Distortion correction using mutual information in 3D diffusion tensor microscopy

Magnetic resonance diffusion tensor imaging (MR-DTI) is often complicated by distortion in the diffusion-sensitizing directions due to the large diffusion gradients. In this work we have corrected for this distortion using mutual information as the cost function, and applied this technique to real 3D diffusion data. We evaluated the performance of various cost functions, along with the effect of different interpolation techniques on convergence to the global maximum. Taking the advantage of the Fourier nature of the data we suggest its use in performing shifts and shears. With these corrections we observed marked improvement in the estimation of diffusion tensor maps, supporting its use in 3D diffusion tensor imaging.