Quantitative device and quantitative method

Abstract

The present invention provides a quantitative device and quantitative method that can rapidly quantify the components contained in a sample. [Solution] The quantitative analysis device 1 comprises a measurement unit 10, a pre-processing unit 20, and an analysis unit 30. The measurement unit 10 includes a light source 11 that emits light, and a spectrometer 12 that measures the spectrum by spectrally analyzing the light emitted from the light source 11 and after passing through the sample S. The pre-processing unit 20 performs pre-processing on the spectrum as measured by the spectrometer 12. The pre-processing unit 20 does not perform processing using the spectrum obtained by spectrally analyzing the light emitted from the light source 11 that reaches the spectrometer 12 without passing through the sample S. The analysis unit 30 analyzes the spectrum after pre-processing to quantify the components contained in the sample S.

Description

[Technical Field]

[0001] The present invention relates to a quantitative measurement device and a quantitative measurement method. [Background technology]

[0002] Absorption spectroscopy is a known technique for quantifying the components contained in a sample. In this quantitative technique, the intensity of light transmitted, reflected, or diffusely reflected by the sample when light is irradiated onto it is measured, and the magnitude of light absorption in the sample is determined based on that intensity to quantify the components contained in the sample (Patent Documents 1-3). The light used to irradiate the sample is light that is absorbed by the components contained in the sample, and in most cases, light in the near-infrared region is used.

[0003] Such quantitative techniques are used in various fields, for example, in quantifying the components contained in pharmaceuticals. In particular, in recent years, the pharmaceutical and formulation industry has been promoting the establishment of continuous production, which integrates multiple manufacturing processes in a continuous manner, with the aim of suppressing human error and improving production efficiency. Along with this, the automation of inspection processes has also become necessary. In these inspection processes, there is a need to perform high-speed quantification of components contained during production and quantification of components contained in the entire pharmaceutical product. [Prior art documents] [Patent Documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2022-52731 [Patent Document 2] Japanese Patent Publication No. 2007-187624 [Patent Document 3] Japanese Patent Publication No. 2019-155423 [Overview of the Initiative] [Problems that the invention aims to solve]

[0005] However, conventional quantitative techniques make it difficult to quantify the contained components at high speed.

[0006] This invention was made to solve the above-mentioned problems, and aims to provide a quantitative device and quantitative method that can perform quantitative determination of components contained in a sample at high speed. [Means for solving the problem]

[0007] A first aspect of the quantitative apparatus of the present invention comprises: (1) a measuring unit including a light source that outputs light and a spectrometer that spectrally analyzes the light output from the light source and after passing through a sample to measure the spectrum; (2) a pre-processing unit that performs pre-processing on the spectrum measured by the spectrometer for the light after passing through the sample; and (3) an analysis unit that analyzes the spectrum after pre-processing by the pre-processing unit to quantify the components contained in the sample.

[0008] In a second aspect of the quantitative apparatus of the present invention, in addition to the first aspect, the preprocessing unit performs one of the following as spectral preprocessing: smoothing, data interpolation, standard normalization, first derivative, or second derivative.

[0009] In a third aspect of the quantitative apparatus of the present invention, in addition to the first or second aspect, the analysis unit performs one of the following methods as spectral analysis: partial least squares method, principal component analysis, principal component regression, and neural network processing.

[0010] A first aspect of the quantitative method of the present invention comprises: (1) a measurement step of spectrally analyzing the light output from a light source and after passing through a sample using a spectrometer and measuring the spectrum; (2) a pre-processing step of pre-treating the spectrum obtained in the measurement step with respect to the light after passing through the sample; and (3) an analysis step of analyzing the spectrum after pre-processing in the pre-processing step to quantify the components contained in the sample.

[0011] In the second aspect of the quantification method apparatus of the present invention, in addition to the first aspect, in the pretreatment step, as the pretreatment of the spectrum, any one of smoothing processing, data interpolation, standard normalization, first-order differentiation, and second-order differentiation is performed.

[0012] In the third aspect of the quantification method of the present invention, in addition to the first aspect or the second aspect, in the analysis step, as the analysis of the spectrum, any one of partial least squares method, principal component analysis, principal component regression, and neural network processing is performed.

Effect of the Invention

[0013] According to the present invention, the quantification of the components contained in the sample can be performed at high speed.

Brief Description of the Drawings

[0014] [Figure 1] FIG. 1 is a diagram showing the configuration of the quantification apparatus 1. [Figure 2] FIG. 2 is a flowchart of the quantification method. [Figure 3] FIG. 3 is a diagram showing the absorbance spectra A(λ) of each of five types of samples assumed in the simulation. [Figure 4] FIG. 4 is a diagram showing the reference spectrum (log10(I0(λ))) assumed in the simulation. [Figure 5] FIG. 5 is a diagram showing the absorption spectra (-log10(I(λ))) of each of five types of samples assumed in the simulation. [Figure 6] FIG. 6 is a diagram showing the absorption spectra (-log10(I(λ))) of each of five types of samples after adding random noise and SNV processing assumed in the simulation. [Figure 7] FIG. 7 is a diagram showing the calibration curve obtained in the simulation. [Figure 8] FIG. 8 is a diagram showing the spectrum of the tablet sample after SNV processing and second-order differentiation obtained in the experiment. [Figure 9] FIG. 9 is a diagram showing the calibration curve obtained in the experiment. [Modes for carrying out the invention]

[0015] Hereinafter, embodiments for carrying out the present invention will be described in detail with reference to the attached drawings. In the description of the drawings, the same elements will be denoted by the same reference numerals, and redundant descriptions will be omitted. The present invention is not limited to these examples, but is indicated by the claims, and all modifications within the meaning and scope equivalent to the claims are intended to be included.

[0016] Figure 1 shows the configuration of the quantitative analysis device 1. The quantitative analysis device 1 comprises a measurement unit 10, a preprocessing unit 20, and an analysis unit 30. The measurement unit 10 includes a light source 11 and a spectrometer 12. The preprocessing unit 20 and the analysis unit 30 can be configured by a computer that includes a calculation unit such as a CPU for performing calculations, a storage unit such as a hard disk drive or memory for storing various data and programs, a display unit such as a liquid crystal display for displaying calculation results, and an input unit such as a keyboard or mouse for receiving instructions from the operator.

[0017] Light source 11 outputs light with a specific wavelength range. The wavelength range of the light output by light source 11 includes the wavelengths of light absorbed by the target component in sample S, for example, the near-infrared range (wavelengths 800 nm to 2500 nm). The light output from light source 11 passes through sample S and is then input to spectrometer 12. The light input to spectrometer 12 may be light transmitted through sample S, or light reflected, totally reflected, or diffusely reflected by sample S. Spectrometer 12 inputs the light that has passed through sample S, performs spectral analysis, and measures the light intensity, i.e., spectrum, for each wavelength λ of the light.

[0018] The preprocessing unit 20 performs preprocessing on the spectrum measured by the spectrometer 12 for light that has passed through the sample S. Preprocessing performed here includes, for example, smoothing to remove noise, data interpolation to make the spectral data equally spaced, standard normalization (SNV), first derivative and second derivative.

[0019] The analysis unit 30 analyzes the spectrum after preprocessing by the preprocessing unit 20 to quantify the components contained in the sample S. The analysis performed here is preferably multivariate analysis, such as partial least squares (PLS), principal component analysis (PCA), principal component regression (PCR), and processing using a neural network (e.g., a convolutional neural network).

[0020] In this embodiment, during pretreatment and analysis, the spectrum measured by the spectrometer 12 after passing through the sample S is used for processing, and no other spectra (for example, the spectrum obtained by spectrally separating light output from the light source 11 that reaches the spectrometer 12 directly without passing through the sample S) are used.

[0021] Figure 2 is a flowchart of the quantitative analysis method. The quantitative analysis method consists of three steps in order: measurement step S1, pretreatment step S2, and analysis step S3. In measurement step S1, the measurement unit 10 measures the spectrum by spectrally analyzing the light output from the light source 11 and after it has passed through the sample S using a spectrometer 12. In pretreatment step S2, the pretreatment unit 20 performs pretreatment on the spectrum measured by the measurement unit 10. In analysis step S3, the analysis unit 30 analyzes the spectrum after pretreatment by the pretreatment unit 20 to quantify the components contained in the sample S.

[0022] In the following explanation, we will assume that the light emitted from the light source 11 and transmitted through the sample S is spectrally analyzed by the spectrometer 12. Let I(λ) be the light intensity at wavelength λ measured by the spectrometer 12 when the sample S is placed in the optical path between the light source 11 and the spectrometer 12. Let I0(λ) be the light intensity at wavelength λ measured by the spectrometer 12 when the sample S is not placed in the optical path between the light source 11 and the spectrometer 12. The absorbance A(λ) of the sample S at wavelength λ is expressed by the following equation (1).

[0023]

number

[0024] If α(λ) is the absorption coefficient of sample S at wavelength λ, and C is the content of the component to be quantified in sample S, then the absorbance A(λ) of sample S at wavelength λ is expressed by equation (2) below. Based on this proportional relationship, a calibration curve can be created in advance. That is, by assuming α(λ) is known and measuring A(λ), the content C can be determined.

[0025]

number

[0026] The absorbance A(λ) measured at multiple wavelengths λ by the spectrometer 12 is called an absorbance spectrum. Since the shape of the absorbance spectrum A(λ) differs depending on the type and amount of components contained in the sample, the type and amount of components contained in the sample can be analyzed using the absorbance spectrum A(λ). In addition, since the absorption coefficient is large at wavelengths with high absorbance, the absorbance spectrum A(λ) is used to select appropriate wavelengths for creating the calibration curve described above.

[0027] For example, when quantifying components contained in actual samples such as pharmaceuticals, it is often difficult to quantify them at a single wavelength using the relationship in equation (2) above. In such cases, it is preferable to perform quantification by multivariate analysis using the absorbance of the entire spectrum or at multiple wavelengths within the spectrum.

[0028] Furthermore, when quantifying components contained in an actual sample, the baseline of the spectrum measured by the spectrometer may fluctuate, and this baseline fluctuation may interfere with subsequent quantitative analysis. Therefore, in order to improve quantitative accuracy, it is preferable to perform pre-processing on the spectrum prior to quantitative analysis. Examples of pre-processing include normalization of the centroid position of the spectrum and baseline correction.

[0029] Equation (1) above can also be expressed as equation (3) below. From equation (3) and equation (2) above, it can also be expressed as equation (4) below. As shown in equation (4), the absorption spectrum expressed as the logarithm of the intensity spectrum I(λ) when there is a sample S (-log 10 (I(λ)) and the reference spectrum (-log) which is the logarithm of the intensity spectrum I0(λ) when there is no sample S. 10 The difference from (I0(λ)) is proportional to the component amount C, with α(λ) being the proportionality constant. From this, the absorption spectrum (-log 10 In addition to finding (I(λ)), we also find the reference spectrum (-log 10 By also determining (I0(λ)) and performing multivariate analysis on the spectrum of the difference between the two spectra (left side of equation (4)), the amount of component C in the sample S can be determined.

[0030]

number

[0031]

number

[0032] However, with such conventional quantitative methods, it is necessary to measure not only the intensity spectrum I(λ) when sample S is present, but also the intensity spectrum I0(λ) when sample S is absent, making it difficult to rapidly quantify the amount of component C in sample S. For example, in the pharmaceutical and drug manufacturing industry, the automation of inspection processes required with the establishment of continuous production necessitates rapid quantification of components during production and quantification of components in the entire drug product, but the above quantitative methods have difficulty meeting this requirement. Furthermore, when measuring the intensity spectrum during the transport of a drug product, a major problem arises as to when to measure the intensity spectrum when the drug product is absent, using the same measurement system.

[0033] In the quantification device and quantification method of the present embodiment described below, in order to solve such problems, the measurement of the intensity spectrum I0(λ) when there is no sample S is made unnecessary, and without using the reference spectrum (-log 10 (I0(λ))), by performing quantitative analysis of the component amount C in the sample S based on the absorption spectrum (-log 10 (I(λ))), the quantification is accelerated.

[0034] In the above formula (4), since I0(λ) is the intensity spectrum measured by the spectroscope 12 when there is no sample S, the values at each wavelength of the reference spectrum (-log 10 (I0(λ))) do not depend on the amount C of the components contained in the sample S. On the other hand, since I(λ) is the intensity spectrum measured by the spectroscope 12 when there is a sample S, the greater the amount C of the components contained in the sample S, the greater the light absorption by that component, and the values at each wavelength of the absorption spectrum (-log 10 (I(λ))) are large. The values at each wavelength of the absorption spectrum (-log 10 (I(λ))) are in a linear relationship of the first order with α(λ) as the coefficient with respect to the component amount C.

[0035] In the present embodiment, this fact is utilized. In the present embodiment, in the pretreatment step S2, the pretreatment unit 20 does not perform processing using the spectrum obtained by splitting the light output from the light source 11 and reaching the spectroscope 12 without passing through the sample S by the spectroscope 12. That is, the pretreatment unit 20 does not perform processing using the intensity spectrum I0(λ) or the reference spectrum (-log 10 (I0(λ))) when there is no sample S.

[0036] Regarding the conditions and results of the simulation performed for the quantification method of the present embodiment, they will be described using FIGS. 3 to 7. In this simulation, five types of samples were assumed in which the content C of the component to be quantified was each value of 0.25, 0.50, 1.00, 1.25, and 1.50. FIG. 3 is a diagram showing the absorbance spectra A(λ) of each of the five types of samples assumed in the simulation. FIG. 4 is the reference spectrum (log10 This is a diagram of (I0(λ)).

[0037] Using the absorbance spectra in Figure 3 and the reference spectra in Figure 4, the absorption spectra of each of the five samples (-log) can be obtained from equation (4) above. 10 (I(λ)) was calculated. Figure 5 shows the absorption spectra (-log) of each of the five types of samples assumed in the simulation. 10 This is a diagram of (I(λ)).

[0038] The absorption spectrum in Figure 5 was subjected to 2% random noise to generate an absorption spectrum closer to that obtained from actual measurements. Furthermore, the absorption spectrum after random noise addition was normalized using standard normalization (SNV) as a preprocessing step. SNV is a process that calculates the mean and standard deviation of the spectral data, subtracts the mean from each data point in the spectrum, and divides the result by the standard deviation. After SNV processing, the mean of the spectral data is 0 and the standard deviation is 1. Figure 6 shows the absorption spectra (-log) of each of the five samples assumed in the simulation after random noise addition and SNV processing. 10 This is a diagram of (I(λ)).

[0039] For each of the five samples, the absorption spectra (-log) obtained by applying random noise and SNV processing as described above were obtained. 10 Six patterns of (I(λ)) were created. Three of these patterns were used as calibration data, and the other three were used as validation data. A quantitative analysis and calibration were performed using the partial least squares (PLS) method as a multivariate analysis.

[0040] Figure 7 shows the calibration curve obtained from the simulation. The horizontal axis represents the assumed content of the target component in the sample, and the vertical axis represents the content predicted by PLS. In the figure, black squares represent calibration data, and white triangles represent validation data. The squared correlation coefficient with respect to the validation data is 0.972, indicating that a good calibration curve has been obtained. As described above, the reference spectrum (-log 10 It was shown that quantification is possible without pretreatment using (I0(λ))).

