Inside Europe’s AI Hospitals: How Olomouc Built a Clinical Innovation Engine

with Pavlina Walter and guest  Antonín Hlavinka

Across Europe, we are watching a rapid divergence between institutions that talk about digital health and those that are actively building it. In healthcare, true transformation rarely originates from the flashiest technologies. It comes from disciplined execution, governance structures that can handle new complexity, and leadership that is willing to experiment with real risks and real constraints.

This episode of Clinical Capital Conversations illustrates this gap clearly. Together with my business partner, Pavlina Walter, I sat down with Antonín Hlavinka, the Deputy Director for Innovation and Digitalization at University Hospital Olomouc. His team has created what I consider one of the most advanced AI-integrated hospital environments in Europe.

This is not a theory. This is not a pilot theatre. This is real adoption, supported by real clinical outcomes, and grounded in operational discipline. For investors and healthcare decision makers, the Olomouc case is a signal of what is coming next across the region.

In this long-form analysis, I will break down how this hospital became a living test bed for clinical AI, what surprised us most in our conversation, and what this means for capital allocation in the next decade.

The Rise of AI Hospitals in Europe.

For years, AI in healthcare was synonymous with radiology. Startups built models for X-ray review or early cancer detection. Hospitals adopted a few isolated tools. The narrative was that AI would support imaging departments and little more.

That era is over.

Today, the leading hospitals have moved far beyond radiology. They are deploying AI in administrative operations, patient journey optimisation, clinical decision support, microbiology, telemedicine, interoperability systems, and multimodal prediction engines. They treat AI not as a software purchase but as infrastructure.

University Hospital Olomouc is one of these environments. It is the fourth largest hospital in the Czech Republic. It serves more than one million patients. It is home to the National Telemedicine Center, a major academic system, and a digital innovation hub that connects clinicians, researchers, government, and industry.

This ecosystem matters because real AI adoption cannot happen in fragmented environments. The story that Antonín Hlavinka shared with us is the story of a system that has aligned technology, regulation, training, and governance around one mission. Increase precision, reduce cost, and build a repeatable model for clinical impact.

Training the Next Generation of Clinicians.

One of the most striking parts of our discussion was the mandatory AI education program for medical students at Palacky University. This is not a side course. It is not an optional seminar. It is a required component of clinical training.

Students learn about AI models, mobile health, eHealth, cybersecurity, assistive technologies, interoperability, and data governance. This matters for a simple reason. If future clinicians do not understand how these systems work, they will never trust them. If they do not trust them, they will not use them.

The early reactions from students reveal a familiar pattern. Some embrace it. Some are hesitant. Most are curious but overloaded. This is normal. What matters is that the foundation is being built. In five to ten years, these graduates will be the first generation of physicians who consider AI literacy as fundamental as understanding imaging or pharmacology.

This shift has long-term strategic consequences. Investors should expect an increasing gap between systems that educate for digital practice and those that do not. That gap will directly influence clinical throughput, cost per patient, and long-term scalability.

Senior Clinicians and the Post-COVID Mindset Shift.

Antonín Hlavinka made an important observation. COVID changed the cultural adoption curve.

Before the pandemic, both clinicians and patients were skeptical of telemedicine and AI. After the COVID-19 pandemic, demand reversed so sharply that the hospital could not keep up. Suddenly, telemedicine was not a futuristic concept. It was a necessity. Suddenly, AI was not a threat. It was a time saver.

Even the most traditional specialists began acknowledging the value of certified diagnostic tools that outperform human accuracy in specific tasks. Radiology has been the clearest example of this. According to Antonín Hlavinka, the best radiologists may achieve around seventy-five percent accuracy on certain diagnostic categories. Narrow AI tools can exceed ninety-five percent.

When the difference is this large, the debate changes. It is no longer a question of preference. It is a question of responsibility. If a tool is safer, faster, and more precise, clinicians must integrate it.

This is the type of mindset shift that investors should watch closely. Adoption accelerates when clinicians see results with their own eyes.

Beyond Radiology: Where AI Delivers Value Inside the Hospital.

The Olomouc team uses more than ten AI tools across their network. These tools fall into two main categories: administrative support and clinical support.

Administrative AI

The hospital uses LLMs for documentation, legislative review, policy interpretation, marketing, and workflow optimisation. They rely on secure European servers for data processing to meet GDPR and cybersecurity standards.

This is a critical point. Most discussions focus on clinical AI, but administrative AI often produces the largest immediate return. A hospital that reduces administrative burden by twenty percent unlocks enormous latent capacity.

Clinical AI

The clinical portfolio is divided into certified tools and experimental models.

Certified models include radiology analysis tools integrated into CT and MRI systems, video analysis for colonoscopy, and microbial detection systems. These are dependable and already delivering measurable accuracy improvements.

Experimental models include:

• AI that predicts post-COVID severity
• Multimodal engines combining imaging, labs, and EHR data
• Microbiology prediction tools that identify bacteria in one day rather than one week
• Telemedicine platforms that integrate real-time ECG or telemetry analysis
• Narrow AI models built in-house for specialty cases

These projects are supported by Horizon grants and national funds, but the real value is the operational infrastructure that makes these experiments possible at scale.

Building Internal Governance: The Innovation Committee

One of the most sophisticated aspects of the Olomouc model is the Innovation Committee. It includes clinicians, lawyers, DPO representatives, IT teams, biomedical engineers, and innovation leads.

