CN110265095A - For HCC recurrence and construction method and the application of the prediction model and nomogram of RFS - Google Patents
For HCC recurrence and construction method and the application of the prediction model and nomogram of RFS Download PDFInfo
- Publication number
- CN110265095A CN110265095A CN201910426920.2A CN201910426920A CN110265095A CN 110265095 A CN110265095 A CN 110265095A CN 201910426920 A CN201910426920 A CN 201910426920A CN 110265095 A CN110265095 A CN 110265095A
- Authority
- CN
- China
- Prior art keywords
- model
- feature
- nomogram
- rfs
- hcc
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
-
- 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
-
- 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
- G16H10/00—ICT specially adapted for the handling or processing of patient-related medical or healthcare data
- G16H10/20—ICT specially adapted for the handling or processing of patient-related medical or healthcare data for electronic clinical trials or questionnaires
-
- 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
- G16H30/00—ICT specially adapted for the handling or processing of medical images
- G16H30/20—ICT specially adapted for the handling or processing of medical images for handling medical images, e.g. DICOM, HL7 or PACS
-
- 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
- G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
-
- 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
- G16H70/00—ICT specially adapted for the handling or processing of medical references
- G16H70/60—ICT specially adapted for the handling or processing of medical references relating to pathologies
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- Epidemiology (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Radiology & Medical Imaging (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Pathology (AREA)
- Databases & Information Systems (AREA)
- Biomedical Technology (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The invention belongs to prediction models and nomogram constructing technology field, disclose it is a kind of for HCC recurrence and RFS prediction model and nomogram construction method and application, using minimal redundancy most relevance algorithm (minimum redundancy maximum relevance algorithm, MRMRA), divide equally you can well imagine from 3 phase of the liver of all patients Enhanced CT image and take out 647 image group features；Image group label is confirmed using minimum absolute retract and Selecting operation symbol (least absolute shrinkage and selection operator, LASSO)-Cox regression model；It is recurred after establishing the curative ablation of HCC using clinical, Pathologic factors joint image group label and the prediction model of RFS and predicts nomogram.The present invention constructs prediction model and prediction nomogram by image group method, for HCC patient, tumor recurrence and RFS to be predicted after the curative ablation of lesion.
Description
Technical field
The invention belongs to prediction model and nomogram constructing technology fields more particularly to a kind of for HCC recurrence and RFS
The construction method and application of prediction model and nomogram.It is specially a kind of to be used for hepatocellular carcinoma (hepatocellular
Carcinoma, HCC) recurrence and recurrence-free survival (recurrence free survival, RFS) prediction model and Nuo Mo
The construction method of figure and application.
Background technique
Currently, the immediate prior art:
It there is no prediction model similarly to the prior art, current with good grounds patient's alpha-fetoprotein (alpha feto
Protein, AFP) result predicts HCC recurrence, once but AFP increase, possible tumour is obvious；Furthermore still have
Quite a few patient's (20% or so) is AFP negative HCC, even if there is tumor recurrence, AFP is still negative, therefore to feminine gender
The unpredictable HCC recurrence of HCC patient-monitoring AFP.Also the method for the pathology differentiation degree prediction recurrence of with good grounds HCC, but it is quick
Perception and specificity are lower, because the low differentiation HCC patient in clinical discovery part after lesion resection/ablation 3 years without recurrence, and phase
When in lesion resection/in ablation later six months tumor recurrence occurs for the middle differentiated HCC patient of ratio, this is HCC biological characteristics
Property and tumour caused by, because biopsy only punctures few part HCC tissue, the biological behaviour of entire HCC cannot be represented.
Also with good grounds patient Italian liver cancer cooperative groups (cancer of the liver Italian program,
CLIP) scoring and liver function Child-pugh scoring and joint AFP carry out the model of recurrence prediction, same sensitivity and specificity
Obvious insufficient, document is specific as follows:
(1)Nakagawa S,Hayashi H,Nitta H,et al.Scoring system based on tumor
markers and Child-Pugh classification for HCC patients who underwent liver
resection.Anticancer Res.2015；35(4):2157-2163.
(2)Zhao WH,Ma ZM,Zhou XR,Feng YZ,Fang BS.Prediction of recurrence and
prognosis in patients with hepatocellular carcinoma after resection by use of
CLIP score.World J Gastroenterol.2002；8(2):237-242.
(3)Tateishi R,Shiina S,Yoshida H,et al.Prediction of recurrence of
hepatocellular carcinoma after curative ablation using three tumor
markers.Hepatology.2006；44(6):1518-1527.
