CN108776962A - A method of the structure good pernicious prediction model of lung neoplasm - Google Patents
A method of the structure good pernicious prediction model of lung neoplasm Download PDFInfo
- Publication number
- CN108776962A CN108776962A CN201810321872.6A CN201810321872A CN108776962A CN 108776962 A CN108776962 A CN 108776962A CN 201810321872 A CN201810321872 A CN 201810321872A CN 108776962 A CN108776962 A CN 108776962A
- Authority
- CN
- China
- Prior art keywords
- lung neoplasm
- prediction model
- good pernicious
- lung
- tumour
- 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
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- 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
- G06T2207/20081—Training; Learning
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30096—Tumor; Lesion
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Quality & Reliability (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Databases & Information Systems (AREA)
- Pathology (AREA)
- Radiology & Medical Imaging (AREA)
- Epidemiology (AREA)
- Primary Health Care (AREA)
- Apparatus For Radiation Diagnosis (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The invention discloses a kind of methods of the structure good pernicious prediction model of lung neoplasm.This method comprises the following steps：(1) lung neoplasm patient's sample is chosen, CT scan (CT) is carried out to the lung areas of lung neoplasm patient, obtains corresponding CT images；(2) the CT images obtained to step (1) are delineated, segmentation pulmonary lesion region, the focal area marked；(3) from the image feature of the focal area Extraction and determination of label；(4) Lasso algorithms are used to carry out feature selecting；(5) using the characteristic selected as input, Logistic is returned using gradient descent algorithm and carries out parameter optimization, finally trains to obtain the good pernicious prediction model of lung neoplasm using Logistic.The method that the present invention builds the good pernicious prediction model of lung neoplasm is simple, and time-consuming short, prediction model accuracy is high, can be applied to the good pernicious etiologic diagnosis of lung neoplasm.
Description
Technical field
The invention belongs to medical image omics technology fields, more particularly to a kind of good pernicious prediction model of structure lung neoplasm
Method.
Background technology
According to the statistics display of American Cancer Society 2016, highest tumor mortality rate is lung cancer, wherein male lung cancer
Incidence and the death rate are highests, and the incidence and the death rate of women account for second.The general onset of lung cancer compares concealment, more
It has been middle and advanced stage when number Patient Detection.5 years survival rates of patient of early detection and process treatment can reach 80% or more, it is seen that
The good pernicious detection of infantile tumour is great to the therapeutic potential of lung cancer.Clinically doctor determines that the main of benign from malignant tumors is examined at present
Disconnected method is by biopsy, it carries out histopathology via the intrusive fraction tissue for performing the operation acquisition tumour and examines
It is disconnected.Biopsy can help clinician to make diagnosis, but the shortcomings that this method is also apparent from.First, biopsy is invasive, disease
People is difficult to receive multiple biopsy and the time is with the treatment that may be delayed patient.Second is that due to the temporal-spatial heterogeneity of tumour,
Biopsy can only react the partial information of tumour, can not react all information of tumour.Therefore it is badly in need of a kind of noninvasive, non-intruding
Property comprehensive assessment tumor information method of early diagnosis.
In clinical practice, what medical image used is more and more extensive.But doctor is also only by the one of image at present
A little signs of lobulation, cavity, the features such as calcification make preliminary judgement, it is necessary to which final inspection can just be provided by compareing biopsy pathology information
Report result.The equipment and technology of acquisition image are also more and more perfect with the development of science and technology, the image data generated
Also increasing.Obvious doctor is this to be difficult to the use of image and processing mode in the big data letter that fully excavation image is included
Breath.Therefore image group is come into being.
Image group is an emerging field, has become the latest development side of International Medical image area research now
To.By medical image, for example, CT, Positron Emission Computed Tomography (Positron Emission Tomography,
PET/CT), magnetic resonance imaging (magnetic resonance imaging, MRI) region of interesting extraction largely quantifies shadow
As feature, these features are screened by machine learning method, are analyzed, most worthy associated with clinical problem is selected
Feature, carry out the diagnosis of tumour using the feature construction model selected and clinical phenotypes predicted.It can be seen that using image
The method that group is learned can solve above-mentioned problem.
