MIT Critical Data, the University Clinical Center in Gdansk and the Medical University of Gdansk, in cooperation with Roche Poland, are initiating a joint project called Poland Healthcare Datathon. The two-day event - a programming competition and conference with the participation of leading experts from Poland and abroad - is aimed at a wide range of enthusiasts and practitioners of the use of AI in healthcare - healthcare system employees, engineers, students, as well as patients themselves. The aim of the Datathon is to develop new methods of analyzing local medical datasets and to test how well modern artificial intelligence (AI) models understand and process information about neurological diseases, using the Polish language. The Poland Healthcare Datathon will take place on September 27-28, 2025 at the UCK in Gdansk, Poland.
Diagnosis of the future: enthusiasts test AI in medicine
Published July 2, 2025 07:19
Artificial intelligence (AI), innovative digital solutions and the use of health data are increasing the availability of medical services and increasingly boldly entering the world of medicine, opening up new opportunities. According to the "Survey of the Degree of Informatization of Entities Performing Medical Activities"[1] conducted by the Center for e-Health, in 2024 13.2% of Polish hospitals used AI in their medical services, a twofold increase compared to 2023. Other medical facilities - those providing inpatient and 24-hour non-hospital services and outpatient health centers - use AI much less frequently: 3.2% and 4.1%, respectively. Among the most common applications of AI in Polish medicine are: diagnostic imaging, including analysis of CT and MRI scans; patient service - chatbots on call centers and virtual assistants on facility websites; and analysis of patient data and support for doctors in diagnosis and therapy selection.
Advances in AI, particularly the development of Large Language Models (LLMs), are making artificial intelligence increasingly applicable to medicine, especially in the area of diagnostics. Modern AI algorithms can analyze huge medical data sets in seconds, which previously required months of work by entire research teams. This makes it possible to make faster and more accurate diagnostic decisions and personalize therapy according to the individual needs of patients. At the same time, the models are known for their so-called hallucinatory nature while providing a convincing tone of response. An additional challenge is their tendency to replicate biases and stereotypes present in the datasets used to create them. LLMs also have the ability to analyze texts and images and draw conclusions based on them.
- The health sector generates about 30% of all digital information worldwide. Data centers make it possible to analyze huge medical data sets, support diagnosis, treatment and drug discovery. Advances in artificial intelligence, and in particular the development of large-scale LMM language models, offer hope for supporting medicine through the ability to simulate human reasoning and generate accurate answers, says Professor Michal Hoffmann, vice-rector for clinical affairs at the Medical University of Gdansk.
The benefits of using AI are enormous, and the health care system could prove to be one of its biggest beneficiaries. Currently, there is a growing use of LLM models by both doctors and patients.
- However, we must keep in mind that despite their enormous potential, LMMs are not without limitations, especially in terms of diagnosis, treatment, or advice on daily functioning. LMMs learn from huge datasets, without programming every detail. They rely on a variety of databases that, as they grow, increase the chance of identifying new diagnostic patterns in many different medical fields. At the same time, the models are known for their so-called hallucinatory nature while providing a convincing tone of response. An additional challenge is their tendency to replicate biases and stereotypes present in the datasets used to create them. Testing the capabilities of LLMs is one of the key directions of AI research today, explains Michal Maciejewski, chief project coordinator on behalf of Roche Informatics.
Responding to the challenges posed by the development of modern AI algorithms, the Poland Healthcare Datathon - a conference and a programming marathon for health and a platform for the exchange of experience between healthcare stakeholders and Polish and international AI experts. The event aims to discover new diagnostic patterns specific to the local population and analyze the effectiveness of LLM in the area of neurological diseases in Poland. Participants will have a real and anonymized dataset from the clinical practice of UCK Gdansk.
- As one of the largest and most modern hospitals in Poland, the University Clinical Center is constantly looking for new solutions that will affect the quality of care, effectiveness of therapy and safety of our patients. Currently, projects based on artificial intelligence are already in operation at UCK. We are testing and implementing more always with the aim of saving lives or improving the health of patients hospitalized with us. This unique project not only in the context of Poland, but also in Europe, gives us the opportunity to reach new ideas that may prove useful in our hospital," says Tomasz Stefaniak, MD, UCK's director of treatment.
Teams of 3-5 people consisting of medical and technical students are eligible to participate in the Poland Healthcare Datathon. The two-day event will begin with a symposium on AI, featuring Polish and international experts, including Dr. Leo Anthony Celi of the Massachusetts Institute of Technology, Dr. Judy Gichoya representing Emory University, and Dr. Stephen Pfohl of Google Research, who represent diverse and rapidly evolving fields of medicine, which, combined with their extensive professional experience, creates a space for an inspiring exchange of insights on disruptive applications of technology in medicine.
- As artificial intelligence permeates tasks and processes at work, promising greater efficiency and cost savings, there is a very real danger of losing critical thinking skills as we begin to hand over data synthesis and decision-making to AI. The way to protect human capabilities and agility is to design AI to support human skills. AI's legacy is to provoke reflection on what really makes us human, what makes us thrive and - ultimately - what gives us happiness. Decades of research on happiness always lead to the same conclusion: happiness comes from a sense of connectedness. Ultimately, the only activities that bring people together are those that foster dialogue and working together. This is the power of datathons," points out Dr. Leo Celi.
On the second day, participants, under the guidance of mentors, will work in two thematic tracks: analysis of a clinical dataset provided by UCK Gdansk and evaluation of large LLM language models in the area of neurological diseases. The new methods of analysis of the local dataset and the results of the LLM evaluation will be presented by the participants in front of the Jury, which will include representatives of the Datathon organizers. There will be financial prizes for the top 3 teams in each track.
Registration for the Poland Healthcare Datathon continues until July 7, 2025 or until the limit of places is reached. Details and regulations are available at: https://www.healthcare-datathon.pl/rules
***
The Poland Healthcare Datathon educational project was created in cooperation between MIT Critical Data, the University Clinical Center in Gdansk, Gdansk Medical University and Roche Poland. Partners and patrons of the event are: Poznan University of Technology, Karol Marcinkowski Medical University in Poznan, Emory University, Google Research and Gdansk University of Technology.
