CN111595816A - Method for distinguishing tablet coating end point based on near-infrared PCA analysis and application thereof - Google Patents
Method for distinguishing tablet coating end point based on near-infrared PCA analysis and application thereof Download PDFInfo
- Publication number
- CN111595816A CN111595816A CN202010720155.8A CN202010720155A CN111595816A CN 111595816 A CN111595816 A CN 111595816A CN 202010720155 A CN202010720155 A CN 202010720155A CN 111595816 A CN111595816 A CN 111595816A
- Authority
- CN
- China
- Prior art keywords
- tablet
- film coating
- pca
- endpoint
- near infrared
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 89
- 238000000513 principal component analysis Methods 0.000 title claims abstract description 67
- 238000009492 tablet coating Methods 0.000 title claims abstract description 9
- 239000002700 tablet coating Substances 0.000 title claims abstract description 9
- 238000009501 film coating Methods 0.000 claims abstract description 121
- 239000007888 film coating Substances 0.000 claims abstract description 121
- 230000008569 process Effects 0.000 claims abstract description 38
- 238000002329 infrared spectrum Methods 0.000 claims abstract description 17
- 239000003826 tablet Substances 0.000 claims description 106
- 238000001228 spectrum Methods 0.000 claims description 26
- 238000000576 coating method Methods 0.000 claims description 19
- 238000004458 analytical method Methods 0.000 claims description 17
- 239000011248 coating agent Substances 0.000 claims description 17
- 238000009826 distribution Methods 0.000 claims description 14
- 238000002474 experimental method Methods 0.000 claims description 14
- 238000011156 evaluation Methods 0.000 claims description 13
- 238000012544 monitoring process Methods 0.000 claims description 12
- 239000007941 film coated tablet Substances 0.000 claims description 9
- 230000009466 transformation Effects 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 7
- 238000007781 pre-processing Methods 0.000 claims description 6
- 238000012937 correction Methods 0.000 claims description 3
- 230000004069 differentiation Effects 0.000 claims description 3
- 238000009499 grossing Methods 0.000 claims description 3
- 230000002159 abnormal effect Effects 0.000 claims description 2
- 239000000463 material Substances 0.000 abstract description 3
- 238000001514 detection method Methods 0.000 abstract description 2
- 239000000523 sample Substances 0.000 description 25
- 238000012360 testing method Methods 0.000 description 21
- 238000012795 verification Methods 0.000 description 15
- 238000003066 decision tree Methods 0.000 description 12
- 230000003595 spectral effect Effects 0.000 description 10
- 238000002203 pretreatment Methods 0.000 description 8
- 230000000875 corresponding effect Effects 0.000 description 7
- 230000008859 change Effects 0.000 description 6
- 238000012797 qualification Methods 0.000 description 6
- 238000012549 training Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 239000000796 flavoring agent Substances 0.000 description 5
- 238000002790 cross-validation Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 235000019634 flavors Nutrition 0.000 description 4
- 230000000052 comparative effect Effects 0.000 description 3
- 230000003247 decreasing effect Effects 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 238000012847 principal component analysis method Methods 0.000 description 3
- 235000014347 soups Nutrition 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 239000003405 delayed action preparation Substances 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 238000010200 validation analysis Methods 0.000 description 2
- 238000004497 NIR spectroscopy Methods 0.000 description 1
- 238000002835 absorbance Methods 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000000137 annealing Methods 0.000 description 1
- 238000000149 argon plasma sintering Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000013211 curve analysis Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 239000008298 dragée Substances 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 239000002662 enteric coated tablet Substances 0.000 description 1
- 238000007667 floating Methods 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000000670 limiting effect Effects 0.000 description 1
- 239000002075 main ingredient Substances 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000036961 partial effect Effects 0.000 description 1
- 239000000825 pharmaceutical preparation Substances 0.000 description 1
- 239000006187 pill Substances 0.000 description 1
- 239000000843 powder Substances 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000002829 reductive effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000007940 sugar coated tablet Substances 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 230000002459 sustained effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 238000005303 weighing Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/359—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using near infrared light
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/3563—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light for analysing solids; Preparation of samples therefor
Landscapes
- Physics & Mathematics (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention relates to a method for judging a tablet coating end point based on near-infrared PCA analysis and application thereof. The method comprises the following steps: near infrared spectra of a film coating standard tablet sample and a tablet film coating sample in the whole process are collected, main component information is extracted after pretreatment and PCA analysis treatment, and a PCA film coating endpoint discrimination model is established, so that the film coating endpoint condition of the tablet film coating sample in the whole process is judged. The end point judgment for the film coating is an indirect method and has the advantages of no damage, rapidness and accuracy. The method has important significance in the aspects of ensuring the batch-to-batch and batch-to-batch uniformity of film coating products, improving the utilization rate of film coating materials, realizing the on-line detection of film coating end points and the like.
Description
Technical Field
The invention relates to a method for judging a film coating end point, in particular to a method for judging a tablet coating end point based on near infrared PCA analysis and application thereof.
