A novel approach of doctors' handwritten medical prescription recognition using a deep neural network and generating digital records of the patients

Although handwritten prescriptions are still frequently used in medical practice, automated recognition is severely hampered by their irregular spacing, variety of handwriting styles, and use of non-standard abbreviations. Current OCR techniques frequently rely on printed text, are insufficiently scalable to handle recently approved drug names, and are restricted to small or single-language datasets. The main research question arises from this gap: How can a multilingual, scalable, and highly accurate framework for the trustworthy recognition of handwritten medical prescriptions be developed? To close this gap, we suggest a hybrid deep learning OCR system that combines multilayer perceptrons (MLPs) and convolutional neural network (CNN) layers. Through GPU-accelerated mini-batch training, the model can be incrementally adjusted, is computationally efficient, and generalizes to various handwriting styles. The framework provides a reliable, scalable, and extremely accurate method for integrating handwritten prescriptions into electronic health records (EHR) systems through digitization. It was trained on a large dataset that contained medical, numerical, and textual elements.