This committee evaluates every AI proposal based on:

• Clinical need
• Feasibility
• Data governance
• Cybersecurity standards
• Cost and potential ROI
• Certification pathway
• KPI definition
• Ethical considerations

No AI tool moves forward without a structured assessment. This is the type of rigorous governance that private equity firms and strategic buyers look for when assessing operational maturity.

The committee also determines whether a POC is required, how it will be tested, and how results will be validated before procurement. This reduces the risk of technology being adopted without evidence.

Real Impact: Two Cases That Changed Minds

Antonín Hlavinka shared two stories that captured the power of AI more effectively than any slide deck.

Case One: A rare diagnosis solved in 22 seconds

A patient had gone months without a correct diagnosis. Clinicians, biochemists, and specialists had exhausted their hypotheses. When the case was fed into a secure LLM, the first recommendation was the correct diagnosis. It took twenty-two seconds.

This was the moment that convinced the team to expand LLM access across departments.

Case Two: A tumour detected before the human eye could see it

During the testing of an AI radiology tool, the model flagged a tiny anomaly that clinicians initially missed. It turned out to be a tumour in an early stage. That early detection saved the patient.

These cases give AI adoption a human face. They turn abstract predictions into real outcomes. They change clinical culture faster than any policy.

The Constraints: Certification, Regulation, and Workforce Dynamics

Europe faces unique headwinds. GDPR, the AI Act, and medical device regulations create a slower certification timeline. Public hospitals face procurement processes that are rigid and sometimes counterproductive. The shortage of evaluators slows development even further.

At the same time, staff worry that AI will replace jobs. Nurses and administrative workers are concerned that efficiency gains will reduce headcount. Clinicians worry about being compared to models that outperform them.

Antonín Hlavinka addressed this openly. Transformation always generates fear. The key is to demonstrate that AI does not replace skills. It replaces bottlenecks. It gives clinicians more time with patients. It eliminates repetitive work. It shifts roles rather than eliminating them.

This is a message that investors should internalise. The systems that manage workforce transitions well will accelerate. Those that do not will face internal resistance for years.

What the Future Looks Like: The First Agent-Led Hospitals

Antonín Hlavinka believes the first fully agent-supported hospitals will appear within five years. I share this view.

This does not mean hospitals without humans. It means environments where:

 • intake is automated
• patient histories are structured by AI
• diagnostics are supported by narrow models
• treatment plans are validated through prediction engines
• microbiology and pathology run on accelerated cycles
• telemedicine is a primary, not secondary, channel
• administrative burden is reduced to a minimum
• clinicians supervise rather than manually operate every step

This is not speculative. The technologies exist. The only barrier is system design.

What Olomouc Shows Investors

For our investor community, the Olomouc model demonstrates several critical points.

1. AI maturity will define asset quality in the next decade

Hospitals, CROs, and medtech systems with embedded AI operations will outperform those without them. This will influence valuations directly.

2. The ability to certify models is a strategic moat

Hospitals that understand MDR pathways, data governance, and long-cycle validation will have a competitive advantage in Europe.

3. Real-world test beds will become valuable investment partners

Investors do not only need products. They need environments where those products can be tested, validated, and scaled.

4. Multimodal prediction engines will redefine clinical timelines

Reducing diagnosis cycles from months to seconds changes the economics of entire specialties.

5. The next decade will reward operators, not theorists

The winners will be those who build infrastructure. Not those who talk about innovation.

Our conversation with Antonín Hlavinka reinforced a simple truth. The future of AI-enabled healthcare is not being built in boardrooms. It is being built in hospitals that are willing to test, fail, iterate, and scale. The transformation of University Hospital Olomouc is a preview of what Europe will look like once digital health finally matures.

For investors, this is the moment to pay attention. The leaders of tomorrow are already showing us their operating systems today.

My thanks to Antonín Hlavinka for his transparency and to Pavlina for co-leading this discussion. Episodes like this are the reason I built Clinical Capital Conversations. Real insights from real operators who are shaping the future of medicine.

Timecode:

00:00 Welcome and Introduction

00:24 Role and Responsibilities in Healthcare

01:31 Adoption of AI in Healthcare

02:27 Impact of COVID-19 on Technology Adoption

04:18 AI Applications in Hospitals

04:49 Certified vs. Experimental AI Models

07:47 AI Tools in Use

14:05 Innovation and Project Management

21:41 Introduction to AI in Radiology

22:10 The Evolution of AI in Healthcare

22:55 Challenges and Rapid Developments

23:30 Microsoft's Game-Changing AI Model

24:42 AI in Medical Practice

27:41 AI in Education

28:55 Balancing AI and Human Cognition

30:55 AI's Impact on Healthcare Workforce

33:30 Global Perspectives on AI Adoption

36:42 Future of AI in Hospitals

39:03 Collaborating with Startups and Innovators

40:33 Conclusion and Future Prospects

Links:

Peter M. Kovacs LinkedIn: https://www.linkedin.com/in/petermkovacs/

Peter M. Kovacs Personal Website:https://www.petermkovacs.com/

PMK Group Website: https://www.pmk-group.com/

Guests:

Antonín Hlavinka: https://www.linkedin.com/in/anton%C3%ADn-hlavinka-b045754b/

Pavlina Walter: https://www.linkedin.com/in/pavlinawalter/

Transcript:

Pavlina Walter: So today I'm extremely happy that we can welcome here at Antonín Hlavinka . I consider him as a one of the biggest expert for artificial intelligence in the Czech Republic. I would love to ask you to introduce you to our audience. So if you can a little bit, say a few words about yourself and about your passion and artificial intelligence.

