Lingsoft is participating in the Medallion project and leads the Patient as an Active Participant use case. Recently, we have been exploring how the content of patient records, such as radiology reports, could be presented in a format that is easier for patients to understand.

Patients already have access to their own health information. However, access does not necessarily mean that the information is easy to understand. Medical language is precise and highly specialised, but from a patient’s perspective this can create uncertainty, concern and additional questions.

Our objective is to investigate solutions that could help patients better understand their health information and take a more active role in their own care. At the same time, it is important that these solutions support healthcare professionals in their work rather than attempting to replace it through fully automated approaches.

This work brings together large language models, structured medical knowledge, terminologies and high-quality Finnish language resources. Fluent text is not enough: patient communication must be clear, natural and easy to understand, while ensuring that the underlying facts remain accurate.

Within the Medallion project, we have been examining two parallel agentic approaches: Lingsoft’s initial proof of concept and AMD Silo AI’s more streamlined solution, optimised for AMD’s GPU computing environment.

The work is still at an early stage, and further testing is required. This is precisely what makes the current phase so interesting: meaningful results do not emerge from a single model or technology, but from collaboration, evaluation and a careful approach to the complexity of medical language. When patients have a better understanding of the information they receive, shared understanding throughout the care process is strengthened as well.

The aim is not to determine which approach is ‘better, but rather to understand what can be learned from different solutions in practice. During this collaboration, Lingsoft has contributed extensive expertise in medical language, terminologies and the processing of specialist texts. AMD Silo AI has developed an agentic patient-friendly translation workflow that combines knowledge graphs and other data sources with Poro2. Read more about AMD Silo AI’s agentic workflow here.