Stephen R. Lowry

An algorithm for a selective nearest neighbor decision rule (Corresp.)

A procedure is introduced to approximate nearest neighbor (INN) decision boundaries. The algorithm produces a selective subset of the original data so that 1) the subset is consistent, 2) the distance between any sample and its nearest selective neighbor is less than the distance from the sample to any sample of the other class, and 3) the subset is the smallest possible.

Polymer-Coated, Tapered Cylindrical ATR Elements for Sensitive Detection of Organic Solutes in Water

Commercially available tapered chalcogenide fibers have been coated with poly(vinyl chloride) (PVC) plasticized with 47% (w/w) chloroparaffin containing 60% Cl by weight. The coating procedure involves applying a drop of a solution containing PVC and the plasticizer in tetrahydrofuran along the fiber while allowing the solvent to evaporate. The coated fibers were exposed to 0.15% (v/v) benzene in water (1479-cm−1 band), 0.40% (v/v) chloroform in water (1216-cm−1), and 0.10% (v/v) nitrobenzene in 1.5% (w/v) methanol/water (1348-cm−1). All three organic solutes gave readily detectable signals with the coated fibers but were not observable when the aqueous solution was sampled with the use of an uncoated, tapered fiber. Detection limits for benzene, chloroform, and nitrobenzene were calculated to be 0.02%, 0.11%, and 0.006% by volume, respectively. These data show that the advantage of using a polymer coating to concentrate the analyte and reduce the water background may be combined with the advantages of using a tapered optical fiber to yield a sensitive method for detecting nonpolar organic solutes in water.

A Comparison of Two Discriminant Functions for Classifying Binary Infrared Data

Some form of information compression is essential if one is to be able to utilize effectively the increasingly large data compilations. One approach is to eliminate the intensity information, leaving spectra packed in a peak/no peak format. This paper reports the comparison of two simple discriminant functions for classifying binary infrared data. For the multicategory problem of 13 classes used in this investigation, random guessing would achieve about 8% correct classification. A dot product calculation produces 49.1% correct classification, while a distance measurement produces 58.7%. The results from this investigation are also qualitatively compared to previous work using infrared data which retained some intensity information. It is found that the binary packing of spectral data shows great promise in the area of infrared analysis.

Use of FT-NIR transmission spectroscopy for the quantitative analysis of an active ingredient in a translucent pharmaceutical topical gel formulation

The objective of this study was to demonstrate the use of transmission Fourier transform near-infrared (FT-NIR) spectroscopy for quantitative analysis of an active ingredient in a translucent gel formulation. Gels were prepared using Carbopol 980 with 0%, 1%, 2%, 4%, 6%, and 8% ketoprofen and analyzed with an FT-NIR spectrophotometer operated in the transmission mode. The correlation coefficient of the calibration was 0.9996, and the root mean squared error of calibration was 0.0775%. The percent relative standard deviation for multiple measurements was 0.10%. The results prove that FT-NIR can be a good alternative to other, more time-consuming means of analysis for these types of formulations.

Pattern Recognition Methods for the Classification of Binary Infrared Spectral Data

Five pattern recognition methods are compared for their ability to classify binary infrared spectra. Included is a discussion of the time vs success balance for each of the techniques. Predictive ability decreases in the order maximum likelihood > distance > Tanimoto similarity ∼ Hamming distance > dot product. The time required for each prediction after the classifier has been developed increases in order maximum likelihood ∼ distance ∼ dot product < Tanimoto similarity ∼ Hamming distance.

Determination of Wavelength Accuracy in the Near-Infrared Spectral Region Based on NIST's Infrared Transmission Wavelength Standard SRM 1921

A major concern with the use of near-infrared (NIR) spectroscopy in many QA/QC laboratories is the need for a simple reliable method of verifying the wavelength accuracy of the instrument. This requirement is particularly important in near-infrared spectroscopy because of the heavy reliance on sophisticated statistical vector analysis techniques to extract the desired information from the spectra. These techniques require precise alignment of the data points between the vectors corresponding to the standard and sample spectra. The National Institute of Standards and Technology (NIST) offers a Standard Reference Material (SRM 1921) for the verification and calibration of mid-infrared spectrometers in the transmittance mode. This standard consists of a 38 μm-thick film of polystyrene plastic. While SRM 1921 works well as a mid-infrared standard, a thicker sample is required for use as a routine standard in the near-infrared spectral region. The general acceptance and proven reliability of polystyrene as a standard reference material make it a very good candidate for a cost-effective NIR standard that could be offered as an internal reference for every instrument. In this paper we discuss a number of the parameters in a Fourier transform (FT)-NIR instrument that can affect wavelength accuracy. We also report a number of experiments designed to determine the effects of resolution, sample position, and optics on the wavelength accuracy of the system. In almost all cases the spectral reproducibility was better than one wavenumber of the values extrapolated from the NIST reference material. This finding suggests that a thicker sample of polystyrene plastic that has been validated with the SRM 1921 standard would make a cost-effective reference material for verifying wavelength accuracy in a medium-resolution FT-NIR spectrometer.

Expert system for interpretation of the infrared spectra of environmental mixtures

Abstract A program for the identification of the principal components of mixtures through interpretation of the infrared mixture spectrum (IntIRpret) was developed. This program, which was developed as a preliminary screening tool for unknown organic mixtures, has five main subroutines: the interferogram processing and peak-selection subroutine (PUSHSUB), the automated knowledge-acquisition subroutine (AUTOGEN), the system optimization subroutine (STO), the interpretation subroutine (PAIRS), and final processing subroutine to subtract spectral similarity (PAIRSPLUS). Principal advantages of this system compared to earlier systems are speed, flexibility and accuracy.

Density Estimations and the Characterization of Binary Infrared Spectra

A truncated orthogonal expansion has been used to represent binary data taken from a multidimensional file of infrared data. The expansion represents an approximation for the true class conditional probability density functions (pdfs). As a first approximation, statistical independence is assumed and the only terms necessary are the estimated class conditional probabilities for each peak. A more accurate estimation of the pdf is attained when a second term, a correlation term, is included in the expansion. The data set consists of 2600 spectra in thirteen mutually exclusive classes with each spectrum represented by 139 dimemqions. Results are obtained for a maximum discriminant function case, as well as for pairwise discrimination among the classes. For the thirteen class problem, correct classification occurs 67.2% of the time by the class conditional probabilities and 87.3% of the time when the correlation terms are inclrlded. For pairwise discrimination, the results are 92.9% and 98.1% respectively.

Application of a text search system based on boolean strategy to mass spectral data identification

Abstract A general algorithm for text searching, operated on a tape-based minicomputer, has already been reported. This paper presents the application of the general text-searching algorithms to the Registry of Mass Spectral Data of 18,806 different entries. The text format allows multi-information input to be used to search the spectral library on the basis of data not necessarily extracted from mass spectra. Two library files have been generated; one is approximately half the size of the other, less important information having been deleted. The shorter library contains all 18,806 entries but enjoys much faster search times. Batch processing of searches is also possible. The text search is shown to be versatile in its operation, as the user can construct searches to be either broad or very selective, depending on the application. The search also has the capability to examine the data base internally and to check certain data for their validity.

Progressive filter network: A general classification algorithm☆

Abstract Mass spectral data and EKG data have been used for the development of a Progressive Filter Network (PFN) algorithm which should be of wide utility. The PFN algorithm was compared to more sophisticated pattern recognition techniques and found to give comparable results. The advantages of the PFN algorithm are computational simplicity and the fact that the decision network generated may be used directly for feature selection.