CN103632162B - Disease-related electrocardiogram feature selection method - Google Patents
Disease-related electrocardiogram feature selection method Download PDFInfo
- Publication number
- CN103632162B CN103632162B CN201310403000.1A CN201310403000A CN103632162B CN 103632162 B CN103632162 B CN 103632162B CN 201310403000 A CN201310403000 A CN 201310403000A CN 103632162 B CN103632162 B CN 103632162B
- Authority
- CN
- China
- Prior art keywords
- class
- feature
- positive
- grader
- negative
- 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.)
- Active
Links
Abstract
An electrocardiogram feature selection method provided by the invention divides electrocardiograms into four classification systems N, S, V and F and decomposes the four classification systems N, S, V and F into six secondary classifiers NvS, NvV, NvF, SvV, SvF and VvF. In each of the secondary classifiers, features are sorted according to scores from high to low to form a candidate feature set, and then the optimal feature subset is selected from each secondary classifier and a to-be-predicated electrocardiogram sample is predicated based on the optimal feature subsets to obtain the category of the to-be-predicated electrocardiogram sample. According to the electrocardiogram feature selection method provided by the invention, features are sorted according to scores from high to low in each secondary classifier to form a candidate feature set, and then the optimal feature subset is selected from each secondary classifier and the to-be-predicated electrocardiogram sample is predicated based on the optimal feature subsets to obtain the category of the to-be-predicated electrocardiogram sample, thus improving the accuracy of prediction.
Description
【Technical field】
The present invention relates to ECG signal sampling technical field, more particularly to the Characteristics of electrocardiogram selecting party that a kind of disease is related
Method.
【Background technology】
Electrocardioscopy is diagnosis arrhythmia cordis, a kind of effective method of myocardial ischemia, and the method has hurtless measure, low
The advantage of cost, has larger portfolio in hospital, especially in the mechanisms such as MEC, remote medical consultation with specialists center, full-time electrocardiogram
Doctor it is daily need interpretation substantial amounts of Electrocardiographic, be the work load for mitigating doctor, computer assisted electrocardio in recent years
Scheme automatic classifying and identifying system to be increasingly taken seriously.
The automatic classifying and identifying system of electrocardiogram of complete set generally comprises following process:Data acquisition, data prediction,
Feature extraction, classifier training/prediction.Due to Electrocardiographic species it is various, for unified and specification electrocardiogram automatic identification system
The interpretational criteria of system, Used In The Regulation of Medical Device In Usa promotes association(Association for the Advancement of Medical
Instrumentation;AAMI)It is five classes by Electrocardiographic category division:(1)N, normal ECG and block class electrocardio
Figure;(S)S, supraventricular exception;(3)V, room sexual abnormality;(4)F, between room sexual abnormality and it is normal between;(5)Q, it is impossible to clearly
Distinguish.In actual classification, because Q classes do not have obvious statistical law, generally just for this four analogous-design patterns classification of NSVF
Device.Can the grader with good generalization ability that a stalwartness be trained directly affect follow-up accuracy of identification, and one
It is the premise for training grader that set can characterize the feature set of various disease.
【The content of the invention】
It is an object of the invention to provide a kind of related Characteristics of electrocardiogram system of selection of disease, the method can be from numerous
Characteristics of electrocardiogram in select those features that can improve classification and recognition, reject redundancy feature, improve Classification and Identification
Precision.
For achieving the above object, the present invention adopts following technical proposals:
A kind of related Characteristics of electrocardiogram system of selection of disease, comprises the steps:
Step S110:According to AAMI evaluation criterions, electrocardiogram is divided into into the class categorizing systems of NSVF tetra-;
Step S120:Based on the rule of OvO, the class categorizing systems of the NSVF tetra- are decomposed into into NvS, NvV, NvF, SvV,
Six two graders of SvF, VvF;
Step S130:In above-mentioned each two grader, each feature is ranked up by score height;
Step S140:All character subsets after above-mentioned score sorts from high to low form candidate feature set;
Step S150:Above-mentioned each two grader are trained using SVM classifier, and selects optimum special from each two grader
Levy subset;And
Step S160:The optimal feature subset is treated survey ecg samples and is predicted, and obtains the electrocardiogram to be measured
The classification of sample.
