Our team focuses on advanced research and development of algorithms for the processing and analysis of medical data, mathematical modelling, and simulation of complex systems. On this page, you will find an overview of our department's research activities.
Computer Analysis of Medical Data
We specialise in the development and implementation of algorithms for processing and analysing medical data, primarily for diagnostic purposes. Our research includes the detection of specific events in physiological signals, the extraction of complex information from limited datasets, correlation analysis, segmentation of cells from microscope images, and the creation of specialised databases for further research use. In the field of physiological signals, we have experience with ECG, EEG, EMG, and intracranial (ICP) and arterial blood pressure (ABP) data. For image data, we primarily work with cell and tissue microscopy. We utilise a wide range of tools for data processing – from classical programming languages (e.g. Python) to Big Data technologies (e.g. Apache Hadoop, Cassandra). We also focus on implementing machine learning methods, including the use of specialised hardware and massively parallel computing approaches.
- OP JAK: DIGITECH, 2025–present.
- FN Bulovka, FN Královské Vinohrady, CTU FBME, UJEP Department of Informatics: PhysioConnect.cz, 2023–present.
- IGA Krajská zdravotní, a.s.: Artificial Intelligence in Radiological Image Segmentation and Assessment (Umělá inteligence v segmentaci a hodnocení radiologického obrazu), 2022–2024.
Predictive Modelling
We apply mathematical modelling methods to analyse and forecast time series. We work with data from business processes and corporate environments, as well as with time series generated within other research activities. Our approach includes both traditional statistical techniques and modern machine learning methods – such as neural networks – or their combinations.
Time Series Modelling
We model time series using classical statistical methods such as autoregressive models (ARMA, ARIMA), their seasonal extensions (SARIMA), or versions with external regressors ([S]ARIMAX). The appropriate model type and its parameters are chosen based on a thorough analysis of the given time series, including its statistical characteristics.
Another group of models includes machine learning methods. We either use existing software packages – such as Prophet, which is based on additive regression and Bayesian principles – or design and implement custom neural networks tailored to specific forecasting tasks. We most commonly work with recurrent neural networks such as LSTM or GRU, implemented using standard libraries like TensorFlow or PyTorch. Successful neural network training requires sufficiently large and high-quality datasets. We also focus on explaining the influence of individual regressors on time series development using interpretable models or sensitivity analyses.
- ADLER Czech, a.s.: Data Analysis at ADLER Czech, a.s. (Datová analýza ve společnosti ADLER Czech, a.s.), 2019–2020.
- Severočeské doly a.s.: Application of Machine Learning Methods for Monitoring and Forecasting Groundwater Conditions in Mining Areas (Analýza využití metod strojového učení pro monitoring a predikci vývoje stavu spodní vody v dobývacím území), 2020–2021.
Computational Simulation of Complex Systems
We specialise in the simulation of complex systems using both particle-based and continuum approaches. In particle modelling, we mainly employ molecular dynamics to describe physicochemical phenomena – such as the simulation of polymer-based composite materials or separation processes in membranes. We also focus on pedestrian movement simulations, using both discrete cellular automata and continuum-based methods within computational fluid dynamics, where pedestrians are modelled as an intelligent fluid.
Simulation of Polymer-Based Composite Materials
The computer design of nanomaterials requires consideration of a wide range of parameters and their possible combinations. Simulations provide an effective tool for systematically exploring and optimising these parameters prior to experimental validation. Given the high number of degrees of freedom, we use coarse-grained models of polymers and other composite material components. To improve computational efficiency, we utilise GPU acceleration. Our research focuses primarily on modelling the self-assembly of copolymer systems – in bulk, on solid surfaces (e.g. nanoparticles), or under non-equilibrium conditions such as under flow.
- Czech Science Foundation (GA ČR): Controlled Association of Polyelectrolyte Copolymers with Surfactants in Aqueous Solutions (Řízená asociace polyelektrolytových kopolymerů se surfaktanty ve vodných roztocích), 2013–2016.
- ERDF/ESF: UniQSurf – Centre of Biointerfaces and Hybrid Functional Materials, 2018–2022.
- Technology Agency of the Czech Republic – Kappa: METAMORPH – Advanced Hybrid Organic-Inorganic Nanofibers for CO₂ Capture and Photocatalysis, 2021–2023.
- Czech Science Foundation (GA ČR): Separation of Racemic Mixtures by Membrane Processes (Separace racemických směsí membránovými procesy), 2017–2019.