Hyperspectral Imaging Techniques for Lyophilization: Advances in Data-Driven Modeling Strategies and Applications

Abstract

Lyophilization, aka freeze drying, is a key process used in the production of biotherapeutic products. The optimization of lyophilization formulations and operations is a slow process that could be accelerated by on-line analytics. In recent years, hyperspectral imaging (HSI) has garnered increasing attention from both academia and industry in biopharmaceutical and food engineering fields. As a non-invasive, rapid, non-destructive, accurate, and automated tool that combines advantages from both spectroscopy and imaging techniques, HSI holds significant potential for analyzing and optimizing lyophilization processes and products. However, the huge and information-rich datasets generated from HSI are difficult to be modeled and interpreted properly. This article reviews and discusses the literature on the application of HSI on lyophilization, and the strategies that use the resulting data to build models. Such strategies include preprocessing, spectral unmixing, classification and regression, and data fusion. From the data modeling and application perspectives, the current challenges and future prospects regarding HSI techniques for lyophilization are addressed. This article is intended to provide guidance and insights for non-specialist researchers and engineers into leveraging HSI and the data-driven modeling strategies for addressing a wide range of lyophilization-related challenges.