TY - JOUR T1 - Classification of structurally related commercial contrast media by near infrared spectroscopy. AU - Yip,Wai Lam, AU - Soosainather,Tom Collin, AU - Dyrstad,Knut, AU - Sande,Sverre Arne, Y1 - 2013/12/07/ PY - 2013/07/10/received PY - 2013/11/27/revised PY - 2013/11/29/accepted PY - 2013/12/31/entrez PY - 2014/1/1/pubmed PY - 2014/10/7/medline KW - %ICC KW - BVE-PLS KW - BiPLS KW - CSMWPLS KW - LV KW - MIPCR KW - MRI KW - MWPLS KW - Main and Interactions of Individual Principal Components Regression KW - Main and Interactions of Individual Principal Components Regression (MIPCR) KW - Multivariate classification KW - NIRS KW - Near infrared spectroscopy (NIRS) KW - PAT KW - PC KW - PCA KW - PLS-DA KW - PLSR KW - Partial least squares discriminant analysis (PLS-DA) KW - RMSECV KW - SCMWPLS KW - SIMCA KW - SNV KW - SVM KW - Soft independent modelling of class analogy (SIMCA) KW - Standard Normal Variate KW - backward interval partial least squares KW - backward variable elimination partial least squares regression KW - changeable size moving window partial least squares KW - iPLS KW - incorrect classification rate KW - interval partial least squares KW - latent variables KW - magnetic resonance imaging KW - moving window partial least squares KW - near infrared spectroscopy KW - partial least squares discriminant analysis KW - partial least squares regression KW - principal component KW - principal component analysis KW - process analytical technology KW - root mean square error of cross validation KW - search combination moving window partial least squares KW - soft independent modelling of class analogy KW - support vector machine SP - 148 EP - 60 JF - Journal of pharmaceutical and biomedical analysis JO - J Pharm Biomed Anal VL - 90 N2 - Near infrared spectroscopy (NIRS) is a non-destructive measurement technique with broad application in pharmaceutical industry. Correct identification of pharmaceutical ingredients is an important task for quality control. Failure in this step can result in several adverse consequences, varied from economic loss to negative impact on patient safety. We have compared different methods in classification of a set of commercially available structurally related contrast media, Iodixanol (Visipaque(®)), Iohexol (Omnipaque(®)), Caldiamide Sodium and Gadodiamide (Omniscan(®)), by using NIR spectroscopy. The performance of classification models developed by soft independent modelling of class analogy (SIMCA), partial least squares discriminant analysis (PLS-DA) and Main and Interactions of Individual Principal Components Regression (MIPCR) were compared. Different variable selection methods were applied to optimize the classification models. Models developed by backward variable elimination partial least squares regression (BVE-PLS) and MIPCR were found to be most effective for classification of the set of contrast media. Below 1.5% of samples from the independent test set were not recognized by the BVE-PLS and MIPCR models, compared to up to 15% when models developed by other techniques were applied. SN - 1873-264X UR - https://www.unboundmedicine.com/medline/citation/24374816/Classification_of_structurally_related_commercial_contrast_media_by_near_infrared_spectroscopy_ DB - PRIME DP - Unbound Medicine ER -