In conclusion problem of the existing technology is:
(1) universality is poor；
(2) sensibility is low；
(3) specificity is low；
(4) accuracy rate is low；
Solve the difficulty of above-mentioned technical problem:
Pass through the research method of image group, using artificial intelligence technology, entire tumour in liver CECT preoperative to patient
Scan data (DICOM format) excavated, extract and process analysis, in a image group features up to a hundred extract and HCC
Closely related feature is recurred, clinic and laboratory in conjunction with patient, pathological examination establish recurrence prediction model and nomogram,
Forecasting accuracy can be significantly improved.
The meaning for solving above-mentioned technical problem is:
Under the premise of melting postoperative recurrence probability compared with Accurate Prediction HCC, clinician can be instructed to adjust therapeutic strategy (pre-
Survey is by the postoperative auxiliary targeting medicine of patient of high relapse rate or traditional Chinese medical herbal treatment prevention recurrence) and follow-up intensity is (again depending on prediction
The height of hair rate is adjusted strong or turns down follow-up intensity, such as high recurrence probability person, monthly carry out 1 CECT or Contrast-enhanced MRI (CEMRI),
Blood drawing looks into tumor markers inspection etc. with early detection recurrent focus, conversely, carrying out within every 3-6 months such as low recurrence probability person
State recurrence).
Summary of the invention
In view of the problems of the existing technology, the present invention provides a kind of for HCC recurrence and prediction model and the promise of RFS
The construction method of mould figure and application.
The invention is realized in this way a kind of construction method packet of prediction model and nomogram for HCC recurrence and RFS
It includes:
The first step, using MRMRA (minimal redundancy most relevance algorithm minimum redundancy maximum
Relevance algorithm, MRMRA), dividing equally from 3 phase of liver CECT scan image, which you can well imagine, takes out 647 image groups spies
Sign；
Second step accords with (least absolute shrinkage and using minimum absolute retract and Selecting operation
Selection operator, LASSO)-Cox regression model confirmation image group label；
Third step establishes HCC after the curative ablation of HCC using clinical, Pathologic factors joint image group label
The prediction model and prediction nomogram of recurrence and RFS.
Further, the step 1 specifically includes:
Totally 647 reflection machines read the image group feature of Features and subtle texture information from the sense divided
It is extracted in region of interest using software Matlab 2014a；Image is carried out to initial three-dimensional tumour level using un-decimated wavelet transform transformation
Filtration processing；The feature extracted from original image or decomposition image can be divided into two classes: textural characteristics and non-grain spy
Sign；Non-grain feature includes: shape, size and density feature；The intuitive feature of shapes and sizes feature acquisition tumour；Density is special
Gather-exhibit shows the histogram feature of lesion；Textural characteristics are based on following four texture matrix and extract, i.e. gray level co-occurrence matrixes
(gray level co-occurrence matrix, GLCM), the long matrix of gray level (gray level run-length
Matrix, GLRLM), gray level band matrix (gray level size zone matrix, GLSZM) and neighborhood grey scale difference square
Battle array (neighborhood gray-tone difference matrix, NGTDM).
Further, the step 2 specifically includes:
(1) clinical, pathological factor analysis, selection
Clinical, pathological factor is analyzed using single factor test Cox risk ratio regression model, and the factor of p value < 0.10 enters
Choosing, selected factor are integrated and enter the multifactor Cox model of successive Regression；In multiplicity, the p value < 0.05 of variable
Be identified as the factor may be related to the RFS of patient and be included into the foundation of model, otherwise i.e. be excluded；
(2) selection of image group feature
The stability of a certain feature is calculated and determined by test-retest method of inspection by ICC；The attribute of a certain feature is related
Coefficient < 0.75 will be excluded；In order to reduce the redundancy and unnecessary complexity of calculating and modeling, MRMRA is used for feature choosing
It selects；The purpose of the algorithm is selection character subset, which can obtain the best feature of difference between two groups, wherein considering one
Item limiting factor, i.e., these features are entirely different but have extremely faint correlation again with Clinical Outcome of interest；And
MRMRA is proved that extracted feature can be made more stable, particularly with even more so for the research of image group；Final basis
The MRMRA of each feature exports score, shares 20 features and is selected to be used to construct model；" irr " R software package is used for
Computation attribute related coefficient；" mRMRe " R software package is used for MRMRA feature selecting.
