Digital Transformation of Medication Label Identification in the Operating Theatre: Technological Evolution

Abstract

Medication errors remain a significant challenge in healthcare, often resulting in adverse patient outcomes. Traditional identification methods, such as barcode scanning and RFID/NFC, though effective in certain contexts, fall short in handling the complexities of curved surfaces and multi-angle visual data associated with cylindrical medication containers. This research explores the potential of computer vision techniques to address these limitations by proposing novel methods for unwrapping and stitching medication labels captured from multiple angles.

The study investigates two approaches: Image Stitching Followed by OCR and OCR Followed by Text Stitching, each designed to enhance the accuracy and efficiency of medication identification. The proposed methods aim to tackle challenges such as label distortion, varying orientations, and inconsistent label designs. A multi-camera system is developed to capture images from different perspectives, followed by the implementation of algorithms for unwrapping and stitching. The performance of the proposed methods is evaluated based on their ability to extract accurate medication information and integrate seamlessly into healthcare workflows.

The results demonstrate that both methods offer high accurate medication identification results. This research contributes to advancing automated medication identification, ultimately reducing errors and improving patient safety. The findings provide a foundation for integrating computer vision into practical healthcare systems, delivering a scalable and reliable solution for medication identification.