Building the foundations of medical Artificial Intelligence using radiological data

For over ten years, Keymaging has been supporting radiologists in their daily practice through our Keydiag® solution, a leading platform dedicated to the creation of structured radiology reports. Today, our ambition goes far beyond simply assisting with writing. In fact, thanks to the data ecosystem built around our tool, we have a major advantage in tackling one of the greatest challenges in medical artificial intelligence: access to reliable, structured, and actionable clinical data.

Radiology lacks actionable data

Artificial intelligence is gradually transforming the healthcare sector. Yet in radiology, a paradox remains: the volume of images is constantly increasing, but the data available to train high-performance AI systems remains limited.

Furthermore, most medical records are still written in free-form text or stored as PDFs. As a result, while billions of pieces of clinical data exist within healthcare systems, they remain difficult for artificial intelligence models to process.

Keydiag®: A Unique Repository of Structured Radiological Data

While many are now searching for data that can be used for AI, Keymaging has been building it for years.

Thus, although initially designed as a writing aid, Keydiag® has gradually evolved into a platform for organizing radiological knowledge.

Today:

• over 1,000 radiologists are actively using the solution
• over 250 guided case studies cover all modalities and specialties
• over 1.3 million reports were generated in 2025

Over time, this expertise has made it possible to build a unique database of structured radiological images, which is updated daily based on the real-world practice of healthcare professionals.

Data: The True Fuel of AI

The development of large language models (LLMs) has demonstrated the considerable potential of generative AI. But the reality remains unchanged: the quality of an AI system depends directly on the quality of the data it is trained on.

In the medical field, this requirement is fundamental.

The structured data generated through Keydiag® provides a solid foundation for developing specialized AI systems based on consistent and validated clinical knowledge.

Our vision: reliable and explainable medical AI

We are convinced that the future of medical AI lies in the combination of language models and domain-specific data.

Approaches based on LLMs and RAG (Retrieval-Augmented Generation) combine these two dimensions: generative reasoning and access to specialized knowledge.

Specifically, thanks to the structured data from the Keydiag® ecosystem, we have the necessary foundation to build a radiology AI system based on real-world, validated clinical data.

Laying the Groundwork for the Future of Radiological AI

Beyond simply reporting findings, our vision is to transform every radiological examination into a source of structured knowledge that contributes to the continuous improvement of artificial intelligence tools.

As data quality becomes the key differentiator in AI projects, Keymaging enjoys a rare advantage: several years’ worth of structured radiological data from clinical practice.

Ultimately, the future of artificial intelligence in medicine will not be built solely on algorithms. It will be built, above all, on high-quality data. And that is precisely what Keymaging is developing every day through Keydiag®.