Enhancement and Intensity Inhomogeneity Correction of Diffusion-Weighted MR Images of Neonatal and Infantile Brain Using Dynamic Stochastic Resonance

Abstract

Imaging of infantile/neonatal brain mandates tailored radio frequency coils (RF coils) to achieve a homogeneous field over a small region of interest (ROI). Most centers, however, perform pediatric imaging using adult RF coils only as procurement of tailored pediatric coils might prove quite expensive. This practice may not scientifically justified, whereas the image post-processing techniques reduces the deleterious effects of magnetic field inhomogeneity due to a small ROI being scanned in a large RF coil. Further, the eccentric placement of ROI within the RF coil perpetuates the field inhomogeneity within the scanned region. Hence, the structures closer to the coil appear brighter than those farther away giving rise to a 'shading artefact'. The effect even more accentuates in weak signal sequences like diffusion-weighted imaging (DWI). The proposed method significantly removes shading artefact of real DWI and synthetic T1 and T2 weighted magnetic resonance images. Dynamic stochastic resonance (DSR) intelligently uses the coefficient of discrete cosine transform of an image for brightness normalization and image enhancement simultaneously. The quality of the output image depends on the bistability parameters associated with the dynamic equation. Particle swarm optimization (PSO) tunes these bistability parameters for the entropy minimization of different group of tissues individually. The proposed algorithm outperforms the post processing based homomorphic filtering, local entropy minimization with spline model and multiplicative intrinsic component optimization methods. The proposed PSO based DSR approach may be helpful in accurate diagnosis.

http://ift.tt/2sxvP7f