Antonín Hlavinka: Yeah. Hello everybody. Uh, I'm director Deputy of the fourth biggest hospital in Czech Republic. It's University Hospital Olomouc. And now I'm at the position, uh, director deputy for innovation and digitalization. Previously I work as, uh, IT deputy, uh, but uh, we have to decide or we decided to split this function because of new innovations, because of new staff, uh, medical staff, um, uh, AI in, uh, hospitals.Uh, it's very a hot topic team, so. We split this function and now are responsible for the innovations digitalization in university hospital and also the director of the department in medical faculty, uh, in university, Palacký in all modes. And, uh, I'm leading the course for, uh, future, uh, future doctors or the medical students, which is, uh, mandatory for them.And we are teaching them, uh, some stuff about ai, about, uh, mHealth, eHealth, uh, cybersecurity, uh, assistive technology and, and, uh, interoperability as well. And, and so on. So it's, uh, also my part of field. 

Pavlina Walter: So I think this subject is totally new for the, um, healthcare professionals. So how they are taking it, do they understand, do they see also the vision as we see the vision also in a future in ai?

Antonín Hlavinka: Yeah, some, some of them, yes. Because it's fourth grade, which is, uh, I think the hardest. Uh, great, uh, for, for that students. And, uh, sometimes, uh, they are accepting it, uh, very, very, um, smoothly. And some, maybe the half of them are, you know, very reserved. About the new stuff because it's not their topic. Uh, so far it's, uh, they find out it's very interesting for them.But, uh, they have, uh, not such a time for, uh, you know, get deeply in this topic. 

Dr. Peter M. Kovacs: And what is the situation with the, with the elder, uh, the more senior professionals, for example, in your hospitals, how they accept these. New technology. 

Antonín Hlavinka: Uh, it's, it's changing. I think the, the COVID was a game changer, uh, because, um, I'm divided the, the, the time before Codi, COVID and after COVID in, in, uh, case of, uh, acting in the new technologies.Uh, because, uh, we are also the National Telemedicine Center in University Hospital. We're a competent center of ministry of health for, for the telemedicine and adoption adoption of, of, um, health generally. And, uh, there is big difference between, uh, between that time. So before COVID. Uh, they were very reserved, all the patients as well as the physicians.And now after COVID, they started to ask and we were not able, you know, to, uh, fulfill their, their, uh, uh, uh, needs. So now it's, uh, I think, um, in the right way, in the right track because they are now, uh, accepting. They are, uh, able to work with the new technologies, especially with ai. I think how, maybe more how of the physicians are using.Personally, the ai, the GPT, the models generally. Mm-hmm. Because, you know, it's, it's helping them and now they know that it's, uh, for example, in decision support and, uh, some, some kind of stuff. It's better than the, the best physician in, in that, uh, particular field. So. I think that now it's, it's, uh, changing and also the, uh, the more, or the, the most conservative doctors, the, the oldest one, are slowly accepting this, uh, uh, uh, this, this new technology, especially in radiology, uh, because there's certified AI and this, this is very acceptable for them to use it because it's saving time.

Yes. And they're more open. They're not all, not all, but I think the, the more than half of of these, uh, physicians are able to use it and not, uh, not afraid of, uh, using this, yeah. Technology. 

Dr. Peter M. Kovacs: It is very well known that, uh, AI was coming into the healthcare sector, especially in the hospitals, in the radiology fields.So, because the AI can. Analyze the, the CT MRI scan X-ray Exactly. Or mammography reports. And, uh, they are, they are very smart and they can spot very early, much earlier than a physician. So it is very well known. But now the AI in a hospital is not, not just radiology and very few people know about it. Can you are explain and give us some, a few examples that where you can use the AI in a hospital?

Antonín Hlavinka: Yes. We have to divide it on a, for example, the administrative part of the GPT or uh, AI generally. And the, the clinical using and the clinical, we have to divide it to the certified one and the un uncertified experimental one. Yes. Uh, so we are the, uh, research company as well, or research organization. So we are developing our own models.We are trying to train the, the new models, but they are not certified device, uh, or, or, uh, models. Not yet. Uh, maybe, you know, generally the, the general LLM models, uh, won't never be the, the, the certified solution because you have to prove it. Uh, what is the data set? You train the, the, the model and you are not able to prove.Uh, the digital model. So that's why, uh, it's very hard to certify this. Maybe it won't be able at all. I don't know. Uh, but if you are using some local models, uh, some special models, the narrow ai, it, it goes narrow ai, and you can train your own model and you can start the certify process. You can, uh, you know, count with the, the future certification and you can start the developing of this, uh, model and the clinical trials, et cetera.Uh, according to the certification standards and needs. So, uh, it, it could be done, but it's a long run. It's for three and three to five years, you know, from start to end of the certification process. And also it depends. Uh, the, the usage of, of these models, because you can use it as a class one, for example.Mm-hmm. It's quite easy. And if it is, uh, some kind of decision support system for clinicians, they, uh, are able to, you know, uh, count with this as a supportive tool. Uh, you have to certify this to be or to a two a. Uh, according, according to MDR, uh, directions, and it's, uh, very complicated. It is very hard and very expensive as well.Mm-hmm. So, uh, and once you finish the certification, your model is, is obsolete. It's out of date. So, so, 

Pavlina Walter: so you have this kind of, uh, models or, um, like devices, softwares, whatever, in a process. And then you have already. Certify probably from suppliers. Uh, yes, 

Antonín Hlavinka: yes, yes. We are using this or certify, which is, uh, mostly the part of the, some modalities.Mm-hmm. For example, the, the CTA and et cetera. Mm-hmm. It's, it's part of this, uh, part of this, uh, equipment. Um, some of them are, for example, the, the check startups, uh, like Carebot and, and others, they are al already certified, but they are software as a service. So we are consuming it as a service and it's already certified.And our, uh, PACs system are, uh, also certified. So it's compatible and we can use it. Uh, quite, uh, you know, without, without, uh, any, any, any scare or something. And can 

Pavlina Walter: you, can you count or can, can you share with us how many tools with AI right now is used in your hospital? 

