Atlas-Powered Deep Learning (ADL) - Application to Diffusion Weighted MRI

Deep learning has a great potential for estimating biomarkers in diffusion weighted magnetic resonance imaging (dMRI). Atlases, on the other hand, are unique tool modeling spatio-temporal variability of biomarkers. In this paper, we propose first framework to exploit both deep and atlases biomarker estimation dMRI. Our relies non-linear tensor registration compute estimate atlas reliability maps. We also use align subject error alignment. atlas, map, alignment addition dMRI signal, as inputs model estimation. our fractional anisotropy neurite orientation dispersion from down-sampled data test cohort 70 newborn subjects. Results show that method significantly outperforms standard methods well recent techniques. is more robust higher measurement down-sampling factors. study shows advantages can be synergistically combined achieve unprecedented accuracy