Currently I am doing shape analysis of white matter tracts in human brain. The data I am analyzing are fractional anisotropy (FA) values of corpus callusom (CC) region extracting from the DTI experiment. The FA value of CC is illustrated as follows:
The corpus callusom can be visualized as a two dimensional manifold lies in a three dimensional space, which is shown in the left panel and mid-right panel above. We developed a principal surface algorithm to extract the main feature of it. The principal surface algorithm finds the center surface that pass through a target data cloud. Some simulation results are shown below:
We illustrate the surface fitting of CC and the flattened surface as follows. The surface on the left is the fitted principal surface of the CC and the flattened one is illustrated on the right, with the average projected FA value on it. Notice that the flattened CCs have the same dimension across all the subjects, so that we could run voxel based analysis based on this.
Another research project involves quantifying the reproducibility of graphical data. Look at the following two groups of graphs. The left one shows a good reproducibility while the right panel illustrate poor reproducibility. Our objective is to quantify the reproducibility. We calculate the intra-class correlation coefficient for repeated graphical measurement for such purpose.
