Medicine box: Doctor's prescription recognition using deep machine learning

Abstract

A Doctor's prescription is a handwritten document written by doctors in the form of instructions that describes list of drugs for patients in time sickness, injuries and other disability problems. While we receiving a new prescription from doctor, it is unable to understand what drug name is prescribed on it. In most cases, however, we wouldn't be able to read it anyway because doctors use Latin abbreviations and medical terminologies on prescriptions that are not understandable by the general persons which make reading it very difficult. According to the National Academy of Sciences estimates that at least 1.5 million peoples are sickened, injured or killed each year by errors while reading prescription. This paper resolves the problems in doctor?s prescriptions through Medicine Box, and Smart phone application that uses Conventional Neural Network (CNN) to recognize handwritten medicine names and return readable digital text. This mobile application uses TensorFlow as the machine learning library, and Custom Repository to match the partial string with the drug name. With Medicine Box, cases of misinterpretation of medicine names can be decreased. This makes the ordinary persons to understand what doctor is prescribed in the prescription and also help for pharmacists.