The advantage for being diagnosed and being predicted to tumour based on image group, the invention discloses a kind of good evils of structure lung neoplasm
The method of property prediction model, this method construct the good pernicious prediction model of lung neoplasm, it can be achieved that lung neoplasm is good based on image group
Pernicious prediction, extends efficient help for clinician.
Invention content
The purpose of the present invention is to provide a kind of methods of the structure good pernicious prediction model of lung neoplasm.It is swollen that the present invention builds lung
The method of the good pernicious prediction model of tumor is simple, takes short, and prediction model accuracy is high, can be applied to that lung neoplasm is good pernicious to determine
Property diagnosis.
To achieve the above object, the present invention takes following technical scheme：
A method of the structure good pernicious prediction model of lung neoplasm, this method comprises the following steps：
(1) lung neoplasm patient's sample is chosen, CT scan is carried out to the lung areas of lung neoplasm patient
(CT), corresponding CT images are obtained；
(2) the CT images obtained to step (1) are delineated, segmentation pulmonary lesion region, the focal area marked；
(3) from the image feature of the focal area Extraction and determination of label；
(4) Lasso (least absolute shrinkage and selection operator) algorithm is used to carry out
Feature selecting；
(5) using the characteristic selected as input, it is excellent that progress parameter is returned to Logistic using gradient descent algorithm
Change, finally trains to obtain the good pernicious prediction model of lung neoplasm using Logistic.
Further, the lung neoplasm patient in the step (1) includes benign tumour patient and malignant tumor patient.
Further, the benign tumour includes hamartoma, inflammatory pseudotumor of lung and sclerosing hemangioma；It is described pernicious swollen
Tumor includes squamous carcinoma and gland cancer.
Tumoral character is broadly divided into four classes：1) tumour strength characteristic converts the three-dimensional data of tumour to single straight
Fang Tu.2) shape of tumor feature includes the size of tumour, shape, volume etc..3) textural characteristics of tumour, describe in tumour
The temporal-spatial heterogeneity in portion.4) wavelet character of tumour, to morphological feature, strength characteristic etc. carries out wavelet decomposition.
Further, the image feature in the step (3) includes tumour strength characteristic, shape of tumor feature, tumour
The wavelet character of textural characteristics and tumour.
Feature is excessive, wherein may include the too big feature of many uncorrelated or correlations, can cause to analyze and train
The overlong time of feature.Feature is excessive, and the model of structure is also more complicated, and generalization ability can also decline, and feature is excessively also easy
Cause " dimension disaster ".And usually medical image is all small-sample learning, feature can excessively be easier to cause over-fitting, reduce and divide
The efficiency of class device, causes classifying quality to be deteriorated.Therefore feature selecting is carried out to reject uncorrelated or redundancy feature.Feature
Selection can reduce analysis feature and build the time of model consumption, improve the generalization ability of model, reduce the complexity of model.
In clinic, doctor is it is desirable that fast and accurately aided diagnosis method.Feature choosing is carried out in the image group feature of extraction
It selects, data basis can be provided to build the efficient good pernicious prediction model of lung neoplasm.
Further, the formula of the Lasso algorithms is：
Wherein λ >=0, in formulaIndicate the degree of models fitting, i.e. loss function,Table
The punishment for showing parameter rejects variable by achieving the purpose that the smaller past zero compression of coefficient.Adjust whole parameter lambda ∈ [0 ,+∞),
When λ reaches certain value, the regression coefficient boil down to 0 of certain variables is played the effect of selection variable by Lasso.Since L1 is punished
Natural quality, the model of Lasso algorithm constructions is sparse, can realize that continuous variable shrinks and automatically select variable.λ
Bigger punishment is also bigger, it is meant that penalized more variables are zero.But it may result in the model variable of structure in this way
Error that is very few and omitting important variable generation bigger.Conversely, λ gets over small punishment with regard to smaller, the change for including in the model of structure
Amount is also more, and it is poor to eventually result in model over-fitting interpretation.Therefore select suitable adjusting parameter most important.
