SIMULTANEOUS RECOVERY OF LEFT VENTRICULAR MOTION AND INPUT FORCES FROM MEDICAL IMAGE SEQUENCES

Since the introduction of biomechanical models into cardiac image analysis, there have been a number of efforts to estimate patient-specific cardiac kinematic functions and material properties, assuming the unobservable driving forces (a.k.a. the input forces) as known or constructible from boundary conditions. In this paper, we present a multiframe estimation framework which simultaneously recovers the cardiac motion parameters and the input forces from periodic medical image sequences. It is realized through a Kalman filter which generates the residual innovation sequences, followed by an on-line recursive least-squares filter that uses the residual innovation sequences to compute the values of input forces and to correct the estimations of cardiac kinematic functions. We demonstrate the ability of the framework on synthetic data and magnetic resonance image sequences