AI-Assisted Scientific Illustration in Biomedical Research: A Practical Taxonomy and Prompt Engineering Guide Based on Analysis of 2,792 Real-World Use Cases

Abstract

The creation of scientific illustrations has traditionally been a time-consuming bottleneck in biomedical research communication. Recent advances in multimodal large language models (LLMs) and text-to-image generation have opened new possibilities for researchers to rapidly produce publication-quality figures. However, the biomedical community currently lacks systematic guidance on what types of scientific figures can be effectively generated with AI assistance and how to craft effective prompts for each use case. In this study, we analyze 2,792 real-world AI-generated biomedical illustrations from SciDraw ( https://sci-draw.com ), a specialized platform for AI-assisted scientific figure generation. Through systematic categorization, we identify nine distinct use-case categories spanning molecular biology, immunology, clinical medicine, pharmacology, microbiology, plant science, experimental workflows, graphical abstracts, and data visualization. For each category, we present representative examples, analyze common prompt patterns, and provide actionable prompt engineering guidelines. Our taxonomy and recommendations aim to serve as a practical reference for biomedical researchers seeking to integrate AI-powered illustration into their publication workflows, while also discussing current limitations and ethical considerations.

Ähnliche Arbeiten