Background
In recent years, near infrared spectroscopy combined with discriminant mode analysis technology is widely applied to process analysis technology in pharmaceutical industry because of the capability of directly performing rapid and nondestructive detection on samples. The technology is already used for the research of the coating endpoint judgment fields of sustained and controlled release preparations, sugar-coated tablets, dropping pills, enteric-coated tablets, traditional Chinese medicine tablet film coating, on-line monitoring and the like. Principal Component Analysis (PCA) is an effective method for data dimension reduction. The method is based on the principle of maximum variance, and carries out characteristic decomposition on original data to obtain new variables (principal components) which are orthogonal to each other. The larger the variance, the larger the information that contains the original data. Therefore, the information of the original data can be contained by using a few principal component variables, and the feature extraction of the data is realized.
In the prior art, the granted patent CN101713731A "an identification method of coating quality of pharmaceutical preparations" discloses that a qualitative identification model of coating quality is rapidly established by combining methods such as near infrared spectrum, interval principal component analysis method and support vector machine method. The interval principal component analysis method adopted in the patent is based on the principle that the correlation between characteristic information and categories is the maximum, the classification result is inaccurate, the influence of pretreatment is large, part of test results are less than 40%, and the accuracy rate of predicting the coating end point is low. In addition, the samples used for identification in the above patents are marked as acceptable and unacceptable, and are only suitable for coating endpoint acceptability determination, but do not exhibit coating process variation, and are therefore not suitable for coating process monitoring.
Therefore, the invention aims to develop a method for determining the film coating endpoint by using a near infrared combined principal component analysis method with higher accuracy and suitability for process monitoring.
Disclosure of Invention
The invention aims to provide a method for quickly, conveniently, simply and accurately judging the coating end point of a tablet film and application thereof. The method utilizes the near infrared technology to collect the obtained near infrared spectrum of the film coating standard tablet and establish a PCA analysis model, thereby carrying out film coating process monitoring and film coating end point judgment on the tablet to be detected.
In order to achieve the purpose, the invention adopts the following technical scheme:
in one aspect, a method for distinguishing tablet coating endpoint based on near infrared PCA analysis and application thereof are provided, comprising the following steps:
s1, collecting a near infrared spectrum: collecting near infrared spectra of samples in the whole process of film coating of the film coating standard tablet and the tablet;
s2, processing a near infrared spectrum: residual analysis is carried out on the near infrared spectrum obtained in the step S1, abnormal spectra which are obviously far away from the average spectrum are detected and removed, and then the main component information is obtained through pretreatment and PCA analysis processing;
s3, establishing a PCA endpoint discrimination model: calculating a principal component score value by using the principal component information obtained in the step S2, carrying out PCA endpoint judgment model modeling, and establishing a PCA score map;
s4, judging an end point: the PCA endpoint judgment model in the step S3 is used for judging the film coating process monitoring and the film coating endpoint condition of the sample in the whole process of the film coating of the tablet to be detected; the judgment standard is as follows: in the PCA score plot, the film coating endpoint was not reached when the first principal component score PC1 did not reach the threshold.
Further, the film coating standard tablet in step S1 is a film coated tablet entering the film coating end point determined by an experiment or an expert, and specifically is a film coated tablet meeting pharmacopeia standards or a film coated tablet entering the film coating end point determined by experiments such as a moisture-proof experiment, a disintegration experiment, taste evaluation, and the like; the samples in the whole process of tablet film coating comprise a tablet core, intermediate tablets with different film coating time but not entering the film coating end point and finished film coating tablets.
Further, in the PCA score chart of step S3, the distribution of the standard tablet, the tablet core and the intermediate tablet at the 1 st principal component score PC1 is significantly different, and the boundary of the first principal component score PC1 of the cluster of the standard tablet is analyzed accordingly to obtain the threshold for endpoint discrimination.
Further, the PCA score plot is presented in the order:
the tablet cores are distributed at the leftmost side of the score map; the standard plates are distributed at the rightmost side of the score map; the middle plate is translated towards the right side along with the time increment in the distribution of the score map; clear boundaries are distributed on the score chart among the standard tablet, the tablet core and the middle tablet.
Further, the preprocessing in step S2 includes one or more of smoothing, multivariate scatter correction, standard positive-variance transformation, derivative differentiation, fourier transformation, and wavelet transformation.
Further, the derivative is differentiated into a second derivative.
On the other hand, the method is applied to film coating process monitoring and film coating end point condition judgment in the pharmaceutical field.
Has the advantages that:
1. the invention finds that when a PCA discriminant analysis model is established, on an obtained PCA score map, observation and calculation find that a near infrared spectrum is subjected to pretreatment and PCA analysis treatment, a first principal component score PC1 obtained by extraction is in positive correlation with film coating time, and a correlation coefficient R =0.987 and a P value =0.02 are obtained by Pearson correlation test, so that the film coating time and the first principal component score PC1 are in high positive correlation, and the result is consistent with the PCA score map display result. And with increasing time, PC1 will increase and the sample distribution will shift to the right, visually representing the entire film coating process. In order to visually reflect the spectral difference between the test sample and the standard tablet, PC1 is combined with a confidence interval to realize the monitoring of the film coating process and the discrimination of the film coating end point. Therefore, compared with the quality control method of the film coating process such as a weighing method, manual visual identification and the like which are commonly used in production, the method has the advantages of rapidness, accuracy and capability of reflecting the film coating process and quality more objectively and accurately.