Antonín Hlavinka: I think it's more than 10. 

Pavlina Walter: Mm-hmm.

Antonín Hlavinka: More than 10 tools because four administrative tool we are using the, the copilot from Microsoft. I don't know if I can tell the, the, sure. Yeah. Okay. It's no problem. Uh, so copilot, uh, we are also have some license from open ai. To the latest, uh, GPT models because of the, you know, best results of, uh, clinician works.

So we have the agreement, all our, or the, these data are stored in a, a European urine servers mm-hmm. Infrastructure and they are some guarantee that they want trained, uh, the new models from our data. So it's very mandatory for us because we are also under the cybersecurity or GDPR in European Union, et cetera.

So that's, uh, one part. It's, uh, the administrative tool. Also, we tried, uh, uh, some special tools for, uh, legislation for, uh, uh, agreement, uh, validation, et cetera. It's some special tool for, for that, uh, that case. And, uh, for marketing, it's some picture and video generators, et cetera. So that, that's the administrative tool.It's, uh, quite normal now, uh, in, in, in the university hospital. And, uh, the clinical part, they are the, as I, as I, uh, told you before, it's uh. Some, uh, MDR certified clinical, uh, models for radiology, for the, for the, uh, CT scans, et cetera. And we tested three of them and compared to between, uh, each other, which is better specific, uh, and, and, and so on.So now we are about to, uh. To, um, call the public tender for, for the provider of that, uh, tool in, in general in, in all, all these fields. So 

Pavlina Walter: you are not only focused on a check provider, but you are like to any global one. Yes. Who is the best tool actually for your services? 

Antonín Hlavinka: Definitely, definitely, uh, because, you know, as I told you that, uh, some kind of, uh, this MDR.

Uh, general or the, sorry. The, the AI tools professional are part of this, uh, modalities. So we are trying also to use this, uh, this modalities for special cases. For example, colonoscopy, they are analyzing from, from videos, uh, et cetera. So this is another part of, of, uh, using gp, uh, ai, sorry, uh, in, in, uh, hospital.

And we also started lots of projects in, in, uh, in history, for example. During COVID, uh, we started the first AI tool which identifies, you know, the cough of, of, of the ill people. The, the distance between the peoples, and it was some automatic system, cameras and microphones in the, in the big department.

And they, they were monitoring all the peoples and the, the security, just checking, you know, the monitor and, uh, counting the, uh, the incident. So, so it was the first usage of. Uh, of the, uh, AI models in, in, in hospital. And then we started, uh, we started some kind of multimodal agent, uh, projects, for example, in, uh, uh, post COVID syndrome.

We combined, it was some, uh, grant from a Czech agency, technic agency, and we started a project which should identify or predict the. Severity of, uh, post COVID syndromes. So we combine all the data laboratory results, uh, cts or, or, uh, uh, uh, x-rays and combine it with the EHR reports and, uh, try to predict, uh, you know, the development of this, of this illness.

It was quite good because we get. Precision, uh, above 80% of the recommendation whether the patient is ready for, uh, for the right treatment. Yeah, so it was, uh, uh, the, I think second big, uh, AI project. And now we are starting the new big project, which is l legalization of telemedicine in Czech Republic.It's from, uh, uh, structure for, for, uh, funds, from from eu. And, uh, this project is not, uh, generally focused on ai. Uh, the secondary effect or secondary, uh, projects are using the AI in a telemedicine. So for example, if I collecting the telemetric data from patients, I can use for, for example, scan a, ECG, or something like that.Uh, via the, uh, the ai, which is, uh, implemented in the, in this, uh, telemedicine platform. So it's, uh, we will start with piloting about 11 or 12, uh, interventions, telemedicine interventions, and the goal is to get the, uh, the codes for, for, uh, insurance. Companies. So that, that's, that's the goal of, of this. And also the, the anchor is, uh, it in, uh, in a c legislation part of it.It's already done. And now we are continuing. So this is another part of ai, which are, you know, start to use and now. We are finished lots of project, uh, the Horizon projects, uh, which are big project from European Union, and now they are calls from, uh, about, about the ai, general AI and medicine, et cetera.And now we are, uh, applying for example, 1, 2, 3, 4, 4 such a projects. One is for devices. Uh, uh, one is, uh, for multimodal agent, uh, agent system, uh, for IBD patients. Uh, one is for, uh, some special, um, ax method to using in, in, in me medical system or medicine. Uh, it's, uh, like some, uh, patented tool from University  Palacký .And we are now testing it, uh, through some project. We are about to testing it if the project will be success, uh, in, uh, in, uh, some special cases of training the staff, for example, uh, predict some kind of neurological or neurodegenerative, uh, illness, et cetera. And, uh, the last is some kind of microbiology system which could predict, uh, uh, ahead with, uh, um, with infections, for example, bacteria infections in some, uh, some, uh, department.And we can, uh, shorten the time. To identify the right bacteria from one week to one, one day, for example. You are 

Pavlina Walter: actually the leaders, I guess, in Czech Republic because you are the really, the hospital, an institution using the most AI in Czech Republic, who is actually the person who is deciding which project is gonna be implemented.Yeah. 

