Development of near infrared spectroscopy methodology for human albumin determination using a new calibration...

Though near infrared spectroscopy (NIRS) has been applied widely in the field of pharmaceutical, there is still a bottleneck which limits its development. The main barrier is that conventional NIRS calibration method is based on experiences and trials, which causes the established model is not stable and difficult to explain. Therefore, a new strategy which was based on design of experiment (DoE) combined with statistical analysis was provided to solve the limitations. A pre-processing method library was set up first and orthogonal experiment design was then introduced to investigate the effects and interactions of different pre-processing methods. Paired t-test was used to select the most suitable pre-processing method. Finally, the pre-processing method selected above and three commonly used variable selection methods (CARS, UVE, VIP) were combined randomly to select the best calibration model. The results showed that the new calibration approach could provide a reasonable way for researchers to establish a more stable, objective calibration model.