Deformed CT reconstruction from limited projection data

In a sequence of CT studies of the same patient, the anatomy can deform but much of the anatomical information remains consistent from one study to the next. Therefore, once an initial CT has been made, it should be possible to approximately reconstruct subsequent 3D representations of the anatomy from less than a full set of CT sinogram data. We present a scheme to do this. A parameterized deformation model is applied to the initial CT. Two or more projection radiographs are then acquired of the anatomy after it has changed shape. Corresponding synthetic radiographs are computed from the initial CT while it is iteratively deformed by the model. The model parameters are adjusted until the synthetic radiographs optimally resemble the actual projections. We show that the deformation model that yields matching real and synthetic projections accurately describes the actual deformation of the anatomy imaged in the radiographs.