A hybrid projection decomposition algorithm for dual-energy computed tomography

Abstract

Dual Energy Computed Tomography (DECT) has recently gained significantly research interest because of its ability of object separation, contrast enhancement, artifact reduction and material composition assessment. Aiming for the nonlinear projection decomposition problem in DECT image reconstruction, a hybrid projection decomposition algorithm is proposed with combination of iterative method and isotransmission fitting method. Firstly, the nonlinear projection aquation for a given polychromatic projection value is transformed to linear equation plus a nonlinear pertubative item for single variable to speed up the iteartion speed and avoid non-convergence. Secondly, a linear approximation fitting model of iso-transmission line is built for high and low energy projection with iso-transmission point pairs solved above. Then the high and low energy projection joint nonlinear solving problem is transformed into solving high and low energy projection iso-transmission line intersection problem, so as to realize the fast decomposition of dual energy projection. Finally, local searching near the intersection of projection matching is used to compensate the error of linear approximation fitting to get more accurate decomposition result. The accuracy and efficiency of the proposed method are demonstrated by numerical simulation.