DocPresRec
S. Prabu, K. Joseph Abraham Sundar · 2023
Medical prescription is a crucial component in the healthcare system. Patients may be interested in knowing the information about their prescribed medicines before purchasing them. Doctors use complicated medical names, medical terminologies, and Latin abbreviations to write prescriptions; most patients and some pharmacists find it challenging to decipher doctor’s handwritten prescriptions. Misinterpreted drug names in the prescriptions affect patient safety. We propose a deep learning model, DocPresRec, to recognize handwritten English medical prescriptions. DocPresRec converts handwritten drug names into readable digital text. We employ an end-to-end trainable text spotter approach named MaskTextSpotter, to detect and recognize handwritten drug names from the prescription. Along with scene text detection and recognition datasets, we train our proposed model with handwritten datasets such as MNIST and IAM-OnDB. The proposed DocPresRec can recognize both printed and handwritten text from the medical prescription, and it achieves a recognition accuracy of 96.4% in the HP Labs dataset. The proposed model outperforms most of the existing methods comfortably. Misinterpretation of drug names can be reduced using DocPresRec. It will assist pharmacists in dispensing correct medicines. It will also benefit patients by learning more about their medications.