NIRS discriminant analysis

In the case of qualitative analysis, the moisture content of the samples is the most important factor that influences the NIR spectra. If there are differences in moisture content among the groups of samples, each group can be easily classified. To avoid this possibility, it is better to perform a moisture control of the samples. In a study on the classification of normal and aged soybean seeds by NIR discriminant analysis, the moisture content of each single kernel was controlled in a desiccator with silica gel at room temperature until each one had a constant moisture content of 14.6% (11). 

A. Candolfi, W. Wu, S. Heuerding and D.L. Massart, Comparison of classification approaches applied to NIR-spectra of clinical study lots. J. Pharm. Biomed. Anal., 16 (1998) 1329-1347. T. Fearn, Discriminant analysis. NIR News, 4 (5) (1993) 4-5. 

W. Wu, Y. Mallet, B. Walczak, W. Penninckx, D.L. Massart, S. Heuerding and F. Erni, Comparison of regularized discriminant analysis, linear discriminant analysis and quadratic discriminant analysis, applied to NIR data. Anal. Chim. Acta, 329 (1996) 257-265. 

Current methods for supervised pattern recognition are numerous. Typical linear methods are linear discriminant analysis (LDA) based on distance calculation, soft independent modeling of class analogy (SIMCA), which emphasizes similarities within a class, and PLS discriminant analysis (PLS-DA), which performs regression between spectra and class memberships. More advanced methods are based on nonlinear techniques, such as neural networks. Parametric versus nonparametric computations is a further distinction. In parametric techniques such as LDA, statistical parameters of normal sample distribution are used in the decision rules. Such restrictions do not influence nonparametric methods such as SIMCA, which perform more efficiently on NIR data collections. 

Rose, J. R. "Discriminant Analysis of NIR Research", 10th Annual Federation of Analytical Chemistry and Spectroscopy Societies meeting, Philadelphia, PA, Sept. 1983 Paper 26. 

From 1982 through 1985, few NIR analyses of dosage forms were published. Since 1986, there have been many articles. The first was a 1986 paper by Ciurczak and Maldacker [33] using NIR for tablet formulation blends, examining the use of spectral subtraction, spectral reconstruction, and discriminant analysis. Blends were prepared where actives—aspirin (ASA), butalbital (BUT), and caffeine (CAF)—were omitted from the formulation or varied over a range from 90 to 110% of label strength. 

As indicated in another chapter in this text, polymers have been analyzed by NIR for some time. In 1985, Shintani-Young and Ciurczak [37] used discriminant analysis to identify polymeric materials used in packaging plastic bottles, blister packaging, and PVC wrap, to name a few. Replacement of the... 

Whitfield [13] was one of the first NIR spectroscopists to discuss the use of NIR diffuse reflection analysis for veterinary products in 1986. Even then, he recognized the need for specificity and included an identification step in the analysis. This discriminant analysis program, named DISCRIM (Technicon, Tarrytown, NY) was an early version of the types of algorithms available commercially now. 

Fourier transform mid-infrared (FTIR), near-infrared (FTNIR), and Raman (FT-Raman) spectroscopy were used for discrimination among 10 different edible oils and fats, and for comparing the performance of these spectroscopic methods in edible oil/fat studies. The FTIR apparatus was equipped with a deuterated triglycine sulfate (DTGS) detector, while the same spectrometer was also used for FT-NIR and FT-Raman measurements with additional accessories and detectors. The spectral features of edible oils and fats were studied and the unsaturation bond (C=C) in IR and Raman spectra was identified and used for the discriminant analysis. Linear discriminant analysis (LDA) and canonical variate analysis (CVA) were used for the disaimination and classification of different edible oils and fats based on spectral data. FTIR spectroscopy measurements in conjunction with CVA yielded about 98% classification accuracy of oils and fats followed by FT-Raman (94%) and FTNIR (93%) methods however, the number of factors was much higher for the FT-Raman and FT-NIR methods. 

Yang, H., Irudayaraj, 1., and Paradkar, M.M. Discriminant analysis of edible oils and fats by FTIR, FT-NIR and FT-Raman spectroscopy. Food Chemistry, 93, 25-32. 2005. 

A spectroscopic NIR imaging system, using a FPA detector, has been developed for remote and on-line measurements on a macroscopic scale. Multivariate statistical techniques are required to extract the important information from the multidimensional spectroscopic images. These techniques include PCA and linear discriminant analysis for supervised classification of spectroscopic image data (178). 

Two studies have suggested that the IR spectra of synovial fluid specimens provide the basis to diagnose arthritis and to differentiate among its variants.A NIR study demonstrated that osteoarthritis, rheumatoid arthritis, and spondyloarthropathy could be distinguished on the basis of the synovial fluid absorption patterns in the range 2000-2400 nm.< In that case, the pool of synovial fluid spectra was subject to principal component analysis, and eight principal component scores for each spectrum were employed as the basis for linear discriminant analysis (LDA). On that basis, the optimal LDA classifier matched 105 of the 109 spectra to the correct clinical designation (see Table 7). 

NIR instruments may not only be calibrated for quantitative analysis, but they may also be trained for qualitative purposes. This process is usually termed discriminant analysis. The criteria noted previously for establishing quantitative calibrations with minimum prediction error are equally applicable to discriminant calibration sets. 

Discriminant analysis of NIR spectra was used in 1986 to assay the level of lincomycin in a pharmaceutical formulation. This was the first NIR analysis to be accepted by this US Food and Drug Administration (FDA). NIR is now used to measure the salicylic acid content of aspirin and pharmaceutical companies use discriminant NIR procedures to check incoming raw materials for drug production. Sample identification can be achieved using at-line NIR fiber optic systems. 

In 1984, Mark introduced the Mahalanobis distance in an algorithm for discriminant analysis of raw materials. The theory behind the software was described in a paper by Mark and Turmell [21] and first applied to pharmaceuticals by Ciurczak [22]. With the advent of 100% testing of incoming raw materials, qualitative analysis of raw materials by NIR became popular quickly. 

Workers at Johns Hopkins University, under the tutelage of Chris Brown, worked on screening PAP smears using NIR spectroscopy [229]. Healthy patients, patients with abnormal cells, and patients with cervical cancer were screened. Using discriminant analysis and principal component analysis, the samples were grouped and employed to examine further samples. It was seen that malignant and healthy tissues were distinctly different, while abnormal tissues carried spectral features from both sets. New developments for the detection of cervical cancer by NIR have been published by Yang et al. [230]. The possibility to detect endometrial cancer was also tested [231]. 

Since the first-ever application of NIR by Hart et al. in 1962 to the determination of moisture in seeds (6), the bibliography of NIRS technology has proliferated until it now numbers over 35,000 entries, many of which describe a very diverse assortment of applications to grains and seeds. The main areas have been composition analysis, analysis for prediction of functionality, and classification by NIR discriminant or classification analysis (NIRCA). Near-infrared spectroscopy has been applied to the analysis of many of the above commodities. Over 30 factors have been successfully predicted in cereals and pulses, and over 20 factors in oleaceous seeds. These applications have recently been comprehensively reviewed by Delwiche (7) and Dyer (8). 

Discriminant analysis is a way of classifying of substances on the basis of their spectra. As this stimulating application of NIR technology develops and becomes more... 

P. Dardenne, R. Biston. Attempt to recognize wheat species by discriminant analysis. In Near Infrared Diffuse Reflectance/Transmittance Spectroscopy, Proceedings of the International NIR/NIT conference, Budapest, Hungary, May 12-16, 1986. J. Hollo, et al co-ed. Akademiai Kiado, Budapest, 1987, p. 51-60. 

When using discriminant analysis on NIR data (X), each sample is allocated or classified into a categorical variable (y). For example, beef or pork. The quality of the estimated model, the prediction error, is typically presented as either correct or noncorrect classification of a single sample. As for regression, aU classification results should always be a result of a validation process. If the number of correctly classified samples is C and the number ofnoncorrectly classified samples is , then C + E = N and the percentage of correct and noncorrect for a set of samples are calculated as lOOC and lOOE/I, respectively. 

In a subsequent publication (35), discriminant analysis showed that the NIR spectra gave good allocation into three classes Tender, intermediate, and tough samples. Up to 60% of the samples were correctly classified for all three classes. Note that one cannot expect very high correct classifications of continuous classification variables. The number of classifications in correct or neighboring subgroups for the two extreme subgroups was equal to 97%. 

T. Ntes, K.I. Hildrum. Comparison of multivariate calibration and discriminant analysis in evaluating NIR spectroscopy for determination meat tenderness. Appl Spectrosc 51 350-357, 1997. 

Isaksson et al. showed that VIS/NIR reflectance could possibly be applied in salmon production plants to classify fillets into broad texture classes before further processing or sale. Spectra of fillets of farmed Atlantic salmon were correlated to Kramer shear force measurement and texture profile analysis (TPA). Samples were analyzed prerigor (2 h after slaughter) and postrigor (6 days after slaughter). Classification using linear discriminant analysis gave up to 79% correct classification into three... 

In near-infrared spectroscopy (NIRS), as we have seen in earlier chapters, one teaches the instru-ment/computer system what to look for in a given type of sample, then expects the hardware/software combination to produce valid answers when it is presented with unknown samples of the same type. But what kind of answers are we seeking Usually, they are quantitative answers, in other words, how much of substance XYZ is present in the sample. Occasionally we use what is called discriminant analysis to obtain a qualitative answer, in other words, is this sample A, B, C, or none of these ... 

The statistical theory of discriminant analysis also defines a linear discriminant function very similarly to Mahalanobis distance. These functions have characteristics that are of interest to NIR spectroscopists the linear discriminant functions are similar in form to regression equations, so that. 

---