Fast, fully automated MRI-based measurement of the human corpus callosum

Timothy J Herron (US Veterans Affairs, Neurology), Xiaojian Kang (UC Davis, Neurology & Neuroscience), And U Turken (US Veterans Affairs Research), David L Woods (US Veterans Affairs & UC Davis)

We present a stand-alone open-source MatLab software package, C8, which measures midsagittal cross-sectional thickness and area of the human corpus callosum from high-resolution T1 in vivo MR images. C8 quickly extracts and measures the callosum using segmented, affine normalized brains as input. Areas are computed for geometrically defined callosal compartments according to several different extant compartment schemes. Thickness is sampled along a median line within the callosum according to three methods of spacing the samples: equal angle, equal-distance, and equal-area. Coregistered accessory input images have values sampled along the median line and near the edges of the callosum. The algorithm’s performance was tested using the high-quality OASIS image database [] and showed C8 producing consistent, reliable callosal measurement estimates over repeated scans and with values comparable to manually-segmented callosa.

We used 1000+ medium quality T1 images from the 1000 Functional Connectomes Project [] to evaluate between group differences in measured callosum values and regional effects of age and gender on thickness measures. We also sampled normalized T1 values along the callosum to capture regional variation.

A second analysis used the OASIS database young normals plus our own high-quality set of T1 images from young normals (~290 subjects total) in order to correlate callosal thickness/area with regional cortical gray matter thickness and area in 34 cortical ROIs defined using FreeSurfer []. We found that regional callosal area (and callosal thickness to a lesser extent), correlate positively with regional cortical area but not with regional cortical thickness.
Fast, fully automated MRI-based measurement of the human corpus callosum
A: 50 Thickness values (standard deviation error bars) for 1039 subjects taken from the 1000 Functional Connectomes Project database. B&C: Partial Spearman correlations of regional callosal thickness with age (B) and being female (C) over the FCP data. D: Spearman correlations of Hofer-Frahm (HF) callosum partition [NeuroImage, 32(3), 2006] area measurements with Freesurfer derived thickness and area measurements over lobes for 292 young, normal subjects.
Preferred presentation format: Poster
Topic: Neuroimaging

Document Actions