Fresh and Expired Medicine Detection using Hyperspectral Imaging and Machine Learning: Real-Time Pharmaceutical Quality Assurance

The detection of expired or degraded pharmaceuticals is critical for safeguarding public health, as consumption of expired drugs may result in adverse health effects or therapeutic failure. Traditional pharmaceutical quality assessment methods are often labor-intensive, time-consuming, and prone to human error. This research explores an innovative, non-destructive framework for real-time identification of expired medicines using Hyperspectral Imaging (HSI) with a Specim FX-10 camera operating in the 400–1000 nm wavelength range, combined with Machine Learning (ML) algorithms. Preprocessing includes dimensionality reduction, region of interest (ROI) selection for feature extraction, and the empirical line method (ELM) to analyze reflectance patterns. Post-processing employs the Savitzky–Golay (SG) filter for noise reduction and spectral smoothing. Multiple ML techniques including Nu-SVM, Polynomial-SVM, Linear-SVM, Reduced Error Tree (RET), Decision Tree models (C4.5, ID3), One-vs-All LDA, One-vs-One (OvO) LDA, ECOC-LDA, Ridge Regression, Lasso Regression, and Logistic Regression were evaluated, with OvO-LDA demonstrating superior performance in distinguishing fresh and expired medicines through discriminative spectral pattern learning. Models were trained and validated using a dedicated dataset of fresh and expired medicine samples. Precision, recall, F1-score, Jaccard index, Cohen’s kappa, Matthews Correlation Coefficient, and accuracy confirmed the effectiveness of the proposed pipeline, achieving a validation accuracy of 97%. Overall, this study establishes a robust multiclass framework for pharmaceutical degradation status classification, offering strong potential for real-time pharmaceutical quality assessment and mitigation of health risks associated with expired medicines.