An Unbiased Iterative Group Registration Template for Cortical Surface Analysis

Accurate alignment of explicit surface representations of human cerebral cortices is necessary in order to compare local individual differences in cortical morphometric measurements (thickness, surface area, gyrification, etc.) in both normal and clinical populations. This paper presents a methodology for developing unbiased, high resolution iterative registration templates from a group of 222 subject hemispheres and shows that the resulting template provides better alignment of a separate set of test data than single-subject templates. It demonstrates that between 30 and 50 subjects are required to generate a stable iterative template. It also explores the way in which fold variants in registration templates affect the quality of registration. Finally, it shows that hemisphere-specific group registration templates systematically better register subject hemispheres of the same laterality, underlining the need to develop templates free of hemisphere bias for asymmetry analysis.