Kathleen A. Garrison PhD, Presenter: Nothing to Disclose

Kevin Chikai Ma BS, Abstract Co-Author: Nothing to Disclose

Corianne Rogalsky PhD, Abstract Co-Author: Nothing to Disclose

Hanna Damasio MD, Abstract Co-Author: Nothing to Disclose

Carolee J. Winstein PhD, Abstract Co-Author: Nothing to Disclose

Lisa S. Aziz-Zadeh PhD, Abstract Co-Author: Nothing to Disclose

Studies in stroke patients often use functional MRI to assess changes in brain function after stroke and relate these changes to behavioral information, to improve clinical outcomes and reduce disability. A key step in group fMRI analysis is spatial normalization, or image registration to a standard template. After stroke, structural brain lesions and brain volume loss may result in a lack of correspondence between the patient and standard template images, confounding the normalization process. Suboptimal normalization may lead to improper alignment and decreased sensitivity within a group, or false attribution of differences between groups to functional not structural anatomy. Spatial normalization is important to lesion analysis, and necessary for comparisons of fMRI data with other patients or healthy subjects. In this study, we evaluate the effects of spatial normalization on fMRI region of interest (ROI) analysis in stroke patients. 

We use a recent protocol by Crinion (2007) to evaluate the effects of an automated method for spatial normalization (Ashburner 2005) on results from an fMRI study on stroke. We compare results from ROI analyses between normalized data, with ROI defined by reference to a template atlas (MNI; Evans, 1993); and data analyzed using individual patient structural anatomy, with ROI defined manually based on neuroanatomical landmarks using the principles described by Allen (2005, 2002).

We find that spatial normalization impacts fMRI ROI analyses within-subjects, e.g. measures of functional signal strength from a given region for a patient may vary if the region is defined in normalized versus native anatomy. However, between-subjects comparisons for this fMRI study, for example, patients versus healthy subjects, are consistent between methods.

These results support the use of spatial normalization in fMRI studies of stroke patients. Automated spatial normalization reduces inter-user variability, and requires less time and labor, as compared to hand-tracing ROI. However, caution is recommended when interpreting ROI analyses that employ spatial normalization, specifically with regard to anatomical localization of functional results.

fMRI after stroke provides information that is important to neurorehabilitation. A common approach in fMRI is ROI analysis. Successful spatial normalization is a key step in group fMRI analysis.

Garrison, K, Ma, K, Rogalsky, C, Damasio, H, Winstein, C, Aziz-Zadeh, L, Spatial Normalization of Lesioned Brains: Impact on fMRI Region of Interest Analyses.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11012680.html