Antonín Hlavinka: Uh, it's a news new, uh, committee we established, uh, last year. Last year we started, uh, this, this project we established, uh, uh, innovation committee in, uh, in, uh, our hospital, which are, uh, you know, part of the, uh, clinical, uh, professionals. There are lawyers, there are a DPO, uh, it guys. Uh, our, um, uh, innovation, innovation guys are from innovations.And, uh, for example, the biomedical engineers, uh, in, in some particular field. And we have some systems. So if someone would like to test or adapt or, uh, buy or something, some AI tool so he, uh, can apply for this project, internal project. And, uh, we, uh, will decide whether the project is, uh, viable or not. And we, if everything is okay, the money, the, the project, the goals, the KPIs, the, the cyber security, et cetera.So we can, uh, we can validate it or we can provide it to the. Uh, to the, uh, the deputies and director. Okay. Yeah. 

Pavlina Walter: You mentioned correctly those many things that the companies or you needs to complain about, like the GDPR cybersecurity. Yeah. And so on, especially cybersecurity. Last week I joined the, uh, AI big conference for cybersecurity for the hospitals.There was a big discussion about the new regulation. Actually, you need a lot of new position people. Do you have those teams? Do you have those experts or you had to totally change the resources? Oh, you use your old one? Yeah. And retrain. Yes. 

Antonín Hlavinka: Yes, but mostly because, um, as I said, we have or we had, uh, the quite big team from national, uh, eHealth or, um, health center.Uh, so we are very close to, uh, AI projects as well. So we are now using the, that guys for the projects, for the validations and for the seeking of new opportunities, for example. Uh, we are in tight connect with, uh, with the clinicians because they are now coming, you know, on their own. And then they would like to ask us, uh, whether they want to buy this one or test it and piloting.So, uh. That's the, that's the committee which, uh, decide whether this project is, uh, viable and, uh, whether we are start A POC, for example. Mm-hmm. So it's very mandatory the POC testing before, uh, before, uh, public tender or something like that. So we are piloting in whether the, the, the results are Okay. So satisfied.Uh, we are satisfied everyone, everything is okay. So now we, uh, you know, already, already know what, what we want so we can start a public procurement. It's, uh, it's a very hard stuff to create the this and to, uh, go through this public procurement because it's, uh. Quite tricky. 

Pavlina Walter: And you have now a lots of projects ongoing and also finishing and coming.Who has the capacity actually where you would like to end up? Because you, I mean like this is never ending story. Yes. So what is the aim of the special committee, innovation committee or your hospital? 

Antonín Hlavinka: Yeah. Uh, the, I think the goal for University Hospital generally is, uh, you know, to, uh, create the, uh, the treatment more effective.

Uh, cost less and mm-hmm. More effective. I think this is the, this is the goal for, for everyone. So we are the test bed piloting and, and, uh, and, uh, the end. We should, uh, try or create some kind of SPI spinoff or, or, or startup or, uh, join with a university, create startups and for example, try to get the certification for, for, for clinical usage, for example, in this special case.This is the goal. Mm-hmm. And we are, uh, on the road, uh, I think in the, in the right direction. And it's, uh, really hard stuff because we are lack of this, uh, public notification bodies in Czech Republic. So there's lack of the professionals for. For the certification process. So it's, it's, it's a very limitation for everyone, I think.Mm-hmm. In, in Czech Republic. 

Dr. Peter M. Kovacs: And how you can, or do you measure the impacts already? The financial impacts, social and medical impact? KPI, we, 

Antonín Hlavinka: yeah. Every project, uh, should, uh, uh, establish in the start KPIs. For example, the, the time shortage, the cost effectiveness as, as you, you mentioned, the specificity and, uh, global, uh, overall effectiveness of, of, of this, uh, of this new model or new AI or new tool, whatever it is.It, it's not, not only about ai, it's uh, about, uh, hel applications, for example. Uh, we have some, some special, uh, mobile application which can automatically, uh, detect, uh, whether the wound, for example, is, is healing properly or not. Mm-hmm. Uh, we can, um, use this mobile application for electronization or the dig digitalization, the paper.Uh, health records, et et cetera. So it's very, very broad field of, of usage innovations in hospital. And 

Dr. Peter M. Kovacs: what was your biggest surprise when you get the project and you get the results that you were really surprised that, oh, I never, never think about it, that we could save so much time. We could save so much money.

Antonín Hlavinka: I think it's not, nothing yet is such a big deal. Yeah. Such a big deal. I think generally the, the LLM models, uh, are very big game changer. Yes. But it is not only in a, in a clinical field, it's also in an administrative field, but lots of guys using the, uh, LLM, the secure LLM models, they are well-trained professionals and for example.Uh, I can get one example. The normal method, for example, was, uh, some, uh, some ill patient and we were not able, the, the clinicians and the biochemists, uh, were not able to, you know, use the, the right diagnose for two months. Yeah. To 

Dr. Peter M. Kovacs: identify it. 

