An explainable hybrid machine learning approach for automated drug identification for safe dosing in smart healthcare systems

Overdoses and improper dosages of medicines are a big challenge in the medical sector, as they may harm and endanger the lives of people. This paper introduces the automatic detection of pharmaceutical compounds directly from images of medicine packaging. The Machine Learning (ML)-enabled model is used to classify medicinal compounds. Additionally, we have applied eXplainable Artificial Intelligence (XAI) to enhance the transparency of the proposed model. Moreover, we have also employed OpenCV-based image preparation techniques, including Gaussian blurring, binarization, and conversion to grayscale, to reduce computational demands. Empirical results achieved 90% to 94% accuracy over multiple executions. The experimental outcomes validate the efficacy of the proposed model, offering transparency and interpretability.