2. Compared with the existing interval PCA discriminant analysis model, the method determines the optimal spectral band according to the principle that the maximum difference of the corresponding characteristic information among the categories is the maximum before the PCA model performs principal component analysis, and then performs the characteristic extraction of the principal component analysis on the information on the spectral band, so that the difference among different categories is the maximum due to the characteristic information, the different categories are easier to distinguish, and the classification result is more accurate. Therefore, the method is more excellent in classification performance; from the accuracy, the internal cross of the interval PCA model analysis is only 90%, and the data of the method is more than 99% under the condition of more data volume; secondly, the result of the interval PCA model analysis is obviously influenced by the pretreatment, and part of test results are less than 40 percent, but the accuracy rate of the method is far higher than that of the method; furthermore, the existing method is only used for identifying the film coating end point qualified line, and the change of the film coating process cannot be monitored. The method of the invention is not only used for judging the end point, but also can effectively reflect the change of the film coating state along with the time, gradually approaches to the film coating end point, and is suitable for process monitoring.
Drawings
FIG. 1 is a near infrared original spectrum of a tablet core of a Digeda-4-flavor decoction piece;
FIG. 2 is a near infrared raw spectrum of a film-coated intermediate tablet of a Digeda-4 taste decoction tablet;
FIG. 3 is a near infrared average spectrum at each time point during film coating of a Digeda-4 flavour soup tablet;
FIG. 4 is a near infrared spectrum of a film coated process tablet of Gegeda-4 flavor decoction pre-treated with a standard tablet;
FIG. 5 is a PCA score plot of a tablet during a film coating process of a Digeda-4 flavored soup tablet versus a standard tablet;
FIG. 6 is a graph of Nc coefficient obtained by the method of the present invention as a function of film coating time;
FIG. 7 is a graph of the rate of pass versus time;
FIG. 8 is a ROC graph.
Detailed Description
The present invention is further described in the following examples, which should not be construed as limiting the scope of the invention, but rather as providing the following examples which are set forth to illustrate and not limit the scope of the invention.
The following is a description of establishing a film coating endpoint discrimination model experiment by specific near infrared PCA analysis, comprising the following steps:
1. selecting the first-grade product and the finished product film-coated tablets above as film-coated standard tablets, performing near-infrared diffuse reflection spectrum, and taking the near-infrared spectrum as a reference spectrum.
The film coating standard tablet described in the present application refers to a film coating tablet entering a film coating end point determined by an experiment or an expert, and specifically refers to a film coating tablet meeting pharmacopeia standards or a film coating tablet entering a film coating end point determined by experiments such as a moisture-proof experiment, a disintegration experiment, taste evaluation, and the like.
2. Sampling from the tablet core, and collecting the spectrum of near infrared diffuse reflection of the sample in the whole process of tablet film coating.
3. The near infrared spectrum is pretreated, and the pretreatment method comprises the following steps: the method comprises the following steps of preprocessing such as smoothing, multivariate scattering correction, standard positive-space variable transformation (SNV), differentiation and Fourier transformation, and selecting spectral bands of the spectrum by using related algorithms such as Genetic Algorithm (GA), interval least square method (iPLS), combined partial least square method (SiPLS), non-information variable elimination method (UVE), annealing algorithm and competitive adaptive re-weighting sampling (CARS).
These pretreatment methods may be used alone or in combination of a plurality of them to achieve the optimum pretreatment effect. In the collection process, factors such as high-frequency random noise, serious baseline drift, over-strong signal background, uneven samples, light scattering and the like influence the spectrum quality, and further influence the accuracy of the model. Therefore, the spectra must be preprocessed before extracting the spectral feature information.
4. And (3) preprocessing the near infrared spectrum of the samples in the whole process of film coating of the film coated standard tablet and the tablet, and then carrying out PCA (principal component analysis) processing on the preprocessed samples to obtain the main component information. And selecting information highly related to the film coating time from the main component information as spectral characteristics, and establishing a film coating end point discrimination model by combining a confidence interval. The specific algorithm flow is as follows: (1) calculating the score value of the main component; (2) determining the score value of the principal component containing the film coating time as the spectral characteristic; (3) calculating a confidence interval of the spectral characteristics of the qualified plate; (4) and (4) judging the test sample, and judging whether the principal component score value is in the confidence interval range in the step (3). The confidence interval for the reference spectrum is calculated as follows:
confidence interval =
（1）
Wherein the content of the first and second substances,
representing confidence level
Corresponding standard score (commonly used)
=0.05 or 0.1),
and
respectively, the mean and standard deviation of the ith principal component score value of the reference spectrum, and n represents the number of the reference spectra. The present study method uses a 95% confidence level, i.e.
。
In order to conveniently judge the qualification of the test sample, the judgment coefficient is provided by combining the formula (1)
And the method is used for judging the qualified slice. See the formulas (2) and (3).
