Extraction of Handwritten Text from Doctors’ Prescriptions

The world today is witnessing the digitalization of every sector of the global economy and the health-care sector is no different. With the rapid development of modern health care, it is important to ensure reliability and accuracy in modern machinery and methods. The legibility and accuracy of doctor’s prescriptions ensure patient wellness and efficient pharmaceutical management. In this paper, a handwritten text recognition model using Convolutional Neural Network which analyzes and enhances English handwritten text from medical prescriptions is presented. In the model, Bidirectional Long Short-Term Memory networks are implemented to avoid vanishing gradients. The study involves the application of Connectionist Temporal Classification to determine handwriting in different alignments and sequences. CTC also acts as a metric to compute the loss quantity incurred during model training. The dataset (Doctor’s Handwritten Prescriptions Dataset, 2024) used in the training is the samples of handwritten medical terms collected from various medical organizations. The study achieved an accuracy of 62.33%.