Volume averaging limitations of computed tomography

Abstract

A simple volume averaging model was shown to be inadequate for predicting effective computed tomographic attenuation values of mixtures of bone or air with soft tissue. Computed tomographic values derived from scanning stratified sub-slice thickness objects were shown theoretically and experimentally to have a non-linear dependency on relative fractional content and a surprisingly large dependency on spatial extent of the stratified substances. The mode was applied to the problem of computed tomographic values in thin, flat structures such as pools of blood in the subarachnoid space. The results show only small deviations from simple volume averaging theory for layers of low contrast substances such as blood and soft tissue, but potentially large deviations for layers of substances with high contrast differences such as bone, air, and tissue. This phenomenon explains certain artifacts and demonstrates rather fundamental problems in the accuracy of analytic reconstruction techniques. It may justify postprocessing correction or iterative approaches.