Image interpolation in 4D CT using a BSpline deformable registration model

Purpose: To develop a method for deriving the phase-binned four-dimensional computed tomography (4D CT) image sets through interpolation of the images acquired at some known phases. Methods and Materials: Four-dimensional computed tomography data sets for 3 patients were acquired. For each patient, the correlation between inhale and exhale phases was studied and quantified using a BSpline deformable model. Images at an arbitrary phase were deduced by an interpolation of the deformation coefficients. The accuracy of the proposed scheme was assessed by comparing marker trajectories and by checkerboard/difference display of the interpolated and acquired images. Results: The images at intermediate phases could be derived by an interpolation of the deformation field. An analysis of marker movements indicated that 3 mm accuracy is achievable by the interpolation. The subtraction of image analysis indicated a similar level of success. The proposed technique was useful also for automatically mapping the organ contours in a known phase to other phases, and for designing patient-specific margins in the presence of respiratory motion. Finally, the technique led to a 90% reduction in the acquired data, because in the BSpline model, a lattice of only a few thousand values is sufficient to describe a CT data set of 25 million pixels. Conclusions: Organ deformation can be well modeled by using a BSpline model. The proposed technique may offer useful means for radiation dose reduction, binning artifacts removal, and disk storage improvement in 4D imaging.