In the present embodiment, in step S130, in above-mentioned each two grader, each feature is carried out by score height
Sequence, specifically, in above-mentioned each two grader, each feature is scored using following formula, and it is high according to score
Low to be ranked up, the formula is:
Wherein, n+For the sample number of positive class, n-To bear the sample number of class,WithI-th positive sample is represented respectively
With k-th feature of j-th negative sample,Represent positive sample k-th feature it is average
Value,The mean value of k-th feature of negative sample is represented,It is two
The mean value of k-th feature of class, F (k) be it is calculated be k-th feature score.
In the present embodiment, in step S150, using SVM classifier above-mentioned each two grader are trained, and from each two
Grader selects optimal feature subset, specifically, with precision as the index of selection optimal feature subset, the precision highest is special
It is optimal feature subset to levy subset, and the precision is designated as Accuracy,
Wherein, TP is the sample number that positive class is correctly divided into positive class, and TN is the quantity that negative class is correctly divided into negative class, and FN is
Positive class is divided into the quantity of negative class by mistake, and FP is the quantity that negative class is divided into positive class by mistake.
In this embodiment, in step S150, above-mentioned each two grader are trained using SVM classifier, and from each two points
Class device selects optimal feature subset, specifically, with the geometrical mean of positive sample sensitivity and negative sample sensitivity to select most
The index of excellent character subset, the maximum character subset of the geometrical mean is optimal feature subset, the positive sample sensitivity
Se (Pos) is designated as, the negative sample sensitivity is designated as Se (Neg), and the geometrical mean is designated as g-mean,
Wherein, TP is the sample number that positive class is correctly divided into positive class, and TN is the quantity that negative class is correctly divided into negative class, and FN is
Positive class is divided into the quantity of negative class by mistake, and FP is the quantity that negative class is divided into positive class by mistake.
Above-mentioned technical proposal is adopted, the beneficial effects of the present invention is:
The Characteristics of electrocardiogram system of selection that the above embodiment of the present invention is provided, by electrocardiogram the classes of NSVF tetra- classification system is divided into
System, and the class categorizing systems of the NSVF tetra- are decomposed into into NvS, six two graders of NvV, NvF, SvV, SvF, VvF, above-mentioned every
In individual two grader, candidate feature set is ranked up and is formed by score height to each feature, then from each two grader
Optimal feature subset is selected, survey ecg samples is treated according to the optimal feature subset and is predicted, obtain described treating thought-read
The classification of electrograph sample.The Characteristics of electrocardiogram system of selection that the present invention is provided, forms feature after feature score is sorted from high to low
Subset, and optimal feature subset is selected from each two grader, using optimal feature subset treat survey ecg samples carry out it is pre-
Survey, the classification of electrocardio pattern to be measured is obtained, so as to improve precision of prediction.
Additionally, the Characteristics of electrocardiogram system of selection that the above embodiment of the present invention is provided, devises the side of a set of feature scoring
Method, calculates percentage contribution of each feature to classification;Meanwhile, feature is ranked up by score height, and using each two points
Class device selects optimal feature subset in character subset, improves the sensitivity of classification.
【Description of the drawings】
The step of Fig. 1 is Characteristics of electrocardiogram system of selection provided in an embodiment of the present invention flow chart.
【Specific embodiment】
In order that the objects, technical solutions and advantages of the present invention become more apparent, below in conjunction with accompanying drawing and it is embodied as
Example, the present invention will be described in further detail.It should be appreciated that specific embodiment described herein is only to explain this
It is bright, it is not intended to limit the present invention.
