Extrapolation and filtering of incomplete marker tracks by singular value decomposition [heart motion measurement]

The authors investigate whether the singular value decomposition (SVD) method can be modified to handle incomplete tracks such that appropriate extrapolations for incomplete marker tracks are obtained in an animal experiment were made incomplete by obscuring a part varying in length from 3% up to 44% of the observation time. Comparison of extrapolations with measurements results in a root mean square (RMS) error value of 0.25 pixel, where the total excursion of a marker is about 20 pixels. Using computer simulation data under similar conditions, the estimations were compared with the known signal values. The RMS error was found to be 1.2σ for the extrapolation in the obscured part and 0.7σ for the filtering in the observed part, where σ² is the variance of the noise