Further, using cross-validation method selection adjusting parameter λ；Preferably, most using the selection of 5 folding cross-validation methods
Excellent adjusting parameter λ.
It needs to do normalized to training set and verification collection before training pattern, because often different evaluation index has
Have different dimensions and dimensional unit, can be likely to influence in this case data analysis as a result, in order to eliminate index
Between this influence, need that data are normalized, to solve the comparativity between data target.Initial data is passed through
It is in the same order of magnitude after data normalization processing, is appropriate for Comprehensive Correlation evaluation.
Further, characteristic is normalized before the step (5).
Further, the normalized formula is：
Wherein, max (x) is the maximum value in sample data, and min (x) is the minimum value of sample data.
Herein optimized parameter is found using gradient descent method.Gradient descent method (gradient descent) is to solve for nothing
Constrained optimization problem most popular method, is a kind of alternative manner, the primary operational often walked be to solve for the gradient of object function to
Amount, using current location negative gradient direction as the direction of search (object function declines most fast in this direction).
The present invention has following technical characterstic：
1) the present invention is based on image group methods to build the good pernicious prediction model of lung neoplasm, which can reflect that tumour is whole
The biological characteristics of body can be applied to be applied to the good pernicious etiologic diagnosis of lung neoplasm, instead of traditional in vivo tissue examination
Invasive and invasive.
2) method of the present invention structure good pernicious prediction model of lung neoplasm extracts four kinds of main tumours using image group
Feature rejects uncorrelated or redundancy feature, can reduce analysis feature and build the time of model consumption, improve model
Generalization ability reduces the complexity of model.
3) method of the present invention structure good pernicious prediction model of lung neoplasm describes tumour, energy using more clarification of objective
Enough improve the accuracy of prediction model.
Description of the drawings
The original image (a) of Fig. 1 pulmonary lesion region segmentation figures；In the focal area (b) of original image label；Label
Focal area (c).
ROC curve of Fig. 2 lung neoplasms prediction model in training set (a) and test set (b).
Specific implementation mode
Following specific examples is the further explanation to method provided by the invention and technical solution, but is not construed as
Limitation of the present invention.
One, the selection of data
Using 80 lung neoplasm patients of Huaxi Hospital Attached to Sichuan Univ CT image datas (benign tumour mainly include paramnesia
Tumor, inflammatory pseudotumor of lung and sclerosing hemangioma；Malignant tumour includes mainly squamous carcinoma and gland cancer).Wherein 45 for building prediction
The training set of model, 35 for verifying the prediction model established.Training set includes 23 malignant tumours and 22 benign tumours.
Verification concentrates malignant tumour to have 17, and benign tumour has 18.
Two, it the acquisition of image and delineates
All patients acquire CT images data and mainly pass through Siemens's Defintion AS scanners.
The acquisition parameter of Siemens's Defintion AS scanners is as follows：Rotational time is 0.5s, and detector collimation is 64*
0.625mm, visual field size are 300*300mm, and the size for rebuilding picture element matrix is 512*512.Resolution ratio model of the image in x-axis
It encloses for 0.53~0.89mm, resolving range on the y axis is 0.53~0.89mm, and the resolution ratio in z-axis is 0.7mm.
CT images use ITK-SNAP (version:3.3.2；Www.itk-snap.org) software is delineated.Delineating is
It is carried out under lung window, the window width delineated is 1500HU, and window position is -400HU.To lump form, leaflet, pleural indentation sign, hair
The observation lung window that thorn sign increases is more relatively sharp than mediastinum window.CT images delineate be image group committed step, lung tumors area
Domain precisely delineate to later feature extraction and structure model it is most important.Pulmonary lesion region segmentation figure is shown in Fig. 1.