Fig. 1 is referred to, Fig. 1 is flow process Figure 100 the step of Characteristics of electrocardiogram system of selection provided in an embodiment of the present invention, from
Visible in Fig. 1, Characteristics of electrocardiogram system of selection 100 comprises the steps:
Step S110:According to AAMI evaluation criterions, electrocardiogram is divided into into the class categorizing systems of NSVF tetra-;
Step S120:Based on the rule of OvO, the class categorizing systems of NSVF tetra- are decomposed into into NvS, NvV, NvF, SvV, SvF,
Six two graders of VvF;
Common sorting algorithm is both for two class classification problems, and the multi-class problem in practical application is usually used a pair
One method(one-versus-one;OvO)Multiple two graders are assembled into into multi-categorizer;It is appreciated that the present invention is according to AAMI
Evaluation criterion, electrocardiogram is divided into into the class categorizing systems of NSVF tetra-, NvS, NvV, NvF, SvV, SvF, VvF can be broken down into
Six two graders.
Step S130:In above-mentioned each two grader, each feature is ranked up by score height;
In the preferred embodiment that the present invention is provided, each feature is scored using following formula, and according to
Height is divided to be ranked up;
Wherein, n+For the sample number of positive class, n-To bear the sample number of class,WithI-th positive sample is represented respectively
With k-th feature of j-th negative sample,Represent positive sample k-th feature it is average
Value,The mean value of k-th feature of the sample number of negative class is represented,
The mean value of k-th feature of two classes, F (k) be it is calculated be k-th feature score.
It is appreciated that above-mentioned formula is represented, if distance is more remote between positive sample and negative sample, positive sample and negative sample
This each internal dispersion degree is less, then this feature is easier causes positive and negative two class separate, and the score of feature is also higher.This
The scoring model for planting feature is that disease category is related, and same feature possible score in different disease contrast groups is different,
That is the scoring of feature is that a certain class disease is different from for another kind of disease.
Step S140:All character subsets after above-mentioned score sorts from high to low form candidate feature set;
Specifically, all features after score is sorted from high to low are added to one by one in order the feature set currently chosen
In, to form candidate feature set.
Step S150:Above-mentioned each two grader are trained using SVM classifier, and selects optimum special from each two grader
Levy subset;
In the preferred embodiment that the present invention is provided, using SVM classifier above-mentioned each two grader are trained, and from every
Individual two grader selects optimal feature subset, specifically adopts the index with precision as selection optimal feature subset, precision highest
Character subset is optimal feature subset, and precision is designated as Accuracy,
Wherein, TP is the sample number that positive class is correctly divided into positive class (True Positive), and TN is correctly divided into for negative class
The quantity of negative class (True Negative), FN is the quantity that positive class is divided into negative class (False Negtative) by mistake, and FP is
Negative class is divided into the quantity of positive class (False Positive) by mistake.
In another preferred embodiment that the present invention is provided, using SVM classifier above-mentioned each two grader are trained, and from
Each two grader selects optimal feature subset, specifically adopts with the geometrical mean of positive sample sensitivity and negative sample sensitivity
To select the index of optimal feature subset, the maximum character subset of the geometrical mean is optimal feature subset, the positive sample
This sensitivity is designated as Se (Pos), and the negative sample sensitivity is designated as Se (Neg), and the geometrical mean is designated as g-mean,
Wherein, TP is the sample number that positive class is correctly divided into positive class (True Positive), and TN is correctly divided into for negative class
The quantity of negative class (True Negative), FN is the quantity that positive class is divided into negative class (False Negtative) by mistake, and FP is
Negative class is divided into the quantity of positive class (False Positive) by mistake.
It is appreciated that can be service precision for positive sample and the class training samples number of negative sample two are classified in a balanced way
Index, and classification unbalanced for quantity uses geometrical mean index.
Step S160:Optimal feature subset is treated survey ecg samples and is predicted, and obtains the classification of electrocardio pattern to be measured.
Specifically, Jing after step S150, each two grader is determined after optimal feature subset, and each two grader uses each
From the optimal feature subset test sample unknown to classification be predicted, the decision-making by way of ballot, many classifications of gained vote
For the classification of prediction.