[0041] Next, the experimental conditions and results of the quantitative method of this embodiment will be explained using Figures 8 and 9. In this experiment, tablet samples consisting of an active ingredient and an additive were used. The weight of each tablet sample was 130 mg, and the content C of the active ingredient to be quantified was 2 mg, 4 mg, 8 mg, and 12 mg. Near-infrared light was irradiated onto the tablet samples, and the intensity spectrum I(λ) of the light transmitted through the tablet samples was measured.

[0042] The absorption spectrum (-log) is the logarithm of this intensity spectrum I(λ). 10 SNV treatment and second differentiation were performed as pretreatment on (I(λ)). Figure 8 shows the spectra of the tablet samples obtained experimentally after SNV treatment and second differentiation. This figure shows the spectra for tablet samples with active ingredient content C of 2 mg and 12 mg, respectively.

[0043] For each of the four types with different active ingredient content C values, 20 tablet samples were used and subjected to the above-described measurements and pretreatment (SNV processing and second derivative) to create 20 spectral patterns as shown in Figure 8. Of these, 10 patterns were used as calibration data and the other 10 patterns as validation data. Quantitative analysis and calibration were performed using the PLS method as a multivariate analysis.

[0044] Figure 9 shows the calibration curve obtained in the experiment. The horizontal axis represents the content of the target component in the tablet sample, and the vertical axis represents the content predicted by PLS. In the figure, black squares represent calibration data, and white triangles represent validation data. The squared correlation coefficient with respect to the validation data is 0.991, indicating that a good calibration curve was obtained. As described above, the reference spectrum (-log 10 Both simulation and experimental results demonstrated that quantitative analysis is possible without pretreatment using (I0(λ)).

[0045] As described above, in this embodiment, the reference spectrum (-log 10 Without using (I0(λ)), the absorption spectrum (-log 10 Since the quantitative analysis of the amount of component C in sample S is performed based on (I(λ)), it is unnecessary to measure the intensity spectrum I0(λ) when sample S is absent, and therefore, the quantitative analysis of components contained in the sample can be performed quickly. [Explanation of Symbols]

[0046] 1...Quantitative device, 10...Measurement unit, 11...Light source, 12...Spectrometer, 20...Preprocessing unit, 30...Analysis unit.

Claims

1. A measuring unit including a light source that emits light, and a spectrometer that spectrally analyzes the light emitted from the light source and after passing through a sample to measure its spectrum, A preprocessing unit that performs preprocessing on the spectrum obtained by the spectrometer after the light has passed through the sample, An analysis unit analyzes the spectrum after pretreatment by the aforementioned pretreatment unit to quantify the components contained in the sample, A quantitative measuring device equipped with the following features.

2. The preprocessing unit performs one of the following as preprocessing of the spectrum: smoothing, data interpolation, standard normalization, first derivative, or second derivative. The quantitative apparatus according to claim 1.

3. The analysis unit performs one of the following methods for analyzing the spectrum: partial least squares method, principal component analysis, principal component regression, and neural network processing. The quantitative apparatus according to claim 1 or 2.

4. A measurement step involves using a spectrometer to spectrally analyze the light emitted from the light source, after it has passed through the sample, and measuring the spectrum. A pre-processing step is performed on the spectrum of the light after it has passed through the sample in the measurement step, The analysis step involves analyzing the spectrum after pretreatment in the aforementioned pretreatment step to quantify the components contained in the sample, A quantitative method comprising the following features.

5. In the aforementioned preprocessing step, the preprocessing of the spectrum may be performed by smoothing, data interpolation, standard normalization, first derivative, or second derivative. The quantitative method according to claim 4.

6. In the analysis step, one of the following methods is performed for the analysis of the spectrum: partial least squares method, principal component analysis, principal component regression, and neural network processing. The quantitative method according to claim 4 or 5.

Images