Antonín Hlavinka: To identify it was very hard. It is very time spending and, uh, I think annoying for the patient.And after that, after that field, we, you know, uh, we decided what. Probably the, the, the, uh, the diagnosis is the right and we tested, uh, we put it on in a new GPT model oh three, and it was in 22nd, right in the first place, the, the right diagnosis. So it was, you know, wow. Game changer. So we decided to, to buy lots of, lots of lines.Yeah, 

Dr. Peter M. Kovacs: it was a, it was a very good example. So I was wondering that you have these kind of examples. 

Antonín Hlavinka: And another one was when we tested the, the care bot. It was special. I think it was carebot in, in radiology and was some case with, uh, with a broken, uh, arm or something. The, the patient and the AI recognized the, the, the, the two small mark in, and it was, it was the two more.And so we saved the patient in very early stage. So it was, uh, life, life saving, definitely. So that's was 1, 2, 2 big examples, uh, which shows the, the AI is really game changer in, in healthcare. 

Pavlina Walter: If you can a little bit move to your person, um, were in and where the passion for artificial intelligence or telemedicine actually became.

Antonín Hlavinka: Yeah, I'm, uh, it guy from many years. Mm-hmm. For many years now, about 30. Five years. I'm a professional in it. And first, uh, I met the ai, it was ai, it was some machine learning in, in, uh, in a university in school, uh, 20, 24 years ago or something like that. And now there was, you know, come nothing more than some special very.Uh, nerd tools, uh, ai, and it was really calm. And now before 3, 3, 4 years, it started maybe 10 years back. 10 years. It started the, the, the radiology systems. Yes. But there was not Ed, definitely. And, uh, then we started, tested the special, um, narrow AI for specialties. As I said, the COVID. Before the LLM uh, times and after the GPT free or something, uh, appears.So it's, uh, starting it. It was, it was amazing. It was something like, I dream of about 10 years. Such a tool would be very perfect. It must be done somehow and 10 years, years later, it's here. It's here, it's here. It's, it's really game changing. And, you know, if you. You are not able to follow the changes.

It's every 14 days is something new. You have to test it because if you start now to test some, uh, some model or something, some, uh, technical specialty. You are, start testing it, you are wasting lots of time. And after two months it's, it's, it's already already done. And it's a part of this general model.So it's really, really hard for every, uh, you know, scientist and et cetera to create something new because it's so, so quick. Uh, this development, for example, now that the Microsoft, I think the, yeah. Microsoft, uh, has the new model and it is, it is, uh. Four times better than the best clinicians for, for, uh, for decision support or from differential diagnosis.It's, it's really game changing and it's also. Safe, uh, safe costs. So for, for the treatments, it's, it's, uh, tremendous. We contacted Microsoft. We would like to be, you know, part of this research team. Yes. We have to test it. It is, it is amazing. So, uh, yeah, we are in the, in the and the list. And do you think, 

Dr. Peter M. Kovacs: do you think that is just a beginning?And where can, this is definitely, 

Antonín Hlavinka: this is the beginning. For example. Imagine if now you know, that the, the precision of the model for our ideology, for example, is. Uh, 95% and the precision of the best radiologist is 75%. Uh, for me it's, it's simple. And the radiologist cost five, 500 euros and, uh, AI costs 50 euros, for example.So I think the decision is, is, uh, very simple, quite, quite simple. 

Pavlina Walter: Okay. You have this comments here. If you would go to your doctors in your hospital, you would tell them how they will react. 

Antonín Hlavinka: Ah, I'm using the, the ai for example, every, every medical report I just, you know, uh, try to analyze with the, with the GPT, with the AI, because mm-hmm.I understand it and I can go, uh, to, to my clinician or to the, to the professional. And they know me. So, uh, it's, it's quite easy. They expected that I will challenge them, uh, with the, with the ai. So, uh, once they see, for example, the results, they are amazed. They are, you know, it, it, it's still, it's something new for them, how precise it could be.How, uh. Time saving. It could be. So it's, uh, I, we, we are not able to satisfy all these ideas from these, uh, from these doctors because they are playing with this and now they got the idea. We can, uh, set it. For example, the application in the ambulance. Uh, it's a QR code in, in the field, and the patient, which is waiting for the treatment, he can scan the qr, it's open, the, the symptom checker, for example, which is AI based, and now they can, uh, check or it's, uh, uh, ask for every symptom every.Um, um, problems that the patient has and the end of the, uh, end of the process is the, uh, is the structure, uh, report for the physician. The physician see it and you just can ask for some, uh, some special cases. So it's very productive. It can save you lots of time for the. Uh, with, with the patient. And the patient, uh, is satisfied because during the waiting, because of the waiting times in university, hospitals quite big.So he can, uh, you know, very, uh, proactive use this, uh, wasting of time by proactive, uh, things. 

Pavlina Walter: You mentioned now, and uh, you gave us several examples of very successful stories, um, and your devices, tools. Is there any failure? 

Antonín Hlavinka: Failure? I think failure is, uh, I think yes, to start the right, proper, the POC and start the procurement.Mm-hmm. I think the, you know, right set of this public procurement, uh, the set, the price for the procurement and all this, uh. Um, needs the properties of, of, of this. It's, it's very, very hard and sometimes, uh, it could happen that, uh. You, you are not able to, uh, to get it because someone, uh, someone just, uh, you know, stop it because, uh, it's, uh, more expensive or is it's less expensive and it's, uh, focus on only one field.I can do this field so they can, you know, there's a conflict, conflicts, conflicts, interest and vr. We are the. Uh, the poor guys, which are facing this, this, this thing. So yeah, this, this is, this is very bad in, in the public. 

