Structural modelling for smart healthcare: quantitative phenotype gaining network (QPGnet) for taxonomic identification of complementary medicines

A primary challenge for medical institutions is management and identification of complementary medicines. Medicinal materials originating from plants, animals, fungi, bacteria, and minerals manifest intricate pharmacognostic traits. Structural modelling is fundamental for digital governance and optimized administration of distribution networks. Crude medicine traits exhibit diverse phenotypic characteristics, including structures, textures, colours, compositions, topologies, and transparencies. These varied combinations can lead to phenotypically similar yet distinct medicines; this similarity hinders identification and compromises classification accuracy. Consequently, high phenotypic similarity does not guarantee taxonomic identity, and large-scale high-similarity medicine identification results in poor consistency and discriminative power when identifying medicines across multiple categories. To ameliorate effects, we propose a structural modelling approach via quantified phenotypic characteristics for identifying crude medicines. This method iteratively extracts and learns from phenotypic characteristics to generate robust representations as decision-supportive discriminative medicinal signatures and substantially improves the efficiency of large-scale taxonomic identification across 172 categories, achieving 97% precision, 96% recall, a 96% F1-score, and 96.5% accuracy.