Three, feature extraction and feature selecting
According to the medical characteristics of tumour and clinical practice, it is extracted 485 in total from the cut zone of CT images label and determines
The image feature of amount.Feature is broadly divided into four classes：1) tumour strength characteristic converts the three-dimensional data of tumour to single straight
Fang Tu.2) shape of tumor feature includes the size of tumour, shape, volume etc..3) textural characteristics of tumour, describe in tumour
The temporal-spatial heterogeneity in portion.4) wavelet character of tumour, to morphological feature, strength characteristic etc. carries out wavelet decomposition.
Feature is excessive, wherein may include the too big feature of many uncorrelated or correlations, can cause to analyze and train
The overlong time of feature.Feature is excessive, and the model of structure is also more complicated, and generalization ability can also decline, and feature is excessively also easy
Cause " dimension disaster ".And usually medical image is all small-sample learning, feature can excessively be easier to cause over-fitting, reduce and divide
The efficiency of class device, causes classifying quality to be deteriorated.Therefore feature selecting is carried out to reject uncorrelated or redundancy feature.Feature
Selection can reduce analysis feature and build the time of model consumption, improve the generalization ability of model, reduce the complexity of model.
In clinic, doctor is it is desirable that fast and accurately aided diagnosis method.Feature choosing is carried out in the image group feature of extraction
It selects, data basis can be provided to build the efficient good pernicious prediction model of lung neoplasm.
The present embodiment extracts four classes totally 485 image group features respectively from the CT images after segmentation.It again will be good pernicious swollen
Tumor is grouped carry out test of difference, has filtered out 350 image group features with otherness, has then further utilized
Lasso methods carry out feature selecting.
Using Lasso (The least absolute shrinkage and selection operator method)
Method carries out feature selecting.The algorithm is that Tibshirani was proposed in 1996.The basic thought of Lasso is regression coefficient
The sum of absolute value is less than under the constraints of a normal number, its residual sum of squares (RSS) is made to minimize, stringent so as to generate some
Null regression coefficient, the model for obtaining to explain using the feature of regression coefficient non-zero.This thought of Lasso algorithms is more
It is suitble to analyze image group.The formula that Lasso is defined：
Wherein λ >=0, first half indicates the degree of models fitting, i.e. loss function, second part expression parameter in formula
Punishment, by by the smaller past zero compression of coefficient achieve the purpose that reject variable.Adjusting parameter λ ∈ [0 ,+∞), when λ reaches
When certain value, the regression coefficient boil down to 0 of certain variables is played the effect of selection variable by Lasso.Due to the nature of L1 punishment
The model of attribute, Lasso algorithm constructions is sparse, can realize that continuous variable shrinks and automatically select variable.
The sparsity of Lasso algorithm solutions is mainly quantified by adjusting parameter, and λ is bigger, and punishment is also bigger, it is meant that more
Penalized more variable is zero.But the model variable that may result in structure in this way is very few and omits important variable and generates more
Big error.Conversely, λ gets over small punishment with regard to smaller, the variable for including in the model of structure is also more, eventually results in model mistake
It is poor to be fitted interpretation.Therefore select suitable adjusting parameter most important.The selection of adjusting parameter generally uses cross validation
Method, the present embodiment use 5 folding cross validations (5-fold cross validation) to select optimal correction parameter lambda.5
The best λ value obtained after folding cross validation is 0.275.
3 and the good pernicious maximally related image group feature of prediction of lung neoplasm are had finally chosen, are respectively：glcm_5_
Difference_entropy, glcm_5_entropy and glcm_7_maximum_probability.
