Abstract: Accurate interpretation of doctor’s handwritten prescription is a critical task in healthcare for the patient safety and minimization of medication errors. Moreover, illegible handwriting poses a great risk, which leads to misinterpretation resulting in adverse health outcomes in many cases. Recent breakthroughs in AI, more so the deep learning segment, introduce robust methods to automate handwritten text interpretations, hence providing a way to streamline processes and improve the accuracy in the healthcare setup. In previous studies, models like CNN, RNN were explored, but the standalone CNN and RNN were giving suboptimal responses. In this study, a Hybrid CNN-BiLSTM model is proposed to predict and analyze words in handwritten prescriptions. This research utilizes CNN for capturing the spatial features and BiLSTM networks for analyzing sequential dependencies, which makes this approach suitable for complex handwriting pattern recognition in medical documents. To evaluate the performance of the proposed model, its performance was compared with Google’s Vision API. It is a machine learning-powered service for image content analysis. The results are indicative of the great potential that lies in the application of the CNN-BiLSTM architecture for advancements in automated prescription analysis to improve patient safety and operational efficiency within healthcare settings.

Keywords: Convolutional Neural Network (CNN), Bidirectional Long-Short Term Memory (Bi-LSTM), Deep Learning, Machine Learning

| DOI: 10.17148/IARJSET.2025.12107