MEDSCANLATION: A Deep Learning-Based AI Scanner and Translation Device for Doctor's Prescription Medicine

A prescription is a legal document given to a pharmacist by a licensed physician to produce or administer pharmacological agents/medications to diagnose, cure, or treat an illness. A written prescription is the most common method of prescribing medications in the Philippines. The unreadability of a doctor's handwriting when prescribing medication leads to medication errors. Use of an incorrect dose of medication, or even death. As a result, doctors must rely on stenography or shorthand writing to quickly write vital information on prescriptions. The study aims to develop a device that helps pharmacists in performing jobs effectively and helps the pharmacy to translate the sloppily written prescriptions. In data gathering, the researchers visited two doctors to collect data in a handwritten prescription samples and two doctors wrote the names of five antibiotics on a piece of paper 200 times, producing 1000 handwritten prescription samples. This study conducted a data gathering process flow to come up with the final dataset with 11,000 iamges. The study showed also the training process of Custom SSD Mobilenet v2 Detector with the pre-process handwritten antibiotic names. The validation of the detection model are 92.4% accuracy and 57.4 milliseconds speed. The detection model was uploaded to the raspberry pi for the researchers to develop a device that can detect antibiotic names. The performance and functionality of the whole system were tested and evaluated. The system accurately detects the 39 antibiotic names out of 45 with an f-score of 87% accuracy. This shows and validates that the system is efficient and effective in detecting antibiotic's names.