Multimodal mri template creation in the ring-tailed lemur and rhesus macaque

We present a multimodal registration algorithm for simultaneous alignment of datasets with both scalar and tensor MRI images. We employ a volumetric, cubic B-spline parametrised transformation model. Regularisation is based on the logarithm of the singular values of the local Jacobian and ensures diffeomorphic warps. Tensor registration takes reorientation into account during optimisation, through a finite-strain approximation of rotation due to the warp. The combination of scalar, tensor and regularisation cost functions allows us to optimise the deformations in terms of tissue matching, orientation matching and distortion minimisation simultaneously. We apply our method to creating multimodal T2 and DTI MRI brain templates of two small primates (the ring-tailed lemur and rhesus macaque) from high-quality, ex vivo, 0.5/0.6 mm isotropic data. The resulting templates are of very high quality across both modalities and species. Tissue contrast in the T2 channel is high indicating excellent tissue-boundary alignment. The DTI channel displays strong anisotropy in white matter, as well as consistent left/right orientation information even in relatively isotropic grey matter regions. Finally, we demonstrate where the multimodal templating approach overcomes anatomical inconsistencies introduced by unimodal only methods.