Automatic point-wise validation of respiratory motion estimation

摘要

A system for validating motion estimation comprising a field unit (110) for obtaining a deformation vector field (DVF) estimating the motion by transforming a first image at a first phase of the motion into a second image at a second phase of the motion, a metric unit (120) for computing a metric of a local volume change at a plurality of locations, and a conformity unit (130) for computing a conformity measure based on the computed metric of the local volume change at the plurality of locations and a local property of the first or second image defined at the plurality of locations. Based on the value of the conformity measure, the DFV estimating the motion is validated. Experiments show that the conformity measure based on the computed metric of a local volume change at a plurality of locations and the local property of the first or second image, defined at the plurality of locations, does not necessarily favor a large weight for the outer force to provide a more accurate registration. One reason for this observation may be that large deformations providing more accurate alignment often lead to deformations resulting in unreasonably large volume changes. DVFs comprising such deformations thus are more likely to be discarded by the system of the invention.

主权项

1. A system for validating motion estimation comprising:
a field unit that obtains a deformation vector field that estimates motion between a first image and a second image by transforming the first image into the second image; a metric unit, executed by a hardware processor, that computes a vector-field-based metric of a local volume change between the first image and the second image at a plurality of locations, based on the deformation vector field; and a conformity unit, executed by the hardware processor, that computes a conformity measure, based on the computed vector-field-based metric; wherein the conformity unit: i) computes the conformity measure further based on an image-intensity-based metric of the local volume change at the plurality of locations, the image-intensity-based metric being computed based on the first and the second image by the metric unit, or ii) determines a neighborhood at each one of the plurality of locations, and computes the conformity measure at each one of the plurality of locations, based on a correlation coefficient computed by fitting a hyperbolic function to a joint distribution of corresponding values of the vector-field-based metric and intensities of the first image within the determined neighborhood at each position.