Minimizing sources of errors in medical image registration

This paper compares the image registration errors associated with traditional unidirectional registration algorithms to the newer inverse consistent image registration algorithms. Inverse consistent registration algorithms jointly estimate the forward and reverse transformation between two images while minimizing the inverse consistency error between these transformations. Unidirectional algorithms estimate the forward and the reverse transformations independently. It is also shown that combining landmark and intensity information produces better registration results than using either type of information independently. In addition to examining the inverse consistency error, the transitivity error associated with each type of registration algorithm is examined. Transitivity of the transformations was evaluated by measuring the difference between the identity mapping and the composition the transformations from image A-to-B, B-to-C, and C-to-A. In all cases, the inverse consistent algorithms out performed the unidirectional algorithms by producing transformations that have less inverse consistency error and less transitivity error.