Bladder wall motion compensation using a new information-theoretic measure

In this paper we introduce a novel non-rigid registration algorithm using an information-theoretic measure, i.e., α-information, which is derived from the α-divergence. The tunable nature of the a -information provides more freedom in choosing the metrics by adjusting the α value, which may help improve registration accuracy. The presented method has been applied to the non-rigid registration of bladder T₁-weighted magnetic resonance images from both volunteer and patient studies. Experimental results demonstrated that the presented registration method can effectively capture the bladder wall motion. Furthermore, better registration results were found compared to the results of mutual information-based registration. The findings suggested that the use of α-information has potential to yield more accurate non-rigid registrations than the well-known mutual information approach.