Pavlina Walter: Also, when we saw each other last time, you mentioned that your kids, uh, are educated and you use AI home regularly, so they still love it.You still love it to do? Definitely. Yeah. 

Antonín Hlavinka: Yesterday I just. Learned the mathematics with, uh, my daughter. And, uh, she was using for two hours, the GPT to create, you know, the geometry, for example. And it creates the examples and the results. And it was quite sophisticated and very, it's very cool. For GPT, you can ask, if you are not able to understand this explanation, for example, the, the mathematical records or something like that, you can ask, explain to me like.I'm a child, for example, and now it's different. It's totally different and I can understand it. Even me, I can understand it and I can explain it to my daughter. So yeah, it's, it's very productive. I think it's, uh, in, in, uh, generally in education is, is, uh, maybe big game changer than in, in, in the clinics because, you know, the personalized, uh, education system, oration plans.Are light. Amazing. 

Dr. Peter M. Kovacs: Yeah. But I think it's very important. I have also small kids struggling with algebra stuff. Yes, yes. But is it important to, to use it as just as a support and not instead of thinking, because it's very important that for these kids, they have to use the brain as we learned. As we did in our time, um, because there are some studies from MIT for example, that Yeah, yeah.People are using Chat GPT, they, the cognitive function is declining very quickly because they then don't try it hard. They just quickly go into ChatGPT and ChatGPT answer everything, and they copy paste. Yes, 

Antonín Hlavinka: yes, yes. That's right. That's right. But I think it depends because, uh, I think the neuroplasticity of our brains are, are very big.And, uh, when the internet, for example, starts, it was the similar things. Okay? For example, when the GPS the navigation start, it was, we, we lost, you know, the, the control. But we are still here and we are more productive and more productive. So I think whether, uh, is, is the, the using is, uh, really like, uh, for example to be more productive.You still need to think, you know, because you have to have some goal. If you not have goal, the GPT or or LM just tell you that should be the right goal, but it is not the right goal because it's, uh, very supportive. It's, uh, uh, you know, the GPT is, is it's only, only machine. It's statistics. It's nothing more.And if you don't need, whether the goal, where the goal is, you are, you know, you are in a, in a bad way. And you are, find out sooner or later you are, you, you are gonna find out and now, uh, so you have to start to, uh, uh, to creativity. I think it could support the creativity because you are not wasting time with the, with the, uh, borings administrative tools, with the researching, uh, et cetera.And you are directly in the, in the center and it's time consuming. Very good for your dopamines because you are directly and very quickly to your goal. And it's, uh, I think it's, you know, every, everything, uh, in, in history like atomic bomb or the fire you can use it for, for help or you can use it for destruction.It, it depends on, on you 

Dr. Peter M. Kovacs: and technical question, just, uh, uh, you use a lot of different tools in your hospital, but there is a general global issue of shortage of human resources, shortage of doctors, and in Europe, mainly shortage of nurses. Do you see that AI could help this mm-hmm. Shortage using just, uh, patient journey management.

So how you, you manage the patient short, improving their, but also this HR factors. Is it possible that, that you can Yes. Even not solve it, just, just improve this? It's 

Antonín Hlavinka: possible. Uh, it's very sensitive question, especially in public sector, because when you start the AI, for example, in, I don't know, in administrative, in accountant, for example.The, the employees are starting to afraid to, to lose their job. So they are fighting against it. It's, it's normal. It's normal. It's the same like some physician physicians are not accepting it because they're afraid. Yes. Because they know or maybe. Um, somehow, uh, they, they, they feel it. That could be better than them and they could be, you know, uh, uh, jeopardizing their job.Yeah, definitely. Definitely. That, that's right. I think it's, uh, it's already happening. Yes. You know, you, if you heard in Albania, there is a first ministry of, of, uh, uh, I don't know if it's a ministry of, uh, laws or something. Uh. Uh, in this first AI model. Mm-hmm. So it's now they fire the, the ministries and set up the, the AI agent.I think it's not wise, but I think it's the trend and we have to can with this, for example, the teachers, I think now. If you have some average or under average teacher, it's way more better to use GPT for, for teaching. It's proved already, for example, India, there are some special cases that, uh, the, the effectiveness of, of, uh, this education with the AI a with a personalized education is about, I don't know, 40% better.So you can. Uh, you can start, uh, the project or you can start a training program, uh, which are personalized for your, uh, level of, of knowledge, knowledge, knowledge, knowledge and capacity. And definitely, definitely. And it's, it's very productive. Some, some, I don't know. I've read something about that. Uh, one year, uh, the university with the GPT could be the same like five years in university without GPT.Yeah. About the productivity knowledge and Yeah. And the results. So I think it's a trend and it's already here. 

Pavlina Walter: If you would compare the situation now in Czech Republic versus the other countries, is it still Denmark? Who is playing the number one in the Europe? How would you consider like a global situation right now?

