On correlated sources of uncertainty in four dimensional computed tomography data sets.

The purpose of this work is to estimate the degree of uncertainty inherent to a given four dimensional computed tomography (4D-CT) imaging modality and to test for interaction of the investigated factors (i.e., object displacement, velocity, and the period of motion) when determining the object motion coordinates, motion envelope, and the confomality in which it can be defined within a time based data series. A motion phantom consisting of four glass spheres imbedded in low density foam on a one dimensional moving platform was used to investigate the interaction of uncertainty factors in motion trajectory that could be used in comparison of trajectory definition, motion envelope definition and conformality in an optimal 4D-CT imaging environment. The motion platform allowed for a highly defined motion trajectory that could be as the ground truth in the comparison with observed motion in 4D-CT data sets. 4D-CT data sets were acquired for 9 different motion patterns. Multifactor analysis of variance (ANOVA) was performed where the factors considered were the phantom maximum velocity, object volume, and the image intensity used to delineate the high density objects. No statistical significance was found for three factor interaction for definition of the motion trajectory, motion envelope, or Dice Similarity Coefficient (DSC) conformality. Two factor interactions were found to be statistically significant for the DSC for the interactions of 1) object volume and the HU threshold used for delineation and 2) the object velocity and object volume. Moreover, a statistically significant single factor direct proportionality was observed between the maximum velocity and the mean tracking error. In this work multiple factors impacting on the uncertainty in 4D data sets have been considered and some statistically significant two-factor interactions have been identified. Therefore, the detailed evaluation of errors and uncertainties in 4D imaging modalities is recommended in order to assess the clinical implications of interaction among the various uncertainty factors.