Bart Jansen is a professor at the department of Electronics and Informatics (ETRO) at the Vrije Universiteit Brussel. He is interested in developing image and signal processing methods and artificial intelligence methods for a variety of applications in the broad biomedical engineering domain, but mainly focussing on rehabilitation engineering.

Bart Jansen graduated in 2001 as a Master of  Science  in  Computer  Science, obtained a master after master in AI in 2003 and obtained a PhD in computer science (VUB Ai-lab, promotor Luc Steels) in 2005, all from the Vrije Universiteit Brussel. Since 2006, he is working at the department of electronics and informatics (ETRO) at VUB. In 2016 he became a full time tenure track professor and since 2020 he also has a 10% appointment with imec as a professor.

"My ultimate goal is to advance the biomedical engineering domain to respond to the needs of people with disabilities."

Research Interests 

Rehabilitation Engineering

Rehabilitation engineering is a term with many different definitions. Some consider it to be a rather broad domain focussing on assessing and responding to the needs of people with disabilities, others (like the EMBS) define it as creating methods and technologies to help patients regain cognitive and/or motor function.

I am driven by the observation that in physical and cognitive rehabilitation there are several important aspects that could improve the rehabilitation process for the patients in a crucial manner.

One of those is the poor quantification of treatment effect: a fine grained monitoring of patients (e.g. suffering from neurodegenerative or other conditions) does not exist. Hence, evolution of for instance motor skills are poorly quantified and the effects of treatment options on these are largely unquantified at an individual basis. This is not surprising, as such an analysis requires specialised tools and examinations as for instance functional assessment at a gait lab. Such assessment is not available nor feasible or desirable on a very frequent basis for a large group of people. Rather, more frequent or even continuous assessment, easily accessible or even ubiquitous, could provide a through quantification of the cognitive and motor functions of the subject. This would allow for a detailed quantification (and later on understanding) of the impact of various rehabilitation schemes. In this domain, we mainly focus on the development of tools and methods for physical assessment by means of low-cost mass market sensor devices, including the Wii balance board, the Microsoft Kinect and others. Besides improving the quality of care and the efficiency of the rehabilitation process, our tools also provide digital biomarkers for pharmaceutical studies focusing on various interventions on subjects with disabilities of movement disorders.

A second issue with current rehabilitation practices is the suboptimal compliance and adherence to optimal rehabilitation exercise doses. Rehabilitation is a slow process requiring a lot of repetitions and practising. Weekly, or even daily, one-on-one sessions with a rehabilitation specialist are hence often not sufficient, driving the prescription of for instance home-based exercises. However, a vast minority of subjects comply to these, as they are perceived as being difficult, painful, boring or even useless. Game-based rehabilitation concerns the use of computer games to present, assess and steer the rehabilitation exercises and might actually improve adherence and compliance. However, whereas real-life therapists excell in adapting therapeutic interventions to the needs of every individual subjects, computer games are fit-for-all and rigid. We investigate the development and use of a flexible and adaptive game based rehabilitation platform.

Neuro Evolution

Neuro Evolution concerns the use of genetic algorithms to evolve neural networks. As this is an optimization approach which is not using gradient information, it is a slow and computationally intensive search process. However, the powerful search strategies provided by contemporary genetic algorithms allows for the joint optimization of network topology and network weights (e.g. NEAT). Our research focuses on integrating feature selection as a third component in this joint optimization process and has resulted in the development of FD-NEAT, a successor of NEAT performing feature selection, network topology optimization and weight optimization.

Augmented Reality

Although seriously hyped, the adoption of augmented reality beyond gadgets is rather poor. Both improvements in terms of hardware (glasses) as well as applications are needed. In this domain, we focus on developing niche applications for instance in medical and look into visualisation of medical image data and object recognition.

  • Affiliated to imec.
  • Member of the etro iof and the iof.
  • Member of steering board of the Brubotics Rehabilitation Research Center.
  • Member of IEEE.
  • Member of ACM and SIGEVO (Special Interest Group on Genetic and Evolutionary Computation) special interest group.
  • Member of FARI, the AI for the Common Good Institute Brussels.