Four, the good pernicious prediction model of lung neoplasm is built
After feature selecting having been carried out using Lasso algorithms, using these features as input, in order to improve predictablity rate,
The characteristic of Lasso selections is normalized.The present embodiment returns Logistic using gradient descent algorithm and carries out
Parameter optimization finally trains to obtain the good pernicious prediction model of lung neoplasm using Logistic, verifies again later on collection to construction
The good pernicious model of lung neoplasm is verified.
Logistic recurrence is a kind of statistical method to grow up on the basis of multiple linear regression, nowadays wide
It is general to be used for the area researches such as social science.1967, Truelt J, Connifield J and Kannel W delivered coronary heart disease because
Logistic is returned the research for medicine direction by the paper of element research earlier.Because of the good pernicious classification prediction of lung neoplasm
Belong to two classified variables, so using two classification analysis of regression model.
It needs to do normalized to training set and verification collection before training pattern, because often different evaluation index has
Have different dimensions and dimensional unit, can be likely to influence in this case data analysis as a result, in order to eliminate index
Between this influence, need that data are normalized, to solve the comparativity between data target.Initial data is passed through
It is in the same order of magnitude after data normalization processing, is appropriate for Comprehensive Correlation evaluation.The normalization formula that the present invention uses for：
Wherein, max (x) is the maximum value in sample data, and min (x) is the minimum value of sample data.
Optimized parameter is found using gradient descent method.Gradient descent method (gradient descent) is to solve for no constraint
Optimization problem most popular method is a kind of alternative manner, and the primary operational often walked is to solve for the gradient vector of object function, will
Current location negative gradient direction is as the direction of search (object function declines most fast in this direction).
Fig. 2 is ROC curve of the lung neoplasm prediction model in training set and test set.Finally the AUC on training set is
0.870 (95% confidence interval：0.760 to 0.978), sensitivity 0.870, specificity 0.818.Verification collection on AUC be
0.853 (95% confidence interval：0.717 to 0.989), sensitivity 0.882, specificity 0.778.It is shown in Table 1.
The accuracy rate of diagnosis of the good pernicious prediction model of 1 lung neoplasm of table
It can be obtained from data above to draw a conclusion：The lung neoplasm image group that the present invention is built predicts disaggregated model application
In the good pernicious etiologic diagnosis of lung neoplasm, effectively doctor can be helped quickly to be diagnosed.
The explanation of above example is only intended to help to understand the method for the present invention and its core concept.It should be pointed out that for
For those skilled in the art, without departing from the principle of the present invention, if can also be carried out to the present invention
Dry improvement and modification, these improvement and modification are also fallen into the claims in the present invention protection domain.
Claims (9)
1. a kind of method of the structure good pernicious prediction model of lung neoplasm, which is characterized in that this method comprises the following steps：
(1) lung neoplasm patient's sample is chosen, CT scan (CT) is carried out to the lung areas of lung neoplasm patient, is obtained
Obtain corresponding CT images；
(2) the CT images obtained to step (1) are delineated, segmentation pulmonary lesion region, the focal area marked；
(3) from the image feature of the focal area Extraction and determination of label；
(4) Lasso (least absolute shrinkage and selection operator) algorithm is used to carry out feature
Selection；
(5) using the characteristic selected as input, Logistic is returned using gradient descent algorithm and carries out parameter optimization, most
It trains to obtain the good pernicious prediction model of lung neoplasm using Logistic afterwards.
2. a kind of method of structure good pernicious prediction model of lung neoplasm as described in claim 1, which is characterized in that the step
(1) the lung neoplasm patient in includes benign tumour patient and malignant tumor patient.
3. a kind of method of structure good pernicious prediction model of lung neoplasm as claimed in claim 2, which is characterized in that described benign
Tumour includes hamartoma, inflammatory pseudotumor of lung and sclerosing hemangioma；The malignant tumour includes squamous carcinoma and gland cancer.
4. a kind of method of structure good pernicious prediction model of lung neoplasm as described in claim 1, which is characterized in that the step
(3) image feature in include tumour strength characteristic, shape of tumor feature, the textural characteristics of tumour and tumour wavelet character.