The Characteristics of electrocardiogram system of selection that the above embodiment of the present invention is provided, by electrocardiogram the classes of NSVF tetra- classification system is divided into
System, and the class categorizing systems of NSVF tetra- are decomposed into into NvS, six two graders of NvV, NvF, SvV, SvF, VvF, it is above-mentioned each two
In grader, candidate feature set is ranked up and is formed by score height to each feature, then select from each two grader
Optimal feature subset, treats survey ecg samples and is predicted according to optimal feature subset, obtains the classification of electrocardio pattern to be measured.
The Characteristics of electrocardiogram system of selection that the present invention is provided, forms character subset after feature score is sorted from high to low, and from each
Two graders select optimal feature subset, treat survey ecg samples using optimal feature subset and are predicted, and obtain treating thought-read
The classification of electrograph sample, improves precision of prediction.
The present invention is expanded on further below by way of specific embodiment, these embodiments are only used for the purpose for illustrating, and
The scope of the present invention is not limited.
Embodiment one
By taking the arrhythmia cordis database of MIT-BIH as an example, the validity of this method is verified.Whole data set has 48 30 points
The lead electrocardiogram (ECG) datas of Holter bis- of clock or so, wherein 4 records are Electrocardiographic, the independent places of needs that placed pacemaker
Reason, remaining 44 are used for the automatic class test of electrocardiogram.
According to AAMI evaluation criterions, above-mentioned two leads electrocardiogram (ECG) data is divided into into the class categorizing systems of NSVF tetra-, refers to the institute of table 1
Show:
Table 1 is MIT-BIH database heart umber of beats amounts
DS1:101、106、108、109、112、114、115、116、118、119、122、124、201、203、205、207、
208、209、215、220、223、230;
DS2:100、103、105、111、113、117、121、123、200、202、210、212、213、214、219、221、
222、228、231、232、233、234
44 records are averagely divided into into DS1 and DS2, wherein, DS1 is used for train classification models, and DS2 gives birth to as test set
Into test report, contrast without feature selecting and the classification difference for having feature selecting.
, according to the rule of OvO, NvS, NvV, NvF, SvV, SvF, VvF six will be decomposed into through the class categorizing systems of NSVF tetra-
Individual two grader;
In test, we in table 2 using listing based on the feature of a phase, the feature based on form and based on area
Feature is used as candidate feature.
Table 2 is candidate feature list
Wherein, 22 records in DS1 carry out the cross validation of 22 foldings using leaving-one method in the training process, i.e., often folding is instructed
When practicing, wherein one is recorded as checking collection, remaining 21 is training set.The confusion matrix that 22 foldings collect be evaluation of classification according to
According to, it is contemplated that the imbalance of categorical measure, we use g-mean as the performance indications of classification, high by score to each feature
It is low to be ranked up, candidate feature set is formed, each two grader is trained using SVM classifier, and feature selecting is carried out, it is determined that
After optimized parameter and optimal feature subset, the g-mean indexs on checking collection are as shown in table 3:
The training result of table 3
To obtain the model of optimum g-mean indexs as optimal models, test set is carried out into the classification results after feature selecting
As shown in table 4:
The test result selected using feature selecting on the test set of table 4
In order to verify the test effect of feature selecting, table 5 has listed file names with the survey of the grader for not using feature selecting
Test result:
The test result of feature selecting is not used on the test set of table 5
With reference to table 4 and the comparing result of table 5, it can be seen that after using feature selecting choosing, per class, the sensitivity of classification is obvious
Improve.
The above, is only presently preferred embodiments of the present invention, and any pro forma restriction is not made to the present invention, though
So the present invention is disclosed above with preferred embodiment, but is not limited to the present invention, any to be familiar with this professional technology people
Member, in the range of without departing from technical solution of the present invention, when making a little change or modification using the technology contents of the disclosure above
For the Equivalent embodiments of equivalent variations, as long as being without departing from technical solution of the present invention content, according to the technical spirit pair of the present invention
Any simple modification, equivalent variations and modification that above example is made, still fall within the range of technical solution of the present invention.
