classification of benign and malignant breast tumors in ultrasound image with texture and morphological features | |
CN107463964A (en) | A kind of tumor of breast sorting technique based on features of ultrasound pattern correlation, device | |
Wei et al. | Multi-feature fusion for ultrasound breast image classification of benign and malignant |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20110713 |