A Machine Learning Algorithms for Drug Identification: Enhancing Medication Safety in Thailand

Medication errors are one of the leading causes of patient harm in the healthcare system. Nowadays, there are many tools and solutions available to reduce the occurrence rate of medication errors that might be caused by Look-Alike, Sound-alike (LASA) drugs. This study focuses on the problem of drug identification by using a machine learning approach. First, a dataset is manually collected of commonly used drugs in Thailand. Then the object detection models are trained and evaluated based on the different version of pretrained YOLO models. The objective of this study is to develop and compare the performance of different versions of the trained model. The results show that YOLOv8-nano achieves the highest accuracy in drug identification, with a mAP score of 99.5%, a recall of 99.7%, and an F-l score of 99.4%, compared to other versions.