Antonín Hlavinka: Yeah, I think it's about the environment. Uh, I think that Denmark, uh, is nothing better, for example, of using this tool. But there is environment which is very friendly, which is very supportive for, uh, developing this AI models for adapting it, for using it in a, in a clinical practice, for example, at et cetera.It's very important in Czech Republic is very lack of this, uh, supportive ecosystem for innovations in, in, uh, uh. Generally, not only in healthcare, but generally it's a very lack of. Of, you know, supportive, uh, initiatives, laws, initiatives, et cetera. It's getting better, but I think in comparation with the other countries, it's, um, it's very bad.Not only in ai, but I can see it in eHealth, in mHealth using of telemedicine, for example. We are, and the in the, if, if you, you can download the DESI index, uh. Uh, within European Union and if you, uh, if you filter the healthcare or digitalization, so we are, I think, uh, last, last second or something like that, the Czech Republic.So we have to, uh, long way to, to get it. And for example, in, in US, they have no, uh uh, uh, no, no obstacles like the GDPR, not, not only GDPR AI ACT and cybersecurity. It's, I think it's, uh, not, not, not good and that because they have not everything under control. And it could, it could be problem in the future, but, uh, I think now for the productivity, for the innovation is supportive and, and, and yes, it's, it's very good.So, uh, lots of countries are, uh, uh, ahead of us. Uh, but I think it's getting better. And now the all European Union just realized that we have to be competitive and we have to start to, to support this. This, uh, innovative ecosystem. So we just established the, the Dig two Health, which is some hub, uh, digital innovation hub.In, in, and there is the, the government, local government, university, hospital, and, uh, Paske University. Mm-hmm. Just, uh, the, the tripod. And we are trying to support, uh, digital medicine, um, projects, system, medical devices and whatever. So test bed, uh, piloting, uh. Uh, support the granting, uh, fundings and, uh, uh, and, uh, uh, creating the, or support the, uh, startups, et cetera.So yeah, we are first in republic, in, in that field. In, in digital healthcare. It's very hard to get funding because we are not. Uh, for example, only focused on cybersecurity or ai or the high computer performance. So we are in the middle and it's, it's quite hard because the system is obsolete. It's not, you know, uh, so supportive for new innovations, for example, in healthcare, which is very important for.

Dr. Peter M. Kovacs: And finally how you see your hospitals in the next five, 10 years in case of the AI tools and services and yourself in this system. 

Antonín Hlavinka: Uh, I, I, I think that we could start some, some AI department. The only AI department I just read about it, I would like to get to see it in China. They, they, they just, uh. Uh, created some kind of hospital, but it's, uh, it's, it's very confidential.So I, we are not able to, to get there, but I think this is, this is the future. The AI robots, or AI tools ar agent systems, which are with the, with the, uh, with the supervising of, of the, of the medical stuff. But still, I think it's, uh, it's a future. You know, the 100%. Agentic or AI hospital, I think this is the future in five years could be the first, uh, first models or first examples, and me in, in five years.Uh, I'm not sure. Maybe I will have some implanted in, in my brain, uh, to, to, to, to be able to understand the, the, the a GI for example. I dunno, 

Pavlina Walter: I have this question because you mentioned this special hospital in a China with only AI and robots. Do you think that the patient would actually love it? They're coming to hospital just because of the doctors to have a personal contact with someone human.So do you think it could be accepted? I mean, of course we Who will love, who love, um, the ai? It depends, depends on generation, I think. Mm-hmm. 

Antonín Hlavinka: Our generation, yeah. We are used to at, uh, the physical contact and face to face with the, with the doctor. It's some kind of satisfying way. It's, it's, our brain is, uh, just learn. Uh, how to behave like that. Mm-hmm. But the gen, gen Z and gen alpha, I think they are completely different. Mm-hmm. They are afraid of physical contact. Mm-hmm. And they display dependent. Yes. Yes. So I think for them it is the future and they will create a future, future for, for themselves. I think that that's the way and.We were on the, on the, on the, um, back and we will just check it and maybe afraid of it and et cetera. So I think that that's the future and we are not able to stop it. It's already here. I think it's the Rubicon has been crossed and now we will see. 

Pavlina Walter: And as we are in a connection with so many, like startups, new companies, developing something of this ai, if they would be interesting to test it in your hospital.So what would be the best approach to contact you? Contact your committee? Yeah. Yeah. 

Antonín Hlavinka: We have some form the innovations, uh, that, uh, our, our hospital sees that mm-hmm. fnol.cz, that sees that and you can, uh, download the, uh, the documents. For example, the, the example. So agreements. And, uh, you can write down the, the, your idea and, uh, the part of the cooperation, and we can get it, we can read it and we can decide it, whether it's okay for us or not.And we can contact it and create some piloting, uh, get some fundings, for example, uh, free capacity of the clinicians. Mm-hmm. Uh, data, uh, from, from it guys, uh, data governance, cybersecurity, everything must be done. Maybe clinical trials. So we have to create also the clinical trials, whether it's, it's some, some mandatory field.And that's it. So it's quite simple, but, but in general, you are open for the such collaborations. But generally we are, we are very open. We are, uh, as, as I said, we have the digital innovation hub, which is the, the, the front door for example, for this new, uh, innovation startups and also the big companies because we are looking also for the big companies, we should support this, uh, digital innovation hub and we could support the young, you know, brains, uh, with, with ideas and, um, new startups.In that field. So yeah, it's uh, very, very likable. 

Pavlina Walter: It's interesting, extremely interesting. 

Dr. Peter M. Kovacs: Thank you so much. It is very promising and congratulations for this nice achievement that you get in this last few years 

Pavlina Walter: and hopefully in a half a year we'll be able to speak about some new projects and the development as everything is going so fast.So we'll be very happy again to welcome you in a few months. 

Antonín Hlavinka: Yeah, thank you very much. I think it will be lots of projects done, maybe lots of projects start. So, uh, I'm looking forward definitely. Thank you. Thank you very much. Thank you for coming. Thank you so much.