Robust filtering strategies for soft tissue Young's modulus characterization

Accurate and robust quantification of soft tissue elasticity has significant clinical implications for disease diagnosis. For imaging-based strategies, the aim is to recover the material parameters of the assumed tissue constitutive model from noisy image-derived measurements on the kinematic states. In this paper, we develop a general material parameter identification formulation based on soft tissue continuum mechanics models and state space representation. Within this unifying formulation, we analyze the widely-used least-square (LS) solution, which does not perform well under reasonably realistic levels of disturbances, and the popular extended Kalman filtering (EKF) strategy, which is also far from optimal and subject to divergence if either the initializations are poor or the noises are not Gaussian. We then present a robust estimation paradigm derived and extended from the /spl Hscr//sub /spl infin// filtering principles. It is particularly powerful for real-world situations where the types and levels of the disturbances are unknown. Experimental results on synthetic and real images demonstrate its superior performance.