Micro-Electro-Mechanical Systems/Near-Infrared Validation of Different Sampling Modes and Sample Sets Coupled with Multiple Models

The aim of the present study was to demonstrate the reliability of micro-electro-mechanical systems/near-infrared technology by investigating analytical models of two modes of sampling (integrating sphere and fiber optic probe modes) and different sample sets. Baicalin in Yinhuang tablets was used as an example, and the experimental procedure included the optimization of spectral pretreatments, selection of wavelength regions using interval partial least squares, moving window partial least squares, and validation of the method using an accuracy profile. The results demonstrated that models that use the integrating sphere mode are better than those that use fiber optic probe modes. Spectra that use fiber optic probe modes tend to be more susceptible to interference information because the intensity of the incident light on a fiber optic probe mode is significantly weaker than that on an integrating sphere mode. According to the test set validation result of the method parameters, such as accuracy, precision, risk, and linearity, the selection of variables was found to make no significant difference to the performance of the full spectral model. The performance of the models whose sample sets ranged widely in concentration (i.e., 1-4%) was found to be better than that of models whose samples had relatively narrow ranges (i.e., 1-2%). The establishment and validation of this method can be used to clarify the analytical guideline in Chinese herbal medicine about two sampling modes and different sample sets in the micro-electro-mechanical systems/near-infrared technique.