Trustworthy artificial intelligence in medicine is not a marketing label, but a prerequisite for safe care. It determines whether a system’s recommendations will be predictable, ethical, and verifiable. The lecture showed how to achieve this, thanks also to the TE HEAL project and the involvement of Slovak partners.
The TE HEAL project and Slovak teams
The central hub of the project is Charité Berlín, and the project is led by Petra Ritter. Several European partners are involved; for Slovakia, the key ones are Nemocnica Martin and Jesseniova lekárska fakulta v Martine. Žilinská univerzita v Žiline is also participating, and in the lecture it presented its role.
Žilinská univerzita focuses on defining the quality of data and algorithms, which is key to accuracy as well as the automation of evaluation. In practice, the emphasis is shifting from neural network architectures to high‑quality data that determine the outcome. Not all parameters are clearly established today, but the right mathematical indicators can reliably describe the expected performance.
From data quality to certification
The project is launching a service for assessing the quality of data and algorithms so that AI development is not the privilege of big players. Many companies run up against a lack of know‑how or hardware and poor data preparation, which fundamentally affects development time. Anonymization is also key: without it, healthcare data cannot be shared, and researchers often end up with only a few CT scans, which are not enough for serious development.
Another track is a training course for obtaining certification, intended to bring the unified European assessment structure closer. For now, the model is the scenario used in France, and a form is being sought that reduces the complexity and long duration of the process. This is why a serious game is being created: in an early beta, an entrepreneur goes through steps from arranging a meeting at a hospital to requesting data and understanding individual requirements.
