Nonrigid multimodal medical image registration using features extracted from the monogenic signal

Multimodal medical image registration is a challenging task as anatomical structures might have different appearances depending on the physical principles that are the foundation of the imaging modality. The task is even more difficult to perform when ultrasound (US) images are used in the registration process because of their low resolution and low signal to noise ratio (SNR). We present a multi-scale method whose aim is to automatically register ultrasound and magnetic resonance (MR) images using features extracted from the monogenic signal and local phase-based images. We use an elastic approach based on free-form deformation (FFD) and mutual information (MI) as similarity measure. Finally we compare our results with pure local phase-based image registration techniques.