Born at Masaryk University, RationAI is a multidisciplinary research group bridging the gap between advanced computer science and clinical medicine. We specialize in transforming "black box" algorithms into Explainable AI (XAI), creating a collaborative ecosystem where data scientists and medical experts can work in lockstep. By integrating automated traceability, robust visualization, and validation metrics co-designed with clinicians, we ensure that every AI-driven insight is transparent, trustworthy, and ready for the high-stakes world of biomedicine. At RationAI, we don't just build powerful models, we build the trust necessary to bring them into the lab and the clinic.
We aren't just developing AI models, we’re teaching them to speak our language. Our focus is on creating Explainable AI (XAI) that transforms "black box" algorithms into transparent, collaborative partners for scientists and doctors.
Digital Pathology
Our main directions contains development of methods for detection of various types of cancerous tissue using cutting-edge deep learning techniques. The system is modular and allows rapid prototyping of new deep learning methods for WSI.
Explainability
We focus on develop cutting-edge AI methods and an appropriate environment that will maximally support cooperation between domain experts and computer science specialists with a focus on explaining the behavior of these AI methods (explainable AI, XAI).
Precision Interfaces
Because not everyone is a computer science expert, we are designing advanced UIs tailored for "normal people", transforming raw output of our machine learning models, to more visually understandable shape which can be used to showcase results.
The Feedback Loop
Thanks to pathologist who cooperate with us and are using our technology. We are able to continuously and iteratively refine our models and UIs, based on their direct critiques and insights.
Regulatory Excellence (IVDR)
We provide the validation evidence required for regulatory compliance, specifically the In Vitro Diagnostic Regulation (EU) 2017/746, by generating trusted data and AI method provenance.
Secure Scaling & FAIRness
To protect large-scale multimodal health data, we implement FAIRness support and develop federated machine learning methods that ensure insights remain clear while data stays decentralized and secure.
Hugging Face (follow us)
Explore our other projects (GitHub)
Explaining Digital Pathology Models Via Clustering Activations.
A. Bajger, J. Obdržálek, V. Kůr, R. Nenutil, P. Holub, V. Musil, T. Brázdil
IEEE 23rd International Symposium on Biomedical Imaging, ISBI. IEEE, 2026.
https://doi.org/10.1109/ISBI61048.2026.11515782
Weakly Supervised Multicenter Nancy Index Scoring in Ulcerative Colitis Using Foundation Models
A. Kukučka, O. Fabián, V. Musil, T. Brázdil
arXiv preprint arXiv:2604.23706, 2026.
https://arxiv.org/abs/2604.23706
LSP-DETR: Efficient and Scalable Nuclei Segmentation in Whole Slide Images
M. Pekár, V. Musil, R. Nenutil, P. Holub, T. Brázdil
arXiv preprint arXiv:2601.03163, 2026.
https://arxiv.org/abs/2601.03163
| phone: | +420 549 49 3922 |
|---|---|
| e‑mail: |
|
| phone: | +420 549 49 7026 |
|---|---|
| e‑mail: |
|
doc. RNDr. Tomáš Brázdil, Ph.D., MBA
doc. RNDr. Petr Holub, Ph.D.
MUDr. Rudolf Nenutil, CSc.
Mgr. Karel Štěpka, Ph.D.
doc. Mgr. Jan Obdržálek, PhD.
RNDr. Lukáš Hejtmánek, Ph.D.
Mgr. Jiří Horák