5. a kind of method of structure good pernicious prediction model of lung neoplasm as described in claim 1, which is characterized in that described
The formula of Lasso algorithms is：
Wherein λ >=0, in formulaIndicate the degree of models fitting, i.e. loss function,Indicate ginseng
Several punishment rejects variable by achieving the purpose that the smaller past zero compression of coefficient.
6. a kind of method of structure good pernicious prediction model of lung neoplasm as claimed in claim 5, which is characterized in that using intersection
Proof method selects adjusting parameter λ.
7. a kind of method of structure good pernicious prediction model of lung neoplasm as claimed in claim 6, which is characterized in that use 5 foldings
Cross-validation method selects optimal correction parameter lambda.
8. a kind of method of structure good pernicious prediction model of lung neoplasm as described in claim 1, which is characterized in that in the step
Suddenly characteristic is normalized before (5).
9. a kind of method of structure good pernicious prediction model of lung neoplasm as claimed in claim 8, which is characterized in that the normalizing
Changing processing formula is：
Wherein, max (x) is the maximum value in sample data, and min (x) is the minimum value of sample data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810321872.6A CN108776962A (en) | 2018-04-11 | 2018-04-11 | A method of the structure good pernicious prediction model of lung neoplasm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810321872.6A CN108776962A (en) | 2018-04-11 | 2018-04-11 | A method of the structure good pernicious prediction model of lung neoplasm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108776962A true CN108776962A (en) | 2018-11-09 |
Family
ID=64033931
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810321872.6A Pending CN108776962A (en) | 2018-04-11 | 2018-04-11 | A method of the structure good pernicious prediction model of lung neoplasm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108776962A (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109583447A (en) * | 2018-12-11 | 2019-04-05 | 山东省肿瘤防治研究院（山东省肿瘤医院） | A kind of image group credit analysis model building method and analysis method |
| CN109686426A (en) * | 2018-12-29 | 2019-04-26 | 上海商汤智能科技有限公司 | Medical imaging processing method and processing device, electronic equipment and storage medium |
| CN110428396A (en) * | 2019-06-25 | 2019-11-08 | 郑州大学第一附属医院 | A kind of feature vectors dimensional down method and system based on CT images |
| CN110706804A (en) * | 2019-08-23 | 2020-01-17 | 刘雷 | Application method of hybrid expert system in classification of lung adenocarcinoma |
| CN111554381A (en) * | 2020-04-23 | 2020-08-18 | 上海市第一人民医院 | Artificial intelligent pathological diagnosis method and diagnosis model for renal clear cell carcinoma based on deep learning |
| CN111598859A (en) * | 2020-05-13 | 2020-08-28 | 吉林大学第一医院 | Chest lymph node standard uptake value detection method based on CT (computed tomography) image omics |
| CN111640503A (en) * | 2020-05-29 | 2020-09-08 | 上海市肺科医院 | Prediction system and method for tumor mutation load of patient with advanced lung cancer |
| CN111767952A (en) * | 2020-06-30 | 2020-10-13 | 重庆大学 | Interpretable classification method for benign and malignant pulmonary nodules |
| CN111932495A (en) * | 2020-06-30 | 2020-11-13 | 数坤（北京）网络科技有限公司 | Medical image detection method, device and storage medium |
| CN112349425A (en) * | 2020-02-10 | 2021-02-09 | 胡秋明 | Novel artificial intelligent rapid screening system for coronavirus infection pneumonia |
| WO2021031279A1 (en) * | 2019-08-20 | 2021-02-25 | 东北大学 | Deep-learning-based intelligent pneumonia diagnosis system and method for x-ray chest radiograph |
