AI Where It Counts: Human–AI Collaboration in Clinical Image Translation

Title

Abstract

Clinical trials generate a wealth of image-based content – figures, annotated charts, labels and regulatory documents – that often requires translation for global submissions and local regulatory review. Traditional manual translation of text embedded in images is time-consuming, error-prone and difficult to scale.

This presentation will introduce a human–AI collaborative approach to clinical image translation, leveraging large language models with vision capabilities to automatically extract, translate and reposition text within images. The workflow preserves the original layout and visual context while enabling human reviewers to interactively edit, adjust and verify all translated content before finalisation. We will demonstrate the practical application of this approach in real-world clinical settings, discuss quality considerations, and share lessons learned in balancing automation with human oversight to ensure accuracy and regulatory compliance.