Automatic localization of AC and PC landmarks in T2-weighted MR volumetric neuroimages

Abstract

A novel method is presented for automatic identification of the anterior commissure (AC) and posterior commissure (PC) in T2-weighted MR volumetric images (MRI). AC and PC are two critical landmarks of human brain. It is important to accurately identify them for brain segmentation, registration, functional neurosurgery, human brain mapping, and particularly for the Talairach transformation. The algorithm of identifying them consists of four steps: (1) Identify the corpus callosum and select the region of interest; (2) Identify the fornix and locate the initial AC; (3) Segment part of the third ventricle and locate the initial PC; (4) Refine the positions of AC and PC. The algorithm has been validated quantitatively with 20 T2-weighted MRI data sets. The mean errors of the localization of them were 1.50 mm (AC) and 1.60 mm (PC). It took about 15 seconds for identifying them on P4 3.0 GHz. This fully automatic algorithm is potentially useful in clinic and for research.