CNN Based Handwritten Prescription Recognition for Medicine Identification

Abstract

Doctors frequently write prescriptions in unreadable handwriting due to the growing demands on healthcare workers, making it difficult to correctly identify the names of the recommended medications. Patients are greatly affected by this problem since they could find it difficult to comprehend the prescription drugs they are meant to take. Because doctors' handwriting styles vary so much, no method has been able to completely address the challenge of recognizing handwritten medicine names despite multiple tries. In this work, we present a solution that uses machine learning techniques to identify handwritten pharmaceutical names. The system is implemented through a mobile application that captures prescription medicine images, preprocesses them with techniques such as image crop, and resizing, gray scaling, normalization and then classifies the images using a Convolutional Neural Network (CNN). The proposed system is evaluated using a dataset of handwritten medicine names, with the CNN model demonstrating an accuracy of 83.53%. By reducing medicine name misinterpretations, this technology helps patients and pharmacists ensure proper prescription consumption.