Wherein the content of the first and second substances,
the film coating end point discrimination coefficient is shown,
represents the ith principal component score discrimination coefficient,
representing the weight of the ith principal component in the selected principal component
，
Representing the mean of the ith principal component score of the reference spectrum,
the i-th principal component score value of the sample to be measured is shown. Whether the film coating endpoint is reached or not is judged whether the main component score of the test sample spectrum falls within the reference spectrum confidence interval. The concrete expression is as follows: corresponding to the main ingredient containing the film coating time information
Weighted sum as endpoint discrimination factor for film coating
Smaller values indicate smaller differences between spectra. When in use
If the test sample reaches the film coating end point, judging the test sample to be a qualified tablet; when in use
The test sample has not reached the end of the film coating and is a non-acceptable tablet. In addition, if
And is
The trend of the change is decreasing, the forward end point of the film coating process is close, otherwise, the forward end point of the film coating process is close
The trend of change is increasing, and increasing distance 1 may be the case where excessive film coating occurs.
5. G-mean was used to evaluate model performance, and F-value measures the classification performance of a small number of classes (end-point classes). AUC values were used to compare the performance of the different models.
The general accuracy will be mainly affected by the majority of the results and the performance of the model cannot be effectively evaluated, therefore, the overall performance of the model is evaluated by using the geometric mean G-mean, while the classification performance of the minority is the F value, considering that the recall ratio and the precision ratio of the minority are equally important, therefore, β =1I.e., the F value. In particular, see formulas (4) and (5), the standard slice class is set as a positive class, the non-end point class is set as a negative class, and P+Is the positive type precision, P-Is the negative class precision, R+For positive type recall, R-And the result is a negative class recall ratio. The closer the G-mean is to 1, the better the classification performance of the model is, and the closer the F value is to 1, the better the classification effect on the end point class is. The AUC value is the area under the ROC curve, the larger the value, the better the model overall performance.
Example 1
The method is applied to the monitoring of the film coating progress and the rapid judgment of the end point of the Di Geda-4 flavor decoction tablets, and the film coating material of the Di Geda-4 flavor decoction tablets is Opadry gastric-soluble yellow film coating powder.
1. Experimental Material
(1) An experimental instrument:
coating a pan: HT/F700 non-porous coating machine (IMMERGAS, Italy);
a near-infrared spectrometer: micro nir Onsite portable near infrared spectrometer (vivai Solutions), spectrum acquisition software: MicroNIR Pro v2.5.1, process software: pycharm 2018.2.4 software;
scanning conditions are as follows: the wavelength range is 908.0-1678.8 nm, the integration time is 7.9 ms, and the scanning times are 100 times; and (3) testing conditions are as follows: the temperature is 23 +/-3 ℃, and the humidity is 25 +/-4%.
(2) Experimental samples:
tablet cores of the Digeda-4 soup tablets, sample intermediate tablets with 12 different film coating times and finished film coating tablets (self-made in a laboratory); the qualified tablets were manually selected as film-coated standard tablets, and the other 6 batches were samples of the tablets during the whole process of film coating, and were numbered 20191126A, 20191126B, 201912101, 201912102, 201912103, 201912104, respectively.
2. Experimental methods and results
2.1 obtaining the near-Infrared original average Spectrum
Obtaining near-infrared original average spectra of the film coating standard tablet, the tablet core and the intermediate tablet according to scanning conditions and testing conditions, as shown in figure 1; the obtained near infrared raw average spectrum is read and displayed by NIR film coating endpoint discrimination software based on PCA, and the reading result is shown in figure 2.
In order to examine the capability of judging the film coating qualification of the model, the sample categories are marked as qualified slices and unqualified slices before modeling. Combining the amount of the film coating solution, the film coating time and the manual identification, 14 points and qualified tablets within the time period of 0-130min are placed in the same map for comparison, as shown in fig. 3. As can be seen from the analysis of FIG. 3, the difference between the 0-100min sample and the qualified tablet is significant in the absorbance at the wavelength of 908.0-1412.2nm, which may be caused by incomplete or uneven coating film on the surface of the tablet core; the spectral difference between the sample after 100min and the qualified plate is not obvious, and part of the sample may be qualified. Therefore, samples from 0-100min were labeled as non-acceptable tablets, while only samples at the end of the film coating were labeled as acceptable tablets.
2.2 pretreatment and PCA analytical processing
The near infrared spectrograms of the film coating standard tablet, the tablet core and the intermediate tablet are preprocessed by adopting an SNV + second derivative preprocessing method, and the result is shown in figure 4. As can be seen from fig. 4, the signal-to-noise ratio is significantly improved compared to the original spectrum, and the resolution of the spectrum is correspondingly improved.
And then the first two main factors (PC 1 and PC 2) are extracted through PCA analysis processing. The film coating standard tablets are set as a data set, main component information of samples in the whole process of film coating of 6 batches of tablets is set as training sets in batches 20191126A and 201912102, and the batches 20191126B, 201912101, 201912103 and 201912104 are set as external test sets to establish models.
2.3 establishing PCA endpoint discrimination model
Calculating the principal component score value of the sample principal component information obtained in the step 2.2, carrying out PCA endpoint judgment model modeling, and establishing a PCA score map; in the PCA score chart, the distribution of the standard tablet, the tablet core and the intermediate tablet in the 1 st principal component score value PC1 is obviously different, and the boundary of the cluster-like first principal component score value PC1 of the standard tablet is analyzed according to the distribution, so that the threshold value for judging the end point is obtained
The 6-batch tablets were film coated throughout the process, and the film coating endpoint was not reached when the first principal component score PC1 did not reach the threshold in the PCA score plot, as shown in fig. 5, with the abscissa of the first principal component score PC1 and the ordinate of the second principal component score PC2, it can be seen that as the film coating time increased, the distribution of the intermediate tablets in the PCA score plot shifted to the right with increasing time, and the standard tablets that reached the film coating endpoint were centered on the right.
Evaluating the performance of the PCA discrimination model:
and performing layered k-fold cross validation (k = 4) on the training set as internal validation, wherein the internal validation is used for preprocessing selection and optimization models and finding the optimal modeling condition. And predicting the external test set by the model obtained from the training set to obtain the mean value and the relative standard deviation of the corresponding indexes as an external test, and verifying the performance and the generalization capability of the model.
And distinguishing F values and G-mean of model performance evaluation indexes, and calculating a performance index mean value and a Relative Standard Deviation (RSD) by adopting layered k-fold internal verification and layered k-fold external verification respectively. The specific situation is shown in table 1: and comparing different pretreatments by utilizing internal cross validation to obtain the optimal pretreatment method. The internal verification result shows that the model has good performance on the training set, and the end point analysis is accurate and stable in performance. The external verification result shows that G-mean is close to 1, F value is good, the model overall performance is good, the end point classification effect is good, the relative standard deviation RSD corresponding to the two indexes is less than 5%, and the model is stable in test result of a plurality of data sets, so that the method is good and stable in performance and has good prediction capability and robustness.
TABLE 1 evaluation index of PCA discriminant model
3. Conclusion
I. In the PCA score plot, the tablets exhibited a certain distribution pattern at different film coating times. As shown in fig. 5, the tablet cores are distributed on the left side of the score chart, the standard tablets are distributed on the right side of the score chart, and the middle tablets are shifted to the right side along with the time increase in the distribution of the score chart, namely, the film coating time is in positive correlation with the PC1 value.
To analyze the film coating process more specifically, the Nc coefficient obtained in the present invention is used to analyze the spectral difference between the tablets and the acceptable tablets during the whole film coating process, so as to obtain the variation trend of the film coating time and the Nc coefficient, and further realize the process analysis of the film coating, as shown in fig. 6. As shown in fig. 6, the Nc value decreased overall with increasing film coating time, indicating that the degree of difference between the sample and the acceptable tablets decreased gradually. The model discrimination parameter Nc coefficient is specifically expressed as: (1) the discrimination parameter distribution of the 0min plain tablet (tablet core) model is concentrated, which shows that the consistency of the plain tablet is good; (2) the difference between the discrimination parameters of the sample model of 10-20min and 0min is small, which shows that most samples are still plain films; (3) the discrimination parameter distribution of the sample model of 30-40min is dispersed, which indicates that the coating film of part of the sample begins to form; (4) the model discrimination parameters of the sample of 50-90min have a descending trend, and the difference between the sample and the qualified sheet is reduced, which indicates that the coating film is gradually formed; (5) the number of acceptable tablets in the sample after 100min increased with time, and at the end of the film coating, the tablets were essentially acceptable. The results show that (1) the plain tablets have good consistency, and the difference between film-coated tablets and plain tablets in production mainly comes from the coating film. Therefore, the near infrared spectrum can effectively reflect the difference between the film coated tablet and the plain tablet caused by the coating film; (2) the model discrimination parameters can accurately discriminate the qualification of the film coating, has certain explanatory power on the forming process of the coating and can be used for process analysis. In addition, for the overall analysis, the model prediction qualified rate and the change trend thereof are adopted to carry out process analysis and film coating end point judgment. The yield of the samples in the test set was calculated using the above model, as shown in fig. 7, where the abscissa is time, the ordinate is the predicted endpoint ratio, and the dotted line is the yield of 95% and is used as the threshold for endpoint discrimination of the film coating. The overall change trend of the yield is as follows: (1) the qualification rate of 0-70min is about 0, and the film is almost a non-qualified film; (2) the qualification rate of 80-90min is less than 25%, and a small amount of qualified slices begin to appear; (3) the good qualification rate is gradually increased at the moment of 100 min-film coating end point.
PCA score plot, the distribution of tablets that did not reach the end of the film coating was clearly distinguishable from the standard tablet distribution. As shown in fig. 5, the vertical line here may serve as a boundary between the point where the film coating is not reached and the point where the film coating is reached. Thus, there is a high correlation between PC1 and film coating time. For verification, correlation between film coating time and PC1 was analyzed by Pearson correlation test, and correlation coefficient R =0.987 and P =0.02 were calculated, demonstrating that film coating time is highly positively correlated with the first principal component score PC1, thus reflecting film coating progress in the PCA score plot-as film coating time increases, PC1 increases, and the samples will gradually approach the film coating endpoint region to the right. The internal verification and the external verification of the PCA discriminant model respectively have F values of 0.838 and 0.835, G-mean is 0.96 and 0.926, and the Relative Standard Deviation (RSD) of the two indexes is less than 5%, so that the PCA discriminant model has good prediction capability and robustness. In addition, the PCA discrimination model is compared with the decision tree discrimination model and the K neighbor discrimination model in performance, and the result shows that the PCA discrimination model is better in performance. Therefore, the model can be used for monitoring the film coating process of unknown samples and quickly judging the film coating end point.
Comparative example 1
Discriminating models using decision trees
Step 2.1 and step 2.2 adopt the same embodiment 1, step 2.3 establishes a decision tree discrimination model, and the performance evaluation of the established decision tree discrimination model is as follows:
evaluating the performance of the decision tree discrimination model:
and distinguishing F values and G-mean of model performance evaluation indexes, and calculating a performance index mean value and a Relative Standard Deviation (RSD) by adopting layered k-fold internal verification and layered k-fold external verification respectively. The specific situation is shown in table 2: and comparing different pretreatments by utilizing internal cross validation to obtain the optimal pretreatment method. The internal verification result shows that the model has good performance on the training set, and the end point analysis is accurate and stable in performance. The external verification result shows that the G-mean is close to 1, the F value is good, the overall performance of the model is good, and the classification effect on the end point is good, but the relative standard deviation RSD corresponding to the two indexes is larger than 13%, and the stability of the test result of the model on a plurality of data sets is poor. Therefore, this method is poor in stability and generalization ability, and cannot be used as an endpoint judgment model.
TABLE 2 decision Tree discrimination model evaluation index
Comparative example 2
Discriminating model adopting K nearest neighbor
Step 2.1 and step 2.2 adopt the same embodiment 1, step 2.3 establishes a K neighbor discriminant model, and the performance evaluation of the established K neighbor discriminant model is as follows:
evaluating the performance of the decision tree discrimination model:
and distinguishing F values and G-mean of model performance evaluation indexes, and calculating a performance index mean value and a Relative Standard Deviation (RSD) by adopting layered k-fold internal verification and layered k-fold external verification respectively. The details are shown in Table 3: and comparing different pretreatments by utilizing internal cross validation to obtain the optimal pretreatment method. The internal verification result shows that the model has good performance on the training set, and the end point analysis is accurate and stable in performance. The external verification result shows that the G-mean is close to 1, the F value is good, the overall performance of the model is good, and the classification effect on the end point is good, but the relative standard deviation RSD of the F value is slightly larger than 5%, and the G-mean is slightly smaller than 5%, which indicates that the model has certain floating on the test results of a plurality of data sets, but has no obvious difference on the results of different test data sets. Thus, the method has good prediction capability, but the stability is general, and further improvement is still needed.
TABLE 3K neighbor discriminant model evaluation index
The evaluation index comparison and AUC values of example 1 and comparative examples 1-2 were used to compare the performance of the PCA discrimination analysis, the decision tree discrimination model and the K-nearest neighbor discrimination model, and the results are shown in table 4 and fig. 8. As can be seen from the index parameters in Table 4, the F value and G-mean have no obvious difference, and the decision tree shows the worst performance in the corresponding relative standard deviation RSD, so the decision tree has the worst performance. When the PCA discrimination model is compared with the k-nearest neighbor model, the k-nearest neighbor method on F values and G-mean indexes is superior to the PCA discrimination model, so that the k-nearest neighbor model is better in prediction capability, and the PCA discrimination model on corresponding relative standard deviation RSD is superior to the k-nearest neighbor method, so that the stability of the PCA discrimination model is better. To compare the model performance synthetically, the AUC values were calculated using ROC curve analysis, as shown in fig. 8. The results of FIG. 8 show the ROC curve with the pseudo-positive class ratio on the abscissa and the true class ratio on the ordinate. The PCA discrimination model is the uppermost curve, the area under the PCA discrimination model comprises a K neighbor method discrimination model, and the K neighbor method discrimination model comprises a decision tree discrimination model, which is particularly characterized in that the AUC value of the PCA discrimination model is larger, so that the PCA discrimination model provided by the method is optimal. The model performance is related to: PCA discrimination model > K nearest neighbor method discrimination model > decision tree discrimination model.
TABLE 4 comparison of evaluation indexes of different discriminant models
Claims (7)
1. The method for distinguishing the coating endpoint of the tablet based on the near infrared PCA analysis is characterized by comprising the following steps:
s1, collecting a near infrared spectrum: collecting near infrared spectra of samples in the whole process of film coating of the film coating standard tablet and the tablet;
s2, processing a near infrared spectrum: residual analysis is carried out on the near infrared spectrum obtained in the step S1, abnormal spectra which are obviously far away from the average spectrum are detected and removed, and then the main component information is obtained through pretreatment and PCA analysis processing;
s3, establishing a PCA endpoint discrimination model: calculating a principal component score value by using the principal component information obtained in the step S2, carrying out PCA endpoint judgment model modeling, and establishing a PCA score map;
s4, judging an end point: the PCA endpoint judgment model in the step S3 is used for judging the film coating process monitoring and the film coating endpoint condition of the sample in the whole process of the film coating of the tablet to be detected; the judgment standard is as follows: in the PCA score plot, the film coating endpoint was not reached when the first principal component score PC1 did not reach the threshold.
