Model-Based Automatic Functional Area Labeling for NTU Standard Brain Template

Most data analyses were developed based on the standard brain template in the field of neuroimaging research, because of the complexity and variability of human brain. For figuring out the relationship between brain function and structure, Talairach set a three-dimensional coordinate system, cooperating with the postmortem brain autopsy, and thus a worldwide adopted standard brain atlas was created.
For the purpose of anatomic representation of general population, Montreal Neurologic Institute (MNI) collected 305 brains MRI image, and adopted image processing methods such as realignment and transformation to create a population-based standard brain template by averaging three hundreds brain dataset and this process became standard method for building template. The MNI also created ICBM_152 and colin27 which both are worldwide adopted nowadays.
However the difference of brain structures between western and eastern people was observed in past experiment [1]. The mismatch issue may lead to bias or inappropriate interpretation in neurocognitive studies. In this thesis, NTU Medical Image Lab recruited ninety-five subjects and developed the National Taiwan University standard brain template (NTU template)[1] based on the process of MNI template. But NTU template lacked of functional area information, which can’t be used on fMRI study directly. To solve this problem, we established an optical flow nonlinear registration algorithm, spatial normalizing ICBM_152 to NTU template and mapped ICBM_152’s functional area information to NTU template. Compared with other fMRI software’s registration function, we found that the method we used can provide better registration result.
Using NTU template in fMRI data analyses can get more precise localization in fMRI experiments. After compared functional area distribution between NTU template and ICBM_152, we found that distribution exist different between eastern and western，which caused by structure’s nonlinear deformation. In the future, we will work on higher resolution template, which may benefit in neuroscience study, human brain mapping and clinical applications.