| CN113284615A (en) * | 2021-06-16 | 2021-08-20 | 北京大学人民医院 | XGboost algorithm-based gastrointestinal stromal tumor prediction method and system |
| CN113889261A (en) * | 2021-09-23 | 2022-01-04 | 之江实验室 | Pathological feature assistance-based PET/CT automatic lung cancer diagnosis classification model training method |
| CN114677378A (en) * | 2022-05-31 | 2022-06-28 | 四川省医学科学院·四川省人民医院 | Computer-aided diagnosis and treatment system based on ovarian tumor benign and malignant prediction model |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150051484A1 (en) * | 2013-08-14 | 2015-02-19 | Siemens Aktiengesellschaft | Histological Differentiation Grade Prediction of Hepatocellular Carcinoma in Computed Tomography Images |
| CN105975443A (en) * | 2016-05-04 | 2016-09-28 | 西南大学 | Lasso-based anomaly detection method and system |
| CN107103591A (en) * | 2017-03-27 | 2017-08-29 | 湖南大学 | A kind of single image to the fog method based on image haze concentration sealing |
-
2018
- 2018-04-11 CN CN201810321872.6A patent/CN108776962A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150051484A1 (en) * | 2013-08-14 | 2015-02-19 | Siemens Aktiengesellschaft | Histological Differentiation Grade Prediction of Hepatocellular Carcinoma in Computed Tomography Images |
| CN105975443A (en) * | 2016-05-04 | 2016-09-28 | 西南大学 | Lasso-based anomaly detection method and system |
| CN107103591A (en) * | 2017-03-27 | 2017-08-29 | 湖南大学 | A kind of single image to the fog method based on image haze concentration sealing |
Non-Patent Citations (1)
| Title |
|---|
| XINZHONG ZHU: "《Radiomic signature as a diagnostic factor for histologic subtype classification of non-small cell lung cancer》", 《EUROPEAN RADIOLOGY》 * |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109583447A (en) * | 2018-12-11 | 2019-04-05 | 山东省肿瘤防治研究院（山东省肿瘤医院） | A kind of image group credit analysis model building method and analysis method |
| CN109686426A (en) * | 2018-12-29 | 2019-04-26 | 上海商汤智能科技有限公司 | Medical imaging processing method and processing device, electronic equipment and storage medium |
| CN110428396A (en) * | 2019-06-25 | 2019-11-08 | 郑州大学第一附属医院 | A kind of feature vectors dimensional down method and system based on CT images |
| WO2021031279A1 (en) * | 2019-08-20 | 2021-02-25 | 东北大学 | Deep-learning-based intelligent pneumonia diagnosis system and method for x-ray chest radiograph |
| CN110706804A (en) * | 2019-08-23 | 2020-01-17 | 刘雷 | Application method of hybrid expert system in classification of lung adenocarcinoma |
| CN110706804B (en) * | 2019-08-23 | 2024-02-02 | 刘雷 | Application method of mixed expert system in lung adenocarcinoma classification |
| CN112349425A (en) * | 2020-02-10 | 2021-02-09 | 胡秋明 | Novel artificial intelligent rapid screening system for coronavirus infection pneumonia |
| CN111554381A (en) * | 2020-04-23 | 2020-08-18 | 上海市第一人民医院 | Artificial intelligent pathological diagnosis method and diagnosis model for renal clear cell carcinoma based on deep learning |
| CN111598859A (en) * | 2020-05-13 | 2020-08-28 | 吉林大学第一医院 | Chest lymph node standard uptake value detection method based on CT (computed tomography) image omics |
| CN111640503B (en) * | 2020-05-29 | 2023-09-26 | 上海市肺科医院 | System and method for predicting tumor mutation load of advanced lung cancer patient |
| CN111640503A (en) * | 2020-05-29 | 2020-09-08 | 上海市肺科医院 | Prediction system and method for tumor mutation load of patient with advanced lung cancer |
| CN111932495A (en) * | 2020-06-30 | 2020-11-13 | 数坤（北京）网络科技有限公司 | Medical image detection method, device and storage medium |
| CN111932495B (en) * | 2020-06-30 | 2021-09-07 | 数坤(北京)网络科技股份有限公司 | Medical image detection method, device and storage medium |
| CN111767952A (en) * | 2020-06-30 | 2020-10-13 | 重庆大学 | Interpretable classification method for benign and malignant pulmonary nodules |
