A Deep Learning Model for Medical Prescription Transcription: Integrating CNN and OCR for Enhanced Accuracy

Abstract

Inaccurate interpretation of handwritten medical prescriptions is a pressing problem in healthcare, often leading to medication errors and adversely affecting patient safety. The complexity of deciphering diverse handwriting styles necessitates an automated, accurate transcription solution. This research addresses the critical question: How can deep learning improve the accuracy and efficiency of medical prescription transcription? We developed an advanced deep learning model combining Convolutional Neural Networks (CNN) and Optical Character Recognition (OCR) to accurately transcribe handwritten prescriptions. The methodology involved curating a dataset of prescription images, preprocessing for optimal deep learning application, and training the model to recognize and transcribe text. Our model achieved a significant breakthrough with an accuracy of 91.80% for English and 87.50% for Bangla scripts, demonstrating a robust ability to handle real-world prescription variability. The results affirm the effectiveness of integrating CNN and OCR in solving the problem of prescription transcription. The goals of enhanced patient safety and streamlined healthcare documentation have been substantially achieved. The practical implementation of this model has the potential to drastically reduce medication errors, contribute to theoretical advances in AI applications in healthcare, and bear significant ethical implications by improving patient outcomes. This research presents a novel approach to prescription transcription, offering a valuable tool for healthcare professionals. It sets a new precedent in medical documentation, paving the way for future innovations and serving as a benchmark for similar applications in healthcare technology.