The seminar entitled Machine learning of dynamics of magnetic nanoparticles for medical technologies will be given by Assoc. Prof. Ondrej Hovorka from University of Southampton, UK on April 2, 2026, at 10:00 am in the IPM lecture hall. The seminar will be held in English.
Magnetic nanoparticles are finding increasingly broad applications in biomedical technologies, for example in tumour imaging and treatment using magnetic particle imaging and hyperthermia. Their wider adoption in clinical practice, however, requires tools capable of rapidly and reliably describing their dynamics and translating complex physical information into a form usable for diagnosis and therapy planning.
The lecture will present an approach based on physics-informed neural networks that enables real-time analysis of magnetic nanoparticles within emerging clinical workflows. It will further show how dimensionality reduction can be used to transform large-scale molecular dynamics simulations or time-resolved measurements into interpretable Markov state models that capture the key physical processes occurring on long timescales. The presented framework also incorporates Boltzmann generators, generative probabilistic models that enable more efficient sampling of structures from thermodynamic distributions, which is important, for example, for mapping particle distributions in tumour treatment planning. Overall, the lecture will demonstrate how advanced machine-learning methods can help bridge the modelling of magnetic nanoparticles with future clinical applications.