2. The method for determining the tablet coating endpoint based on near infrared PCA analysis according to claim 1, wherein the film coating standard tablet in the step S1 is a film coated tablet entering the film coating endpoint determined by experiments or experts, specifically a film coated tablet meeting pharmacopeia standards or a film coated tablet entering the film coating endpoint determined by experiments such as moisture-proof experiments, disintegration experiments, taste evaluation and the like; the samples in the whole process of tablet film coating comprise a tablet core, intermediate tablets with different film coating time but not entering the film coating end point and finished film coating tablets.
3. The method for distinguishing the coating endpoint of a tablet according to claim 1 based on near infrared PCA analysis, wherein in the step S3 PCA score chart, the distribution of the 1 st principal component score value PC1 of the standard tablet, the tablet core and the middle tablet is obviously different, and the boundary of the cluster first principal component score value PC1 of the standard tablet is analyzed to obtain the threshold value for distinguishing the endpoint.
4. The method for determining tablet coating endpoint based on near infrared PCA analysis of claim 3 wherein the PCA score plot is presented in the order:
the tablet cores are distributed at the leftmost side of the score map; the standard plates are distributed at the rightmost side of the score map; the middle plate is translated towards the right side along with the time increment in the distribution of the score map; clear boundaries are distributed on the score chart among the standard tablet, the tablet core and the middle tablet.
5. The method for determining tablet coating endpoint based on near infrared PCA analysis according to claim 1 wherein said preprocessing of step S2 comprises one or a combination of smoothing, multiple scatter correction, standard positive-variance transformation, derivative differentiation, fourier transformation, and wavelet transformation.
6. The method for determining tablet coating endpoint based on near infrared PCA analysis of claim 5 wherein said derivative differential is a second derivative.
7. Use of a method for the determination of the end-point of the coating of a tablet according to any of claims 1 to 6 based on near infrared PCA analysis for the monitoring of the progress of the film coating and the determination of the end-point of the film coating in the pharmaceutical field.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010720155.8A CN111595816B (en) | 2020-07-24 | 2020-07-24 | Method for distinguishing tablet coating end point based on near-infrared PCA analysis and application thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010720155.8A CN111595816B (en) | 2020-07-24 | 2020-07-24 | Method for distinguishing tablet coating end point based on near-infrared PCA analysis and application thereof |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111595816A true CN111595816A (en) | 2020-08-28 |
| CN111595816B CN111595816B (en) | 2020-11-20 |
Family
ID=72191911
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010720155.8A Active CN111595816B (en) | 2020-07-24 | 2020-07-24 | Method for distinguishing tablet coating end point based on near-infrared PCA analysis and application thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111595816B (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2010112887A (en) * | 2008-11-07 | 2010-05-20 | Astellas Pharma Inc | Principal component analysis method, principal component analyzer, different kind article detection device, principal component analysis program, and recording medium for recording the principal component analysis program |
| CN101713731A (en) * | 2009-11-09 | 2010-05-26 | 北京中医药大学 | Method for distinguishing coating quality of medicine preparation |
| US20110216313A1 (en) * | 2010-03-02 | 2011-09-08 | Weiyong Li | Supervised principal component analysis |
| CN103913433A (en) * | 2014-04-04 | 2014-07-09 | 浙江大学 | Online detection method for double-effect concentration process of danhong injection |
| CN105092520A (en) * | 2015-08-06 | 2015-11-25 | 浙江大学 | Zhengtian pill mixing process end point determination method |
| CN107843574A (en) * | 2016-09-21 | 2018-03-27 | 北京中医药大学 | Chinese medicinal granule mixed process terminal on-line monitoring method |
| CN108318442A (en) * | 2018-02-06 | 2018-07-24 | 江苏康缘药业股份有限公司 | A kind of detection method suitable for Chinese medicine suspending system |
| CN108318440A (en) * | 2017-12-22 | 2018-07-24 | 江西中医药大学 | A method of quickly judging tablet thin film coating terminal using near-infrared fingerprint pattern technology |
| CN109856084A (en) * | 2019-03-26 | 2019-06-07 | 河南中医药大学 | A kind of stomach pill of aucklandia and amomum fruit condensed pill preparation process quality monitoring method |
| CN109883988A (en) * | 2019-01-18 | 2019-06-14 | 江苏康缘药业股份有限公司 | A kind of detection method of solid pharmaceutical preparation coating qualification testing model |
| CN110887810A (en) * | 2019-11-29 | 2020-03-17 | 朗天药业(湖北)有限公司 | Method for evaluating quality consistency of Xuesaitong dropping pills based on near infrared spectrum technology |
-
2020
