Non-parametric tracking of shift and shape functions in medical images
Jeffrey A. Fessler, Albert Macovski · 2003
Summary form only given, as follows. Several important estimation problems, in particular the quantification of blood vessel position and radius from projections, involve tracking of dynamics shift band shape parameters. The authors present an alternative algorithm for tracking shift and shape parameters that is based on nonparametric cubic-spline smoothing. Rather than requiring a known Gauss-Markov model, the algorithm assumes only that the shift and shape functions be smoothly varying in a sense defined. They discuss the physical motivation for their (global) optimality criterion, derive an efficient algorithm for computing the optimal estimates, and demonstrate the performance on angiographic data. The performance of the algorithm is demonstrated on simulated angiogram data.