In digital health, the first obstacle is rarely the one founders expect. It’s not the market. It’s not even funding. It’s proving, early and credibly, that an idea will work in clinical reality.
“That first barrier is validation,” says Anja Barešić, coordinator of AI4HEALTH.CRO (EDIH) and Head of the Laboratory for Computational Biology and Translational Medicine at the Ruđer Bošković Institute. A strong pitch, she adds, is not the same as a product ready for hospitals.
That is precisely the gap AI4HEALTH.CRO was designed to close, giving health innovators access to structured “test before invest” support, often expensive, fragmented, or simply unavailable in healthcare, so they can de-risk development before scaling.
The user: a Rijeka startup rethinking radiology workflows
One of AI4HEALTH.CRO’s users, Medusa Technologies, is a Rijeka-based startup founded in 2024 by Diego Ivanović and Leonard Martinis. Their mission is ambitious but practical: to reduce friction and error risk in radiology by tackling a problem clinicians experience every day — software fragmentation.
“Radiologists have to jump between multiple programs,” Ivanović explains, describing diagnostic workflows built on disconnected systems. That constant switching slows work and increases the risk of mistakes—while also contributing to burnout pressure across the profession.
Medusa’s answer is not “another tool.” The team is developing an operating system for radiology that consolidates applications and software tools radiologists currently use separately, aiming to modernise the infrastructure rather than add another standalone layer.
The bottleneck: clinical validation and access to real data
For AI startups in healthcare, the hardest step often comes long before pilots or procurement: access to authentic clinical data and credible validation conditions.
Through AI4HEALTH.CRO’s Test Before Invest services, Medusa gained access to anonymised radiology data from a clinical environment, enabling the team to train AI models on real X-ray images while adhering to GDPR requirements. For many early-stage teams, this is the difference between theoretical performance and clinically relevant development.
“In normal circumstances, access to real clinical data is extremely complicated and almost unavailable to young startups,” Ivanović says. “Through AI4HEALTH.CRO, we could train our models on authentic X-ray images.”
What AI4HEALTH.CRO delivered: not just testing, but traction
The value wasn’t limited to datasets. Medusa highlights three concrete outcomes from participating in AI4HEALTH.CRO’s pathway:
-
Connections to Croatian hospital centres, creating routes into real clinical contexts
-
Consultations with top radiologists, helping the product align with real workflows and expectations
-
Additional anonymised data collection, strengthening model testing and improvement beyond the team’s existing database
This is the point where “validation” becomes more than a checkbox. In healthcare, credible evidence and fit-for-workflow are what unlock conversations with hospitals, investors, and (eventually) regulators.
The broader impact: a growing pipeline, not one-off support
Medusa’s story is one example within a wider pattern. In roughly two years, AI4HEALTH.CRO’s Test Before Invest service has been used by around 40 organisations—from startups and hospitals to associations and research institutes.
The hub’s support model spans modules that are frequently costly or hard to access in digital health, digital maturity assessment, concept verification, digital strategy development, piloting, usability testing, and more combined with education, networking, and guidance on access to finance.
And demand is increasingly practical: according to consortium partners, the most requested services include proof of concept and feasibility studies, software usability testing, and testing/experimentation/piloting in clinical environments, the unglamorous work that determines whether a solution is safe, usable, and adoptable.
Why this matters for hospitals and the ecosystem
Anja Barešić frames Test Before Invest as more than technical assistance: it plays a strategic role in healthcare and the economy, helping teams surface limitations early, before time and money are sunk into assumptions that don’t hold in practice. In digital health, even one wrong assumption can be expensive.
Medusa’s progress shows what “de-risking” looks like when done properly: real clinical-grade data access, clinical expert input, and structured iteration, so a product can move from concept to something that can credibly earn its place in a radiology department.
You can read the full feature story on Medusa Technologies and AI4HEALTH.CRO in Poslovni dnevnik (Josipa Ban, 22 September 2025).