Claims (3)
1. the related Characteristics of electrocardiogram system of selection of a kind of disease, it is characterised in that comprise the steps:
Step S110:According to AAMI evaluation criterions, electrocardiogram is divided into into the class categorizing systems of NSVF tetra-;
Step S120:Based on the rule of OvO, the class categorizing systems of the NSVF tetra- are decomposed into into NvS, NvV, NvF, SvV, SvF,
Six two graders of VvF;
Step S130:In above-mentioned each two grader, each feature is ranked up by score height;
Step S140:All character subsets after above-mentioned score sorts from high to low form candidate feature set;
Step S150:Above-mentioned each two grader are trained using SVM classifier, and optimal characteristics is selected from each two grader
Collection;And
Step S160:The optimal feature subset is treated survey ecg samples and is predicted, and obtains the ecg samples to be measured
Classification;
In step S130, in above-mentioned each two grader, each feature is ranked up by score height, specifically, upper
In stating each two grader, each feature is scored using following formula, and be ranked up according to score height, the public affairs
Formula is:
Wherein, n+For the sample number of positive class, n-To bear the sample number of class,WithI-th positive sample and jth are represented respectively
K-th feature of individual negative sample,The mean value of k-th feature of positive sample is represented,The mean value of k-th feature of negative sample is represented,It is two classes
The mean value of k-th feature, F (k) be it is calculated be k-th feature score.
2. Characteristics of electrocardiogram system of selection according to claim 1, it is characterised in that in step S150, using svm classifier
Device trains above-mentioned each two grader, and selects optimal feature subset from each two grader, specifically, with precision to select most
The index of excellent character subset, the precision highest character subset is optimal feature subset, and the precision is designated as Accuracy,
Wherein, TP is the sample number that positive class is correctly divided into positive class, and TN is the quantity that negative class is correctly divided into negative class, and FN is positive class
The quantity of negative class is divided into by mistake, FP is the quantity that negative class is divided into positive class by mistake.
3. Characteristics of electrocardiogram system of selection according to claim 1, it is characterised in that in step S150, using svm classifier
Device trains above-mentioned each two grader, and selects optimal feature subset from each two grader, specifically, with positive sample sensitivity
With the index that the geometrical mean of negative sample sensitivity is selection optimal feature subset, maximum feature of the geometrical mean
Integrate as optimal feature subset, the positive sample sensitivity is designated as Se (Pos), and the negative sample sensitivity is designated as Se (Neg), described
Geometrical mean is designated as g-mean,
Wherein, TP is the sample number that positive class is correctly divided into positive class, and TN is the quantity that negative class is correctly divided into negative class, and FN is positive class
The quantity of negative class is divided into by mistake, FP is the quantity that negative class is divided into positive class by mistake.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310403000.1A CN103632162B (en) | 2013-09-06 | 2013-09-06 | Disease-related electrocardiogram feature selection method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310403000.1A CN103632162B (en) | 2013-09-06 | 2013-09-06 | Disease-related electrocardiogram feature selection method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103632162A CN103632162A (en) | 2014-03-12 |
CN103632162B true CN103632162B (en) | 2017-05-03 |
Family
ID=50213189
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310403000.1A Active CN103632162B (en) | 2013-09-06 | 2013-09-06 | Disease-related electrocardiogram feature selection method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103632162B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104783782B (en) * | 2015-04-13 | 2017-09-01 | 深圳市飞马与星月科技研究有限公司 | Electrocardiosignal automatic testing method and device |
CN107451417B (en) * | 2017-09-08 | 2021-07-06 | 北京蓬阳丰业科技有限公司 | Dynamic electrocardiogram analysis intelligent diagnosis system and method |
CN108030494B (en) * | 2017-11-08 | 2019-11-15 | 华南理工大学 | Electrocardiosignal error flag training sample recognition methods based on cross validation |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6584456B1 (en) * | 2000-06-19 | 2003-06-24 | International Business Machines Corporation | Model selection in machine learning with applications to document clustering |