(3) foundation of image group label
For training set data, 20 features of selection are carried out in Cox proportional hazard model using LASSO method deep
Level variables choice；Final choice preferably has the feature of minimum cross validation log portion likelihood；Show feature and existence
The nonzero coefficient of risk ratio is confirmed as the weight of each selection feature；The image group label of each patient passes through institute
Feature and the product of respective coefficient is selected to generate；For CECT image, training set data is utilized using LASSO-Cox model
It establishes model and is verified in verifying the set pair analysis model prediction efficiency；The predictive value of image group label is commented by KM curve
Estimate；Patient in group is divided into high risk group and low-risk group by the median of image group label；Pass through logarithm rank sum test ratio
Compared with KM curve；LASSO-Cox variables choice and model foundation are carried out using glmnet R software package；Using R software package to existence
Curve is compared；
20 features of final choice are entered LASSO-Cox model；According to a scarce cross-validation method, the use being finally included in
In the arterial phase, Portal venous phase and the image feature of period of delay of establishing image group model；Single factor test Cox proportional hazard model with
The image group label of arterial phase, Portal venous phase, period of delay in 3 phase CECT establishes prediction model respectively, and multifactor Cox ratio
Example risk model then integrates 3 phase CECT whole image group labels and establishes joint forecast model.
Further, the step 3 specifically includes:
(1) clinical, pathological factor is incorporated into above-mentioned 4 images group prediction model, portal venous phase images group student's federation closes pre-
It surveys model and shows optimum prediction ability in validation group；
(2) nomogram is established: the portal venous phase images group conjunctive model because being integrated with clinical, pathological factor has best
Predictive ability, therefore nomogram is established based on this model；Nomogram calibration curve is drawn simultaneously for training group and validation group.
Another object of the present invention is to provide it is a kind of by the construction method construct for HCC recurrence and RFS it is pre-
Survey model and nomogram.
Another object of the present invention is to provide the models and nomogram that are used to predict HCC recurrence and RFS described in one kind to exist
HCC locally carries out the application of tumor recurrence and RFS prediction after curative ablation.
In conclusion advantages of the present invention and good effect are as follows:
It is recurred provided by the present invention for hepatocellular carcinoma (hepatocellular carcinoma, HCC) and without recurrence
Construction method and the application of the prediction model and nomogram of existence (recurrence free survival, RFS), using minimum
Redundancy most relevance algorithm (minimum redundancy maximum relevance algorithm, MRMRA), from whole
Divide equally you can well imagine in 3 phase of the liver Enhanced CT image of patient and takes out 647 image group features；Using minimum absolute retract and
Selecting operation accords with (least absolute shrinkage and selection operator, LASSO)-Cox regression model
Confirm image group label；It is established after the curative ablation of HCC again using clinical, Pathologic factors joint image group label
The prediction model and prediction nomogram of hair and RFS.The present invention constructs prediction model and prediction nomogram by image group method,
For to HCC patient, tumor recurrence and RFS to be predicted after the curative ablation of lesion.
The present invention constructs prediction model and prediction nomogram by image group (radiomics) method, is used for HCC patient
Tumor recurrence and RFS are predicted after lesion locally curative ablation.
The present invention is used for HCC by what the construction method of the prediction model and nomogram for HCC recurrence and RFS constructed
The prediction model and nomogram of recurrence and RFS, be can after the curative ablation of Accurate Prediction HCC recurrence probability tool.
The prediction model and nomogram that the present invention is used for HCC recurrence and RFS for HCC, in lesion, locally control by curative ablation
After treatment HCC recurrence and recurrence-free survival prediction in application.It can be according to the image group number of a certain patient using the nomogram
Value, clinical and pathological characters directly read its recurrence probability in 1/2/3.
Detailed description of the invention
Fig. 1 is the construction method stream of the prediction model and nomogram provided in an embodiment of the present invention for HCC recurrence and RFS
Cheng Tu.
Fig. 2 is the technology path of the prediction model and nomogram provided in an embodiment of the present invention for HCC recurrence and RFS
Figure.
Fig. 3 is the specific implementation step schematic diagram that the present invention establishes prediction model and nomogram.
After Fig. 4 is clinical, pathological characters and portal venous phase images group label integrated optimization provided in an embodiment of the present invention
Prediction nomogram.
Fig. 5 is training group conjunctive model prediction nomogram correction graph provided in an embodiment of the present invention.
Fig. 6 is validation group conjunctive model prediction nomogram correction graph provided in an embodiment of the present invention.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, below in conjunction with example is implemented, to this hair
It is bright to be further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and do not have to
It is of the invention in limiting.
Application principle of the invention is explained in detail with reference to the accompanying drawing.
