Mikael Brudfors
I work on developing image processing software for segmentation, super-resolution and registration that should be applied to hospital grade brain scans. The framework I use involves modelling the observed data in a statistical setting, defining a forward model, data likelihood and prior. This type of modelling have previously been shown to generalise well to handle a wide distribution of healthy neuroimaging data. The idea is that it should also be a good candidate to deal with its more heterogenic counterpart – images that can be found in hosptial databases.
Projects
Image SynthesisPublications
Flexible Bayesian Modelling for Nonlinear Image Registration
Brudfors, M., Balbastre, Y., Flandin, G., Nachev, P. and Ashburner, J., 2020. arXiv preprint arXiv:2006.02338.
Nonlinear markov random fields learned via backpropagation
Brudfors, M., Balbastre, Y. and Ashburner, J., 2019. In International Conference on Information Processing in Medical Imaging (pp. 805-817). Springer, Cham.
Empirical Bayesian Mixture Models for Medical Image Translation
Brudfors, M., Ashburner, J., Nachev, P. and Balbastre, Y., 2019, October. In International Workshop on Simulation and Synthesis in Medical Imaging (pp. 1-12). Springer, Cham.
MRI super-resolution using multi-channel total variation
Brudfors, M., Balbastre, Y., Nachev, P. and Ashburner, J., 2018, July. In Annual Conference on Medical Image Understanding and Analysis (pp. 217-228). Springer, Cham.