Timo Deist

Full Name
T.M. Deist
Researcher - Unknown
+31 20 592 4364
Life Sciences and Health


He received his PhD at Maastricht Unversity (NL) in 2019 on the topic of ‘Distributed learning and prediction modelling in radiation oncology’ supervised by prof. dr. Philippe Lambin and prof. dr. Andre Dekker. As part of his PhD research, he joined Dr. David Craft’s team at the Massachusetts General Hospital/Harvard Medical School (USA) as a visiting researcher. Prior to that, he studied Econometrics and Operation Research at Tilburg University (NL), and BioHealth Computing at the University of Barcelona (ES) and the Université Joseph Fourier (FR).


Timo Deist’s research focuses on two domains:

  • multi-objective continuous optimization/machine learning,
  • distributed/federated machine learning for healthcare applications.

Currently, he investigates gradient-based multi-objective optimization techniques to support decision-making processes. Complex decision-making problems with conflicting objectives can often be modeled as multi-objective optimization problems. In many cases, informed decision-making is only possible when multiple alternatives with various trade-offs are presented to the decision-maker. This means that the Pareto front of the underlying optimization problem needs to be estimated. Timo develops gradient-based optimizers to speed-up the approximation of Pareto fronts (arxiv). Furthermore, he investigates how this class of optimization problems can converted to a data-driven multi-objective learning problem to generate Pareto front approximations (arxiv). These concepts are developed as part of the the MODIR project to support decision-making of medical professionals when processing medical images using deformable image registration.

While in Maastricht, he contributed to the design and roll-out of a global federated/distributed learning infrastructure across oncology institutes allowing privacy-preserving machine learning research. This technology is part of the Dutch Personal Health Train initiative (homepage).
Together with his colleague Frank Dankers, they received the ESTRO-Varian Award 2019 by the European SocieTy for Radiotherapy and Oncology for their work ‘Distributed learning on 20 000+ lung cancer patients’ which highlighted the capabilities of this infrastructure.

Google Scholar Link


Current projects with external funding

  • Multi-Objective Deformable Image Registration (MODIR) – An Innovative Synergy of Multi-Objective Optimization, Machine Learning, and Biomechanical Modeling for the Registration of Medical Images with (MODIR)