CN101853394A (en) * | 2010-05-14 | 2010-10-06 | 中国农业大学 | Image feature optimization selection method and system based on improved ant colony optimization |
CN101937440A (en) * | 2009-06-30 | 2011-01-05 | 华为技术有限公司 | Feature selection method and device |
CN101968851A (en) * | 2010-09-09 | 2011-02-09 | 西安电子科技大学 | Medical image processing method based on dictionary studying upsampling |
CN102819738A (en) * | 2012-07-24 | 2012-12-12 | 东南大学 | License plate identification system based on OpenMP multithreading framework |
CN102930286A (en) * | 2012-09-18 | 2013-02-13 | 重庆大学 | Image-based early diagnosis system for senile dementia |
-
2013
- 2013-09-06 CN CN201310403000.1A patent/CN103632162B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6584456B1 (en) * | 2000-06-19 | 2003-06-24 | International Business Machines Corporation | Model selection in machine learning with applications to document clustering |
CN101937440A (en) * | 2009-06-30 | 2011-01-05 | 华为技术有限公司 | Feature selection method and device |
CN101853394A (en) * | 2010-05-14 | 2010-10-06 | 中国农业大学 | Image feature optimization selection method and system based on improved ant colony optimization |
CN101968851A (en) * | 2010-09-09 | 2011-02-09 | 西安电子科技大学 | Medical image processing method based on dictionary studying upsampling |
CN102819738A (en) * | 2012-07-24 | 2012-12-12 | 东南大学 | License plate identification system based on OpenMP multithreading framework |
CN102930286A (en) * | 2012-09-18 | 2013-02-13 | 重庆大学 | Image-based early diagnosis system for senile dementia |
Non-Patent Citations (2)
Title |
---|
ROC曲线及AUC评价指标;它它;《http://lansesky23.blog.163.com/blog/static/3572462720135189549672/》;20130618;第1-4页 * |
SVM多类分类---多个二值分类combine;Yode;《http://blog.sina.com.cn/s/blog_4c98b96001009b8d.html》;20080414;全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN103632162A (en) | 2014-03-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111444747B (en) | Epileptic state identification method based on migration learning and cavity convolution | |
Hattie et al. | Procedures for assessing the validities of tests using the" known-groups" method | |
CN109785976A (en) | A kind of goat based on Soft-Voting forecasting system by stages | |
Grajski et al. | Classification of EEG spatial patterns with a tree-structured methodology: CART | |
CN108511055B (en) | Ventricular premature beat recognition system and method based on classifier fusion and diagnosis rules | |
CN104887224A (en) | Epileptic feature extraction and automatic identification method based on electroencephalogram signal | |
CN109934089A (en) | Multistage epileptic EEG Signal automatic identifying method based on supervision gradient lifter | |
CN106778865B (en) | A kind of semi-supervised EEG signals sleep stage method under multi-domain characteristics | |
CN109846472A (en) | Beat classification method based on BiLSTM-Attention deep neural network | |
CN107463766A (en) | Generation method, device and the computer-readable recording medium of blood glucose prediction model | |
Andersson et al. | Comparison of clinicians and an artificial neural network regarding accuracy and certainty in performance of visual field assessment for the diagnosis of glaucoma | |
CN104615894A (en) | Traditional Chinese medicine diagnosis method and system based on k-nearest neighbor labeled specific weight characteristics | |
CN107095671A (en) | Driving fatigue recognition methods and system | |
CN108304887A (en) | Naive Bayesian data processing system and method based on the synthesis of minority class sample | |
CN108090509A (en) | A kind of adaptive electrocardiogram sorting technique of data length | |
CN103632162B (en) | Disease-related electrocardiogram feature selection method | |
CN107273841A (en) | A kind of electric sensibility classification method of the brain based on EMD and gaussian kernel function SVM | |
Jones et al. | Developing an empirically based typology of delinquent youths | |
CN113349780A (en) | Method for evaluating influence of emotional design on online learning cognitive load | |
CN106601271A (en) | Voice abnormal signal detection system | |
CN113749619A (en) | Mental fatigue assessment method based on K-TRCA | |
CN108078563A (en) | A kind of EEG signal analysis method of integrated classifier | |
Fridel et al. | The quantitative study of serial murder: Regression is not transgression | |
Kleinmuntz | Sign and seer: Another example. | |
CN110010246A (en) | A kind of disease Intelligent Diagnosis Technology based on neural network and confidence interval |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |