CN100529731C - Drug distinguishing near infrared spectrum analysis method and apparatus - Google Patents
Drug distinguishing near infrared spectrum analysis method and apparatus Download PDFInfo
- Publication number
- CN100529731C CN100529731C CNB2006100671849A CN200610067184A CN100529731C CN 100529731 C CN100529731 C CN 100529731C CN B2006100671849 A CNB2006100671849 A CN B2006100671849A CN 200610067184 A CN200610067184 A CN 200610067184A CN 100529731 C CN100529731 C CN 100529731C
- Authority
- CN
- China
- Prior art keywords
- cognition
- model
- modeling sample
- near infrared
- sample
- 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
Images
Landscapes
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The present invention relates to drug distinguishing near infrared spectrum analysis method and apparatus, and is especially non-destructive distinguishing method and apparatus of judging the identity between the drug and marked name by means of combined near infrared spectrum analysis method and chemometric method.
Description
Technical field
The present invention relates to a kind of method and apparatus that utilizes near-infrared spectral analytical method identification medicine.In particular to a kind ofly use near-infrared spectral analysis technology, in conjunction with chemometrics method for the whether consistent method and apparatus that can't harm identification of medicine with its sign title.
Background technology
The sale of counterfeit drug, substandard drug and the problem of bringing thus are global.The common form of counterfeit drug mainly shows as with non-medicine pretends to be medicine, or pretends to be the high price medicine with the low value medicine; Substandard drug then mainly shows as the requirement that indexs such as effective ingredient content do not reach drug standard.
Existing pharmacopeia advises with methods such as thin-layered chromatography (TLC), liquid phase chromatography, vapor-phase chromatography, infra-red sepectrometry, mass spectroscopy, color reactions medicine being differentiated usually.The advantage of these methods is to have obtained general approval as official method, and shortcoming then is that the discriminating of being carried out need spend the long time.For example, used methods such as comprising thin-layer chromatography (TLC), color reaction in the medicine quick testing kit that The World Health Organization (WHO) is recommended, but because the relatively poor said method utilization rate that causes of detection effect is not high.Except that detecting this deficiency of weak effect, above method also all exists to be needed medicine to be measured is carried out the destructive defective of handling.At present, the medication management department of many countries and drug manufacture producer are devoted to ban vacation, substandard drug thing, but their work rests on substantially in the identification false, that the substandard drug thing is packed.
Over past ten years, the development of near-infrared spectrum technique, Chemical Measurement and computer software technology, and the combination of these technology, make near-infrared spectral analysis technology rapid at the application development of aspects such as drug quality analysis, detection, and quick, easy and simple to handle and characteristics such as test sample is harmless are come into one's own day by day with its mensuration.Also there are some drug manufacture producers this technology to be used for the quality testing of medicine.
When setting up corresponding discriminating model, be used for the existing near-infrared spectral analysis technology of medicine identification, need usually to consider as of the influence of factors such as granularity, auxiliary material, production technology, temperature, humidity to a certain concrete medicine.The model of cognition of Jian Liing is merely able to realize the evaluation to the single source medicine in view of the above.Therefore this class technology mainly is used for the online quality monitoring of a certain specific product by drug manufacture producer.
Summary of the invention
After deliberation, we find, use the qualitative model of cognition of single near-infrared spectrum analysis, a certain kind near infrared spectrogram differs greatly in meeting appearance medicine to be identified and the model of cognition modeling sample, but but differentiated and be the situation of same substance, this has seriously limited the accuracy of near infrared quantitative model, and will address this problem, and often needs to spend a large amount of human and material resources.
For addressing this problem, through thinking and research in many ways repeatedly, the inventor is surprised to find, unite and use two near-infrared spectrum analysis qualutative models, for example, near-infrared spectrum analysis model of cognition I of the present invention (hereinafter to be referred as " model of cognition I ") and near-infrared spectrum analysis model of cognition II of the present invention (hereinafter to be referred as " model of cognition II "), whether can accurately differentiate sample to be identified consistent with the medicine of its sign, so solved effectively to false, the substandard drug thing carries out accurately, fast, the harmless problem of discerning.
Therefore, the invention provides a kind of method of using near infrared spectroscopic method identification medicine, more particularly, the invention provides a kind of near infrared spectroscopic method that uses and determine the method that sample to be identified is whether consistent with its medicine that is indicated, promptly, a kind of near infrared spectroscopic method that uses determines whether sample to be identified contains the method for its reactive compound that indicates, and this method may further comprise the steps:
A. collect modeling sample;
B. set up and adjust described model of cognition I according to described modeling sample;
C. set up and adjust described model of cognition II according to described modeling sample;
D. use described model of cognition I and II to discern respectively for described sample to be identified;
The recognition result of e. more above-mentioned two kinds of model of cognition is determined described sample to be identified.
When the recognition result of above-mentioned two kinds of model of cognition is identical, and when identical, judge that described sample to be identified is medicine that it indicated with described sample marker title to be identified.
When the recognition result of above-mentioned two kinds of model of cognition not simultaneously, judge that described sample to be identified is different with the medicine that it is indicated.
Modeling sample of the present invention should be the marketed drugs that meets the pharmacopeia code requirement, indicate the identical medicine of chemical constitution of the reactive compound that contains comprising: its contained reactive compound and sample to be identified, with and the different medicine of chemical constitution of the reactive compound that contains of contained reactive compound and sample marker to be identified.
Preferably, the medicine that chemical constitution is identical or structure is close of the reactive compound that contains for its contained reactive compound and sample marker to be identified of described modeling sample.For example, cefalexin and cefadroxil, they are regarded as the close compound of structure in the present invention.
Modeling sample of the present invention preferably has identical technology characteristics and packaged form with described sample to be identified.Wherein, the present invention's alleged " identical technology characteristics " is meant that pharmaceutical dosage form is identical.Identical packaged form is meant that the medicine external packing is identical.For example, when described sample to be identified when being aluminum-plastic packaged, the modeling sample identical with its external packing is aluminum-plastic packaged.
The preferred packaged form of method of the present invention is aluminum-plastic packaged.
The formulation that method of the present invention preferably is suitable for is the constant compound preparation of tablet, capsule, powder ampoule agent for injection, parenteral solution, ointment, supensoid agent, sugar coated tablet or formula rate; More preferably be applicable to the identification and the affirmation of tablet, powder ampoule agent for injection and capsule medicine.
When described modeling sample or described sample to be identified are sugar coated tablet, should remove its sugar-coat before gathering its near infrared spectrum data.
But the concentration of contained reactive compound all should be in the near infrared spectroscopic method sensing range in modeling sample of the present invention and the sample to be identified.
Model of cognition I of the present invention should be able to finish for described each modeling sample is interracial and be distinguished from each other, discerns.When needed, described model of cognition I is distinguished from each other, discerns by the multilayer recognition method and finish for described modeling sample is interracial.
The alleged multilayer identification of the present invention is meant, the near-infrared spectrum analysis model of cognition of on a certain near infrared spectrum spectral coverage, setting up, can not finish when it and to discern each other, to distinguish all modeling samples are interracial, and can only make wherein a part discern each other, when distinguishing, can select other near infrared spectrum spectral coverage to set up a model of cognition again, with that part of modeling sample kind of distinguishing, the last model of identification can not be distinguished, discern.Such process can repeat repeatedly, until all modeling sample kinds all can being distinguished from each other, discerning.
When using described multilayer recognition method, model of cognition I of the present invention is included as and finishes a plurality of model of cognition that aforementioned multilayer identification is set up.
Model of cognition I of the present invention is 100% for the interracial differentiation of described modeling sample, recognition accuracy.Differentiation, recognition accuracy for sample to be identified are not less than 90%, preferably are not less than 95%.
Set up in the prior art near infrared spectroscopic method with the method for adjusting the near infrared qualitative analysis model and all can be used for setting up and adjusting model of cognition I of the present invention.
Be preferably based on following principle and set up model of cognition I of the present invention:
At first use principal component analysis (PCA) (Principal Component Analysis, hereinafter to be referred as the PCA method) carry out dimension-reduction treatment for the near infrared spectrum data of the modeling sample that records, utilize the calculating principle to be pattern-recongnition method (Pattern Recognition) then, the different pieces of information processing mode in the near-infrared spectrum analysis software of preferred PCA techniques of discriminant analysis (PCA Discriminant Analysis) is set up model of cognition I of the present invention.
Principal component analysis (PCA) of the present invention is meant, former variable is changed, and makes the less new variables of number become the method for the linear combination of former variable, and wherein, new variables should characterize the data characteristics of former variable to greatest extent, and drop-out not.
The alleged pattern-recongnition method of the present invention is a kind of multivariate analysis method, is mainly used in the discriminant classification of sample.Its discloses is rule and implicit character inside the matters, a kind of complex art of finishing by mathematical method and computer technology.Pattern-recongnition method to characterized things or phenomenon various forms of (numerical value, literal with logical relation) information handles and analyzes, being the process that things or phenomenon are described, recognize, classify and explain, is the important component part of information science and artificial intelligence.
PCA techniques of discriminant analysis of the present invention is a kind of of pattern-recongnition method, is after utilizing described PCA method to the data dimension-reduction treatment, sample is carried out the method for discriminant classification.
Model of cognition I of the present invention is an index with described modeling sample near infrared collection of illustrative plates to the distance between the average near infared spectrum of this modeling sample kind, carries out cluster analysis (ClusterAnalysis), thereby determines the threshold value of described modeling sample.
The present invention preferably uses Euclidean distance (Euclidian Distance) to carry out cluster analysis as basis of classification.
The alleged cluster analysis of the present invention is meant, according to correlativity between the near infrared collection of illustrative plates or similarity described collection of illustrative plates is sorted out, and similar collection of illustrative plates is classified as a class, and a kind of multivariate analysis method that makes a distinction with the big collection of illustrative plates of difference.
Near-infrared spectrum analysis software of the present invention comprises the subsidiary at random qualitative analysis software of near infrared spectrometer that German BRUKER company produces.
Preferably set up model of cognition I of the present invention by following step:
Collection is also extracted the relevant near infrared light spectrum information of representing its contained reactive compound in described each modeling sample kind near infrared spectrum, and according to these information, the described model of cognition I of described modeling sample can be distinguished, be discerned to foundation.
When setting up model of cognition I of the present invention, described modeling sample preferably contains the medicine of 3 or 3 above different cultivars; 5 or 5 are more preferred to raise variety; Each types of drugs is preferably collected the product of 3 or 3 above producers.
Model of cognition II of the present invention sets up by following steps:
A. near infrared spectrum data collection
Use near infrared spectrometer to measure, gather the near infrared light spectrum information of described each modeling sample kind respectively;
B. set up and the described model of cognition II of adjustment:
At each kind of modeling sample, different data processing methods is set up a plurality of near-infrared analysis rudimentary models in the near-infrared spectrum analysis software of use prior art;
Select in these rudimentary models for the rudimentary model of the powerhouse of each modeling sample kind recognition capability as corresponding model of cognition II;
Set up and adjust the rudimentary model of described model of cognition II according to the method for setting up and adjust the qualitative model of cognition of near infrared in the prior art near infrared spectroscopic method, with the model of cognition II of resulting model as corresponding modeling sample kind.
Set up the preferred described PCA techniques of discriminant analysis of the employed pattern-recongnition method of model of cognition II of the present invention.
Set up in the described model of cognition II process, the present invention alleged for the strongest being meant of modeling sample kind recognition capability, between the average near infared spectrum of per two modeling sample kinds (for example kind A and B) distance (D
AB) with the threshold value (DT of corresponding modeling sample kind
A, DT
B) sum poor, with the standard deviation (SDev of two that use described computed in software gained corresponding modeling sample kind near infrared collection of illustrative plates apart from corresponding kind average near infared spectrum distances
A, SDev
B) merchant's maximum of sum, that is, following formula calculates the income value maximum:
The preferred Euclidean distance that uses carries out cluster analysis as the basis of classification of described model of cognition II.
Distance is meant the Euclidean distance between the average near infared spectrum of two modeling sample kinds between the average near infared spectrum of two modeling sample kinds of the present invention.
Threshold value of the present invention is meant, the ultimate range between the model of cognition average near infared spectrum of actual near infrared collection of illustrative plates of modeling sample and corresponding foundation is with certain multiple (as the correction term) sum of standard deviation.
As one of optimal way of the present invention, for same modeling sample kind, the near infrared spectrum spectral coverage of setting up described model of cognition II should be different from, and for example, is wider than the spectral coverage of setting up corresponding described model of cognition I.
The average near infared spectrum distance D of model of cognition II of the present invention
MAs follows with the computing method of standard deviation S Dev:
Average near infared spectrum distance D of the present invention
MBe meant the mean value of distance of the average near infared spectrum of each modeling sample near infrared collection of illustrative plates of a modeling sample kind and this kind modeling sample, its computing method are:
N represents the number of this kind modeling sample near infrared collection of illustrative plates, and i represents the near infrared collection of illustrative plates of i modeling sample of this kind, and i is 1,2,3 ... the integer of n, D represents Euclidean distance.
Euclidean distance (D between the near infrared collection of illustrative plates of different modeling sample kind (for example kind A and kind B)
AB) calculate by following mode:
In the formula, vector a
kBe the ordinate of A kind average near infared spectrum k data point, vector b
kBe the ordinate of B kind average near infared spectrum k data point, k is illustrated in counting of image data in the corresponding collection of illustrative plates, in this calculating to selected total data point summation.
SDev represents that (for kind A, this value is SDev to modeling sample
A) standard deviation of distance of each near infrared collection of illustrative plates and this kind average near infared spectrum, computing method are as follows:
Wherein, i represents this kind i original near infrared collection of illustrative plates, and n represents the number of original near infrared collection of illustrative plates; I is 1,2,3 ... n, Di represent the Euclidean distance of i near infrared collection of illustrative plates of this kind apart from this kind average near infared spectrum.
When setting up model of cognition II of the present invention, the types of drugs that described modeling sample comprises medicine to be identified at least and indicated; Preferably, described modeling sample comprises types of drugs and the auxiliary material thereof that medicine to be identified indicates, for example starch.
When setting up the near infrared spectrum qualutative model, need carry out the calculating of threshold value.The computing method of prior art are as follows:
DT＝Maximum Hit+0.25 SDev
DT is a threshold value; Hit is the distance of the equal near infrared collection of illustrative plates of original near infrared collection of illustrative plates anomaly
Maximum Hit represents the ultimate range of the equal near infrared collection of illustrative plates of original near infrared collection of illustrative plates anomaly.
When needing, for example when the modeling sample number of a certain kind more after a little while, the threshold value of model of cognition I of the present invention is adjusted by following step a-e, the threshold value of model of cognition II is adjusted by following step a:
Step a:
To the kind that can realize in the modeling sample kind being distinguished from each other, discerning, under situation about not obscuring with other kind, adjust its threshold value and be:
DT
A=Mean hitA+3SD (99% confidence limit);
A represents in the modeling sample kind and can be distinguished, arbitrary kind of identification.
DT
AThreshold value for kind A.
Mean hitA is the mean value of all sample near infrareds of kind A collection of illustrative plates to this kind average near infared spectrum distance.
SD represents the standard deviation of all near infrared collection of illustrative plates of distance and this kind of the equal near infrared collection of illustrative plates of modeling sample near infrared collection of illustrative plates anomaly to the difference of the mean value of average near infared spectrum distance, and computing method are as follows:
I represents this kind i original near infrared collection of illustrative plates, and n represents the number of the original near infrared collection of illustrative plates of this kind; I is 1,2,3 ... n, Xi represent the Euclidean distance of i original near infrared collection of illustrative plates to the average near infared spectrum, and Xm represents the mean value of all near infrared collection of illustrative plates of this kind to this kind average near infared spectrum Euclidean distance.
Step b:
To the kind that can be distinguished from each other, discern in the modeling sample kind, kind A for example may cause with other modeling sample kind and keeps the threshold value of other kind modeling sample constant when obscuring when adjusting threshold value as stated above, and the threshold value of this kind A is adjusted into:
DT
A=Mean hitA+2SD (95% confidence limit)
Wherein, described each symbol implication is with step a;
Step c:
To the kind that can realize in the modeling sample kind being distinguished from each other, discerning, kind A for example causes with other kind and to keep the threshold value of other kind constant when obscuring if be adjusted into behind the Mean hitA+2SD, and the threshold value of kind A is adjusted into:
DT
A=Mean hitA+1.65SD (90% confidence limit)
Wherein, described each symbol implication is with step a;
Steps d:
If the threshold value of kind A still may be obscured with other kind after being adjusted into Mean hitA+1.65SD, then no longer adjust its threshold value, this kind A is put into down proceed identification in one deck model of cognition.
Wherein, described each symbol implication is with step a.
But when causing sample representative not enough, the threshold value that adopts the aforementioned calculation method to be obtained is on the low side, causes for the discrimination of sample to be identified on the low side then when modeling sample quantity is few.In order to address this problem, when the representative deficiency of modeling sample, the present invention adjusts the threshold value of described model of cognition I and II according to following steps:
Step e:
The sample number that is used to set up described model when a certain kind is less, and is representative when not enough, is the threshold value of the kind close with its structure with its threshold setting.
Need to prove, when using model of cognition of the present invention, also unrestricted to the use order of described model of cognition I and II.In other words, can use described model of cognition I earlier, re-use described model of cognition II sample to be identified is discerned, also can use described model of cognition II earlier, and then use described model of cognition I that sample to be identified is discerned.
Any near infrared spectrometer, for example, German BRUKER company's near infrared spectrometer (MATRIX-F), and subsidiary at random near-infrared analysis software all can be used for the collection of near infrared data required for the present invention and the foundation of corresponding model.
A further object of the present invention is to provide a kind of instrument that can detect, discern for medicine, for example near-infrared spectrometers.Described near-infrared spectrometers device also is equipped with near infrared model of cognition I and/or the model of cognition II that is used for determining pharmaceutical active compounds of the present invention except that the function with existing near-infrared spectrometers.
Another purpose of the present invention is, a kind of vehicles that can detect, discern for pharmaceutical active compounds are provided, the instrument that the present invention detects, discerns for medicine is installed, as described in the present invention near-infrared spectrometers on these vehicles.
Description of drawings
The model of cognition I that accompanying drawing 1 is set up for the specific embodiment of the invention.There is shown the recognition sequence of this model of cognition I for modeling sample.
Embodiment
Technology contents disclosed according to the present invention, those skilled in the art can clearly learn other embodiment of the present invention, therefore, the following embodiment of the present invention is only as example of the present invention rather than restriction.Under the situation of not violating purport of the present invention and scope, can carry out various changes and improvements to the present invention, but all these changes and improvements, all should be within protection domain of the present invention.
Embodiment: the identification of the aluminum-plastic packaged tablet class medicine of macrolide antibiotics
1. collection modeling sample
For setting up macrolide antibiotics aluminum-plastic packaged tablet model of cognition I and model of cognition II, the modeling sample of collection is listed in the table below 1:
Table 1: the aluminum-plastic packaged tablet modeling sample of macrolide antibiotics
| Kind | The quantity of manufacturer |
| Azithromycin | 5 |
| Erythromycin | 7 |
| Acetyl spiramycin | 2 |
| Acetylkitasamycin | 1 |
| Erythromycin Ethylsuccinate | 25 |
| Medecamycin | 1 |
| Meleumycin | 2 |
| Roxithromycin | 11 |
| CLA | 7 |
| Kitasamycin | 4 |
2. the foundation of model of cognition I
Adopt German BRUKER company's near infrared spectrometer (MATRIX-F) to carry out the collection of near infrared spectrum data, use the subsidiary at random qualitative analysis software of this near infrared spectrometer to calculate.
A. the collection of modeling sample near infrared spectrum data
Adopt German BRUKER company's near infrared spectrometer (MATRIX-F), indium gallium arsenic (InGaAs) detecting device.
Condition determination: solid fibre-optical probe diffuse reflection scanning method; Resolution is 8cm
-1Number of background scan 64 times; Number of sample scan 64 times; Sweep limit is 12000-4000cm
-1Every batch of modeling sample is got 6 scannings respectively.
B. the foundation of model of cognition I and adjustment
Select 4500～6800cm
-1With 7300～10000cm
-1As the modeling spectral coverage of setting up aluminum-plastic packaged macrolide antibiotics model of cognition I, then, on this spectral coverage, 6 kinds of spectrogram preprocess methods that utilize the appended software of aforementioned BRUKER company's near infrared spectrometer to provide, only change preprocess method, but do not change other modeling condition and set up the qualitative model of cognition of different near infrareds, the model of cognition of selecting the most difficult generation between different modeling sample kinds to intersect to obscure is as the rudimentary model of model of cognition I, and according to the method for being put down in writing in the prior art this model is adjusted, its integral body for described aluminum-plastic packaged macrolides modeling sample is distinguished, recognition correct rate equals 100%.
In view of the above the model of cognition I of Jian Liing be divided into two-layer, its for the recognition sequence of modeling sample referring to accompanying drawing 1.
3. the foundation of model of cognition II
A. the foundation of preliminary model of cognition II
According to the PCA analysis principle, in conjunction with the concrete computing method of OPUS software (BRUKER company is the software of near infrared spectrometer preparation), at 4200～6000cm
-1Set up three kinds of tentative programmes of model of cognition II on the spectral coverage:
Scheme one: the near infrared collection of illustrative plates for all samples spectrogram near infrared collection of illustrative plates in the modeling sample of each kind and this kind reactive compound carries out principal component analysis (PCA), by above-mentioned OPUS software tear open two major components, first principal component has been represented the main information of this kind reactive compound, Second principal component, then comprises the information of auxiliary material and other measuring error, selects first principal component to be used for qualitative analysis.
Scheme two: the near infrared collection of illustrative plates for the other drug of all the near infrared collection of illustrative plates in this kind modeling sample and all non-modeling sample medicines of collecting carries out principal component analysis (PCA), wherein, with described other drug as auxiliary material information.By above-mentioned OPUS software tear open two major components, first principal component has been represented the main information of this kind auxiliary material, Second principal component, then comprises the information of reactive compound and measuring error, selects second major component to be used for qualitative analysis.
Scheme three: carry out principal component analysis (PCA) for the near infrared collection of illustrative plates of all samples near infrared collection of illustrative plates in this kind modeling sample, this kind reactive compound and the near infrared collection of illustrative plates of modeling sample auxiliary material, by above-mentioned OPUS software tear open three major components, preceding two major components have contained the bulk information of this kind reactive compound and auxiliary material, the 3rd major component is measuring error information, selects preceding two major components to be used for qualitative analysis.
With the non-plastic-aluminum tablet mould of Roxithromycin is example, analyzes the quality of three sets of plan.Three models are all selected the principal character spectral coverage (4200～6000cm of this kind
-1) as the modeling spectral coverage, and adopt first order derivative (level and smooth)+normalized preprocess method of vector at 5.Relatively the actual near infrared collection of illustrative plates in each model is apart from standard deviation (SDev), the averaged spectrum distance (D of model average near infared spectrum distance
M) and parameter (referring to table 2) such as threshold value.
The comparison of three kinds of modeling schemes of table 2 model of cognition II
With the average near infared spectrum of different cultivars near infrared collection of illustrative plates as this kind, in each conclusive evidence scheme model, calculate its distance (referring to table 3) to Roxithromycin tablet average near infared spectrum, aminophylline still may be mistaken as Roxithromycin in the discovery scheme two, and has some other potential possibility of obscuring; And scheme one is suitable with the recognition capability of scheme three; Can select as the case may be in the reality to use.But replace auxiliary material near infrared collection of illustrative plates with the starch near infrared collection of illustrative plates that does not contain active component in the scheme three, uncertainty is more, and only use the reference substance near infrared collection of illustrative plates of reactive compound in the scheme one, therefore selection scheme one is as the rudimentary model of model of cognition II, and use method of the prior art that it is adjusted, will adjust the back model as model of cognition II.
Roxithromycin sheet average near infared spectrum is apart from the distance of other kind average near infared spectrum among the different preliminary model of cognition II of table 3
4。Detection for sample to be identified
Use present embodiment model of cognition I and model of cognition II, detect different with modeling sample, 10 kinds of medicines (aluminum-plastic packaged tablet form) that contain the different activities compound.When the recognition result of above-mentioned model of cognition I and II is identical, with the reactive compound of compound shown in this recognition result as sample to be identified.With this result and testing result such as following table 4:
Table 4: embodiment 1 model of cognition is for the identification result of 10 kinds of medicines
| Kind | The accuracy * that model of cognition I and model of cognition II are used in combination |
| Azithromycin | 99.56％ |
| Roxithromycin | 97.55％ |
| Erythromycin | 99.96％ |
| Kitasamycin | 99.60％ |
| Erythromycin Ethylsuccinate | 99.78％ |
| Acetyl spiramycin | 100％ |
| Acetylkitasamycin | 100％ |
| Meleumycin | 100％ |
| Medecamycin | 100％ |
| CLA | 100％ |
Annotate: with respect to the accuracy rate of CNS method.These samples are with the check of CNS method.
Claims (27)
1. method of using near infrared spectroscopic method identification medicine, this method may further comprise the steps:
A. collect modeling sample;
B. set up and adjust near-infrared spectrum analysis model of cognition I according to described modeling sample;
C. set up and adjust near-infrared spectrum analysis model of cognition II according to described modeling sample;
D. use described near-infrared spectrum analysis model of cognition I and II to discern respectively for sample to be identified;
The recognition result of e. more above-mentioned two kinds of model of cognition is determined described sample to be identified: when the recognition result of above-mentioned two kinds of model of cognition is identical, and when identical with described sample marker title to be identified, judge that described sample to be identified is medicine that it indicated; When the recognition result of above-mentioned two kinds of model of cognition not simultaneously, judge that described sample to be identified is different with the medicine that it is indicated;
Wherein,
Described model of cognition I is 100% for the interracial differentiation of modeling sample, recognition accuracy, is not less than 90% for differentiation, the recognition accuracy of sample to be identified;
Described model of cognition II sets up by following steps:
A. near infrared spectrum data collection
Use near infrared spectrometer to measure, gather the near infrared light spectrum information of described each modeling sample kind respectively;
B. set up and the described model of cognition II of adjustment:
At each kind of modeling sample, different data processing method in the near-infrared spectrum analysis software of use prior art is set up a plurality of near-infrared analysis rudimentary models;
Select in these rudimentary models for the powerhouse of each modeling sample kind recognition capability, as the rudimentary model of corresponding model of cognition II;
According to setting up in the prior art near infrared spectroscopic method and the method for adjusting the qualitative model of cognition of near infrared, set up and adjust the rudimentary model of described model of cognition II, with the model of cognition II of resulting model as corresponding modeling sample kind;
Wherein, described for the strongest being meant of modeling sample kind recognition capability, distance and the threshold value sum of corresponding modeling sample kind is poor between the average near infared spectrum of per two modeling sample kinds, with the merchant maximum of two that use described computed in software gained corresponding modeling sample kind near infrared collection of illustrative plates apart from the standard deviation S Dev sum of corresponding kind average near infared spectrum distances;
Distance is meant the Euclidean distance between the average near infared spectrum of modeling sample kind between the average near infared spectrum of described modeling sample kind.
2. according to the method for claim 1, it is characterized in that described modeling sample is the marketed drugs that meets the pharmacopeia code requirement, its contained reactive compound is identical or close with the chemical constitution of the reactive compound that sample marker to be identified contains.
3. according to the method for claim 1 or 2, it is characterized in that when setting up described model of cognition I, the types of drugs of described modeling sample should be more than 3 or 3.
4. according to the method for claim 3, it is characterized in that when setting up described model of cognition I, the kind of described modeling sample should be more than 5 or 5, and the product of each 3 or 3 above producer of variety collection.
5. according to the method for claim 1 or 2, it is characterized in that described modeling sample has identical formulation and external packing form with described sample to be identified.
6. according to the method for claim 1 or 2, it is characterized in that when setting up described model of cognition II, described modeling sample comprises types of drugs and the auxiliary material thereof that sample to be identified indicates.
7. according to the method for claim 1, it is characterized in that described model of cognition I can finish for the interracial differentiation of described each modeling sample, identification.
8. according to the method for claim 7, it is characterized in that when needed, described model of cognition I finishes by the multilayer recognition method for the interracial differentiation of described modeling sample, identification.
9. according to the method for claim 1, it is characterized in that described model of cognition I is not less than 95% for differentiation, the recognition accuracy of sample to be identified.
10. according to the method for claim 1, it is characterized in that, according to setting up with the method foundation of adjusting the near infrared qualitative analysis model in the prior art near infrared spectroscopic method and adjusting described model of cognition I.
11. the method according to claim 9 is characterized in that, described model of cognition I sets up by following step:
At first use principal component analysis (PCA) to carry out dimension-reduction treatment for the near infrared spectrum data of the modeling sample that records;
Utilize the calculating principle to set up model of cognition I of the present invention then for the different pieces of information processing mode in the near-infrared spectrum analysis software of pattern-recongnition method.
12. the method according to claim 11 is characterized in that, described pattern-recongnition method is the PCA techniques of discriminant analysis.
13. method according to claim 12, it is characterized in that, described PCA techniques of discriminant analysis is index with described modeling sample near infrared collection of illustrative plates to the distance between the average near infared spectrum of this modeling sample kind, carries out cluster analysis, and and then determines the threshold value of described modeling sample.
14. the method according to claim 13 is characterized in that, described cluster analysis with Euclidean distance as basis of classification.
15. the method according to claim 1 is characterized in that, the average near infared spectrum distance D of described model of cognition II
MComputing method as follows:
N represents the number of this kind modeling sample near infrared collection of illustrative plates, and i represents the near infrared collection of illustrative plates of i modeling sample of this kind, and i is 1,2,3 ... the integer of n, D are represented described Euclidean distance,
Wherein, the Euclidean distance D between the near infrared collection of illustrative plates
ABComputing method are as follows:
In the formula, vector a
kBe the ordinate of A kind average near infared spectrum k data point, vector b
kBe the ordinate of B kind average near infared spectrum k data point, k is illustrated in counting of image data in the corresponding collection of illustrative plates, in this calculating to selected total data point summation;
SDev represents the standard deviation of each near infrared collection of illustrative plates of modeling sample apart from this kind average near infared spectrum distance, and computing method are as follows:
I represents this kind i original near infrared collection of illustrative plates, and n represents the number of original near infrared collection of illustrative plates; I is 1,2,3 ... n, Di represent the Euclidean distance of i near infrared collection of illustrative plates of this kind apart from this kind average near infared spectrum.
16. the method according to claim 1 is characterized in that, for same modeling sample kind, the near infrared spectrum spectral coverage of setting up described model of cognition II should be different from the spectral coverage of setting up corresponding described model of cognition I.
17. the method according to claim 16 is characterized in that, for same modeling sample kind, the near infrared spectrum spectral coverage of setting up described model of cognition II should be wider than the spectral coverage of setting up corresponding described model of cognition I.
18. the method according to claim 1 is characterized in that, described model of cognition I and II can use in random order.
19. the method according to claim 1 is characterized in that, adjusts the threshold value of described model of cognition I according to following steps a-e, adjusts the threshold value of described model of cognition II according to following steps a:
Step a:
To the kind that can realize in the modeling sample kind being distinguished from each other, discerning, under situation about not obscuring with other kind, adjust its threshold value and be:
DT
A=Mean hitA+3SD (99% confidence limit);
A represents in the modeling sample kind and can be distinguished, arbitrary kind of identification,
DT
ABe the threshold value of kind A,
Mean hitA is the mean value of all sample near infrareds of kind A collection of illustrative plates to this kind average near infared spectrum distance,
SD represents the standard deviation of all near infrared collection of illustrative plates of distance and this kind of the equal near infrared collection of illustrative plates of modeling sample near infrared collection of illustrative plates anomaly to the difference of the mean value of average near infared spectrum distance, and computing method are as follows:
I represents this kind i original near infrared collection of illustrative plates, and n represents the number of the original near infrared collection of illustrative plates of this kind; I is 1,2,3 ... n, Xi represent the Euclidean distance of i original near infrared collection of illustrative plates to the average near infared spectrum, and Xm represents the mean value of all near infrared collection of illustrative plates of this kind to this kind average near infared spectrum Euclidean distance;
Step b:
To the kind that can realize in the modeling sample kind being distinguished from each other, discerning, may cause with other modeling sample kind and keep the threshold value of other modeling sample kind constant when obscuring when adjusting threshold value as stated above, the threshold value of kind A is adjusted into:
DT
A=Mean hitA+2SD (95% confidence limit);
Step c:
To the kind that can realize in the modeling sample kind being distinguished from each other, discerning, cause with other modeling sample kind and to keep the threshold value of other modeling sample kind constant when obscuring if be adjusted into behind the MeanhitA+2SD, the threshold value of kind A is adjusted into:
DT
A=Mean hitA+1.65SD (90% confidence limit);
Steps d:
After if the threshold value of kind A is adjusted into Mean hitA+1.65SD, still may obscure with other modeling sample kind, then no longer adjust its threshold value, this kind is put into down proceeded identification in one deck model of cognition;
Step e:
To be used to set up the sample number of described model less, representative when not enough when a certain modeling sample kind, is the threshold value of the kind close with its structure with its threshold setting.
20. the method according to claim 1 is characterized in that, the formulation of described sample to be identified is the constant compound preparation of tablet, capsule, powder ampoule agent for injection, parenteral solution, ointment, supensoid agent, sugar coated tablet or formula rate.
21. the method according to claim 20 is characterized in that, the formulation of described sample to be identified is tablet, powder ampoule agent for injection and capsule.
22. the method according to claim 20 is characterized in that, described tablet is aluminum-plastic packaged.
23. the method according to claim 20 is characterized in that, before the near infrared spectrum data of gathering sugar coated tablet, should remove the sugar-coat of described tablet.
24. an instrument of discerning medicine is characterized in that, described model of cognition I of one of claim 1-23 and model of cognition II are installed in the described instrument.
25. the instrument according to claim 24 is characterized in that, described instrument is a near infrared spectrometer.
26. vehicles of discerning medicine is characterized in that, the described instrument of claim 24 is installed on the described vehicles.
27. the vehicles according to claim 26 is characterized in that, described instrument is a near infrared spectrometer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2006100671849A CN100529731C (en) | 2005-04-05 | 2006-04-05 | Drug distinguishing near infrared spectrum analysis method and apparatus |
Applications Claiming Priority (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNA2005100630702A CN1696660A (en) | 2005-04-05 | 2005-04-05 | Method and equipment for identifying medication by using near infrared spectral analysis |
| CN200510063070.2 | 2005-04-05 | ||
| CNB2006100671849A CN100529731C (en) | 2005-04-05 | 2006-04-05 | Drug distinguishing near infrared spectrum analysis method and apparatus |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1847828A CN1847828A (en) | 2006-10-18 |
| CN100529731C true CN100529731C (en) | 2009-08-19 |
Family
ID=37077468
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB2006100671849A Active CN100529731C (en) | 2005-04-05 | 2006-04-05 | Drug distinguishing near infrared spectrum analysis method and apparatus |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100529731C (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101398418B (en) * | 2007-09-30 | 2014-08-13 | 杜也兵 | Filtrated layer pollute detection and alarm method for water purifier |
| CN101738991B (en) * | 2008-11-26 | 2014-09-10 | 国立成功大学 | Method for detecting out-of-specification quality of product and estimating actually measured value of product |
| US9581552B2 (en) * | 2011-04-06 | 2017-02-28 | Klein Medical Limited | Spectroscopic analyser |
| CN106124451B (en) * | 2016-08-31 | 2019-09-27 | 上海创和亿电子科技发展有限公司 | A method of update the system and detection through the detection of packaging bag On-line near infrared analyzer |
-
2006
- 2006-04-05 CN CNB2006100671849A patent/CN100529731C/en active Active
Non-Patent Citations (4)
| Title |
|---|
| 近红外光谱法在药学上的应用. 王晋，张汝成，马成禹.中国医药工业杂质，第30卷第1期. 1999 |
| 近红外光谱法在药学上的应用. 王晋，张汝成，马成禹.中国医药工业杂质，第30卷第1期. 1999 * |
| 近红外光谱法在药物分析中的应用. 冯艳春，胡昌勤.中国药事，第17卷第5期. 2003 |
| 近红外光谱法在药物分析中的应用. 冯艳春，胡昌勤.中国药事，第17卷第5期. 2003 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN1847828A (en) | 2006-10-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Been et al. | Profiling of counterfeit medicines by vibrational spectroscopy | |
| US5121338A (en) | Method for detecting subpopulations in spectral analysis | |
| De Maesschalck et al. | Decision criteria for soft independent modelling of class analogy applied to near infrared data | |
| CN1696660A (en) | Method and equipment for identifying medication by using near infrared spectral analysis | |
| Chen et al. | Discrimination of Ganoderma lucidum according to geographical origin with near infrared diffuse reflectance spectroscopy and pattern recognition techniques | |
| Krakowska et al. | Chemometrics and the identification of counterfeit medicines—A review | |
| Zontov et al. | Chemometric aided NIR portable instrument for rapid assessment of medicine quality | |
| CN104316491B (en) | Method for detecting urea doped in milk based on synchronous-asynchronous two-dimensional near-infrared correlation spectrum | |
| CN102179375A (en) | Nondestructive detecting and screening method based on near-infrared for crop single-grain components | |
| Downey | Tutorial review. Qualitative analysis in the near-infrared region | |
| US10151688B2 (en) | Methodology for the identification of materials through methods of comparison of the spectrum of a sample against a reference library of spectra of materials | |
| CN104713845B (en) | Mixture component identification method based on terahertz absorption spectrum datum processing | |
| WO2004079347A1 (en) | Method of analysis of nir data | |
| CN100529731C (en) | Drug distinguishing near infrared spectrum analysis method and apparatus | |
| CN101620178B (en) | Method for quickly detecting additive chemical component in Chinese patent medicine, health-care food or food based on near-infrared spectrum technique | |
| Xu et al. | Integrated sensor array optimization with statistical evaluation | |
| CN109374572B (en) | Terahertz time-domain spectroscopy object classification method based on neural network | |
| Bakker et al. | Current challenges in the detection and analysis of falsified medicines | |
| CN108760677A (en) | A kind of rhizoma pinellinae praeparata based on near-infrared spectrum technique mixes pseudo- discrimination method | |
| Ríos-Reina et al. | Application of hierarchical classification models and reliability estimation by bootstrapping, for authentication and discrimination of wine vinegars by UV–vis spectroscopy | |
| Chen et al. | Rapid identification of three varieties of Chrysanthemum with near infrared spectroscopy | |
| Moros et al. | New cut-off criterion for uninformative variable elimination in multivariate calibration of near-infrared spectra for the determination of heroin in illicit street drugs | |
| CN101566569B (en) | System and method for identifying a plurality of fluorescence spectrum mixed materials through characteristic parameter | |
| CN104596983A (en) | Zopiclone tablet rapid identification method based on near infrared spectrum | |
| Bakeev et al. | Pros and cons of using correlation versus multivariate algorithms for material identification via handheld spectroscopy |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |