No full text
Poster (Scientific congresses and symposiums)
How to analyse big data in hyperspectral imaging? Application for the elucidation of medicine composition
Coic, Laureen; Sacre, Pierre-Yves; Dispas, Amandine et al.
2020Chimiométrie 2020
 

Files


Full Text
No document available.

Send to



Details



Keywords :
Out of memory issue; big data; hyperspectral imaging; Spectral unmixing algorithm
Abstract :
[en] Nowadays, there is an increase of interest regarding hyperspectral imaging in the pharmaceutical field because of the highly valuable information provided. It is a powerful tool to get the composition of samples and to give the spatial distribution and/or homogeneity of each compound without destructing the sample. However, in the case of tablets, the surface to analyse can be wide (> 1 cm²) which provides huge file size (approximately 50 GB). The analysis of such kind of data is a challenge either on a common computer or on an affordable workstation because of “out of memory” issues. Those can sometimes be overcome by parallelizing functions but some interesting algorithms are not parallelizable. Moreover, central processing unit (CPU) computing is quickly limited. For example, spectral unmixing algorithms are rarely applicable on such amount of data without reducing it and, when possible, results are very long to obtain. Some strategies have to be developed to be able to analyse such data with a usual software, such as MATLAB®, on an affordable workstation. The objective of the study was to evaluate several spectral unmixing algorithms in combination with data reduction strategies to allow the elucidation of tablet composition. Two tablets of artemether/lumefantrine formulation were used for this study. One was a genuine Combiart® (manufacturer: Strides Arco Labs, batch 7227669, expiry date: June 2018) further called G1. The other one was a suspected falsified medicine Combiart® (manufacturer: Strides Arco Labs, batch 7225500, expiry date: August 2019) further called S1. The samples were gathered from a previous study [1]. The pharmaceutical tablets were analyzed with a FT-IR Cary 670/620 Agilent series microscope (Agilent Technologies) equipped with a 15x infrared objective, with a numerical aperture (NA) of 0.62 and a FPA (64x64) detector. The chemometric tools used for the study and that will be compared for this application are MCR-ALS, Vertex Component Analysis (VCA) and Pixel Purity Index (PPI). Moreover, two data reduction approaches will be evaluated: Kennard-Stone sub-sampling together with cube splitting and cube splitting only. Each pure spectrum was then matched with a homemade database using the hit quality index (HQI). The proposed approaches allowed analyzing data without “out of memory” issue and without losing any relevant information. Indeed, a preliminary test on the G1 sample with the different algorithms shown that the VCA algorithm elucidated the composition with an analysis time much smaller compared to other algorithms [1]. Moreover, the Kennard-Stone subsampling provided interesting results, probably because the information analyzed was much targeted and the spectral redundancy reduced. The workflow was then applied on the S1 sample and showed that the results were also improved compared to conventional protocols. Thanks to the development of the proposed workflow, it was possible to analyze big data to elucidate the composition of pharmaceutical tablets. It showed that easy-to-develop functions and cube splitting significantly decreased the memory consumption and analysis time (from 2 days to 3 hours) to obtain the same result for G1. In addition, it can be envisaged to use MCR-ALS algorithm in a second step after removing the pure spectra from VCA or PPI, to unmix the pixel information and thus, obtain the complete medicine composition elucidation.
Disciplines :
Pharmacy, pharmacology & toxicology
Physical, chemical, mathematical & earth Sciences: Multidisciplinary, general & others
Author, co-author :
Coic, Laureen  ;  Université de Liège - ULiège > Département de pharmacie > Chimie analytique
Sacre, Pierre-Yves  ;  Université de Liège - ULiège > Département de pharmacie > Chimie analytique
Dispas, Amandine  ;  Université de Liège - ULiège > Département de pharmacie > Analyse des médicaments
Dumont, Elodie ;  Université de Liège - ULiège > Département de pharmacie > Chimie analytique
De Bleye, Charlotte  ;  Université de Liège - ULiège > Département de pharmacie > Département de pharmacie
Fillet, Marianne ;  Université de Liège - ULiège > Département de pharmacie > Analyse des médicaments
Hubert, Philippe  ;  Université de Liège - ULiège > Département de pharmacie > Chimie analytique
Ziemons, Eric  ;  Université de Liège - ULiège > Département de pharmacie > Chimie analytique
Language :
English
Title :
How to analyse big data in hyperspectral imaging? Application for the elucidation of medicine composition
Publication date :
January 2020
Event name :
Chimiométrie 2020
Event organizer :
University of Liège (ULiège), CIRM, Vibra-Santé Hub
Event place :
Liège, Belgium
Event date :
du 27 janvier au 29 janvier 2020
Audience :
International
Funders :
FEDER - Fonds Européen de Développement Régional [BE]
F.R.S.-FNRS - Fonds de la Recherche Scientifique [BE]
Available on ORBi :
since 31 January 2020

Statistics


Number of views
134 (11 by ULiège)
Number of downloads
0 (0 by ULiège)

Bibliography


Similar publications



Contact ORBi