Prospective multi-centre Voxel Based Morphometry study employing scanner specific segmentations: Procedure development using CaliBrain structural MRI data
1 The Division of Psychiatry, Centre for Clinical Brain Sciences (CCBS), School of Molecular and Clinical Medicine, University of Edinburgh, Edinburgh, UK
2 Aberdeen Biomedical Imaging Centre, Division of Applied Medicine University of Aberdeen, Aberdeen, UK
3 The Department of Clinical Physics and Bioengineering, NHS Greater Glasgow South University Hospitals Division, Glasgow, UK
4 SFC Brain Imaging Research Centre, SINAPSE Collaboration http://www.sinapse.ac.uk, Division of Clinical Neurosciences, University of Edinburgh, Western General Hospital, Edinburgh, UK
5 Sackler Institute of Psychological Research, Faculty of Medicine, University of Glasgow, Glasgow, UK
6 Centre for Neuroscience, Division of Medical Sciences, University of Dundee, Dundee, UK
BMC Medical Imaging 2009, 9:8 doi:10.1186/1471-2342-9-8Published: 15 May 2009
Structural Magnetic Resonance Imaging (sMRI) of the brain is employed in the assessment of a wide range of neuropsychiatric disorders. In order to improve statistical power in such studies it is desirable to pool scanning resources from multiple centres. The CaliBrain project was designed to provide for an assessment of scanner differences at three centres in Scotland, and to assess the practicality of pooling scans from multiple-centres.
We scanned healthy subjects twice on each of the 3 scanners in the CaliBrain project with T1-weighted sequences. The tissue classifier supplied within the Statistical Parametric Mapping (SPM5) application was used to map the grey and white tissue for each scan. We were thus able to assess within scanner variability and between scanner differences. We have sought to correct for between scanner differences by adjusting the probability mappings of tissue occupancy (tissue priors) used in SPM5 for tissue classification. The adjustment procedure resulted in separate sets of tissue priors being developed for each scanner and we refer to these as scanner specific priors.
Voxel Based Morphometry (VBM) analyses and metric tests indicated that the use of scanner specific priors reduced tissue classification differences between scanners. However, the metric results also demonstrated that the between scanner differences were not reduced to the level of within scanner variability, the ideal for scanner harmonisation.
Our results indicate the development of scanner specific priors for SPM can assist in pooling of scan resources from different research centres. This can facilitate improvements in the statistical power of quantitative brain imaging studies.