NIR calibration

M. Blanco, J. CoeUo, H. Iturriaga, S. Maspoch and J. Pages, NIR calibration in non-linear systems different PLS approaches and artificial neural networks, Chemom. Intell. Lab. Syst, 50, 75-82 (2000). 

B. Tenge and D.E. Honigs, The effect of wavelength searches on NIR calibrations. Abstracts of Papers of the American Chemical Society, 193(APR), 112 (1987). 

There is a failure to recognize the plant-site requirements for NIR calibration and validation, snch as the existence of appropriate sampling valves, well-designed sampling protocols, good laboratory reference methods, and variability in the analyte concentrations of interest. More details on these chemometric calibration issues can be found in Chapter 12. 

Table 15.6 NIR. calibration model performance for MMA and DMAAm in a reaction mixture...

Martens, H., Jensen, S. A. Partial least squares regression A new two-stage NIR calibration method in Proceedings of the 7-th World Cereal and Bread Congress, Prague, 1982, Elsevier, Amsterdam, in press... 

NIR calibration curves were prepared by measuring the key compact attributes on rectangular surrogate tablets, prepared on the laboratory Carver press. The preparation of the calibration curve requires samples with attributes spanning the range of values likely to be encountered during roller compaction. Hence, drug content (for content uniformity), relative humidity... 

As discussed above, the greatest source of error in NIR calibration is usually reference laboratory error, sample nonhomogeneity, and nonrepresentative sampling in the learning (training) set or calibration set population. Instrument quality and equation selection account for only a fraction of the variance or error attributable to NIR analytical technique in current routine application. 

Finally, however, it is the advantages (3) and (4) which really have a marked effect on NIR applications. The wavelength precision and reproducibility achievable with a good FTIR analyser design translates into long-term spectroscopic stability both within one analyser and across multiple analysers, so that measurement errors between analysers are in the region of a single milliabsorbance unit. For the untroubled development of NIR calibration models, and their maintainability over time, this of itself is critically important. 

Tenge, B. and Honigs, D.E., The Effect of Wavelength Searches on NIR Calibrations Abstracts of Papers of the American Chemical Society 1987, 193(April), 112. 

Blanco, M., Coello, J., Iturriaga, H., Maspoch, S. and Pages, J., NIR Calibration in Non-linear Systems Different PLS Approaches and Artificial Neural Networks Chemometrics Intell. Lab. Syst. 2000, 50, 75-82. 

Summary Statistics for NIR Calibration of Water in Water-Methanol Mixtures Using One Wavelength and a Nonzero Intercept... 

To minimize the number of costly GC measurements and obtain better prediction of the naphthalene and aromatic content of the process feedstock, NIR calibration models were used for prediction. The usual practice was followed by building an initial model using a data set acquired over a limited period of time. To keep the model up to date, a campaign for collecting new data was periodically organized, and a new model was built to replace the old one. A typical example of a data set is shown in Figure 8.31. The same autoscaled data is shown in Figure 8.32, where filled circles indicate an E-optimal subset of ten points. The complete set consists... 

In pharmaceutical applications, NIRS is more often used as a secondary analytical technique than as a primary tool. As a secondary method, a reference method is required to determine the reference component values that are to be used in the NIR calibration. 

The initial calibration process for a substance or a product can be quite detailed. A calibration equation is needed for each constituent in the sample. NIR calibrations must be formulation-specific. The accuracy of the NIR method cannot be better than the reference method from which it was built. Ruggedness of the models improves when all expected types of variation are included in the model. Careful selection of representative samples is imperative to the successful performance of the calibration. The choice of mathematical models depends highly on the character of the sample. 

The application of NIR for predicting the endpoint of clinical or production scale batches based on the NIR calibration models developed in the laboratory may prove to be very important to the pharmaceutical manufacturer. When drugloading or polymer-coating efficiencies vary, as they do, from lab scale equipment to the larger equipment used for manufacturing, the availability of a rapid on-line or at-line measurement for prediction of process endpoint may save many millions of dollars in time and materials. Although the standard error of prediction may not match the values obtained when... 

In 1982, Ciurczak and Torlini published on the analysis of solid and liquid dosage forms [74], They contrasted NIR calibrations for natural products versus those for pharmaceuticals. Samples prepared in the laboratory are spectrally different from production samples. Using laboratory samples for calibration may lead to unsatisfactory results production samples for calibration are preferred for calibration. 

Actual drug content of the samples was determined by HPLC water content was determined by Karl Fischer moisture analysis. For prediction of the drug content, three NIR calibrations using MLR were developed. 

That same year, Jensen et al. used NIR in the analysis of amiodarone tablets [86]. The film coating was scraped off and the tablets glued to an anodized aluminum plate. Six wavelengths were chosen for the calibration. Tablets ranging from 47 to 63% of the active ingredient were prepared in increments of 2%. The NIR calibration provided r2 = 0.996 and SEE = 0.45. 

Martens H, Jensen SA. Partial least Squares Regression A new two-stage NIR calibration methods. In Holas and Kratochvil, editors, Proceedings of the World Cereal and Bread Congress, Prague, June 1982. Amsterdam Elsevier, 1982. p. 607-47. 

Table 2 NIRS calibration statistics for com silage used at the University of Wisconsin for broad-based prediction equations used to estimate neutral detergent fiber (NDF), in vitro trae digestibility (IVTD), protein, and starch of com silage.

Some of the early work on amino acid NIR calibrations was performed by Pazdernik et al. (1997) (Table 6.10). They developed NIR equations to measure 17 amino acids in ground and whole soybean samples on a NIRSystems 6500. They had 90 samples in their calibration set and 26 in the validation set derived from 408 soybean lines. Samples were grown in Minnesota in 1994 and 1995. Samples were analyzed first as whole beans and then ground in a Fred Stein mill for 1 min. Results are reported in grams per kilograms of crude protein. 

Table 6.9. Summary of Amino Acids With initial Nir Calibration Data Set Results 1 ...

Amino Acid NIR Calculation (from Bajjalieh data set) NIR Calibration Bluebook ... 

To obtain a proper NIR calibration, experimental-design aspects also need to be considered, i.e. both the spectral variability as well as the whole concentration gradient of the... 

NIR calibration for direct analysis of sediment sample composition... 

One use of NIRS on sediments is to replace more expensive and slower classical chemical analyses. Successful NIR calibrations for carbon, carbonate, nitrogen, and phosphorus concentrations in a sediment core were recently reported by Malley et al. (1999), which demonstrates that a rapid, non-destructive, simultaneous analysis of these elements can be achieved by the use of NIRS. In this study on the eutrophication history of the lake Arendsee, Germany, an initial attempt was also made to develop calibrations for total diatoms, as well as the relative frequencies of three dominant diatom species. A similar study was also conducted on sediment cores from four Canadian lakes. In this study, the explained variance between results from conventional chemical analysis and NIR-predicted concentrations ranged between 97 and 99% for total C, organic C, and N (Malley et al., 2(XX)). 

It is well documented in the literature that NIR calibration curves for multi-component formulations should be assembled by scanning standards which contain all of the components varying the level of only one component in each sample. The relative contributions of each of the components may vaiy considerably as is illustrated in the following test data. 

---