Abstract:

Acquisition of brain magnetic resonance imaging (MRI) data usually starts with a localizer for properly positioning the field of view based on a prior knowledge of brain anatomy and setting corresponding localization parameters for subsequent scans. We propose an automatic localization method that references directly to the brain atlas. The procedure first quickly acquires a 3D localization image at a median spatial resolution, and then calculates its registration parameter to the atlas and uses these parameters to position the subsequent scans, which therefore ensures the scanning configurations for different subjects are consistent with the atlas. The proposed method benefits inter-subject comparisons and referencing, in that it can help investigators locating abnormal structure, tumors or other regions-of-interest more quickly and easily, and therefore using the data in voxel based analysis more efficiently. We also propose an iterative method for automatic localizing individual subject in multiple independent follow-up scans. By iterating “scan, registration, automatic localization” steps several passes, it progressively minimizes the error of the image registration algorithm. Experiments showed that our atlas-based automatic localization method achieved high consistency of spatial location both in imaging data acquired from different subjects and in multiple separate scans from a single subject, and the localization error between multiple scans of a single subject was less than 1.0 mm and 1.0 degree.

Key words: MRI, automatic localization, registration, atlas, voxel based analysis