Non-rigid motion compensation in free-breathing myocardial perfusion magnetic resonance imaging

Breathing movements during the image acquisition of first-pass gadolinium enhanced, magnetic resonance imaging (MRI) hinder a direct automatic analysis of the myocardial perfusion. In addition, a qualitative readout by visual tracking is also more difficult as well. Non-rigid registration can be used to compensate for these movements in the image series. Because of the local contrast and intensity change over time, the registration method needs to be chosen carefully. We propose to make use of the periodicity of the breathing movement when patients are allowed to breath freely during image acquisition. Specifically, we propose to first identify a subset of the images that corresponds to the same phase of the breathing cycle and register these to compensate for the residual differences. By using a combination of normalised gradient fields and the sum of squared differences we circumvent the problems arising from the change of intensity. Then, for each of the remaining images, reference images of a similar intensity distribution are created by a linear combination of images from the align subset. In the last step, registration is achieved by minimising the sum of squared differences. Our first experiments show that this approach is well suited to compensate for the breathing movements.