Variable clustering and spectral angle mapper-orthogonal projection method for Raman mapping of compound detection in tablets

Attila Farkas, Brigitta Nagy, Balázs Démuth, Attila Balogh, Hajnalka Pataki, Zsombor K. Nagy, György Marosi

Research output: Contribution to journalArticle

2 Citations (Scopus)

Abstract

Raman mapping and chemometrics are proposed to accurately characterize the composition of tablets. The most critical step of the state-of-art curve resolution methods (such as multivariate curve resolution-alternating least squares [MCR-ALS]) is the determination of the number of constituents, when chemical imaging is coupled with multivariate data analysis. However, it is usually performed in a considerably subjective way. We propose a variable clustering approach for the identification of the main dimensionality of vibrational spectral data. The method was tested on a Raman map of a complex pharmaceutical tablet that contained 4 major components with high spectral resemblance, and a low-dose lubricant was also added for tableting purposes. Using a variable clustering algorithm called VARCLUS we were able to construct clusters from the Raman mapping data corresponding to the real constitution of the sample. The modeled clusters were analyzed by the “sum of ranking differences” method. All 4 major components could be identified. The potential of the clustering algorithm was further assessed by applying MCR-ALS and spectral angle mapper-orthogonal projection methods. We have shown that variable clustering corresponded with MCR-ALS results and that it can be used to characterize the qualitative composition of an unknown pharmaceutical sample by combining the clustering algorithm with a pure component resolution method. Therefore, this method is well applicable to analyze and interpret the curve resolution of complex samples. Testing of the previously studied spectral angle mapper-orthogonal projection method, which relies on spectral reference libraries and even the low-dose lubricant (approximately 1% w/w), was identified through the chemical imaging.

Original languageEnglish
Article numbere2861
JournalJournal of Chemometrics
Volume31
Issue number1
DOIs
Publication statusPublished - Jan 1 2017

Fingerprint

Orthogonal Projection
Raman
Projection Method
Clustering algorithms
Tablets
Multivariate Curve Resolution
Alternating Least Squares
Clustering
Angle
Drug products
Curve Resolution
Lubricants
Clustering Algorithm
Pharmaceuticals
Imaging techniques
Dose
Chemical analysis
Pharmaceutical Preparations
Imaging
Multivariate Data Analysis

Keywords

  • chemical imaging
  • constituent number detection
  • spectral angle mapper
  • sum of ranking differences
  • variable clustering

ASJC Scopus subject areas

  • Analytical Chemistry
  • Applied Mathematics

Cite this

Variable clustering and spectral angle mapper-orthogonal projection method for Raman mapping of compound detection in tablets. / Farkas, Attila; Nagy, Brigitta; Démuth, Balázs; Balogh, Attila; Pataki, Hajnalka; Nagy, Zsombor K.; Marosi, György.

In: Journal of Chemometrics, Vol. 31, No. 1, e2861, 01.01.2017.

Research output: Contribution to journalArticle

Farkas, Attila ; Nagy, Brigitta ; Démuth, Balázs ; Balogh, Attila ; Pataki, Hajnalka ; Nagy, Zsombor K. ; Marosi, György. / Variable clustering and spectral angle mapper-orthogonal projection method for Raman mapping of compound detection in tablets. In: Journal of Chemometrics. 2017 ; Vol. 31, No. 1.
@article{2ca7239d9c4b453b9bed3a81c706b4ff,
title = "Variable clustering and spectral angle mapper-orthogonal projection method for Raman mapping of compound detection in tablets",
abstract = "Raman mapping and chemometrics are proposed to accurately characterize the composition of tablets. The most critical step of the state-of-art curve resolution methods (such as multivariate curve resolution-alternating least squares [MCR-ALS]) is the determination of the number of constituents, when chemical imaging is coupled with multivariate data analysis. However, it is usually performed in a considerably subjective way. We propose a variable clustering approach for the identification of the main dimensionality of vibrational spectral data. The method was tested on a Raman map of a complex pharmaceutical tablet that contained 4 major components with high spectral resemblance, and a low-dose lubricant was also added for tableting purposes. Using a variable clustering algorithm called VARCLUS we were able to construct clusters from the Raman mapping data corresponding to the real constitution of the sample. The modeled clusters were analyzed by the “sum of ranking differences” method. All 4 major components could be identified. The potential of the clustering algorithm was further assessed by applying MCR-ALS and spectral angle mapper-orthogonal projection methods. We have shown that variable clustering corresponded with MCR-ALS results and that it can be used to characterize the qualitative composition of an unknown pharmaceutical sample by combining the clustering algorithm with a pure component resolution method. Therefore, this method is well applicable to analyze and interpret the curve resolution of complex samples. Testing of the previously studied spectral angle mapper-orthogonal projection method, which relies on spectral reference libraries and even the low-dose lubricant (approximately 1{\%} w/w), was identified through the chemical imaging.",
keywords = "chemical imaging, constituent number detection, spectral angle mapper, sum of ranking differences, variable clustering",
author = "Attila Farkas and Brigitta Nagy and Bal{\'a}zs D{\'e}muth and Attila Balogh and Hajnalka Pataki and Nagy, {Zsombor K.} and Gy{\"o}rgy Marosi",
year = "2017",
month = "1",
day = "1",
doi = "10.1002/cem.2861",
language = "English",
volume = "31",
journal = "Journal of Chemometrics",
issn = "0886-9383",
publisher = "John Wiley and Sons Ltd",
number = "1",

}

TY - JOUR

T1 - Variable clustering and spectral angle mapper-orthogonal projection method for Raman mapping of compound detection in tablets

AU - Farkas, Attila

AU - Nagy, Brigitta

AU - Démuth, Balázs

AU - Balogh, Attila

AU - Pataki, Hajnalka

AU - Nagy, Zsombor K.

AU - Marosi, György

PY - 2017/1/1

Y1 - 2017/1/1

N2 - Raman mapping and chemometrics are proposed to accurately characterize the composition of tablets. The most critical step of the state-of-art curve resolution methods (such as multivariate curve resolution-alternating least squares [MCR-ALS]) is the determination of the number of constituents, when chemical imaging is coupled with multivariate data analysis. However, it is usually performed in a considerably subjective way. We propose a variable clustering approach for the identification of the main dimensionality of vibrational spectral data. The method was tested on a Raman map of a complex pharmaceutical tablet that contained 4 major components with high spectral resemblance, and a low-dose lubricant was also added for tableting purposes. Using a variable clustering algorithm called VARCLUS we were able to construct clusters from the Raman mapping data corresponding to the real constitution of the sample. The modeled clusters were analyzed by the “sum of ranking differences” method. All 4 major components could be identified. The potential of the clustering algorithm was further assessed by applying MCR-ALS and spectral angle mapper-orthogonal projection methods. We have shown that variable clustering corresponded with MCR-ALS results and that it can be used to characterize the qualitative composition of an unknown pharmaceutical sample by combining the clustering algorithm with a pure component resolution method. Therefore, this method is well applicable to analyze and interpret the curve resolution of complex samples. Testing of the previously studied spectral angle mapper-orthogonal projection method, which relies on spectral reference libraries and even the low-dose lubricant (approximately 1% w/w), was identified through the chemical imaging.

AB - Raman mapping and chemometrics are proposed to accurately characterize the composition of tablets. The most critical step of the state-of-art curve resolution methods (such as multivariate curve resolution-alternating least squares [MCR-ALS]) is the determination of the number of constituents, when chemical imaging is coupled with multivariate data analysis. However, it is usually performed in a considerably subjective way. We propose a variable clustering approach for the identification of the main dimensionality of vibrational spectral data. The method was tested on a Raman map of a complex pharmaceutical tablet that contained 4 major components with high spectral resemblance, and a low-dose lubricant was also added for tableting purposes. Using a variable clustering algorithm called VARCLUS we were able to construct clusters from the Raman mapping data corresponding to the real constitution of the sample. The modeled clusters were analyzed by the “sum of ranking differences” method. All 4 major components could be identified. The potential of the clustering algorithm was further assessed by applying MCR-ALS and spectral angle mapper-orthogonal projection methods. We have shown that variable clustering corresponded with MCR-ALS results and that it can be used to characterize the qualitative composition of an unknown pharmaceutical sample by combining the clustering algorithm with a pure component resolution method. Therefore, this method is well applicable to analyze and interpret the curve resolution of complex samples. Testing of the previously studied spectral angle mapper-orthogonal projection method, which relies on spectral reference libraries and even the low-dose lubricant (approximately 1% w/w), was identified through the chemical imaging.

KW - chemical imaging

KW - constituent number detection

KW - spectral angle mapper

KW - sum of ranking differences

KW - variable clustering

UR - http://www.scopus.com/inward/record.url?scp=85010341356&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85010341356&partnerID=8YFLogxK

U2 - 10.1002/cem.2861

DO - 10.1002/cem.2861

M3 - Article

VL - 31

JO - Journal of Chemometrics

JF - Journal of Chemometrics

SN - 0886-9383

IS - 1

M1 - e2861

ER -