Non‐Destructive Analysis of Phenolic and Flavonoid Contents in Medicinal Plant Powders Using Hyperspectral Imaging and Variable Selection

ABSTRACT

Phenolic and flavonoid contents in medicinal plants are essential to their growth and development and provide numerous health benefits, yet their quantification using traditional wet chemistry is labor‐intensive and time‐consuming. This study utilized the combination of two benchtop hyperspectral imaging (HSI) systems, namely short‐wave infrared (SWIR) and visible near‐infrared, for the non‐destructive quantitative estimation of the phenolic and flavonoid contents present in 12 powder samples of medicinal plants. For quantitative analysis, partial least squares regression (PLSR) prediction models used several spectral preprocessing techniques. To further enhance performance, variable importance in projection (VIP) selected the most informative spectral variables and combined them with PLSR to improve predictions. The optimized VIP‐PLSR models further improved the predictive performance, yielding coefficients of determination ( R ² ) of 0.985 and 0.996 and even lower standard root mean square error of prediction values of 0.165 and 1.150 mg/g for the respective compounds. Additionally, chemical imaging provided a clear spatial visualization of both compounds within the samples. These results underscore the potential of SWIR‐HSI combined with chemometrics as an effective tool for screening and quality control in medicinal plant powders.