Abstract
Deformable image registration is a very important field of research inmedical imaging. Recently multiple deep learning approaches were published inthis area showing promising results. However, drawbacks of deep learningmethods are the need for a large amount of training datasets and theirinability to register unseen images different from the training datasets. Oneshot learning comes without the need of large training datasets and has alreadybeen proven to be applicable to 3D data. In this work we present an one shotregistration approach for periodic motion tracking in 3D and 4D datasets. Whenapplied to 3D dataset the algorithm calculates the inverse of a registrationvector field simultaneously. For registration we employed a U-Net combined witha coarse to fine approach and a differential spatial transformer module. Thealgorithm was thoroughly tested with multiple 4D and 3D datasets publiclyavailable. The results show that the presented approach is able to trackperiodic motion and to yield a competitive registration accuracy. Possibleapplication are the use as a stand-alone algorithm for 3D and 4D motiontracking or in the beginning of studies until enough datasets for a separatetraining phase are available.