Image processing and deep learning applications in herbal microscopy: a systematic review

Abstract

The integration of image processing and deep learning techniques into herbal microscopy holds the potential to revolutionize the identification and analysis of herbal specimens, facilitating rapid assessment of medicinal plants, detection of toxic species, and quality control in the herbal drug industry. Despite growing interest, a comprehensive overview of current methodologies and applications in this specialized field is lacking. This systematic review aims to identify which deep learning tasks are currently employed in herbal or botanical microscopic image analysis, determine the methods used for these tasks, and evaluate their effectiveness. Adhering to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 (PRISMA 2020) guidelines, a systematic literature search was conducted primarily using the Web of Science database. The search targeted studies published from January 2014 up to August 2025 and employed a complex strategy with Boolean operators, wildcards, and composite terms. Inclusion criteria encompassed peer-reviewed, English-language articles applying image processing and/or deep learning techniques to herbal microscopy for purposes such as identification, classification, or analysis. Resulting sections summarize the findings of the selected studies, focusing on the types of image processing and deep learning methods employed, their effectiveness, and their applications within herbal microscopy, including their roles in the quick identification of medicinal plants, assessment of toxic plants, and quality control in the herbal drug industry. By identifying existing advancements and research gaps, this review aims to guide future studies and support the development of more advanced, automated techniques in the field, ultimately enhancing the efficiency and accuracy of herbal drug analysis and safety assessments.