Hyperspectral imaging sensor for optimization of small molecule formulations

Abstract

In this study, hyperspectral imaging (HSI) sensor was used to rapidly estimate the content of an active pharmaceutical ingredient (API) in powder blend samples in order to optimize small molecule formulation. Small molecule powder blend samples containing excipients and varying API concentrations were prepared using a blender. The spectrum of each powder blend was obtained using a short‐wave infrared hyperspectral imaging (SWIR HSI) system over a wavelength range of 1,000–2,500 nm. The use of the SWIR HSI method to predict API concentration in the powder blend samples was validated against that of a high‐performance liquid chromatography method. Partial least squares (PLS) regression and least squares support vector machine (LS‐SVM) analyses were used to build calibration models for predicting API concentration in the powder samples. Both the PLS and LS‐SVM models yielded high coefficients of determination of 0.99 and low errors (root‐mean‐square error of prediction) for API content prediction, which were 0.73 and 0.60 mg, respectively. Furthermore, image processing algorithms were developed to visualize the predicted API concentration in each pixel of the powder surface. Concentration map and binary images were also used to visualize the API concentration in the powder samples. The results suggest that the HSI technique permits the quantification and visualization of pharmaceutical ingredients and could be easily used during manufacturing for the non‐destructive formulations optimization and quality control of products.