High-Throughput eSWIR Hyperspectral Imaging for Pharmaceutical Classification at Industrial Line Speeds

Real-time quality control of pharmaceutical dosage forms requires rapid, non-destructive analytical technologies that provide chemical specificity while maintaining high throughput. Here, an extended short-wave infrared (eSWIR, 1000-2500 nm) line-scan hyperspectral imaging system (L-EOS 2.5e, Photon Etc.) was evaluated for spatially resolved classification of pharmaceutical tablets. The system was evaluated at integration times of 2, 0.2, and 0.02 ms using Standard Normal Variate normalization, Savitzky-Golay second-derivative preprocessing, and an error-correcting output codes support vector machine classifier. Product-level classification of 11 pharmaceutical products achieved accuracies of 99.99%, 100%, and 97.6% at 2, 0.2, and 0.02 ms, respectively. A more challenging manufacturer-level discrimination task involving three 500 mg acetaminophen products containing the same active pharmaceutical ingredient and labeled dose achieved accuracies of 99.7%, 98.6%, and 89.3% at the corresponding integration times. These results demonstrate the potential of eSWIR hyperspectral imaging for rapid, spatially resolved pharmaceutical product discrimination and high-throughput inspection.