Deformable multimodal registration with gradient orientation based on structure tensors

Deformable registration of images obtained from different modalities remains a challenging task in medical image analysis. This paper addresses this problem and proposes a new similarity metric for multi-modal registration. The metric is based on the orientation of image gradients. In contrast to earlier work, our metric is based on the structure tensor of the spatial image gradients and can robustly estimate the local threedimensional orientation of image features. The measure has been implemented in a non-rigid diffusion-regularized registration framework and is optimized using Gauss-Newton. It has been applied to align CT breathing cycle scans with simulated contrast uptake and challenging clinical MRI and CT chest scans. Experimental results demonstrate its advantages over the classical gradient based orientation measure and the most commonly used multimodal similarity metric — mutual information, in terms of improved alignment of anatomical landmarks.