Poster D80, Monday, March 26, 8:00-10:00 am, Exhibit Hall C
Localized test-retest reliability of fMRI task activity
Johan Jansma1, Geert-Jan Rutten1; 1ETZ Elisabeth Hospital, Department of Neurosurgery, Tilburg, the Netherlands
Introduction. Scientific applications of fMRI tend to focus on finding regions that show a significant difference. However, this approach may not always be optimal for clinical use of fMRI. Information about absence of a change in activation during a task can be as valuable to a clinician as a high increase or decrease. Furthermore, clinicians are typically interested in activation in small regions of interest. In this study we used a correlation method to estimate the pattern reproducibility of a well-known and often applied presurgical language protocol. Method. 17 subjects performed a verb generation protocol twice with a seven week apart. A GLM was performed to calculate percentage signal change per voxel, at unsmoothed resolution. We calculated a voxel by voxel correlation as well as intra class correlation value (ICC, consistence) within subject as well as with a apttern average over subjects, in a small 2x2x2cm region of interest (64 voxels)around Broca. Results. Subjects showed a within subject correlation value of r = 0.84 (sem: 0.1) and a ICC of 0.79 (0.03)Correlation with an average pattern was -0.03 (0.08). Discussion. Test-retest fMRI pattern reproducibility in a ROI with a relevant size for presurgical applications appears to be sufficient to produce valuable clinical information. In addition, single subject fMRI patterns in small ROIs appear to be unique, and cannot be reliably estimated using data from other subjects. These results support more extensive use of fMRI for clinical applications.
Topic Area: METHODS: Neuroimaging