| CN111767952B (en) * | 2020-06-30 | 2024-03-29 | 重庆大学 | Interpretable lung nodule benign and malignant classification method |
| CN113284615A (en) * | 2021-06-16 | 2021-08-20 | 北京大学人民医院 | XGboost algorithm-based gastrointestinal stromal tumor prediction method and system |
| CN113889261A (en) * | 2021-09-23 | 2022-01-04 | 之江实验室 | Pathological feature assistance-based PET/CT automatic lung cancer diagnosis classification model training method |
| CN113889261B (en) * | 2021-09-23 | 2022-06-10 | 之江实验室 | Pathological feature assistance-based PET/CT automatic lung cancer diagnosis classification model training method |
| CN114677378A (en) * | 2022-05-31 | 2022-06-28 | 四川省医学科学院·四川省人民医院 | Computer-aided diagnosis and treatment system based on ovarian tumor benign and malignant prediction model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108776962A (en) | A method of the structure good pernicious prediction model of lung neoplasm | |
| Lee et al. | Image based computer aided diagnosis system for cancer detection | |
| Messay et al. | A new computationally efficient CAD system for pulmonary nodule detection in CT imagery | |
| Lu et al. | A novel computer-aided diagnosis system for breast MRI based on feature selection and ensemble learning | |
| US8111896B2 (en) | Method and system for automatic recognition of preneoplastic anomalies in anatomic structures based on an improved region-growing segmentation, and commputer program therefor | |
| Elizabeth et al. | Computer-aided diagnosis of lung cancer based on analysis of the significant slice of chest computed tomography image | |
| CN109035160A (en) | The fusion method of medical image and the image detecting method learnt based on fusion medical image | |
| Peng et al. | Segmentation of lung in chest radiographs using hull and closed polygonal line method | |
| Suárez-Cuenca et al. | Application of the iris filter for automatic detection of pulmonary nodules on computed tomography images | |
| Selvathi et al. | Automatic segmentation and classification of liver tumor in CT images using adaptive hybrid technique and contourlet based ELM classifier | |
| Sammouda | Segmentation and analysis of CT chest images for early lung cancer detection | |
| Reddy et al. | Intelligent deep learning algorithm for lung cancer detection and classification | |
| Nardelli et al. | Deep-learning strategy for pulmonary artery-vein classification of non-contrast CT images | |
| Ghatak et al. | Fractal dimension-based infection detection in chest X-ray images | |
| Ren et al. | High-performance CAD-CTC scheme using shape index, multiscale enhancement filters, and radiomic features | |
| Kumaraswamy et al. | A review on cancer detection strategies with help of biomedical images using machine learning techniques | |
| Zhou et al. | HAUNet-3D: a novel hierarchical attention 3D UNet for lung nodule segmentation | |
| Pereyra et al. | Fractal analysis for computer-aided diagnosis of diffuse pulmonary diseases in HRCT images | |
| Ye et al. | Automated segmentation of mass regions in DBT images using a dilated DCNN approach | |
| Dolejší et al. | Automatic two-step detection of pulmonary nodules | |
| Malathi et al. | Segmentation of CT lung images using FCM with active contour and CNN classifier | |
| Bisht et al. | Hybrid Deep Learning Algorithm for Pulmonary Nodule Detection using Body Sensors | |
| Amritha et al. | Liver tumor segmentation and classification using deep learning | |
| Bandyopadhyay et al. | Artificial-intelligence-based diagnosis of brain tumor diseases | |
| RAM et al. | Intensified Computer Aided Detection System for Lung Cancer Detection using MEM algorithm. |
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 |