- 2020-07-24 CN CN202010720155.8A patent/CN111595816B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2010112887A (en) * | 2008-11-07 | 2010-05-20 | Astellas Pharma Inc | Principal component analysis method, principal component analyzer, different kind article detection device, principal component analysis program, and recording medium for recording the principal component analysis program |
| CN101713731A (en) * | 2009-11-09 | 2010-05-26 | 北京中医药大学 | Method for distinguishing coating quality of medicine preparation |
| US20110216313A1 (en) * | 2010-03-02 | 2011-09-08 | Weiyong Li | Supervised principal component analysis |
| CN103913433A (en) * | 2014-04-04 | 2014-07-09 | 浙江大学 | Online detection method for double-effect concentration process of danhong injection |
| CN105092520A (en) * | 2015-08-06 | 2015-11-25 | 浙江大学 | Zhengtian pill mixing process end point determination method |
| CN107843574A (en) * | 2016-09-21 | 2018-03-27 | 北京中医药大学 | Chinese medicinal granule mixed process terminal on-line monitoring method |
| CN108318440A (en) * | 2017-12-22 | 2018-07-24 | 江西中医药大学 | A method of quickly judging tablet thin film coating terminal using near-infrared fingerprint pattern technology |
| CN108318442A (en) * | 2018-02-06 | 2018-07-24 | 江苏康缘药业股份有限公司 | A kind of detection method suitable for Chinese medicine suspending system |
| CN109883988A (en) * | 2019-01-18 | 2019-06-14 | 江苏康缘药业股份有限公司 | A kind of detection method of solid pharmaceutical preparation coating qualification testing model |
| CN109856084A (en) * | 2019-03-26 | 2019-06-07 | 河南中医药大学 | A kind of stomach pill of aucklandia and amomum fruit condensed pill preparation process quality monitoring method |
| CN110887810A (en) * | 2019-11-29 | 2020-03-17 | 朗天药业(湖北)有限公司 | Method for evaluating quality consistency of Xuesaitong dropping pills based on near infrared spectrum technology |
Non-Patent Citations (4)
| Title |
|---|
| C.-V.MOLTGEN ET AL.: "A novel in-line NIR spectroscopy application for the monitoring of tablet film coating in an industrial scale process", 《TALANTA》 * |
| HUIQUAN WU ET AL.: "Integration of Near-Infrared Spectroscopy and Mechanistic Modeling for Predicting Film-Coating and Dissolution of Modi fi ed Release Tablets", 《INDUSTRIAL & ENGINEERING CHEMISTRY RESEARCH》 * |
| YUSUKE HATTORI ET AL.: "Real-time monitoring of the tablet-coating process by near-infrared spectroscopy - Effects of coating polymer concentrations on pharmaceutical properties of tablets", 《JOURNAL OF DRUG DELIVERY SCIENCE AND TECHNOLOGY》 * |
| 陈红英 等: "近红外光谱技术用于正天丸混合过程终点的判断", 《中国实验方剂学杂志》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111595816B (en) | 2020-11-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Moscetti et al. | Near infrared spectroscopy is suitable for the classification of hazelnuts according to Protected Designation of Origin | |
| CN107478599B (en) | method for detecting illegal additive melamine formaldehyde resin in feed raw materials | |
| CN113720797B (en) | Online rapid quality measuring and liquor picking method for distilled liquor | |
| CN110672578A (en) | Model universality and stability verification method for polar component detection of frying oil | |
| CN111445469A (en) | Hyperspectrum-based apple damage parameter lossless prediction method after impact | |
| CN113418869B (en) | Polarized light-based rice grain doping or adulteration nondestructive rapid quantitative detection method | |
| CN113030007B (en) | Method for rapidly testing quality stability of tobacco essence based on similarity learning algorithm | |
| CN110084227A (en) | Mode identification method based on near-infrared spectrum technique | |
| CN111595816B (en) | Method for distinguishing tablet coating end point based on near-infrared PCA analysis and application thereof | |
| CN113484278A (en) | Tomato comprehensive quality nondestructive testing method based on spectrum and principal component analysis | |
| CN113310934A (en) | Method for quickly identifying milk cow milk mixed in camel milk and mixing proportion thereof | |
| CN111624317A (en) | Nondestructive testing method for judging freshness of baby cabbage | |
| CN106198423A (en) | A kind of method differentiating ham sausage grade based on visible and near infrared spectrum analytical technology | |
| CN113916824B (en) | Alpha in milk s1 Intermediate infrared rapid batch detection method of casein | |
| CN107238557A (en) | A kind of method of utilization near infrared spectroscopy quick detection calcium carbonate particle diameter distribution | |
| CN111595815B (en) | Method for judging tablet film coating end point and application thereof | |
| CN113049526B (en) | Corn seed moisture content determination method based on terahertz attenuated total reflection | |
| CN111595814A (en) | Method for monitoring tablet coating end point based on cluster analysis and application thereof | |
| CN104297205A (en) | Method for quickly and nondestructively identifying edible oil | |
| CN113324941A (en) | Method for rapidly identifying preservation time of raw milk | |
| CN113791049B (en) | Method for rapidly detecting freshness of chilled duck meat by combining NIRS and CV | |
| CN113310933A (en) | Spectrum identification method for number of days for storing raw buffalo milk | |
| CN111693484A (en) | Sesame oil variety identification method based on terahertz time-domain spectroscopy | |
| CN111735796B (en) | Method for near-infrared recognition of tablet film coating progress and application thereof | |
| JPH0640069B2 (en) | Estimation method of taste value by near infrared |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |