The University Hospital in Martin is building the first Hospital AI Laboratory in Slovakia. Its aim is not to chase another percent of model accuracy, but to safely and unobtrusively integrate artificial intelligence into everyday clinical practice. The talk outlined specific uses, who the laboratory will serve, and the challenges it faces.
How to integrate AI into the hospital
Unlike big technology companies or startups, this is not about hunting for another half percent of accuracy. The priority is integrating AI into the real hospital environment so that healthcare professionals do not have to change their procedures. Staff are busy, they do not have time for lengthy training or to use a multitude of new tools, so the goal is minimal change in practice.
A typical scenario: a CT image goes to the server and to the physician; with AI it also goes in parallel into the model, which within a few seconds offers a segmentation preview. The physician may or may not display this layer – the decision remains with them. Such integration into existing solutions is more time-consuming than delivering a completely new tool, but it respects the established clinical workflow.
Who the laboratory is for
Patients will receive faster diagnosis, more precise treatment, and the possibility of continuous monitoring of their health status using AI. Doctors will have their work streamlined, and AI will provide them with a second opinion even when another colleague is not present. The goal is to save time on low value-added tasks and strengthen decision-making where it matters.
The laboratory is also a research center with access to anonymized datasets and computing power that is continuously improving. Researchers thus have a unique environment for consultations with top experts in the field and for testing ideas. Students gain a safe space for experimentation and real-world experience with AI in healthcare.
From small tasks to systemic decisions
Among the “small” solutions is the automation of time-consuming yet routine tasks – for example, counting brown and blue cells in images. Such automation frees up physicians’ time that they can devote to patients. The hospital also produces approximately 60 TB of new data per month and has access to experts for consultations, creating unique conditions for research.
On a larger scale, AI can predict how many medications will be needed in the upcoming season to avoid shortages as well as waste. It can also search for optimal procedures for how to use the new hospital efficiently. The speaker outlined the technical architecture of the solution, but the details would require a separate space.
