I just moved to Amsterdam to pick up a tenure track position at the Centrum Wiskunde & Informatica (CWI). With the group for Computational Imaging, headed by Joost Batenburg, I will work on dynamic computerized tomography (CT). Their facilities include a world class X-ray lab that allows you to design a wide range of interesting challenging experiments. Check out the news story about the opening of the lab including a pretty cool animation
Blending deep learning and iterative image reconstruction has shown great promise to obtain high quality reconstructions from noisy, sub-sampled data and is therefore hot topic in inverse problems at the moment. We adopted an particular approach to enhance the reconstruction of blood vessel structures from sub-sampled, limited-view 3D photoacoustic tomography (PAT) in vivo. Many thanks to Andreas Hauptmann, who did the main work for this exciting project. The paper with all the results can be found on arXiv.
My colleague Ville Rimpiläinen wrote a nice conference proceedings about our work on using the Bayesian approximation error (BAE) approach to compensate for errors in EEG source reconstructions caused by the inherent uncertainty in the skull conductivity. It can be found on arXiv and he will hopefully present this work at the European Medical and Biological Engineering Conference (EMBEC) in Tampere, Finland.
A couple of collaborators from the Photoacoustic Imaging Group at UCL went to the big SPIE Photonics West conference in San Francisco to present some of our work. Now all proceedings have been published in this collection. Big thanks to Martina Fonseca for writing “Three-dimensional photoacoustic imaging and inversion for accurate quantification of chromophore distributions”, Rob Ellwood for writing “Photoacoustic imaging with a multi-view Fabry-Pérot scanner”, and Nam Huynh for writing “Sub-sampled Fabry-Perot photoacoustic scanner for fast 3D imaging”!
We had an interesting collaboration with a group at the Institute of Statistics, Ruhr-Universität Bochum about the statistical properties of a certain class of parameter choice rules that became popular in ill-posed inverse problems recently: Methods based on Stein’s unbiased risk estimator (SURE) choose a regularization parameter by minimizing an estimate of a risk function that cannot be minimized directly as it depends on the true, unknown solution. By a mix of theoretical and numerical studies, we could show that the quality of such estimators can severely deteriorate if the ill-posedness of the problem increases, which is unfortunately a natural asymptotic limit in many inverse problems scenarios. The full results can be found in a paper we recently uploaded to arXiv. Big thanks to all the co-authors!