Probabilistic Atlas Construction of Human Cardiac Fiber Structure in DT-MRI

We investigate the problem of constructing statistical atlas of human cardiac fiber structure from limited number of datasets. We introduced a Parzen-Gaussian model to construct the probabilistic atlas of human cardiac fiber structure due to the fact that it can generate a general atlas using very small number of samples. Experimental results showed that atlas using Parzen-Gaussian model can avoid FA/MD collapse, which can be introduced by atlas from registered DW images and registered tensor fields.