As shown in Figure 1, the present invention implements the structure for the prediction model and nomogram for HCC recurrence and RFS that example provides
Construction method the following steps are included:
S101: using MRMRA, divides equally you can well imagine from 3 phase of the liver CECT scan image of all patients and takes out 647 images
Group learns feature；
S102: image group label is confirmed using LASSO-Cox regression model；
S103: it is multiple to combine HCC after image group label establishes the curative ablation of HCC using clinical, Pathologic factors
The prediction model and RFS of hair and RFS predict nomogram.
Application principle of the invention is further described combined with specific embodiments below.
As shown in Figure 1-3, the present invention implements the prediction model and nomogram for HCC recurrence and RFS that example provides
Construction method the following steps are included:
Step 1 is divided equally you can well imagine from 3 phase of the liver CECT scan image of all patients and is taken out 647 shadows using MRMRA
As group learns feature；
Step 2 confirms image group label using LASSO-Cox regression model；
Step 3 establishes HCC after the curative ablation of HCC using clinical, Pathologic factors joint image group label
The prediction model and RFS of recurrence and RFS predict nomogram.
Further, the step 1 specifically includes:
Totally 647 reflection machines read the image group feature of Features and subtle texture information from the sense divided
It is extracted in region of interest (ROI) using software Matlab 2014a (MathWorks, Natick, MA, USA).Using un-decimated wavelet transform
It converts (original image can be decomposed into 8 components) and image filtration processing is carried out to initial three-dimensional HCC level.From original image or
Two classes: textural characteristics and non-grain feature can be divided by decomposing the feature extracted in image.Non-grain feature includes: shape
Shape, size and density feature.The intuitive feature of shapes and sizes feature acquisition HCC.The histogram that density feature shows lesion is special
Point.Textural characteristics are based on following four texture matrix and extract, i.e. GLCM, GLRLM, GLSZM, NGTDM.The class of whole features
Type and name are shown in Table 1.
The details of feature selected by CECT each phase in 1. image group model of table
X is the intensity value ofthe original image.
XLLL,XLLH,XLHL,XLHH,XHLL,XHLH,XHHL,and XHHH are the intensity value
of the transformation images from the original image by eight three-
dimensional wavelet filters.L:low-pass filter；H:low-pass filter.For example,
XLHL represent the intensity value resulting from directional filtering of X
with a low-pass filter along the x-direction,a high pass filter along the y-
direction and a low-pass filter along the z-direction.
is median intensity value ofX.
R(i,j|θ)is the value of row i and column j in the Gray-Level Run-
Length Matrix for a directionθ.
C(i,j)is the value of row i and column j in the Gray-Level Co-
Occurrence Matrix.
Z(i,j)is the value of row i and column j in the Gray-Level Size Zone
Matrix.
Ngis the number of discrete intensity values in the image.
Nris the number of different run lengths.
Nzis the size of the largest homogeneous region.
Npis the number of voxels in the image.
μX(i)is the mean of row i.
μy(j)is the mean of column j.
σX(i)is the standard deviation of row i.
σy(j)is the standard deviation of column j.
Further, the step 2 specifically includes:
(1) clinical, pathological factor analysis, selection
Clinical, pathological factor is analyzed using single factor test Cox risk ratio regression model, and the factor of p value < 0.10 enters
Choosing, selected factor is integrated and enters the multifactor Cox model of successive Regression.In multiplicity, the p value < of variable
0.05 be identified as the factor may be related to the RFS of patient and be included into the foundation of model, otherwise i.e. be excluded.
(2) selection of image group feature
Attribute related coefficient (intra-class correlation coefficient, ICC) passes through test- in group
The stability of a certain feature is calculated and determined in retest method of inspection.The attribute related coefficient < 0.75 of a certain feature will be excluded.
It is used for feature selecting to reduce redundancy and unnecessary complexity, the MRMRA of calculating and modeling.The purpose of the algorithm is choosing
Select character subset, which can obtain the best feature of difference between two groups, among these consider a limiting factor, i.e., these
Feature is entirely different but has extremely faint correlation again with Clinical Outcome of interest.And MRMRA is proved to make to be mentioned
The feature taken is more stable, particularly with even more so for the research of image group.Finally exported according to the MRMRA of each feature
Score shares 20 ergastic features of tool and is selected to be used to construct model." irr " R software package is used for computation attribute
Related coefficient." mRMRe " R software package is used for MRMRA feature selecting.
(3) foundation of image group label
For training set data, using deep LASSO method to 20 features of selection in Cox proportional hazard model
Carry out profound variables choice.Final choice preferably has the feature of minimum cross validation log portion likelihood.Show feature
It is confirmed as the weight of each selection feature with the nonzero coefficient of survival risk ratio.The image group label of each patient
It is generated by selected feature and the product of respective coefficient.For CECT image, training is utilized using LASSO-Cox model
Collection data are established model and are verified in verifying the set pair analysis model prediction efficiency.The predictive value of image group label passes through KM song
Line (Kaplan-Meier Curve, kaplan-Meier curve) assessment.The median point that patient in group passes through image group label
For high risk group and low-risk group.Compare KM curve by logarithm rank sum test (log-rank test).It is soft using glmnet R
Part packet carries out LASSO-Cox variables choice and model foundation.Survivorship curve is compared using R software package.
20 features of final choice are entered LASSO-Cox model.According to a scarce cross-validation method (leave-one-
Out cross-validation), that is finally included in is used to establish the arterial phase, Portal venous phase and period of delay of image group model
Image feature.Selected sign is shown in Table 2.3 phase CECT image group labels calculation formula be shown in Table 3.Single factor test Cox ratio
Example risk model establishes prediction model with arterial phase, Portal venous phase, the image group label of period of delay in 3 phase CECT respectively,
And multifactor Cox proportional hazard model then integrates 3 phase CECT whole image group labels and establishes joint forecast model.
Each phase feature selecting of table 2.CECT
Note: using a cross-validation method is lacked, feature selecting is carried out to CECT 3 phase scan image by LASSO model
Each phase image group label calculation formula of table 3.CECT
Further, the step 3 specifically includes:
(1) when clinical, pathological factor being incorporated above-mentioned 4 images group prediction model, portal venous phase images group student's federation is closed
Prediction model shows optimum prediction ability in validation group, and the more simple clinical pattern predictive ability of the model also has clear improvement
(ANOVA, p < 0.0001) (table 4).
(2) nomogram is established: the portal venous phase images group conjunctive model because being integrated with clinical, pathological factor has best
Predictive ability, therefore nomogram (Fig. 4) is established based on this model.Joint nomogram is drawn simultaneously for training group and validation group
Calibration curve (Fig. 5, Fig. 6), wherein y-axis represents practical RFS ratio；X-axis represents a possibility that prediction RFS；Diagonal dashed lines represent
The perfect prediction curve of perfect forecast nomogram).The evaluation of Hosmer-Leme Fitting optimization index shows the nomogram in training group
(p=0.791) and validation group (p=0.471) shows good consistency, without significant statistical difference.
Each model RFS prediction efficiency of table 4. compares
Simple clinical pattern prediction efficiency is very low, and simple image group model prediction model is higher, joint clinic, pathology because
The portal venous phase images group joint forecast model prediction efficiency of element is best, is specifically shown in Table 4.
The above description is only the preferred embodiment of the present invention, is not intended to limit the invention, all of the invention
Made any modifications, equivalent replacements, and improvements etc., should all be included in the protection scope of the present invention within spirit and principle.
Claims (6)
1. a kind of construction method of prediction model and nomogram for HCC recurrence and RFS, which is characterized in that described to be used for HCC
Recurrence and the prediction model of RFS and the construction method of nomogram include:
The first step is divided equally you can well imagine from 3 phase of liver CECT scan image and is taken out 647 image group features using MRMRA；
Second step confirms image group label using LASSO-Cox regression model；
Third step is combined HCC after image group label establishes the curative ablation of HCC using clinical, Pathologic factors and is recurred
And the prediction model of RFS and prediction nomogram.
2. being used for the prediction model of HCC recurrence and RFS and the construction method of nomogram as described in claim 1, which is characterized in that
The step 1 specifically includes:
Totally 647 reflection machines read Features and the image group feature of subtle texture information is interested from what is divided
It is extracted in area using software Matlab 2014a；Image filtration is carried out to initial three-dimensional tumour level using un-decimated wavelet transform transformation
Processing；The feature extracted from original image or decomposition image can be divided into two classes: textural characteristics and non-grain feature；It is non-
Textural characteristics include: shape, size and density feature；The intuitive feature of shapes and sizes feature acquisition tumour；Density feature exhibition
The histogram feature of existing lesion；Textural characteristics are based on following four texture matrix and extract, i.e. gray level co-occurrence matrixes, gray level
Long matrix, gray level band matrix and neighborhood grey scale difference matrix.
3. being used for the prediction model of HCC recurrence and RFS and the construction method of nomogram as described in claim 1, which is characterized in that
The step 2 specifically includes:
(1) clinical, pathological factor analysis, selection
Clinical, pathological factor is analyzed using single factor test Cox risk ratio regression model, and the factor of p value < 0.10 is selected, is entered
The factor of choosing is integrated and enters the multifactor Cox model of successive Regression；In multiplicity, the p value < 0.05 of variable is should
Factor may be related to the RFS of patient and be included into the foundation of model, otherwise i.e. be excluded；
(2) selection of image group feature
The stability of a certain feature is calculated and determined by test-retest method of inspection by ICC；The attribute related coefficient of a certain feature
< 0.75 will be excluded；MRMRA is used for feature selecting；Character subset is selected, which can obtain difference between two groups
Best feature finally exports score according to the MRMRA of each feature, shares 20 features and selected for constructing model；irrR
Software package is used for computation attribute related coefficient；MRMReR software package is used for MRMRA feature selecting；
(3) foundation of image group label
For training set data, 20 features of selection are carried out in Cox proportional hazard model using LASSO method profound
Variables choice；Final choice preferably has the feature of minimum cross validation log portion likelihood；Show feature and survival risk
The nonzero coefficient of ratio is confirmed as the weight of each selection feature；The image group label of each patient passes through selected
Feature and the product of respective coefficient generate；For CECT image, established using LASSO-Cox model using training set data
Model is simultaneously verified in verifying the set pair analysis model prediction efficiency；The predictive value of image group label passes through KM curve assessment；Enter
Group patient is divided into high risk group and low-risk group by the median of image group label；Compare KM song by logarithm rank sum test
Line；LASSO-Cox variables choice and model foundation are carried out using glmnet R software package；Using R software package to survivorship curve into
Row compares；
20 features of final choice are entered LASSO-Cox model；According to a scarce cross-validation method, that is finally included in is used to build
Arterial phase, Portal venous phase and the image feature of period of delay of vertical image group model；Single factor test Cox proportional hazard model is with 3 phases
The image group label of arterial phase, Portal venous phase, period of delay in CECT establishes prediction model respectively, and multifactor Cox ratio
Risk model then integrates 3 phase CECT whole image group labels and establishes joint forecast model.
4. being used for the prediction model of HCC recurrence and RFS and the construction method of nomogram as described in claim 1, which is characterized in that
The step 3 specifically includes:
(1) clinical, pathological factor is incorporated into above-mentioned 4 images group prediction model, portal venous phase images group associated prediction mould
Type shows optimum prediction ability in validation group；
(2) nomogram is established: the portal venous phase images group conjunctive model because being integrated with clinical, pathological factor has optimum prediction
Ability, therefore nomogram is established based on this model；Nomogram calibration curve is drawn simultaneously for training group and validation group.
5. a kind of use constructed as described in claim 1 for the construction method of the prediction model and nomogram of HCC recurrence and RFS
In HCC recurrence and the prediction model and nomogram of RFS.
6. a kind of model and nomogram as claimed in claim 5 for predicting HCC recurrence and RFS disappears HCC is locally curative
The application that tumor recurrence and RFS are predicted is carried out after melting treatment.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910426920.2A CN110265095A (en) | 2019-05-22 | 2019-05-22 | For HCC recurrence and construction method and the application of the prediction model and nomogram of RFS |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910426920.2A CN110265095A (en) | 2019-05-22 | 2019-05-22 | For HCC recurrence and construction method and the application of the prediction model and nomogram of RFS |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110265095A true CN110265095A (en) | 2019-09-20 |
Family
ID=67914994
