Designing Algorithms to Agree with Experts for Medical Imaging and Analysis of Biological Networks

Biomedical Informatics Candidate -
In this talk, leveraging a user's expert knowledge along with an automatic algorithm's efficiency for tedious tasks will be discussed. First, a principled solution for interactive medical image registration will be presented. For this purpose, a closed-loop control system that couples a user and a traditional, open-loop registration algorithm is designed. Second, a stochastic approach for constrained optimization, applied to point cloud and image registration will be described. It has the capacity to capture large deformations while respecting constraints placed by the user on the deformation. Finally, including prior information in the analysis of biological networks will be explored. The analysis concentrates on discovering network elements that contribute to the formation and proliferation of cancer cells.