Applying hyperspectral imaging to continuous processing of pharmaceuticals

Dziewierz, Jerzy and McGinty, John and Macfhionnghaile, Pol and Svoboda, Vaclav and Sefcik, Jan and Gachagan, Anthony and Nordon, Alison and Marshall, Stephen and Cleary, Alison (2016) Applying hyperspectral imaging to continuous processing of pharmaceuticals. In: Solving problems with spectral imaging, 2016-02-04 - 2016-02-04, London.

[img]
Preview
Text (Dziewierz-etal-2015-Applying-hyperspectral-imaging-to-continuous)
Dziewierz_etal_2015_Applying_hyperspectral_imaging_to_continuous.pdf
Final Published Version

Download (1MB)| Preview

    Abstract

    Introduction and objectives: In moving from batch to continuous manufacturing of pharmaceutical products, knowledge of all experimental variables is required to help control and achieve a stable system that yields a consistent product with the desired attributes. Spectroscopic tools are often used to provide point measurements at key points in the process. Here we demonstrate the applicability of Hyperspectral Imaging (HSI) to continuous nuclaetion processes. The objective of the work described here was to show how HSI can be used to monitor mixing processes by providing spatially discriminated near-infrared spectra, yielding vital process information that has only previously been estimated using simulation techniques such as computational fluid dynamics (CFD). Methods: Water (antisolvent) and an IPA/water mixture (solvent) containing dissolved paracetamol seed crystals were mixed in a tube by introducing the antisolvent jet at different flow rates, and the resulting mixing in the tube was imaged with an InnoSpec RedEye camera in the spectral range 950 to 1700 nm. Microscope images of the resultant crystals from each different flow rate were taken to confirm that differences in the final crystal product were observed. Machine learning techniques in the form of Support Vector Machine (SVM) analysis were used to analyse and automatically separate the spectral data of the solvent/antisolvent mix into the different components. Results: The antisolvent jet, which resulted in nucleation near the point of injection, could clearly be identified and therefore monitored after application of the SVM. Differences between different flow rates and concentrations were observed from the hyperspectral images obtained, and these differences carried through to the shape and size of the final crystals obtained. Conclusions: We have demonstrated the applicability of HSI and advanced data processing techniques to the monitoring of mixing dynamics, in particular those used in continuous pharmaceutical processing such as solvent/antisolvent crystallisation.