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910426920.2A Pending CN110265095A (en) | 2019-05-22 | 2019-05-22 | For HCC recurrence and construction method and the application of the prediction model and nomogram of RFS |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110265095A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111311558A (en) * | 2020-02-09 | 2020-06-19 | 华中科技大学同济医学院附属协和医院 | Construction method of imaging omics model for pancreatic cancer prediction |
| CN111814868A (en) * | 2020-07-03 | 2020-10-23 | 苏州动影信息科技有限公司 | Model based on image omics feature selection, construction method and application |
| CN111862079A (en) * | 2020-07-31 | 2020-10-30 | 复旦大学附属肿瘤医院 | High-grade serous ovarian cancer recurrence risk prediction system based on imagomics |
| CN111951962A (en) * | 2020-07-27 | 2020-11-17 | 中山大学孙逸仙纪念医院 | Bevacizumab curative effect prediction model for treating radioactive brain injury and construction method thereof |
| CN112185569A (en) * | 2020-09-11 | 2021-01-05 | 中山大学孙逸仙纪念医院 | Breast cancer patient disease-free survival period prediction model and construction method thereof |
| CN112216395A (en) * | 2020-09-11 | 2021-01-12 | 中山大学孙逸仙纪念医院 | Axillary lymph node metastasis prediction model for breast cancer patient and construction method thereof |
| CN112635066A (en) * | 2020-12-02 | 2021-04-09 | 中山大学孙逸仙纪念医院 | Model for predicting inhalation pneumonia of radioactive posterior group cranial nerve paralysis patient and construction method thereof |
| CN112651507A (en) * | 2020-12-22 | 2021-04-13 | 福建医科大学附属第一医院 | Method for constructing microvascular invasion prediction model of hepatocellular carcinoma and probability prediction method |
| CN112819765A (en) * | 2021-01-25 | 2021-05-18 | 西南交通大学 | Liver image processing method |
| CN113378879A (en) * | 2021-05-06 | 2021-09-10 | 上海美杰医疗科技有限公司 | Postoperative tumor assessment method and device and computer storage medium |
| WO2023226217A1 (en) * | 2022-05-25 | 2023-11-30 | 中国医学科学院肿瘤医院 | Microsatellite instability prediction system and construction method therefor, terminal device, and medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106355023A (en) * | 2016-08-31 | 2017-01-25 | 北京数字精准医疗科技有限公司 | Open quantitative analysis method and system based on medical image |
| US20180028531A1 (en) * | 2015-02-12 | 2018-02-01 | BeyondSpring Phamaceuticals, Inc. | Use of plinabulin in combination with immune checkpoint inhibitors |
| CN108470337A (en) * | 2018-04-02 | 2018-08-31 | 江门市中心医院 | A kind of sub- reality Lung neoplasm quantitative analysis method and system based on picture depth feature |
| CN109259780A (en) * | 2018-07-24 | 2019-01-25 | 南方医科大学南方医院 | The aided assessment system and method that gastric cancer prognosis and chemotherapy based on enhancing CT images group benefit |
| 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 |
| CN109671499A (en) * | 2018-10-22 | 2019-04-23 | 南方医科大学 | A kind of rectum toxicity prediction system constituting method |
-
2019
- 2019-05-22 CN CN201910426920.2A patent/CN110265095A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180028531A1 (en) * | 2015-02-12 | 2018-02-01 | BeyondSpring Phamaceuticals, Inc. | Use of plinabulin in combination with immune checkpoint inhibitors |
| CN106355023A (en) * | 2016-08-31 | 2017-01-25 | 北京数字精准医疗科技有限公司 | Open quantitative analysis method and system based on medical image |
| CN108470337A (en) * | 2018-04-02 | 2018-08-31 | 江门市中心医院 | A kind of sub- reality Lung neoplasm quantitative analysis method and system based on picture depth feature |
| CN109259780A (en) * | 2018-07-24 | 2019-01-25 | 南方医科大学南方医院 | The aided assessment system and method that gastric cancer prognosis and chemotherapy based on enhancing CT images group benefit |
| CN109671499A (en) * | 2018-10-22 | 2019-04-23 | 南方医科大学 | A kind of rectum toxicity prediction system constituting method |
| 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 |
Non-Patent Citations (1)
| Title |
|---|
| YUAN CHUNWANG等: "Prediction early recurrence of hepatocellular carcinoma eligible for curative ablation using a Radiomics nomogram", 《CANCER IMAGING》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111311558A (en) * | 2020-02-09 | 2020-06-19 | 华中科技大学同济医学院附属协和医院 | Construction method of imaging omics model for pancreatic cancer prediction |
| CN111814868A (en) * | 2020-07-03 | 2020-10-23 | 苏州动影信息科技有限公司 | Model based on image omics feature selection, construction method and application |
| CN111951962B (en) * | 2020-07-27 | 2022-03-11 | 中山大学孙逸仙纪念医院 | Bevacizumab curative effect prediction model for treating radioactive brain injury and construction method thereof |
| CN111951962A (en) * | 2020-07-27 | 2020-11-17 | 中山大学孙逸仙纪念医院 | Bevacizumab curative effect prediction model for treating radioactive brain injury and construction method thereof |
| CN111862079A (en) * | 2020-07-31 | 2020-10-30 | 复旦大学附属肿瘤医院 | High-grade serous ovarian cancer recurrence risk prediction system based on imagomics |
| CN111862079B (en) * | 2020-07-31 | 2023-09-22 | 复旦大学附属肿瘤医院 | High-grade serous ovarian cancer recurrence risk prediction system based on image histology |
