Friday November 14, 2014 — “Medical Image Synthesis Methods and Applications”
2650 SMBB Auditorium, 11:45 am
Speaker: Jerry L. Prince, PhD, William B. Kouwenhoven Professor, Electrical and Computer Engineering, Johns Hopkins University
Abstract:
There are no truly calibrated magnetic resonance images. Something is always just a little bit different when acquiring data from a different scanner or even the same scanner after an upgrade. Manufacturers have different strategies for optimizing their image quality and MR techs can change a parameter to try to improve the image quality on any given day. We have been exploring image synthesis methods to change this state of affairs. At least for the purpose of analyzing images using automatic algorithms, we are hopeful that through synthesis, we will be able to obtain more consistent results, and through this effort the results of automatic image analysis methods applied to MR images might become a more important part of clinical practice in the future. Four methods and a variety of results are presented in this talk. Sparse reconstruction is an important theme throughout, and image segmentation and registration are key methods that serve to demonstrate improvements. Although our results are very promising, this new area of research is controversial, and its future impact is uncertain. The talk concludes with some ideas about future directions and some thoughts about what might be possible in the future.