Abstract
Artificial intelligence (AI)-based pill-counting applications have been introduced into clinical practice to streamline the work of pharmacists. However, no study has evaluated the safety of these drugs. The purpose of this study was to evaluate AI's ability to recognize Yakushimarukazuko(®) (NeoX Inc., Tokyo, Japan) and identify operational issues. The results confirmed that the transparency, overlap, and direction of the supplements used in the assessment were important factors in AI-based recognition. Accurately identifying and addressing such operational issues with AI-based pill-counting applications will contribute to minimizing the risk of errors and, in turn, enhance medical safety.