Cardiac MRI intervention and diagnosis via deformable collaborative tracking

The high contrast and lack of ionizing radiation, renders Magnetic Resonance Imaging (MRI) a suitable modality for continuous intra-operative imaging. Tracking the motion of key locations in cardiac MRI is of paramount importance in control and guidance in emerging robot-assisted interventions. Tracking can also be used to assess myocardial wall motion for diagnostic purposes. This article presents an expanded collaborative tracking algorithm to facilitate both interventions and diagnosis in MRI. Specifically, the network of trackers not only follows anatomical landmarks on the beating heart but also computes its evolving deformable surface on a specific plane. Experimental investigations with both CINE and real-time MRI demonstrate that the collaborative tracker network achieves robust real-time performance over long periods, outperforming the MIL tracker. Pilot experimental results also demonstrate that the evolution of the network's deformation mesh can be used for blood volume estimation and computation of the ejection fraction - both of great diagnostic value.