Huwei Tan

Transfer of multivariate calibration models: a review

Multivariate calibration models are of critical importance to many analytical measurements, particularly for spectroscopic data. Generally, considerable effort is placed into constructing a robust model since it is meant to be used for extended periods of time. A problem arises, though, when the samples to be predicted are measured on a different instrument or under differing environmental factors from those used to build the model. The changes in spectral variations between the two conditions may make the model invalid for prediction in the new system. Various standardization and preprocessing methods have been developed to enable a calibration model to be effectively transferred between two systems, thus eliminating the need for a full recalibration. This paper presents an overview of the different methods used for calibration transfer and a critical assessment of their validity and applicability. The focus is on methods for transfer of near-infrared (NIR) spectra.

Detection of Lipid Core Coronary Plaques in Autopsy Specimens With a Novel Catheter-Based Near-Infrared Spectroscopy System

OBJECTIVES This study sought to assess agreement between an intravascular near-infrared spectroscopy (NIRS) system and histology in coronary autopsy specimens. BACKGROUND Lipid core plaques cannot be detected by conventional tests, yet are suspected to be the cause of most acute coronary syndromes. Near-infrared spectroscopy is widely used to determine the chemical content of substances. A NIRS system has been developed and used successfully in 99 patients. METHODS Scanning NIRS was performed through blood in 212 coronary segments from 84 autopsy hearts. One histologic section was analyzed for every 2 mm of artery. Lipid core plaque of interest (LCP) was defined as a lipid core >60 degrees in circumferential extent, >200-microm thick, with a mean fibrous cap thickness <450 microm. The first 33 hearts were used to develop the algorithm; the subsequent 51 validation hearts were used in a prospective, double-blind manner to evaluate the accuracy of NIRS in detecting LCP. A NIRS-derived lipid core burden index for an entire artery was also validated by comparison to histologic findings. RESULTS The LCPs were present in 115 of 2,649 (4.3%) sections from the 51 validation hearts. The algorithm prospectively identified LCP with a receiver-operator characteristic area of 0.80 (95% confidence interval [CI]: 0.76 to 0.85). The lipid core burden index detected the presence or absence of any fibroatheroma with an area under the curve of 0.86 (95% CI: 0.81 to 0.91). A retrospective analysis of lipid core burden index conducted in extreme artery segments with either no or extensive fibroatheroma yielded an area under the curve of 0.96 (95% CI: 0.92 to 1.00), confirming the accuracy of spectroscopy in identifying plaques with markedly different lipid content under ideal circumstances. CONCLUSIONS This novel catheter-based NIRS system accurately identified lipid core plaques through blood in a prospective study in coronary autopsy specimens. It is expected that this novel capability will be of assistance in the management of patients with coronary artery disease.

Piecewise orthogonal signal correction

Abstract A novel signal-processing method that performs orthogonal signal correction (OSC) in a piecewise manner, namely piecewise OSC (POSC), is developed and applied to two near-infrared (NIR) data sets of multivariate calibration. Partial least squares (PLS) regression models were constructed for the POSC-corrected spectra, and the results were compared with those obtained by the Wise and Fearn OSC algorithms. It is shown that performing POSC prior to calibration yields regression models that are more parsimonious (fewer latent variables) and with better predictive power than models obtained by the above methods. The removal of orthogonal components from the response matrix is greatly facilitated simply by considering localized spectral features.

Robust Calibration with Respect to Background Variation

The application of linear regression on wavelet coefficients for robust calibration of spectral data with highly variable background was successfully demonstrated with synthetic and real data. A Monte Carlo study was made to investigate the performance of the methods in both the cases where the background variation in the prediction set was the same as in the calibration set and where the variation was different. Multivariate linear regression on wavelet coefficients proved to be competitive in the first case and superior in the second case with respect to partial least squares (PLS) calibration. Results on real near-infrared (NIR) data confirmed the simulation study. As a study of regression on wavelet coefficients, this is the first application study of regression on wavelet coefficients that shows how the wavelets property of vanishing moments can be used for reducing the effects of varying background. As a background correction method, the proposed approach avoided errors introduced in the estimation process. In addition, the strategy proposed here can be applied to data collected by various other analytical techniques as well.

Multivariate calibration of spectral data using dual-domain regression analysis

To date, few efforts have been made to take simultaneous advantage of the local nature of spectral data in both the time and frequency domains in a single regression model. We describe here the use of a novel chemometrics algorithm using the wavelet transform. We call the algorithm dual-domain regression, as the regression step defines a weighted model in the time-domain based on the contributions of parallel, frequency-domain models made from wavelet coefficients reflecting different scales. In principle, any regression method can be used, and implementation of the algorithm using partial least squares regression and principal component regression are reported here. The performance of the models produced from the algorithm is generally superior to that of regular partial least squares (PLS) or principal component regression (PCR) models applied to data restricted to a single domain. Dual-domain PLS and PCR algorithms are applied to near infrared (NIR) spectral datasets of Cargill corn samples and sets of spectra collected on batch chemical reactions run in different reactors to illustrate the improved robustness of the modeling.

Improvement of a Standard-Free Method for Near-Infrared Calibration Transfer

Previously, a standard-free method using the finite impulse response (FIR) filter was successfully employed to transfer the NIR spectra of caustic brines, analgesics, and terpolymer resins. This paper carries the FIR transfer method one step further, leading to an improved algorithm that makes the transfer more robust and general by avoiding transfer artifacts in the filtered spectra. Investigations of the theoretical aspects and application examples of diffuse reflectance NIR datasets show that in comparison with our previous version the improved method is much easier to use and gives artifact-free transferred spectra. The improved method also compares favorably with other current signal correction methods for calibration transfer. With the additional advantage of not requiring a subset of standards to be measured on main and remote instruments, the proposed method is a very useful alternative for calibration transfer.

Improved Piecewise Orthogonal Signal Correction Algorithm

Piecewise orthogonal signal correction (POSC), an algorithm that performs local orthogonal filtering, was recently developed to process spectral signals. POSC was shown to improve partial least-squares regression models over models built with conventional OSC. However, rank deficiencies within the POSC algorithm lead to artifacts in the filtered spectra when removing two or more POSC components. Thus, an updated OSC algorithm for use with the piecewise procedure is reported. It will be demonstrated how the mathematics of this updated OSC algorithm were derived from the previous version and why some OSC versions may not be as appropriate to use with the piecewise modeling procedure as the algorithm reported here.

Wavelet packet denoising robust regression applied to estimation of equivalent circuit parameters for thickness‐shear‐mode acoustic wave sensor

An Erratum has been published for this article in Journal of Chemometrics 14(1) 2000, 47.

A catheter-based near-infrared scanning spectroscopy system for imaging lipid-rich plaques in human coronary arteries in vivo

Although heart disease remains the leading cause of death in the industrialized world, there is still no method, even under cardiac catheterization, to reliably identify those atherosclerotic lesions most likely to lead to heart attack and death. These lesions, which are often non-stenotic, are frequently comprised of a necrotic, lipid-rich core overlaid with a thin fibrous cap infiltrated with inflammatory cells. InfraReDx has developed a scanning, near-infrared, optical-fiber-based, spectroscopic cardiac catheter system capable of acquiring NIR reflectance spectra from coronary arteries through flowing blood under automated pullback and rotation in order to identify lipid-rich plaques (LRP). The scanning laser source and associated detection electronics produce a spectrum in 5 ms at a collection rate of 40 Hz, yielding thousands of spectra in a single pullback. The system console analyzes the spectral data with a chemometric model, producing a hyperspectral image (a Chemogram, see figure below) that identifies LRP encountered in the region interrogated by the system. We describe the system architecture and components, explain the experimental procedure by which the chemometric model was constructed from spectral data and histology-based reference information collected from autopsy hearts, and provide representative data from ongoing ex vivo and clinical studies.