Active Sensing for Model Based Anatomical Target Tracking

This research presents a proof-of-concept study for actively controlling the MRI scanner to select the image plane that maximizes expected information gain during the anatomical target tracking. The proposed approach is based on greedy entropy minimization. It calculates the expected entropy for every possible image plane configuration and selects the one that is expected to increase the knowledge of the system the most. Performance of the active sensing algorithm is explored in terms of the accuracy of the predictions and a comparison to a passive algorithm which randomly selects the image plane.

Manuscript in preparation

• Active Sensing for Model Based Anatomical Target Tracking
  E. E. Tuna, D. Franson, N. Seiberlich, and M.C. Cavusoglu
  In Preparation.