Elsevier

Computers & Chemistry

Volume 22, Issue 1, 20 February 1998, Pages 119-122
Computers & Chemistry

Application of modern computer methods for recognition of chemical compounds in NIRS

https://doi.org/10.1016/S0097-8485(97)00042-9Get rights and content

Abstract

Near infrared spectroscopy (NIRS) facilitates both quantitative and qualitative analysis. The study presents an example of utilization of spectral NIR data for qualitative analysis of pharmaceutical samples.

Introduction

Near infrared spectroscopy (NIRS) is a recognized analytical method with a wide range of applications in agriculture, pharmacy and other branches of industry. NIR (700–2500 nm) spectral features arise from molecular absorption of overtones (700–1800 nm) and combination (1800–2500 nm) bands from fundamental vibration bands in the mid-infrared region. Near infrared spectra contain information relating to differences in bond strengths and chemical species. Its spectral features include aliphatic C–H, aromatic C–H, hydroxyl O–H amines and amides N–H (Goddu and Delker, 1960; Whetsel, 1968; Kelly and Callis, 1990). Historically, the NIR method was first applied for quantitative determination of protein and water in feeds (Norris and Ben, 1968). The first commercial apparatus was based on filters as cheap alternatives of the spectrophotometer and application was restricted, almost exclusively, to assays of essential constituents of feeds and plant materials.

Instrument manufactures have developed a number of different optical geometries for NIR measurement. Each of these geometries has advantages and disadvantages, but all provide adequate data for NIR analysis. Two basic instrument types are now used in food and feed industries, instruments with gratings (monochromators) and instruments with filters. The success of NIRS technology directly parallels advances in computer technology. A scanning monochromator can generate 1050 data points for a sample in less than 30 s (Shenk and Westerhaus, 1995). Different software have been developed to manage NIRS data (NSAS, ISI, IDAS, Pirouette, etc.), all of them allowing the use of mathematical, statistical and computer science methods for improving the extraction of useful information from chemical measurement data. Those methods are known as chemometrics (Geladi, 1996). Chemometrics is one of the cornerstones of the NIR analytical technique. NIR has been a major vehicle for the development of chemometric calibration technology, introducing multiple linear regression (MLR), principal component regression (PCR) and partial least-squares regression (PLS) as chemometric tools for the quantitative NIR community. This acceptance of computerized data analysis using statistical modeling techniques has now extended to qualitative analysis including particulary material identification in pharmaceutical industry (Stark, 1996). The IQ2 is one of the programs used for NIR qualitative analysis. The analysis is based on the comparison of the spectrum of an unkown substance with the library of models created earlier and which can be developed on the basis of the entire spectrum or only on the basis of selected subregions.

Today, the NIR technique has been applied in the determination of composition of food (Osborne and Fearn, 1986), feed (Barton, 1987), natural products (Mroczyk et al., 1992; Mroczyk and Michalski, 1995) as well as in monitoring processes (Kemeny, 1992) or in pharmaceutical analysis (Ciurczak, 1992).

It often happens that a substance of unknown composition is submitted for analysis; for example, in a pharmacy, many raw materials occur in the form of white powders which can easily be mistaken with others. Especially, when the label attached to the packaging of the raw material is destroyed, it is necessary to conduct costly qualitative analyses.

For a long time now, qualitative analysis of compounds have been carried out with the help of apparatus operating in the IR range and which utilize Fourier transformation. A growing interest in FT-NIR has been observed in recent years (Liu et al., 1995).

Since NIR is based on overtones derived from the spectrum in classical infrared, they carry data from the range of basic band absorption. A standard NIR spectrum differs considerably from the IR spectrum—first of all, its range is significantly shorter and it contains substantial data redundancy. In this short range, all bands interleave with one another, overlap and are masked with the matrix. Data extraction from the NIR spectrum according to the method employed in IR spectroscopy is impossible for reasons mentioned above. The objective of present study is to give an example of use of NIR spectral data for qualitative analysis of pharmaceutical samples.

Section snippets

Materials and methods

In order to develop the library, NIR spectra were taken from a set of samples of substances (in solid form—powders and crystals) representative of a cross-section of raw materials of a typical pharmaceutical plant. The above-mentioned substances comprised, among other things, such active ingredients as ibuprofen, paracetamol and its derivatives, vitamins, amino acids, sugars, minerals (sodium carbonate), fillers (calcium pantothenate)—60 substances in all. At least three spectra were taken from

Results

Raw spectra (1/R) (Fig. 1, l-alanine example, five repetitions) shows a baseline shift mainly due the sample granulation. To remove this unwanted effect, all the spectra were transferred to second derivatives (d2od, gap 20 segment 5). Resulting spectra are shown in Fig. 2. This kind of spectra was used to construct library, with the parameter: identyfing correlation threshold=0.95. Validation of the library shows no errors, so the library was fully suitable for qualifications analysis. The

Conclusions

In contrast to FT-IR, sample recognition requires no initial preparation, provided the library has been constructed on samples of the same physical form. The spectrum library designed in the way described above makes it possible to recognize the samples supplied, classify them with regards to their purity and, having the appropriate quantitative calibration pinned to the given product in the library, can also allow quantitative assay. The above-described method eliminates the risk of using

References (16)

  • W.B. Mroczyk et al.

    Computers and Chemistry

    (1995)
  • Barton, F. E. (1987) Analytical application to fibrous foods and commidities. In NIR Technology in Agricultural and...
  • Ciurczak, E. W. (1992) NIR analysis of pharmaceuticals. In Handbook of Near-Infrared Analysis, ed. P. A. Burns and E....
  • Geladi, P. (1996) An overview of multivariate spectral analysis. In Near Infrared Spectroscopy: The Future Waves, ed....
  • Gemperline, P. J., Webber, L. D. and Cox, F. O. (1989) Raw materials testing using soft independent modeling of class...
  • R.F. Goddu et al.

    Analytical Chemistry

    (1960)
  • J.J. Kelly et al.

    Analytical Chemistry

    (1990)
  • Kemeny, G. J. (1992) Process analysis. In Handbook of Near Infrared Analysis, ed. P. A. Burns and E. W. Ciurczak, p....
There are more references available in the full text version of this article.

Cited by (0)

View full text