Development of a data fusion strategy combining FT-NIR and Vis/NIR-HSI for non-destructive prediction of critical quality attributes in traditional Chinese medicine particles

Abstract

This study explores the complementary capabilities of Fourier Transform Near Infrared Spectroscopy (FT-NIR) and Visible/Near Infrared Hyperspectral Imaging (Vis/NIR-HSI) in developing a data fusion strategy to predict the critical quality attributes (CQAs) of Traditional Chinese Medicine Particles (TCMP). The research emphasizes integrating these techniques into an advanced process analytical technology (PAT) platform. By leveraging the unique strengths of FT-NIR for molecular characterization and Vis/NIR-HSI for spatial quality assessment, the study evaluates multiple data fusion strategies to enhance prediction accuracy. Twenty batches of TCMP were produced using fluidized bed granulation, and their properties were characterized using FT-NIR and Vis/NIR-HSI. Comparative analysis revealed that FT-NIR outperformed Vis/NIR-HSI in standalone predictions of moisture content and particle size. Advanced fusion schemes were then developed to combine the complementary information from both spectral ranges, resulting in partial least squares (PLS) models. Among the three fusion levels evaluated, the high-level fusion strategy achieved the most accurate predictions for flowability, particle size, and moisture content. This study demonstrates that high-level fusion of FT-NIR and Vis/NIR-HSI data can significantly improve the efficiency and accuracy of CQAs prediction for TCMP. Moreover, the proposed approach facilitates rapid and non-destructive quality analysis of granular medicines, enables real-time online monitoring, and offers practical insights into advancing automated drug safety process control.