22. European Stroke Conference 245 Brain imaging The role of measurement errors in quantitative brain MRI for epidemiologic studies H. Wersching1, A. Teuber2, H. Kugel3, M. Deppe4, J. Minnerup5, K. Berger6, A. Kemmling7 University of Muenster, Muenster, GERMANY1, University of Muenster, Muenster, GERMA-NY2, University of Muenster, Muenster, GERMANY3, University of Muenster, Muenster, GERMA-NY4, University of Muenster, Muenster, GERMANY5, University of Muenster, Muenster, GERMA-NY6, University of Hamburg, Hamburg, GERMANY7 Background Brain volumes are imaging outcomes increasingly assessed in large scale epidemiologic studies. Sources of error occur at different levels of data acquisition, processing and analysis. Knowledge about the extent, the classification (random versus systematic) and the necessity of these errors is of particular interest in the planning of multisite epidemiologic studies. Methods Eight healthy middle-aged subjects were measured twice in the same 3 T scanner (Philips Intera) to quantify intra-scanner variability using the same imaging protocol (3D-T1w, FLAIR). Twelve sub-jects were measured triply in a 1.5 T Intera, a 3 T Intera and a 3 T Trio (Siemens) scanner to evalu-ate inter-scanner variability. All T1 images were segmented automatically using the VBM8 toolbox r435 for SPM8 to obtain gray matter, white matter and cerebrospinal fluid (CSF) volumes for the whole brain and for the frontal, temporal, parietal and occipital lobe separately. For comparison of segmentation algorithms 40 images were additionally segmented with FSL v5.0. Results The intra-scanner comparison showed a non-systematic difference of global and regional white, gray and CSF volumes of <1%. For inter-scanner variability we found a systematic difference towards higher brain parenchyma volumes and lower CSF volumes across all brain regions in the Trio com-pared to the Intera 3 T scanner. Post hoc analyses revealed failed segmentation of 1.5 T data if de-fault settings are kept. SPM and FSL segmentations non-systematically varied in volume measures of up to 14%. Conclusion Segmentation is comparable across repeated measurements and different 3 T scanners, but not seg-mentation programs. When using 1.5 T scanners specific adjustments of MR protocols and/ or image processing algorithms must be taken into account. 416 © 2013 S. Karger AG, Basel Scientific Programme 246 Brain imaging Probabilistic mapping the location of white matter hyperintensity and subcortical infarcts T.G. PHAN1, J. CHEN2, R. Beare3, J. Ly4, H. Ma5, B. Clissold6, V. Srikanth7 STROKE AND AGING RESEARCH, CLAYTON, AUSTRALIA1, STROKE AND AGING RE-SEARCH, CLAYTON, AUSTRALIA2, STROKE AND AGING RESEARCH, CLAYTON, AUS-TRALIA3, STROKE AND AGING RESEARCH, CLAYTON, AUSTRALIA4, STROKE AND AG-ING RESEARCH, CLAYTON, AUSTRALIA5, STROKE AND AGING RESEARCH, CLAYTON, AUSTRALIA6, STROKE AND AGING RESEARCH, CLAYTON, AUSTRALIA7 Background & aims: Differentiating between white matter hyperintensity (WMH) and stroke can be rather difficult if stroke patients are imaged in the chronic phase. We hypothesise that knowledge of the topographic distribution of WMH and subcortical stroke may help with differentiating these two entities. Methods: The map of the WMH was created from a cohort of community dwelling healthy elderly subjects. Patients with subcortical infarcts on magnetic resonance imaging (MRI) admitted to our institution between 2009 and 2011 were included. These images were aligned to a common stereo-tactic coordinate to facilitate comparison. The disease frequency at each voxel for each group was calculated using: f1 = m1/n1 for WMH and f2 = m2/n2 for subcortical stroke. Where f is frequency, m is the number of disease samples at a location and n is the total number samples in the group. The intersection, or overlap, of the non-zero components of the frequency map of each group was com-puted. These frequency images were then nomalized as p1=f1/(f1+f2) and p2 =f2/(f1+f2). Results: There were 384 subjects (213 males) in TASCOG group and 57 (33 males) subjects in stroke group. The mean age of TASCOG group is 72.1 (SD 7.0) and of stroke group is 64.3 (SD 14.4). WMH predominate around the poles of the lateral ventricles while lesions adjacent to the body of the lateral ventricles and deep grey matter nuclei have a higher probability of being infarcts. Conclusion; The probabilistic map can be used to help differentiating between WMH and subcorti-cal strokes.
Karger_ESC London_2013
To see the actual publication please follow the link above