Erasmus School of Social and Behavioural Sciences

At the Netherlands eScience center, I was line manager of a team of research engineers and project manager of eScience research projects. The eScience center collaborates with all scientific disciplines from astronomy to social sciences and life sciences, with the aim to advance academic research using the latest digital technologies. As a researcher, my expertise is in computer vision in biomedical imaging and benchmark design. I think challenges (or algorithm performance benchmarks) are a great way to advance science. I organized two challenges in the medical image analysis field (MRBrainS, NEATBrainS), organized several tutorials on challenge and benchmark design (IEEE ISBI '16, MICCAI '17, IEEE eScience '18) and was co-chair of the challenge workshops at the IEEE ISBI from 2016 until 2018. I led a project at the eScience center in collaboration with SURF, to set-up a platform to support researchers from all scientific disciplines to set-up benchmarks: Enhance-Your-Research Alliance (EYRA) benchmark platform, and was the main organizer of the NeurIPS 2019 CiML Workshop "Machine Learning Competitions for All" and co-organizer of the NeurIPS 2020 Challenges in Machine Learning (CiML) Workshop.

Worked on the brainbox project: Quantitative analysis of MR brain images for cerebrovascular disease management.

Worked on image quality improvement of XperCT and CT scans in cooperation with Philips Healthcare.

Title PhD thesis: "Noise Reduction and Vascular Enhancement in 4D CT Perfusion Scans". Date PhD defense: December 14th 2010. Promotors: M.A. Viergever, W.M. Prokop. Supervisors: E.P.A. Vonken, B. v. Ginneken.During my PhD research I developed a method (Mendrik et al. Med Phys, 2010) to derive 3D angiograms, arteriograms and venograms from 4D CT perfusion scans. This method automatically suppresses bone by taking the absolute area under the first temporal derivative curve (total bolus extraction (ToBE)) and thus enhances the vasculature, after which the arteries and veins are separated using the time to peak. I showed that the CTP derived arteriograms show improved arterial visualization as compared to 3D CTA, due to suppression of bones and veins in the proximal area (Mendrik et al. AJNR 2012). The ToBE method was also evaluated at the Leiden University Medical Center in the Netherlands (Ghariq, Mendrik et al. 2014) for deriving dynamic CTAs from whole-brain CT perfusion scans, where it proved to better visualize the small arteries than standard dynamic CTA. Next to the vessel enhancement method, I developed the time-intensity profile similarity (TIPS) bilateral filter (Mendrik et al. Phys Med Biol, 2011) and an anisotropic diffusion filter (Mendrik et al. STIA'14) for 4D CT perfusion scans. These filters reduce noise while preserving the time-intensity profiles which are essential to assess the blood flow.

Title MSc thesis: “Noise reduction in coronary CTA images”. Supervisor: E.P.A. Vonken.During my master research I developed a noise reduction method (hybrid diffusion with continuous switch, HDCS) for CT scans based on anisotropic diffusion (Mendrik et al. IEEE TMI, 2009).