| CN112185569A (en) * | 2020-09-11 | 2021-01-05 | 中山大学孙逸仙纪念医院 | Breast cancer patient disease-free survival period prediction model and construction method thereof |
| CN112185569B (en) * | 2020-09-11 | 2022-02-25 | 中山大学孙逸仙纪念医院 | Breast cancer patient disease-free survival period prediction model and construction method thereof |
| CN112216395A (en) * | 2020-09-11 | 2021-01-12 | 中山大学孙逸仙纪念医院 | Axillary lymph node metastasis prediction model for breast cancer patient and construction method thereof |
| CN112635066A (en) * | 2020-12-02 | 2021-04-09 | 中山大学孙逸仙纪念医院 | Model for predicting inhalation pneumonia of radioactive posterior group cranial nerve paralysis patient and construction method thereof |
| CN112651507A (en) * | 2020-12-22 | 2021-04-13 | 福建医科大学附属第一医院 | Method for constructing microvascular invasion prediction model of hepatocellular carcinoma and probability prediction method |
| CN112651507B (en) * | 2020-12-22 | 2024-03-29 | 福建医科大学附属第一医院 | Construction method and probability prediction method of microvascular invasion prediction model of hepatocellular carcinoma |
| CN112819765A (en) * | 2021-01-25 | 2021-05-18 | 西南交通大学 | Liver image processing method |
| CN113378879A (en) * | 2021-05-06 | 2021-09-10 | 上海美杰医疗科技有限公司 | Postoperative tumor assessment method and device and computer storage medium |
| WO2023226217A1 (en) * | 2022-05-25 | 2023-11-30 | 中国医学科学院肿瘤医院 | Microsatellite instability prediction system and construction method therefor, terminal device, and medium |
| US12027255B2 (en) | 2022-05-25 | 2024-07-02 | Cancer Hospital, Chinese Academy Of Medical Sciences | System for predicting microsatellite instability and construction method thereof, terminal device and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110265095A (en) | For HCC recurrence and construction method and the application of the prediction model and nomogram of RFS | |
| Sharma et al. | Dermatologist-level classification of skin cancer using cascaded ensembling of convolutional neural network and handcrafted features based deep neural network | |
| Sharma et al. | [Retracted] Enhanced Watershed Segmentation Algorithm‐Based Modified ResNet50 Model for Brain Tumor Detection | |
| Shrivastava et al. | Bone cancer detection using machine learning techniques | |
| Fang et al. | Brain tumor segmentation based on the dual-path network of multi-modal MRI images | |
| Xu et al. | PA‐ResSeg: A phase attention residual network for liver tumor segmentation from multiphase CT images | |
| Akter et al. | Lung cancer detection using enhanced segmentation accuracy | |
| de Sousa Costa et al. | Classification of malignant and benign lung nodules using taxonomic diversity index and phylogenetic distance | |
| Abdelhafeez et al. | Skin cancer detection using neutrosophic c-means and fuzzy c-means clustering algorithms | |
| Mittapalli et al. | Multiscale CNN with compound fusions for false positive reduction in lung nodule detection | |
| Dai et al. | TD-Net: Trans-Deformer network for automatic pancreas segmentation | |
| Avinash et al. | Early detection of malignant tumor in lungs using feed-forward neural network and k-nearest neighbor classifier | |
| Maaref et al. | Predicting the response to FOLFOX-based chemotherapy regimen from untreated liver metastases on baseline CT: a deep neural network approach | |
| Zhou et al. | [Retracted] Ultrasonic Intelligent Diagnosis of Papillary Thyroid Carcinoma Based on Machine Learning | |
| Liu et al. | Automated classification of cervical Lymph-Node-Level from ultrasound using depthwise separable convolutional swin transformer | |
| Hou et al. | Mass segmentation for whole mammograms via attentive multi-task learning framework | |
| Jain et al. | A deep learning-based feature extraction model for classification brain tumor | |
| Levenson et al. | Advancing precision medicine: algebraic topology and differential geometry in radiology and computational pathology | |
| Saumiya et al. | Unified automated deep learning framework for segmentation and classification of liver tumors | |
| CN111027548A (en) | Brain glioma molecular marker nondestructive prediction method and system based on image omics | |
| Li et al. | A dual attention-guided 3D convolution network for automatic segmentation of prostate and tumor | |
| Bhalerao et al. | Cuckoo search based multi-objective algorithm with decomposition for detection of masses in mammogram images | |
| Sobecki et al. | MRI imaging texture features in prostate lesions classification | |
| Wang et al. | Automated machine-learning framework integrating histopathological and radiological information for predicting IDH1 mutation status in glioma | |
| Elazab et al. | Computer-aided diagnosis system for grading brain tumor using histopathology images based on color and texture features |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190920 |